Protein and energy supplementation in elderly people at risk from malnutrition

Abstract

Background

Evidence for the effectiveness of nutritional supplements containing protein and energy, often prescribed for older people, is limited. Malnutrition is more common in this age group and deterioration of nutritional status can occur during illness. It is important to establish whether supplementing the diet is an effective way of improving outcomes for older people at risk from malnutrition.

Objectives

This review examined trials for improvement in nutritional status and clinical outcomes when extra protein and energy were provided, usually as commercial 'sip‐feeds'.

Search methods

We searched The Cochrane Library, MEDLINE, EMBASE, Healthstar, CINAHL, BIOSIS, CAB abstracts. We also hand searched nutrition journals and reference lists and contacted 'sip‐feed' manufacturers.

Selection criteria

Randomised and quasi‐randomised controlled trials of oral protein and energy supplementation in older people, with the exception of groups recovering from cancer treatment or in critical care.

Data collection and analysis

Two reviewers independently assessed trials prior to inclusion and independently extracted data and assessed trial quality. Authors of trials were contacted for further information as necessary.

Main results

Sixty‐two trials with 10,187 randomised participants have been included in the review. Maximum duration of intervention was 18 months. Most included trials had poor study quality. The pooled weighted mean difference (WMD) for percentage weight change showed a benefit of supplementation of 2.2% (95% confidence interval (CI) 1.8 to 2.5) from 42 trials. There was no significant reduction in mortality in the supplemented compared with control groups (relative risk (RR) 0.92, CI 0.81 to 1.04) from 42 trials. Mortality results were statistically significant when limited to trials in which participants (N = 2461) were defined as undernourished (RR 0.79, 95% CI 0.64 to 0.97).

The risk of complications was reduced in 24 trials (RR 0.86, 95% CI 0.75 to 0.99). Few trials were able to suggest any functional benefit from supplementation. The WMD for length of stay from 12 trials also showed no statistically significant effect (‐0.8 days, 95% CI ‐2.8 to 1.3). Adverse effects included nausea or diarrhoea.

Authors' conclusions

Supplementation produces a small but consistent weight gain in older people. Mortality may be reduced in older people who are undernourished. There may also be a beneficial effect on complications which needs to be confirmed. However, this updated review found no evidence of improvement in functional benefit or reduction in length of hospital stay with supplements. Additional data from large‐scale multi‐centre trials are still required.

Plain language summary

Protein and energy supplementation in elderly people at risk from malnutrition

Much emphasis is placed on the importance of good diet, usually in relation to concern about the health risks of obesity. However it has been generally agreed that the risk of undernutrition rather than overnutrition is the main cause for concern in elderly people, particularly those who are hospitalised or institutionalised. Malnutrition has been shown to have important effects on recovery in a broad range of patients and conditions. It has been associated strongly with impaired immune response, impaired muscle and respiratory function, delayed wound healing, overall increased complications, longer rehabilitation, greater length of hospital stay and increased mortality. Oral protein and energy supplements are potentially safer and easier to administer than nasogastric enteral feeds and are therefore particularly suited to elderly people and are also widely used. However, there may be problems with the willingness and ability of older people to consume oral supplements, and supplements may not be used effectively. Even if supplements are prescribed, they may not always be given, or are given but not consumed. In addition to taste, the composition and timing of administration in relation to meals may be important. Efforts also need to be made to provide normal meals and snacks which meet the needs of elderly people and to provide assistance with feeding if required.

A total of 10,187 randomised participants from the 62 trials has been included. Maximum duration of intervention was 18 months. The reviewers suggest that supplementation appears to produce a small but consistent weight gain. There was no evidence in this updated review of a beneficial effect on mortality overall, but there may be a beneficial effect on mortality in people who are undernourished. Supplementation may also reduce the number of complications. The reported acceptance of supplements was variable between trials. Some adverse effects such as nausea or diarrhoea were reported. However, there were problems of study design and quality. More studies are required to confirm the beneficial effect on the number of complications, to establish whether there is a beneficial effect on mortality for undernourished elderly people and to provide evidence about whether protein and energy supplements can improve morbidity and functional status in frail older people.

Background

Description of the condition

Much emphasis is placed on the importance of good diet, usually in relation to concern about the health risks of obesity. However it has been generally agreed that the risk of undernutrition rather than overnutrition is the main cause for concern in elderly people, particularly those who are hospitalised or institutionalised (DoH 1992; Potter 1988). There have been recent UK and international initiatives to improve practice in this area of health care (Council Europe 2002; NHSQIS 2003; NICE 2006).

There is no universally accepted clinical definition of malnutrition or undernutrition (the terms are used interchangeably here). However, Allison 2000 defined undernutrition as "a state of energy, protein or other specific nutrient deficiency which produces a measurable change in body function and is associated with worse outcome from illness as well as being specifically reversed by nutritional support".

Increased length of hospital stay is associated with malnutrition. Studies have reported the presence of malnutrition in a substantial proportion of hospital patients both on admission and during hospital stay in the USA, Norway, Ireland, UK, Sweden, The Netherlands and Australia (Bistrian 1976; Bruun 1999; Corish 2000; Edington 2000; Flodin 2000; Kruizenga 2003; Zador 1987). This is a particular problem for elderly people as a) over 40% of hospital admissions are elderly people who have longer periods of illness and longer hospital stay (HCUP 2002), and b) data show that elderly patients are more at risk of malnutrition than others (Gallager‐Allred 1996; Kruizenga 2003; McWhirter 1994).

Reasons for poor nutritional status in older people are multi‐faceted and include the physiological, psychological and social changes associated with aging which affect food intake and body weight, exacerbated by the presence of illness. Five to ten percent of elderly people in the community may also be malnourished (McCormack 1997). Elderly people who are already malnourished at home may be at a disadvantage if admitted to hospital for treatment. Within Europe and the USA, nutritional status has been shown to decline with hospital stay due to a lack of adequate nutritional intake during hospitalisation (Corish 2000; Larsson 1990; McWhirter 1994; Sullivan 1999). This is because the poor nutritional state of many patients often goes unrecognised and there may be a lack of awareness of malnutrition by health professionals who receive little training on nutritional issues (Elia 2001). Disease or treatment such as surgery may also increase nutritional demands, so patients who have a poor appetite or difficulty eating will lose weight.

Measurement of nutritional status

Commonly used methods to measure nutritional status are body mass index (BMI) (weight in kg / height in m²), anthropometry such as triceps skin fold thickness and arm muscle circumference, and history of recent weight loss. Serum albumin has also been used as a measure of nutritional status as malnutrition causes a decrease in the rate of synthesis of albumin, but serum albumin levels are known to be affected by changes in fluid balance and illness itself. Illness, injury and age may all therefore confound the measurement of nutritional status. Improved tools for nutrition risk screening have been developed, which include subjective measures of recent weight loss and inadequate intake (BAPEN 2003). The difficulty of defining and measuring nutritional status may help explain some of the wide variation in the reported prevalence of malnutrition in hospitalised adults of between 11% and 40% (Corish 2000).

Effects of malnutrition

Malnutrition has been shown to have important effects on recovery in a broad range of patients and conditions. It impacts on both physiological and biochemical systems and has been associated strongly with impaired immune response, impaired muscle and respiratory function, delayed wound healing, overall increased complications, longer rehabilitation, greater length of hospital stay and increased mortality (Kelly 1984; Potter 1995; Robinson 1987; Sullivan 1990; Windsor 1988). Apathy, depression, fatigue and a loss of will to recover have been demonstrated following weight loss in experimental volunteers (Keys 1950).

Costs

The economic consequences of malnutrition are also considerable. In 1992 the economic cost to the United Kingdom National Health Service of preventable malnutrition was estimated to be £266 (297 EUR, 2009 currency translation) million a year, mainly due to increased length of bed occupancy and associated treatment costs (Lennard‐Jones 1992). More recently, it has been estimated that the annual additional health care cost of malnutrition and associated disease is over £5.3 (5,9 EUR, 2009 currency translation) billion in the UK (Elia 2005). However more studies which gather information regarding the cost‐effectiveness of nutritional support are required.

Description of the intervention

Oral supplements are potentially safer and easier to administer than nasogastric enteral feeds and are therefore particularly suited to elderly people and are also widely used. However, there may be problems with the willingness and ability of older people to consume oral supplements, and supplements may not be used effectively. Even if supplements are prescribed, they may not always be given, or are given but not consumed (Peake 1998). In addition to taste, the macronutrient composition and timing of administration in relation to meals may be important (Wilson 2002). Efforts also need to be made to provide normal meals and snacks which meet the needs of elderly people and to provide assistance with feeding if required.

Why it is important to do this review

A systematic review in 1998 examined the effects of oral and enteral protein and energy supplementation in adults from thirty eligible trials which were identified up to the end of 1996 (Potter 1998). Outcomes assessed were change in body weight and arm muscle circumference, and case fatality. There were indications that nutritional supplementation was associated with improvements in outcomes assessed. However, uncertainties remained, because of the poor quality of included trials.

The present Cochrane review of older adults, when last updated and published in January 2005 included 49 trials with 4790 randomised participants. Most trials had poor study quality. Results suggested a beneficial effect of supplementation for percentage weight change from 34 trials (weighted mean difference (WMD) 2.3% (95% confidence interval (CI) 1.9 to 2.7) and a reduced mortality in the supplemented groups compared to the control groups from 32 trials (relative risk (RR) 0.74, 95% confidence interval (CI) 1.9 to 2.7).

Other reviews have included a review of randomised and non randomised studies in different diagnostic groups with chronic non malignant disorders (Akner 2001) suggesting that patients with certain disorders (such as hip fracture) may be more likely to benefit than others. Stratton 2003 and colleagues also extensively reviewed the evidence base for nutritional support in a recently published book, including a review of 166 randomised and non‐randomised trials of oral nutritional support published up to 2002 in all ages, across different disease groups, both in hospital and in the community.

A recent update of the systematic review for the Cochrane Collaboration (Avenell 2006), of nutritional supplementation for hip fracture aftercare in older people, included 21 trials involving 1727 participants. There was some evidence that oral protein and energy feeds (evaluated by eight trials), reduced unfavourable outcome (death or complications), but there was no demonstrable effect on deaths alone in participants recovering from hip fracture. However overall, the evidence was still weak due to defects in the reviewed studies, particularly inadequate size, methodology and outcome assessment.

A Cochrane systematic review of dietary advice for illness‐related malnutrition in adults of all ages has also been carried out (Baldwin 2008). Thirty‐six studies (37 comparisons) met the inclusion criteria with 2714 randomised participants. No comparison showed a significant difference in mortality. There were several significant results for change in weight and other nutritional indices favouring nutritional intervention, but it remains uncertain whether nutritional supplements and dietary advice produce the same effects. There was insufficient evidence to draw conclusions about clinical outcomes and cost. For specific information on dietary advice for illness related malnutrition, the reader is referred to Baldwin 2000.

Elderly people who are ill and malnourished may be expected to benefit more from supplementation. Providing higher energy supplements over a longer duration may also be expected to be associated with greater benefit. The present review includes a more comprehensive search for randomised trials to specifically examine the effectiveness of oral protein and energy supplements for elderly people.

Objectives

To assess the effects and acceptability of oral dietary supplements in both hospitalised elderly people and elderly people in the community, irrespective of setting:

• to test the null hypothesis that there was no difference in outcomes between participants who were given oral nutritional supplements compared to those participants who were given no intervention, a placebo, or an alternative supplement with a different amount of calories and protein.

• to carry out subgroup analysis in order to assess whether participants who were malnourished, were ill, were aged 75 years or over, were given supplements of 400 kcal or more or who had longer duration (35 days or more) of supplementation showed most benefit.

Methods

Criteria for considering studies for this review

Types of studies

Ideally studies were randomised controlled trials, but we also considered quasi‐randomised controlled trials (for example allocation by day of week, date of birth, alternation). We only accepted trials that had a minimum duration of two weeks (with a minimum duration of intervention of one week).

Types of participants

To be defined as elderly, groups of study participants had to have a minimum average age of 65 years. All groups were included, with the exception of groups exclusively of older people in critical care or recovering from cancer treatment who may have had specific nutritional needs relating to their condition. Mixed groups of patients, where some were recovering from cancer and some had been undergoing critical care were included. Efforts were made to obtain data for the groups of interest from the authors.

Types of interventions

Interventions were aimed at improving the intake of protein and energy using only the normal oral route. Protein was provided together with non‐protein energy sources such as carbohydrate and fat, and with or without added minerals and vitamins. We were interested in supplements in the form of:

• commercial sip feeds;

• milk based supplements;

• via the fortification of normal food sources.

Studies of dietary advice alone were not included in this review. We also excluded studies of specially designed immunomodulatory supplements or supplements of specific amino acids. The comparison intervention was 'usual practice' (for example using no supplement or an alternative supplement with a different amount of calories and protein) or a placebo (for example a low energy drink). Although protein‐only supplementation has occasionally been used for experimental purposes, it would not be considered for routine use and was therefore excluded.

Types of outcome measures

Primary outcomes

(for all participants unless otherwise stated)

• all cause mortality;

• morbidity, number of people with complications (for example pressure sores, deep vein thrombosis, respiratory and urinary infections);

• functional status (for example cognitive functioning, muscle functioning, mobility, ability to perform activities of daily living).

Secondary outcomes

• participants' perceived quality of life, ideally using a validated scale;

• length of hospital stay (hospital patients only);

• number of primary care contacts (non‐hospital participants only);

• adverse effects of nutritional supplementation;

• level of care and support required;

• number of hospital / care home admissions / re admissions;

• nutritional status (change in anthropometry, for example percentage weight change, percentage change arm muscle circumference);

• percentage change in dietary intake (energy and protein intake from food and supplements);

• compliance with intervention (proportion of the supplement provided which is consumed, alone or with assistance);

• economic outcomes.

Desired timing of outcome measures

The outcome measurements were evaluated at the last available time point of the studies. Short term outcomes were defined as up to three months, medium term outcomes 3 to 6 months and long term outcomes over six months.

Search methods for identification of studies

Electronic searches

• Cochrane Central Register of Controlled Trials (Issue 4, 2007);

• MEDLINE (until November 2007);

• EMBASE (until December 2007);

• Healthstar (until March 2001);

• CINAHL (until November 2007);

• BIOSIS (until December 2007);

• CAB abstracts (until October 2007).

The MEDLINE search strategy was adapted for the other electronic databases searched.

The nutrition search strategy was based on the strategy used in a Cochrane review by one of the authors (Avenell 2004). Some additional terms relating to malnutrition and food sources were also included. Phases one and two of the search strategy for randomised controlled trials developed by the United Kingdom Cochrane Centre were used (Alderson 2004). For a detailed search strategy see Appendix 1.

Databases of registered trials were also searched:

• Current Controlled Trials (www.controlled‐trials.com), December 2007.

The results were double‐checked with trials identified by two of the authors for previous systematic reviews: trials of routine protein energy supplementation in adults identified between February 1979 and July 1996 using MEDLINE (Potter 1998), and more recently, trials of nutritional supplementation for hip fracture aftercare in the elderly (Avenell 2004), searching The Cochrane Library, MEDLINE, EMBASE, Healthstar, CINAHL, BIOSIS and CAB abstracts.

Searching other resources

Handsearching

The following journals were hand searched:

• Journal of Human Nutrition: Applied Nutrition: Vol 36A(1) 1982 ‐ Vol 41A(6) 1987;

• Journal of Human Nutrition: Clinical Nutrition: Vol 36C(1) 1982 ‐ Vol 41C(6) 1987;

• Journal of Human Nutrition and Dietetics: Vol 1(1) 1988 ‐ Vol 20(3) 2007;

• Clinical Nutrition: Vol 1(1) 1982 ‐ Vol 26(6) 2007;

• Journal of Parenteral and Enteral Nutrition: Vol 5(1) 1981 ‐ Vol 31(6) 2007;

• Proceedings of the Nutrition Society: Vol 50(2) 1991 ‐ Vol 53(3) 1994 and Vol 57(1) 1998 ‐ Vol 66 2007;

• Journal of the American Dietetic Association Vol 90(1) 1990 ‐ Vol 107(7) 2007;

• American Journal of Clinical Nutrition Vol 62(10) 1995 ‐ Vol 86(4) 2007;

• Australian Journal of Nutrition and Dietetics, which became Nutrition and Dietetics,1989 ‐ Vol 64(4) 2007.

The references of all retrieved studies and reviews were searched for additional trials. Books relating to geriatric medicine and nutrition were searched. Authors of published trials, colleagues, and manufacturers of nutritional supplements were contacted for overlooked, unpublished and ongoing trials.

Data collection and analysis

Selection of studies

For the present update, one reviewer (AA) carried out the search by scanning the titles, abstract sections and keywords of every record retrieved. Full articles were then retrieved for further assessment by two reviewers if the information given suggested that the study:

• used random allocation to the comparison groups;

• compared a protein and energy supplement with no intervention, a placebo or an alternative supplement;

• involved participants who were over 65 years old;

• assessed one or more relevant clinical outcome measure.

Articles were also retrieved if there was some doubt about eligibility. If necessary, trialists were contacted for further information on methodology and data. If no clarification had been provided, and there had been disagreement about eligibility for inclusion, the review group editorial base would have been consulted.

Data extraction and management

Information was independently extracted by two reviewers (either AA and AV or AM and JP). All differences in data extraction were resolved by discussion with a third reviewer, referring back to the original article.

Information gathered included:

• location;

• participant description;

• inclusion and exclusion criteria;

• details and duration of intervention;

• baseline characteristics of the individuals studied;

• participant flow;

• relevant outcome measures recorded.

If any data were missing in a published report (see data extraction list), trialists were contacted for further information.

Assessment of risk of bias in included studies

Methodological quality was assessed by two reviewers (either AA and AV or AM and JP) all differences were resolved by discussion with a third reviewer if necessary. A sensitivity analysis was carried out based on the quality assessment. The assessment protocol scored each item between nil and two as described below. In addition, risk of pre‐allocation disclosure of assignment was rated A, B or C according to the Cochrane Handbook 1997. The following aspects of internal and external validity was reported and assessed:

a) Was the assigned treatment adequately concealed prior to allocation? 
 2 = method did not allow disclosure of assignment (A) 
 1 = chance of disclosure of assignment or states random but no description (B) 
 0 = quasi‐randomised (C)

b) Were the outcomes of patients who withdrew included in the analysis (intention to treat)? 
 2 = intention to treat analysis based on all cases randomised possible or carried out 
 1 = states number and reasons for withdrawal but intention to treat analysis not carried out 
 0 = withdrawals not mentioned, intention to treat analysis not possible

c) Were the outcome assessors blinded to treatment status? 
 2 = action taken to blind assessors, or outcomes such that bias was unlikely 
 1 = chance of unblinding of assessors 
 0 = not mentioned

d) Were the treatment and control group comparable at entry? 
 2 = good comparability of groups, or confounding adjusted for in analysis 
 1 = confounding possible; mentioned but not adjusted for 
 0 = large potential for confounding, or not discussed

e) Were care programmes, other than the trial options, identical? 
 2 = care programmes identical 
 1 = differences in care programmes but unlikely to influence study outcomes 
 0 = not mentioned or differences in care programmes likely to influence study outcomes

f) Were the inclusion and exclusion criteria clearly defined? 
 2 = clearly defined 
 1 = inadequately defined 
 0 = not defined

g) Were the interventions clearly defined (including estimates of nutritional value)? 
 2 = clearly defined interventions were applied with a standardised protocol 
 1 = clearly defined interventions were applied but the application protocol was not standardised 
 0 = intervention and/or application protocol were poorly or not defined

h) Were the participants blind to assignment status following allocation? 
 2 = effective action taken to blind subjects 
 1 = small or moderate chance of unblinding subjects 
 0 = not mentioned (unless double‐blind), or not done

i) Were the treatment providers blind to assignment status? 
 2 = effective action taken to blind treatment providers 
 1 = small or moderate chance of unblinding of treatment providers 
 0 = not mentioned (unless double‐blind), or not done

j) Was the overall duration of surveillance clinically appropriate? 
 2 = optimal (six months or more) 
 1 = adequate (one up to six months) 
 0 = not defined, or not adequate

Data synthesis

Data were combined for meta‐analysis for dichotomous variables mortality and number of patients with complications as described in the protocol. In those studies where the data were reported as the total number of complications instead of the number of affected patients, it was assumed that there was one outcome per patient. For each study relative risks and 95% confidence limits were calculated, the results were combined using fixed‐effect models and presented with 95% confidence limits. Where there was evidence of heterogeneity a random‐effects model was applied. Heterogeneity between comparable trials was explored using the I² test (Higgins 2003) using more than 50% as the cut‐off for significant heterogeneity. A funnel plot to assess small study bias for mortality data was also carried out.

Data for length of hospital stay were combined for meta‐analysis as a continuous variable. Data were combined for meta‐analysis from studies which provided length of stay data as mean number of days and standard deviation. If the data were provided as the median and interquartile range, the median was used instead of the mean and the standard deviation estimated from the interquartile range. Weighted mean differences and 95% confidence intervals were calculated using a fixed‐effect model which assumes the same underlying effect in all studies and considers any heterogeneity between trials to be due to random errors. A random‐effects model was also used if there was any evidence of heterogeneity.

Data were also combined for meta‐analysis for percentage weight change and arm muscle circumference (AMC) and as described in the protocol. Weighted mean difference and 95% confidence intervals were calculated using fixed‐ and random‐effects models as appropriate for changes in weight and anthropometry measures. The trials reported body weight and anthropometric measures in several ways. For meta‐analysis the same method was used to standardise data as used previously by Potter 1998. The mean and standard deviation of the percentage change in body weight during the trial period was selected as the measurement of choice because of its clinical relevance (Potter 1998). Where percentage weight change was not available the difference was calculated between the initial and final body weight, expressed as a percentage of baseline weight and a standard deviation of 10% inferred. This standard deviation was conservative, and at the upper limit of any of the observed results. As in Potter 1998, if baseline weight was not reported, a standard value of 60 kg was assumed, which applied to all patients regardless of their nutritional status. Assumptions made regarding standard deviations were checked by restricting the analysis to those trials where no inferences were made. As in Potter 1998, arm muscle circumference (AMC) was chosen as the anthropometry measure as it is both a measure of fat and muscle. Where this was not described in a trial it was derived from the mid‐arm circumference or mid‐upper arm circumference (MAC / MUAC) and triceps skinfold (TSF) using a standard formula (Gurney 1973). Anthropometry data were then pooled as weight data.

Data were also combined for meta‐analysis for change in handgrip strength where this was provided or could be calculated from the data provided. Weighted mean difference and 95% confidence intervals were again calculated using the fixed‐ and random‐effects model as appropriate.

Subgroup analysis and investigation of heterogeneity

Subgroup analysis was carried out on the basis of:

• baseline nutritional status (nourished, undernourished);

• health status (healthy volunteers or ill patients);

• mean age (less than 75 years, 75 years or more);

• amount of kilocalories provided in supplement (less than 400 kcal, 400 kcal or more);

• duration of intervention (less than 35 days, 35 days or more).

Additional post hoc subgroup analyses

As the result of comments resulting from the previous version of this review, additional subgroup analyses were carried out in order to provide a breakdown of the meta‐analyses on the basis of diagnostic group:

• mortality by diagnostic group;

• complications by diagnostic group;

• length of hospital stay by diagnostic group;

• percentage weight change by diagnostic group;

• percentage arm muscle circumference change by diagnostic group.

Sensitivity analysis

Sensitivity analyses were carried out in order to explore the influence of the following factors on effect size:

• repeating the analysis taking into account of study quality, as specified above.

• repeating the analysis excluding a large study to establish how much it dominated the results.

• repeating the analysis excluding studies using industry funding.

Results

Description of studies

In addition to the twenty‐one thousand titles / abstracts which were identified using the above database search strategies and through hand searching and reference list searching, in 2001, another twelve thousand titles / abstracts were identified for the 2005 update and an additional 28 studies identified for the present update (20 full reports, one ongoing study and seven abstracts). From the database search, reading the abstracts or reading the full article allowed most to be eliminated because they clearly did not meet the inclusion criteria. This left an additional 18 potentially relevant trials. Two reviewers independently assessed these trials and as a result of mutual agreement, 13 additional trials have been included to date (see Characteristics of included studies). The remaining were excluded (Characteristics of excluded studies) for a variety of reasons (for example no outcomes of interest, did not meet age or intervention criteria).

Contacts with authors

Requests for further information have been made. Of the 62 included trials, information on outcomes of interest and study quality was requested for 31 trials and has been obtained for 15 (Banerjee 1978; Brown 1992; Bruce 2003; Hankey 1993; Hankins 1996; Jensen 1997; Krondl 1999; Kwok 2001; Lauque 2000; MacFie 2000; Payette 2004; Potter 2001; Saudny 1997; Schols 1995; Yamaguchi 1998).

Trial design

Included studies were all randomised or quasi‐randomised controlled trials. The large study by Bourdel 2000 with 672 participants was cluster randomised and has not been included in the meta‐analysis, but the results have been included in the narrative part of the review. Young 2004 is also a cluster randomised crossover trial which following discussion with a statistician has been included in the meta‐analysis. Rosendahl 2006 is included in the meta‐analysis as it is clustered for exercise and individually randomised for supplementation.

Six of the included trials also examined the effects of exercise (Bonnefoy 2003; Daniels 2003; Fiatarone 1994; Meredith 1992; Rosendahl 2006; Schols 1995), with the same exercise component in both the supplemented and control groups. In the study by Bonnefoy 2003 a factorial design was used with participants receiving either exercise or memory training. In the study by Fiatarone 1994 there were sufficient data available to include the no exercise group only. In the studies by Schols 1995 and Tidermark 2004, groups of patients randomised to receive nandrolone decanoate has been excluded from the analysis. In the trial of surgical patients (Jensen 1997), there was an analysis of a subgroup of patients over 75 years, which has been used in this review.

Participants

A total of 10,187 randomised participants from the 62 trials has been included. Studies were carried out in Europe, USA, Canada, Australia and Hong Kong. Approximately 55% of participants were female (no information on gender was provided in seven studies). The mean age reported in studies varied from 65 to 88 years (not reported in seven studies). The number of participants in trials varied greatly between 10 (Brown 1992) and 4023 (FOOD trial 2005), 42 trials had fewer than 100 participants.

Although studies took place in a variety of settings, most participants (71%, 26 studies) were hospitalised in‐patients with acute conditions. Other participants were either in long‐stay / care of the elderly / continuing care wards or nursing homes (14%, 15 studies), or at home in the community (15%, 21 studies). Forty studies (48% participants) included older people with no specified disease or condition, except some trials where some or all patients had Alzheimer's disease. Other studies included patients with hip fracture (Brown 1992; Bruce 2003; Daniels 2003; Delmi 1990; Hankins 1996; Madigan 1994; Stableforth 1986; Tidermark 2004), stroke patients (FOOD trial 2005, Gariballa 1998), patients with congestive heart failure (CHF) (Broqvist 1994), patients with chronic obstructive pulmonary disease (COPD) (Deletter 1991; Knowles 1988; Saudny 1997; Schols 1995; Steiner 2003, Vermeeren 2004), older surgical patients (Jensen 1997; MacFie 2000) and patients at home with diabetic foot ulcer (Eneroth 2004). Apart from diagnosis being considered as a marker of nutritional risk (for example post‐surgery, COPD, hip fracture), 60% of participants in the included trials underwent screening and were then classified as actually being undernourished or at nutritional risk. The definition for this (for example weight, BMI, unable to eat independently) varied between studies and was often not provided, very few studies used weight loss as an indicator of nutritional risk. Separate data were provided for those well nourished and undernourished in two studies (Larsson 1990; Potter 2001) and these have been analysed separately as appropriate in this review.

Interventions

The interventions used in the trials aimed to provide between 175 additional kcal/day and up to a maximum of 1350 additional kcal/day. Additional protein was between 10 g protein/day and 50 g protein/day. Less than 400 kcal/day was provided in 20 trials, 400 kcal/day or more in 32 trials, and the energy supplemented was not known for ten trials.

Thirty‐five trials reported using supplements with at least some vitamins and minerals, or both, one gave equivalent vitamins to the control group (Carver 1995), one also gave extra vitamins to some patients in a factorial design (Vlaming 2001), two trials gave calcium and vitamin D supplements to both the intervention and control group (Hampson 2003; Tidermark 2004), 27 trials did not report vitamin content or it was unclear, although the majority of commercial supplements do provide vitamins and minerals. MacFie 2000 had a control group and three groups receiving supplements (pre‐operative supplements only, post‐operative supplements only, and both pre and post‐operative supplements), those receiving supplements have been grouped together.

When reported, supplements were given twice daily for around 31% of participants, but this could also be between one and four times or any number of times for the remaining participants. Thirty‐nine trials used named commercial supplements, the others did not specify a manufacturer. Commercial supplements may have been provided by the manufacturer free of charge, although this was not often explicitly stated. Extra milk alone was provided for one study (Barr 2000), and low lactose milk powder was used in one study (Kwok 2001).

The minimum time period of the intervention was 10 days, the maximum was 18 months. The length of time of the intervention was less than 35 days for 17 trials, and 35 days or more for 37 trials, from admission to discharge in five trials and the intervention period was unclear for two trials. The duration of follow‐up was generally the same as the duration of the intervention, and varied from one week to 18 months.

Outcomes assessed

Most trials assessed nutritional outcomes, particularly weight change and dietary intake, but also change in anthropometry. Mortality was not the principle outcome for most trials. Morbidity and complications and length of hospital stay were provided in a limited number of trials. The majority of trials also included a measure of functional status, however these were often disease specific and too diverse for meta analysis. Sixteen studies measured the effect of supplementation on quality of life. Only one trial provided data on cost effectiveness (Edington 2004).

Funnel plot

The funnel plot of the comparison "oral protein and energy versus routine care ‐ outcome: mortality" appeared to be symmetrical (Figure 1).

1.

Funnel plot of comparison: 1 Oral protein and energy versus routine care, outcome: 1.1 Mortality.

Risk of bias in included studies

For details see Appendix 2. The method of scoring used is described above (quality assessment of trials). Concealment of allocation was confirmed in only 19 studies which used computer allocation or sealed opaque envelopes. A clear 'intention to treat' analysis was only carried out in 24 studies. The quality was poorest with regard to blinding. Action to ensure blinding of outcome assessors was reported in only 12 studies, action to blind participants in 16 studies, and blinding of treatment providers was reported in only 14 studies.

Effects of interventions

Primary outcomes

Mortality

Deaths were reported in 50 trials. Data from Banerjee 1978; Bonnefoy 2003; Rosendahl 2006; Saudny 1997; Schols 1995; Woo 1994 and Wouters 2005, have been excluded as further clarification from the authors was not received. Data on mortality from Bourdel 2000 have been excluded from the meta‐analysis because the trial was cluster randomised, however Bourdel 2000 found no significant difference in the incidence of death during the 15 day follow‐up: 25 in the nutritional intervention group and 22 in the control group (P = 0.18). The relative risk by the end of follow‐up from the remaining 42 trials (8031 participants) did not show a reduced mortality in supplemented compared with control groups (relative risk (RR) 0.92; 95% confidence interval (CI) 0.81 to 1.04), with no significant statistical heterogeneity (I² = 0%). The results of fourteen of these trials fall out of the analysis because the effect measure could not be calculated for zero events (no deaths).

Subgroup analyses

The subgroup analyses suggested that the results were statistically significant or approaching statistical significance when limited to trials in which participants (N = 2461) were defined as undernourished (RR 0.79; 95% CI 0.64 to 0.97), and when 400 kcal or more was offered per day in the supplement (N = 7307), (RR 0.89; 95% CI 0.78 to 1.00).

Results were not significant, when analyses were limited to participants who were at least 75 years old (N = 2967), (RR 0.85; 95% CI 0.69 to 1.05), when supplementation was continued for 35 days or more (N = 2454), (RR 0.97, 95% CI 0.77 to 1.24), and when participants were unwell (N = 7636), (RR 0.92; 95% CI 0.81 to 1.04), and when participants were in hospital or in a nursing home (N = 6582), (RR 0.91; 95% CI 0.80 to 1.04), all with evidence of little heterogeneity.

The results of mortality were also not statistically significant when limited to trials when participants were not defined as undernourished (N = 5403), (RR 0.78; 95% CI 0.53 to 1.15 ‐ RR 0.98; 95% CI 0.83 to 1.14), when less than 400 kcal were offered per day in the supplement (N = 858), (RR 1.09; 95% CI 0.59 to 1.98), participants were less than 75 years old (N = 8049 ‐ N = 5082), (RR 0.91; 95% CI 0.80 to 1.03 ‐ RR 0.94; 95% CI 0.80 to 1.11), when supplementation was continued for less than 35 days (N = 5054), (RR 0.92; 95% CI 0.78 to 1.07), when participants had not been defined as unwell (N = 393), (RR 0.98; 95% CI 0.25 to 3.78), and when living in the community (N = 966), (RR 0.99; 95% CI 0.62 to 1.59).

Sensitivity analyses

The results were consistent when analysis was restricted to 15 trials (N = 6604) with clearly concealed randomisation (RR 0.91; 95% CI 0.79 to 1.03).

For six trials, there was co‐authorship with an employee of the manufacturer or full funding of the trial by a manufacturer of oral supplements (Edington 2004; Eneroth 2004; Krondl 1999; Lauque 2000; Lauque 2004; Salas‐Salvado 2005; Vermeeren 2004; Wouters 2002; Wouters 2003; Wouters 2006). The meta‐analysis of mortality data was also therefore carried out with the exclusion of these trials in order to explore the influence on effect size. Results with the remaining 29 trials, suggest that this had no demonstrable effect (RR 0.90; 95% CI 0.79 to 1.03).

Post‐hoc subgroup analyses for mortality based on diagnostic group suggested that the results were statistically significant when including only trials in patients with a variety of geriatric conditions (N = 2701), (RR 0.78; 95% CI 0.62 to 0.98), where most of these participants were in hospital. However, there was no demonstrable benefit for patients with hip fracture (N = 437), (RR 0.91; 95% CI 0.50 to 1.66) from eight small trials. The limited data from other diagnostic groups such as stroke patients and those with chronic obstructive pulmonary disease (COPD) did not suggest a reduction in mortality.

Morbidity

Data on morbidity were available from 28 trials. Data from Wouters 2005 were unsuitable for analysis without additional information from the authors. Data on development of pressure sores from Benati 2001 and Bourdel 2000 were also unsuitable for meta‐analysis. The results from Bourdel 2000 showed an increased risk of developing pressure ulcer in the control group versus the intervention group in 672 elderly patients (RR 0.57; 95% CI 1.03 to 2.38). Data from the remaining 24 trials (N = 6225), have been combined for meta‐analysis (Broqvist 1994; Delmi 1990; Gariballa 1998; Gariballa 2006; Potter 2001; Tidermark 2004 (infective complications); FOOD trial 2005; Hankins 1996; Larsson 1990 (total pressure sores); Hampson 2003; MacFie 2000; Madigan 1994; Saudny 1997; Wouters 2003 (patients too ill to continue used as a measure of complications); Steiner 2003 (exacerbation of COPD); Stableforth 1986 (anaesthetic, surgical infection, gastrointestinal, urinary); Daniels 2003; Lauque 2004; Vermeeren 2004; Young 2004 (hospital readmission); Collins 2005; Eneroth 2004; (incomplete wound healing); Price 2005 (prescription of antibiotics); Salas‐Salvado 2005 (total severe adverse events). Data on complications from Lauque 2000 have not been included because further clarification from the authors is required. The risk of complications by the end of follow‐up in supplemented groups was statistically significantly different from the control groups (RR 0.86; 95% CI 0.75 to 0.99), with no significant statistical heterogeneity. Subgroup analyses for complications based on diagnostic group suggested that there may be a reduced risk of complications with supplementation for hip fracture patients only (RR 0.60; 95% CI 0.40 to 0.91).

Functional status

Functional status measures were very diverse, few were able to suggest any functional benefit with supplementation. One or more of the following measures of mobility were included in 14 studies: number of falls, activity rating, mobility, physical activity, walking, stair climbing, timed up and go (Bonnefoy 2003; Brown 1992; Deletter 1991; Fiatarone 1994; Gray‐Donald 1995; Hampson 2003; Larsson 1990; Madigan 1994; Payette 2004; Schols 1995; Saudny 1997; Steiner 2003; Tidermark 2004; Wouters 2003). In mixed groups of elderly people Gray‐Donald 1995 reported that the number of falls was lower among supplemented participants than controls (0% versus 21%; P = 0.05). Larsson 1990 reported a significant improvement in the activity rating in the supplemented group at eight weeks compared to the control group (P < 0.05) due mainly to improvement in the initially well nourished patients, however the number of patients studied was not clear. Muscle function and mobility were measured by Bonnefoy 2003, there was a short term improvement in quadriceps muscle power at three months with supplementation (56.8%; P = 0.03) but this was not sustained at nine months. Bonnefoy 2003 found no statistically significant effect on six meters walk, five time chair rise, or six stair climb at 3 and 6 months. Wouters 2003 also found no significant effect of supplementation on a timed 'up and go' test (Podsiadlo 1991), although Payette 2004 found a trend towards improvement in this test (P = 0.057).

In patients following hip fracture Madigan 1994 found that patients given supplements did significantly less well. The number of patients unable to reach goal two of physio‐independent mobility was significantly greater in the intervention group (5 versus 1; P < 0.01). Tidermark 2004 also failed to demonstrate any beneficial effect of supplementation on mobility in women with hip fracture. Hampson 2003 found no difference in level of physical activity in elderly community‐living women with osteoporosis given supplements.

Walking distance or velocity was assessed in four studies of patients with COPD, (Deletter 1991; Saudny 1997; Schols 1995; Steiner 2003) with no statistically significant improvement with supplementation reported, however a non significant trend towards improvement in 12 minute walking distance after nine weeks in the supplemented group compared to the control group was noted by Deletter 1991(65 m versus 16 m; P > 0.05). 
 
 Activities of daily living (ADL) (Katz 1963; Mahoney 1965) was measured in 11 studies (Barr 2000; Bruce 2003; Gariballa 1998; Hankins 1996; Lauque 2004; Potter 2001; Tidermark 2004; Volkert 1996; Woo 1994; Wouters 2002; Wouters 2006). Overall only one study demonstrated a significant improvement at end of follow‐up. Woo 1994 reported a significant difference between the groups with a lower level of functional ability after three months in the control group in patients following chest infection (20 versus 19.5; P < 0.01). Tidermark 2004, however found an improvement at six months (P < 0.05) but no difference at 12 months (number of participants remaining independent 11/16 in the control group versus 14/17 in the intervention group (from graph)). Potter 2001 reported a significant improvement with supplementation only in a subgroup of very malnourished patients (17 versus 11; P < 0.04). Volkert 1996 found an improvement in the ADL score from admission to six months only in the subgroup with good acceptance of the supplement (72% versus 39%; P < 0.05). There was no improvement in frail elderly functional capacity (FEFA) in the study by Manders 2006.

No statistically significant effect of supplementation was reported for hand grip strength in 13 studies (Edington 2004; Gray‐Donald 1995; Kwok 2001; Lauque 2000; MacFie 2000; Manders 2006; Payette 2002; Price 2005; Saudny 1997; Steiner 2003; Tidermark 2004; Wouters 2003; Vermeeren 2004). There was a trend towards improvement in grip strength for COPD patients in the study by Steiner 2003 (0.64 kg force versus ‐0.05; P = 0.06). There was also a trend towards short term improvement in the study by Edington 2004 after eight weeks (1.2 versus ‐0.5; P = 0.055) but with no difference at 24 weeks. Price 2005 reported that the intervention group (when analysed by intention‐to ‐treat) showed a greater increase in handgrip strength over 12 weeks compared to the control group which of borderline statistical significance (P = 0.055). Seven studies (Manders 2006; Payette 2002; Price 2005; Steiner 2003; Tidermark 2004; Wouters 2003; Vermeeren 2004) (535 participants) provided data on the change in handgrip which could be combined for meta‐analysis. Results suggest that supplementation had no demonstrable effect (weighted mean difference (WMD) 0.06; 95% CI ‐0.60 to 0.72).

There was no evidence of an improvement in dynamic strength (maximum weight lifted on a 'thigh‐knee' machine) (Meredith 1992), both the supplemented and control groups gained dynamic strength (P < 0.001) with no effect of diet. There was also no significant effect of the supplement on balance, gait or lower limb strength (Rosendahl 2006). Calf circumference was significantly improved at 24 weeks of supplementation in the study by Manders 2006.

Four studies involving older patients with COPD measured changes in lung function with supplementation. There was a statistically significant improvement in lung function [maximal inspiratory mouth pressure measured in kiloPascals (kPa)] in non tissue depleted patients with COPD between pre and post intervention, which was not seen in the control group (Schols 1995) (0.8 kPa versus 0.5 kPa; P < 0.05), but only in the first four weeks, and in Saudny 1997 forced vital capacity (percentage predicted) improved in the supplemented group as compared with the control group (8.7% versus ‐3.5%; P = 0.015). Deletter 1991 and Vermeeren 2004 however found no evidence of change in ventilatory performance in the supplemented group.

There was no evidence of an improvement in cognitive function between groups with supplementation (Collins 2005; Gariballa 2006; Lauque 2004; Salas‐Salvado 2005; Young 2004).

Secondary outcomes

Health‐related quality of life

Quality of life was ascertained using a variety of measures (general and disease specific well‐being and self perceived health questionnaires (for example SF36, hospital anxiety and depression score (HADS), Nottingham Health Profile (NHP), EQ5D and Self Reported Chronic Respiratory Questionnaire) in 16 studies (Barr 2000; Collins 2005; Edington 2004; FOOD trial 2005; Gariballa 2006; Price 2005; Hampson 2003; Krondl 1999; MacFie 2000; Payette 2002; Saudny 1997; Scorer 1990; Steiner 2003; Tidermark 2004; Woo 1994; Wouters 2002). Collins 2005 was unusable because result between groups was not reported. Gariballa 2006 reported SF36 in a subgroup of patients suggesting significant improvement in physical and social score. Hampson 2003 reported that more women 'felt better' in the supplemented group (48% versus 20%; P = 0.029). Saudny 1997 reported a greater improvement in well‐being in the supplemented group that was not statistically significant (12 points versus ‐10 points; P = 0.07). In the study by Edington 2004, although there was no effect on overall EQ5D score or for the visual analogue scale, the supplemented group reported fewer mobility problems at 24 weeks (P = 0.022). In the study by Krondl 1999, scores for vitality and general health perception increased more from baseline to termination in the supplemented group (P < 0.01), however it was not clear whether these were within or between group differences. No other meaningful between group differences were found.

Length of hospital stay

Length of hospital stay was measured in 12 studies (Brown 1992; Bruce 2003; Delmi 1990; FOOD trial 2005; Gariballa 1998; Gariballa 2006; Hankins 1996; MacFie 2000; Madigan 1994; Potter 2001; Tidermark 2004; Vlaming 2001). Data were analysed separately for three groups in the study by Potter 2001(severely malnourished, moderately malnourished and adequately nourished). Data from Gazzotti 2003 require further clarification from the authors before inclusion. Data were combined for meta‐analysis from studies which provided length of stay data as mean number of days and standard deviation (SD) (Brown 1992; Bruce 2003; FOOD trial 2005; Gariballa 2006; Hankins 1996; Madigan 1994; Vlaming 2001), the SD was assumed to be 10 days in one study (MacFie 2000), based on the SDs for length of stay from the other studies. The median was used instead of the mean and SD estimated from the interquartile range for the four remaining studies (Delmi 1990; Gariballa 1998; Potter 2001; Tidermark 2004). The pooled weighted mean difference for length of stay using a random‐effects model showed no benefit from supplementation ‐0.8 days (‐2.8 to 1.3) with significant heterogeneity (chi‐square 25.53; df 13; P = 0.02; I² = 49%). Subgroup analyses for length of stay were too limited to suggest any difference between diagnostic groups.

Adverse effects

Eighteen trials discussed adverse effects from supplementation, in most cases no comparison with the control group was performed, six reported no adverse effects (Delmi 1990; McWhirter 1996; Potter 2001; Saudny 1997; Tidermark 2004; Wouters 2002). Problems with tolerance and side‐effects were reported in 12 studies: Eneroth 2004 reported nausea, vomiting and diarrhoea in the intervention group. FOOD trial 2005 reported 28% stopped their supplements before discharge (refusal, weight gain, unwanted, nausea), Gariballa 2006 reported 20% nausea in both groups. Hankins 1996 reported dysphagia, nausea, diarrhoea and fatigue as reasons for drop‐out from the study in four out of 17 patients; Fiatarone 1994 reported diarrhoea in two out of 49 participants; Gazzotti 2003 reported loss of appetite, nausea or diarrhoea in five out of 39 patients; Price 2005 reported intolerance to supplements as reason for withdrawal in 20% of participants and significantly more gastro‐intestinal adverse events in the intervention group. Ovesen 1992 excluded ten out of 37 participants from the study because they refused to continue due to gastro‐intestinal discomfort attributed to the supplements; Vermeeren 2004 reported that nausea caused dropouts in three of 29 patients in the supplemented group and in one out of 27 patients in the control group. Stableforth 1986 stated that "intolerance of the supplements proved to be a handicap in correcting the deficits in many patients".

Nutritional status

Weight change

Measures of weight were converted into percentage weight change to allow data from 42 trials with 3058 participants to be included in the meta‐analysis. Percentage change in body mass index (BMI) was used as a proxy measure for weight in one study (Bonnefoy 2003). A standard deviation of 10% was assumed in 20 studies as described previously.

There was a mean weight loss during the trial period for the supplemented group in seven trials, this contrasts with a mean weight loss in 23 trials for the control group. The pooled weighted mean difference for percentage weight change showed a benefit of supplementation of 2.2% (1.8 to 2.5) with no significant heterogeneity (chi‐square 52.35; df 43; P = 0.16; I² =17.5%). This would mean an average weight gain of 1.2 kg for a person weighing 55 kg.

Sensitivity analysis of weight change data

When analysis was restricted to 18 trials where no inference was made regarding standard deviations the results remained consistent for weighted mean difference 2.1% (1.7 to 2.5).

Subgroup analyses for weight change

Subgroup analyses for weight change based on diagnostic group confirmed a significant increase in weight for mixed group of patients with geriatric conditions, weighted mean difference 2.7% (2.2 to 3.1), and for patients with chest conditions, weighted mean difference 1.6% (1 to 2.2). This could not be confirmed for other patient groups with the limited data available.

Arm muscle circumference (AMC)

Fifteen trials with 1382 participants reported a measure of AMC which was, or could be calculated from mid‐upper arm circumference and triceps skinfold thickness (Gurney 1973) and combined in a meta‐analysis. The pooled weighted mean difference for percentage AMC change using a fixed‐effect model showed a benefit of supplementation of 1.2% (0.5 to 2), with no significant heterogeneity (chi‐square 20.75; df 15; P = 0.15; I² = 27.7%).

Subgroup analyses for percentage arm muscle circumference

Subgroup meta‐analyses for percentage AMC change based on diagnostic group were unable to confirm a statistically significant difference in AMC for any diagnostic group where more than one trial could be included.

Intake

Thirty‐two studies using a variety of methods for example dietary recall and weighed intake, reported that supplementation increased daily protein intake and energy intake, or both (Banerjee 1978; Barr 2000; Bourdel 2000; Broqvist 1994; Brown 1992; Deletter 1991; Delmi 1990; Edington 2004; Gariballa 1998; Gazzotti 2003; Hampson 2003 (within group change); Hankey 1993; Hankins 1996; Jensen 1997; Knowles 1988; Krondl 1999; Lauque 2000; MacFie 2000 (pre‐operative only); McWhirter 1996; Meredith 1992; Ovesen 1992; Payette 2002; Payette 2004; Potter 2001; Price 2005; Saudny 1997; Stableforth 1986; Steiner 2003 (within group change); Vermeeren 2004; Woo 1994; Yamaguchi 1998; Young 2004). Wouters 2003 reported no compensation of energy intake from usual diet had occurred with additional supplements. Volkert 1996 found that a statistically significant increase in protein intake during hospital stay was limited to 55% of patients in the intervention group with good acceptance of the supplement. In one study (Fiatarone 1994), supplementation was associated with significant reductions in total energy and protein from habitual diet, and energy intake was only significantly increased in exercising participants who also received nutritional supplementation. Six studies did not find that intake was significantly increased (Gray‐Donald 1995; Kwok 2001; Lauque 2004; Madigan 1994; Salas‐Salvado 2005, Wouters 2006), although reasons for this were not clear. Intake was not reported or not clear in the remaining 21 studies.

Compliance (acceptance of the supplement)

Acceptance was reported to be good in 17 studies although this was often not or variously defined (Barr 2000; Broqvist 1994; Carver 1995; Collins 2005; Delmi 1990; Fiatarone 1994; Gazzotti 2003; Krondl 1999; Kwok 2001; Lauque 2000; Lauque 2004; McWhirter 1996; Potter 2001; Rosendahl 2006; Steiner 2003; Vermeeren 2004; Wouters 2002). Bonnefoy 2003 reported 54% compliance at nine months, FOOD trial 2005 reported 76% compliance, Manders 2006 reported 67% compliance in the 111 out of 176 who completed the study. Payette 2002 found that 55% of participants were compliant at four months, Price 2005 reported 62% compliance overall, Scorer 1990 excluded patients who were unable to consume two cans supplement per day. Vlaming 2001 reported that 63% of 222 participants took 50% or more of the sip‐feed supplement in hospital and Wouters 2003 found no difference in compliance between the intervention and placebo products over six months: 85% (SD 36%) versus 94% (SD 24%) respectively. Bruce 2003 reported that poor compliance in patients after hip fracture had limited the effectiveness of the supplements despite encouragement and strategies offered by the dietitian (mean compliance 20.6/28 cans) due to 'taste problems'. Problems with acceptance were reported in the study by Gray‐Donald 1995, where 36% of potentially eligible participants refused to participate mainly because they did not wish to take a nutritional supplement; of those that did take part, compliance was realised by 68% (measured by counting supplements during a home visit on a weekly basis). Larsson 1990 found that 39 out of 197 patients refused the supplement and were withdrawn from the study and therefore not included in the analysis. Volkert 1996 reported data from 45% of participants who had poor acceptance of the supplements, but stated that "if taken they were well tolerated".

Other outcomes

Insufficient data were provided from these trials to examine other outcomes listed in the protocol, which included number of primary care contacts, level of care and support required. Edington 2004 measured heath care professionals services and social service costs over a period of 24 weeks following hospitalisation with or without eight weeks of supplementation of elderly malnourished patients on discharge from hospital. There was no reduction in health care costs with supplementation.

Discussion

Summary of main results

Supplementation appears to produce a small but consistent weight gain. There was no beneficial effect on mortality overall, but there was a reduction in mortality in undernourished groups and in general geriatric populations which were mainly hospitalised patients. There was a beneficial effect on the number of complications, particularly in hip fracture patients. There was no difference in the length of hospital stay. There was little evidence of benefit to functional outcomes or quality of life.The updated review was dominated by the large international FOOD trial 2005 which tested the hypothesis that adding oral protein‐energy supplements to standard hospital diet until discharge would improve outcome at six months after stroke. The trial contributed 4023 of the 10,187 randomised participants in this review, with only a minority (8%) of the FOOD trial 2005 participants classified as undernourished at baseline. The results of the trial suggest that it is unlikely that routine oral supplementation in well nourished stroke patients is useful. The trial did not answer the question about whether routine oral supplements may have a role to play in the management of undernourished patients.Care must be taken with the inclusion criteria for participants in order to achieve external validity. Trials ideally should have included a representative sample of elderly people who would under normal circumstances be eligible for oral nutrition support. Severely malnourished patients who are likely to benefit most have not been included in many trials for ethical reasons. Furthermore, many trials did not adequately screen participants for nutrition risk or malnutrition. The inclusion of marginal candidates for nutrition support in trials may mask the benefits of treatment (Wolfe 1997).

Mortality

The subgroup comparison of 'undernourished' versus 'not undernourished' was based on a definition of undernourished that was not the same for all trials. Unfortunately there were not enough data provided to stratify the subgroups using a standard measure of nutrition risk such as body mass index or recent weight loss. No single trial was adequately powered or long enough to investigate mortality as a primary outcome.

Morbidity, functional status and quality of life

The definition of complications was also not the same for all trials although most complications were infection related.

Few studies were able to provide data on improvements in functional status or quality of life in general, apart from handgrip data. Measures were too diverse or too limited to combine for meta‐analyses.

Length of stay

The FOOD trial 2005 with a slightly longer average stay for supplemented stroke patients dominated the results for the meta‐analysis of length of hospital stay with supplementation, although non significant, the trend is still towards a slightly shorter stay for supplemented patients.

Nutritional status

The pooled weighted mean difference for percentage weight change in this study showed a small but consistent benefit from supplementation, and there was no evidence that assuming a standard deviation in so many trials biased the results. The evidence from most trials suggested that supplements, if taken, produce weight gain. However we do not know the composition of that weight gain, which could be a gain in fat mass, muscle mass or both. In terms of providing functional benefit, a gain in fat mass may have cosmetic benefit but will not improve muscle strength. Results from Fiatarone 1994 suggested that exercise is also required to produce a significant improvement in muscle strength and function. Trials involving patients living at home provided supplements for periods of between six weeks and six months, with quite modest gains in weight. In the absence of more evidence of benefit for older people in the community, long‐term supplementation at home may not be cost‐effective.

The pooled weighted mean difference for percentage change in arm muscle circumference (which is a measure of both lean and fat tissue) in this study suggested a very small benefit of supplementation.

Intake

It has been suggested that nutritional supplementation may significantly reduce the intake of ordinary diet (Bastow 1983), and this obviously would reduce the effectiveness of supplements. However, the majority of studies reported that supplementation significantly increased total daily protein intake, energy intake or both. It should, however, be remembered that it is very difficult to measure nutrient intake with any degree of accuracy, and few assessors of intake were reported as being blinded to treatment status.

There was one notable exception to the findings above. Results of the trial by Fiatarone 1994 suggested that after 10 weeks of supplementation the supplemented participants did not increase their total energy intake significantly compared to the controls, despite high compliance with the supplements, which was offset by a reduction in normal food intake.

We also do not know yet which component(s) of the supplement may be providing the beneficial effect if any, specifically, whether it is actually the energy and protein that is important or the provision of extra vitamins and minerals from the supplement or all of these. Assessing this was made difficult because it was not clear in many studies whether vitamins and mineral were actually included in the supplements.

At least 10 trials also included some dietary advice as part of the intervention. Furthermore, participants at home received regular phone calls or visits. It is not known what influence this might have had, although it seems likely that this would improve compliance and outcomes. A recent systematic review of dietary advice alone to patients with illness related malnutrition (Baldwin 2008), would suggest that it is difficult to disentangle the different effects of advice and supplements.

Compliance (acceptance of the supplement)

The literature suggests that under normal conditions acceptance of supplements can be a problem for elderly people. This review does not include studies examining only the acceptability in terms of taste of different kinds of supplements for this age group. Problems with acceptance or adverse events such as nausea, gastro‐intestinal discomfort were reported in a few studies. From this review there was no evidence that participants at home were less compliant with supplementation than those in hospital or long term care.

Potential biases in the review process

The review was limited because the quality rating of included trials as reported was poor, particularly for blinding of outcome assessors, participants and treatment providers. Blinding of participants and treatment providers cannot be done without a placebo, and it can be difficult to have an untreated arm for ethical reasons in some trials. Without a placebo group bias may result from supplemented patients receiving a higher standard of care and attention from treatment providers. It should be possible however, to design a trial where the outcome assessors are blinded to the treatment allocation. The fact that this was only done (or at least reported) in 17 studies was a major deficiency of the review and may bias the results towards finding a more beneficial effect.

Analysis of outcomes on an 'intention‐to‐treat basis' was also deficient and this also potentially represents a source of bias. There was inadequate reporting of numbers of participants who were allocated and assessed, and reasons for losses to follow‐up were often not reported. Some patients were excluded from the analysis because they were unable or willing to take the supplements. If the outcomes from all patients had been included in the analysis, then the results would be more representative of the effectiveness of supplements under real life conditions.

The beneficial effect of supplementation on mortality was still evident when only higher quality trials with adequate allocation concealment were included, and also when trials with substantial proprietary involvement were excluded from the analysis. There was therefore no evidence that the commercial trials that did report on mortality, were more likely to report favourable outcomes. But there is a general concern that selective reporting of outcomes such as mortality, could have introduced bias.

Authors' conclusions

Implications for practice.

The results of the present review, which included more than twice the number of patients than the previous version of this review, supports the findings of the previous review in that there is a small weight gain, but no longer supports the finding that there is a beneficial effect on mortality overall. However, mortality in undernourished patients may be reduced. There is more evidence of a reduction in complications than in the previous review. Results however still require to be substantiated as there are doubts due to many included trials having poor study quality.

Although proprietary sip feeds have become a widely accepted means of improving nutritional status, it is not enough to provide supplements and hope for the best. Under normal circumstances patients should have a variety of options for increasing intake. In hospital or long‐term care, at the very least, a choice of attractive and acceptable food should be offered along with dietary advice if required. Furthermore, elderly people may become more malnourished because they do not get assistance with feeding on a busy ward, and encouragement and assistance may be all that they require. There were too few randomised trials that had considered other methods of supplementation such as altering the nutrient density and diversity of the diet, which may be preferable to sip‐feeds for some elderly people, and also few trials which had tried to improve the way that the sip‐feeds were provided in order to improve acceptability and reduce wastage. However proprietary protein and energy supplements used appropriately with nutritionally 'at risk' patients have a useful role to play as part of a raft of measures which should be used to improve the intakes and nutritional status of older people in hospital or long‐term care.

For older people in the community who are at risk, it is also important to consider both dietary ways (for example meals‐on‐wheels) and non‐dietary ways (for example treatment of depression, correction of dental problems and exercise regimes) of improving intakes and nutritional status before they are admitted to hospital. More evidence of benefit from oral nutritional supplements for older people at risk of malnutrition in the community is still required.

Implications for research.

Large scale multi‐centre pragmatic trials of interventions to improve nutritional status of elderly people, particularly malnourished elderly people from clearly defined patient groups (such as patients with hip fracture) are still required. Most individual studies in this review had an intervention time that was too short to have a realistic chance of detecting differences in morbidity, functional status or quality of life. Future trials need to have sufficient statistical power and length of follow‐up to be able to detect any beneficial effects. Future trials of nutritional supplementation should also have properly concealed allocation, blinding and follow‐up of all participants to ensure that those who are not able to consume the supplements, are included (intention‐to‐treat analysis). Trials should also focus more on primary outcomes of relevance to patients such as improvement in function or quality of life measures.

What's new

Date	Event	Description
30 November 2007	New citation required and conclusions have changed	The last updated and published review in November 2004 included 49 trials with 4790 randomised participants. Most trials had poor study quality. Results suggested a beneficial effect of supplementation for percentage weight change from 34 trials (weighted mean difference (WMD) 2.3% (95% confidence interval (CI) 1.9 to 2.7) and a reduced mortality in the supplemented groups compared to the control groups from 32 trials (relative risk (RR) 0.74, 95% confidence interval (CI) 1.9 to 2.7). Sixty‐two trials with 10,187 randomised participants have been included in the current update. Most included trials had poor study quality. The pooled weighted mean difference (WMD) for percentage weight change showed a benefit of supplementation of 2.2% (95% confidence interval (CI) 1.8 to 2.5) from 42 trials. There was no significant reduction in mortality in the supplemented compared with control groups (relative risk (RR) 0.92, CI 0.81 to 1.04) from 42 trials. Mortality results were statistically significant when limited to trials in which participants (N = 2461) were defined as undernourished (RR 0.79, 95% CI 0.64 to 0.97). The risk of complications was reduced in 24 trials (RR 0.86, 95% CI 0.75 to 0.99). Few trials were able to suggest any functional benefit from supplementation. The WMD for length of stay from 12 trials also showed no statistically significant effect (‐0.8 days, 95% CI ‐2.8 to 1.3).

History

Protocol first published: Issue 4, 2001
 Review first published: Issue 3, 2002

Date	Event	Description
14 October 2004	New search has been performed	First update: New studies found and included or excluded: 4/1/04 Conclusions changed: 10/9/04

Appendices

Appendix 1. Search strategy

Search terms

Unless otherwise stated, search terms were free text terms; exp = exploded MeSH: Medical subject heading (Medline medical index term); the dollar sign ($) stands for any character(s); the question mark (?) = to substitute for one or no characters; tw = text word; pt = publication type; sh = MeSH: Medical subject heading (Medline medical index term); adj = adjacency. MEDLINE 1. nutrition [MeSH, all subheadings included] 2. nutri* (textword) 3. maln* (textword) 4. undernutr* (textword) 5. under‐nutr* (textword) 6. undernourish* (textword) 7. under‐nourish* (textword) 8. protein‐energy malnutrition [MeSH, all subheadings included] 9. protein‐energy malnutrition (textword) 10. nutritional status [MeSH, all subheadings included] 11. nutrition disorders [MeSH, all subheadings included] 12. food,fortified [MeSH, all subheadings included] 13. food,formulated [MeSH, all subheadings included] 14. diet [MeSH, all subheadings included] 15. diet therap* (textword) 16. dietary supplements [MeSH, all subheadings included] 17. (diet* or nutri*) near supplement* (textword) 18. enteral nutrition [MeSH, all subheadings included] 19. dietary proteins [MeSH, all subheadings included] 20. energy intake [MeSH, all subheadings included] 21. randomized controlled trial.pt. 22. controlled clinical trial.pt. 23. randomized controlled trials.sh. 24. random allocation.sh. 25. double‐blind method.sh. 26. single‐blind method.sh. 27. or/1‐26 28. limit 27 to animal 29. limit 27 to human 30. 28 not 29 31. 27 not 30 32. clinical trial.pt. 33. exp clinical trials/ 34. (clinic$ adj25 trial$).tw. 35. ((singl$ or doubl$ or trebl$ or tripl$) adj (mask$ or blind$)).tw. 36. placebos.sh. 37. placebo$.tw. 38. random$.tw. 39. research design.sh. 40. (latin adj square).tw. 41. or/32‐40 42. limit 41 to animal 43. limit 41 to human 44. 42 not 43 45. 41 not 44 46. comparative study.sh. 47. exp evaluation studies/ 48. follow‐up studies.sh. 49. prospective studies.sh. 50. (control$ or prospectiv$ or volunteer$).tw. 51. cross‐over studies.sh. 52. or/46‐51 53. limit 52 to animal 54. limit 52 to human 55. 53 not 54 56. 52 not 55 57. 31 or 45 or 56 58. obesity [MeSH, all subheadings included] 59. critical care [MeSH, all subheadings included] 60. 58 or 59 61. or/1‐20 62. 61 not 60 63. 62 and 57 64. limit 63 to (newborn infant or infant <1 to 23 months> or preschool child <2 to 5 years> or child <6 to 12 years> or adolescence <13 to 18 years> or adult <19 to 44 years>) 65. 63 not 64

Appendix 2. Risk of bias

Study	a) A/C b) ITT	c) assessor blind	d) comparability	e) identical care	f) in/exclusion	g) interventions	h) participant blind	i) carers blind	j) duration
Banerjee 1978	1 2	0	0	1	0	1	0	0	1
Barr 2000	1 1	0	2	1	2	0	0	0	0
Benati 2001	1 0	0	0	0	1	0	0	0	0
Bonnefoy 2003	2 1	0	1	2	1	2	2	2	2
Bourdel 2000	1 2	1	2	0	1	1	0	0	0
Broqvist 1994	1 2	2	1	0	2	1	2	2	1
Brown 1992	0 2	2	1	0	2	0	0	0	0
Bruce 2003	0 1	0	2	1	2	1	0	0	2
Carver 1995	0 1	1	2	2	2	2	2	0	1
Collins 2005	1 0	1	2	2	2	2	2	2	1
Daniels 2003	2 2	1	2	0	2	1	0	0	1
Deletter 1991	1 2	0	2	0	2	0	0	0	1
Delmi 1990	1 1	0	2	0	2	2	0	0	2
Edington 2004	2 1	0	0	0	2	2	0	0	1
Eneroth 2004	1 1	2	0	2	2	0	2	2	2
Fiatarone 1994	1 2	0	2	2	2	2	2	2	1
FOOD 2005	2 2	0	2	0	2	1	0	0	2
Gariballa 1998	2 2	1	1	2	1	0	0	0	1
Gariballa 2006	2 2	2	2	0	2	2	2	2	2
Gazzotti 2003	1 1	0	2	0	2	0	0	0	1
Gegerle 1986	1 2	0	2	0	0	2	0	0	0
Gray‐Donald 1995	1 2	1	2	0	2	1	0	0	1
Hankey 1993	1 0	0	0	2	1	1	0	0	1
Hankins 1996	2 2	0	2	1	2	2	0	0	1
Hampston 2003	2 1	0	1	0	2	1	0	0	2
Hubsch 1992	1 2	0	0	0	0	0	0	0	1
Jensen 1997	2 1	1	1	0	2	1	0	0	1
Knowles 1988	1 2	2	1	0	2	1	0	0	1
Krondl 1999	0 0	0	1	2	2	2	0	0	1
Kwok 2001	0 1	2	0	2	2	2	0	0	1
Larsson 1990	2 0	0	0	0	1	1	0	0	1
Lauque 2000	1 1	0	2	0	2	2	0	0	1
Lauque 2004	1 1	0	2	0	2	1	0	0	2
MacFie 2000	2 1	0	2	2	2	0	0	0	1
Manders 2006	1 1	2	2	0	2	2	2	2	1
McEvoy 1982	1 2	0	0	0	1	1	0	0	0
McWhirter 1996	1 1	0	2	0	1	2	0	0	1
Madigan 1994	1 0	0	0	0	2	1	0	0	1
Meredith 1992	1 1	0	2	2	2	1	0	0	1
Ovesen 1992	1 1	1	0	2	2	1	2	2	0
Payette 2002	1 2	1	2	0	2	0	0	0	1
Payette 2004	1 0	1	0	0	2	1	0	0	1
Potter 2001	2 2	1	1	2	2	2	0	0	0
Price 2005	2 2	0	1	0	2	1	0	0	1
Rosendahl 2006	2 2	2	2	1	2	2	2	2	2
Salas‐Salvado 2005	1 1	0	2	0	2	1	0	0	1
Saudny 1997	2 1	1	1	0	2	1	0	0	1
Schols 1995	1 1	2	0	2	2	2	0	0	1
Scorer 1990	1 0	0	2	0	2	0	1	0	1
Stableforth 1986	1 1	0	0	2	2	1	0	0	0
Steiner 2003	2 2	1	2	2	2	1	2	2	1
Tidermark 2004	2 2	1	2	2	2	0	0	0	2
Vermeeren 2004	1 1	1	2	2	2	0	2	2	1
Vlaming 2001	2 2	1	2	0	2	2	2	2	0
Volkert 1996	1 1	0	2	0	2	2	0	0	2
Woo 1994	1 2	2	2	2	2	0	0	0	1
Wouters 2002	1 1	0	1	2	2	2	2	2	1
Wouters 2003	0 1	0	1	2	2	2	2	2	2
Wouters 2005	2 1	2	2	0	2	2	2	2	1
Wouters 2006	1 1	0	2	0	2	1	0	0	1
Yamaguchi 1998	2 0	0	2	2	1	1	2	0	2
Young 2004	2 2	0	1	0	2	2	0	0	1
Footnotes: allocation concealment: A ‐ adeequate, B ‐ unclear, C ‐ not adequate; ITT: intention‐to‐treat; 0 = worst, 2 = best

Data and analyses

Comparison 1. Oral protein and energy versus routine care.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality	42	8031	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.81, 1.04]
2 Mortality: Subgroup analysis for nutritional status	40	7869	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.80, 1.03]
2.1 Undernourished	25	2466	Risk Ratio (M‐H, Fixed, 95% CI)	0.79 [0.64, 0.97]
2.2 Nourished	16	5403	Risk Ratio (M‐H, Fixed, 95% CI)	0.98 [0.83, 1.14]
3 Mortality: Subgroup analysis for kcal offered per day	38	8165	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.79, 1.01]
3.1 400 kcal or more/day	24	7307	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.78, 1.00]
3.2 <400 kcal/day	14	858	Risk Ratio (M‐H, Fixed, 95% CI)	1.09 [0.59, 1.98]
4 Mortality: Subgroup analysis for age category	40	8049	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.80, 1.03]
4.1 Mean age 75 years or more	30	2967	Risk Ratio (M‐H, Fixed, 95% CI)	0.85 [0.69, 1.05]
4.2 Mean age <75 years	12	5082	Risk Ratio (M‐H, Fixed, 95% CI)	0.94 [0.80, 1.11]
5 Mortality: Subgroup analysis for period of supplementation	38	7608	Risk Ratio (M‐H, Fixed, 95% CI)	0.93 [0.82, 1.06]
5.1 <35 days supplementation	12	5154	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.78, 1.07]
5.2 35 days or more of supplementation	26	2454	Risk Ratio (M‐H, Fixed, 95% CI)	0.97 [0.77, 1.24]
5.3 Mean age <75 years	0	0	Risk Ratio (M‐H, Fixed, 95% CI)	0.0 [0.0, 0.0]
6 Mortality: Subgroup analysis for wellness	41	8029	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.81, 1.04]
6.1 Well	6	393	Risk Ratio (M‐H, Fixed, 95% CI)	0.98 [0.25, 3.78]
6.2 Unwell	35	7636	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.81, 1.04]
7 Mortality: subgroup analysis for hospital or community	35	7548	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.81, 1.04]
7.1 In‐patients	21	6582	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.80, 1.04]
7.2 Community	14	966	Risk Ratio (M‐H, Fixed, 95% CI)	0.99 [0.62, 1.59]
8 Mortality: Sensitivity analysis	41		Risk Ratio (M‐H, Fixed, 95% CI)	Subtotals only
8.1 Category 'A' concealment of allocation	15	6604	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.79, 1.03]
8.2 Exclusion of Larsson 1990	40	7584	Risk Ratio (M‐H, Fixed, 95% CI)	0.96 [0.84, 1.09]
8.3 Exclusion of trials with known commmercial involvement	29	7190	Risk Ratio (M‐H, Fixed, 95% CI)	0.90 [0.79, 1.03]
9 Mortality: Subgroup analysis by diagnostic group	38	7496	Risk Ratio (M‐H, Fixed, 95% CI)	0.89 [0.78, 1.01]
9.1 Geriatric conditions	23	2701	Risk Ratio (M‐H, Fixed, 95% CI)	0.78 [0.62, 0.98]
9.2 Hip fracture	8	437	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.50, 1.66]
9.3 Congestive heart failure	1	22	Risk Ratio (M‐H, Fixed, 95% CI)	1.44 [0.10, 20.21]
9.4 Chest conditions	2	120	Risk Ratio (M‐H, Fixed, 95% CI)	3.07 [0.13, 73.30]
9.5 Perioperative	1	100	Risk Ratio (M‐H, Fixed, 95% CI)	1.33 [0.16, 11.38]
9.6 Stroke	2	4063	Risk Ratio (M‐H, Fixed, 95% CI)	0.93 [0.79, 1.10]
9.7 Diabetic conditions	1	53	Risk Ratio (M‐H, Fixed, 95% CI)	1.04 [0.07, 15.75]
10 Participants with complications	24	6225	Risk Ratio (M‐H, Fixed, 95% CI)	0.86 [0.75, 0.99]
11 Participants with complications: Subgroup analysis by diagnostic group	22	5727	Risk Ratio (M‐H, Fixed, 95% CI)	0.87 [0.76, 1.00]
11.1 Geriatric conditions	8	1026	Risk Ratio (M‐H, Fixed, 95% CI)	0.91 [0.77, 1.08]
11.2 Hip fracture	6	298	Risk Ratio (M‐H, Fixed, 95% CI)	0.60 [0.40, 0.91]
11.3 Chest conditions	3	165	Risk Ratio (M‐H, Fixed, 95% CI)	1.25 [0.58, 2.73]
11.4 Perioperative	1	100	Risk Ratio (M‐H, Fixed, 95% CI)	2.11 [0.68, 6.54]
11.5 Stroke patients	2	4063	Risk Ratio (M‐H, Fixed, 95% CI)	0.65 [0.40, 1.03]
11.6 Congestive heart failure	1	22	Risk Ratio (M‐H, Fixed, 95% CI)	7.0 [0.38, 130.56]
11.7 Diabetic conditions	1	53	Risk Ratio (M‐H, Fixed, 95% CI)	0.86 [0.54, 1.35]
12 % Weight change	45	3058	Mean Difference (IV, Fixed, 95% CI)	2.15 [1.80, 2.49]
13 % Weight change: Subgroup analysis by diagnostic group	44	3056	Mean Difference (IV, Fixed, 95% CI)	2.15 [1.80, 2.49]
13.1 Geriatric conditions	32	2387	Mean Difference (IV, Fixed, 95% CI)	2.65 [2.19, 3.10]
13.2 Hip fracture	4	235	Mean Difference (IV, Fixed, 95% CI)	0.58 [‐1.04, 2.19]
13.3 Chest conditions	5	284	Mean Difference (IV, Fixed, 95% CI)	1.58 [0.99, 2.17]
13.4 Perioperative	1	100	Mean Difference (IV, Fixed, 95% CI)	‐1.90 [‐6.43, 2.63]
13.5 Stroke patients	1	31	Mean Difference (IV, Fixed, 95% CI)	1.58 [‐5.55, 8.71]
13.6 Congestive heart failure	1	19	Mean Difference (IV, Fixed, 95% CI)	1.43 [‐7.89, 10.75]
14 % Weight change sensitivity analysis	19	1599	Mean Difference (IV, Fixed, 95% CI)	2.07 [1.68, 2.46]
14.1 No inferences required	19	1599	Mean Difference (IV, Fixed, 95% CI)	2.07 [1.68, 2.46]
15 % Arm muscle circumference change	16	1382	Mean Difference (IV, Fixed, 95% CI)	1.20 [0.45, 1.96]
16 % Arm muscle circumference change: Subgroup analysis by diagnostic group	16	1382	Mean Difference (IV, Random, 95% CI)	1.25 [0.22, 2.28]
16.1 Geriatric conditions	11	1216	Mean Difference (IV, Random, 95% CI)	1.00 [‐0.20, 2.21]
16.2 Hip fracture	1	10	Mean Difference (IV, Random, 95% CI)	2.92 [0.16, 5.68]
16.3 Chest conditions	2	106	Mean Difference (IV, Random, 95% CI)	3.46 [‐3.40, 10.32]
16.4 Perioperative	0	0	Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
16.5 Stroke patients	1	31	Mean Difference (IV, Random, 95% CI)	0.86 [‐6.27, 7.99]
16.6 Congestive heart failure	1	19	Mean Difference (IV, Random, 95% CI)	‐0.39 [‐9.71, 8.93]
17 Length of Stay	14	5735	Mean Difference (IV, Random, 95% CI)	‐0.75 [‐2.84, 1.34]
18 Length of stay: Subgroup analysis by diagnostic group	13	5290	Mean Difference (IV, Random, 95% CI)	‐1.17 [‐3.90, 1.57]
18.1 Geriatric conditions	4	875	Mean Difference (IV, Random, 95% CI)	‐0.80 [‐6.49, 4.89]
18.2 Hip fracture	6	263	Mean Difference (IV, Random, 95% CI)	‐2.14 [‐7.71, 3.42]
18.3 Chest conditions mean age >65 years	0	0	Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
18.4 Perioperative mean age >65 years	1	100	Mean Difference (IV, Random, 95% CI)	‐2.0 [‐6.53, 2.53]
18.5 Stroke patients mean age >65 years	2	4052	Mean Difference (IV, Random, 95% CI)	‐6.50 [‐25.88, 12.88]
18.6 Congestive heart failure mean >65 years	0	0	Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]
19 Handgrip	7	535	Mean Difference (IV, Fixed, 95% CI)	0.06 [‐0.60, 0.72]

1.1. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 1 Mortality.

1.2. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 2 Mortality: Subgroup analysis for nutritional status.

1.3. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 3 Mortality: Subgroup analysis for kcal offered per day.

1.4. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 4 Mortality: Subgroup analysis for age category.

1.5. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 5 Mortality: Subgroup analysis for period of supplementation.

1.6. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 6 Mortality: Subgroup analysis for wellness.

1.7. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 7 Mortality: subgroup analysis for hospital or community.

1.8. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 8 Mortality: Sensitivity analysis.

1.9. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 9 Mortality: Subgroup analysis by diagnostic group.

1.10. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 10 Participants with complications.

1.11. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 11 Participants with complications: Subgroup analysis by diagnostic group.

1.12. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 12 % Weight change.

1.13. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 13 % Weight change: Subgroup analysis by diagnostic group.

1.14. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 14 % Weight change sensitivity analysis.

1.15. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 15 % Arm muscle circumference change.

1.16. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 16 % Arm muscle circumference change: Subgroup analysis by diagnostic group.

1.17. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 17 Length of Stay.

1.18. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 18 Length of stay: Subgroup analysis by diagnostic group.

1.19. Analysis.

Comparison 1 Oral protein and energy versus routine care, Outcome 19 Handgrip.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Banerjee 1978.

Methods	Method of randomisation: not stated Assessor blinding: dietary assessment not blinded Intention to treat: carried out Lost to follow‐up: intervention 2 withdrawn because refused to continue,1 withdrawn due to increased blood urea; control 0 Timing of intervention: 14 weeks
Participants	Location: long‐stay wards University Hospital, South Manchester 63 patients Inclusion criteria: informed consent, not receiving Complan for 3 months prior to study Exclusion criteria: failure to gain consent Sex: 42 female, 21 male Age: mean age 81y
Interventions	a: two drinks Complan: 265 kcal, 18.6g protein, 26.4g carbohydrate, 9.6 g fat per day in addition to normal food intake b: normal food intake Allocated: 33/30 Assessed: 26/24
Outcomes	Main outcomes: Mortality Additional outcomes: Measures of nutritional status ‐ changes in skinfold thickness from the first non‐supplemented to the second (supplemented) period Measures of dietary intake ‐ changes in mean food intake from the first (non‐supplemented) period to the second (supplemented) period compared in the supplemented and control groups (measured over 5 days)
Notes	Further details (method of randomisation, blinding, similarity of care programmes) obtained from trialist 18/4/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Barr 2000.

Methods	Method of randomisation: not stated Assessor blinding: only reported for blood pressure Intention to treat: 204 randomised, results presented for 200, intention to treat analysis not possible Lost to follow‐up: intervention 3; control 1 Timing of intervention: 12 weeks
Participants	Location: 6 medical centres, USA 204 patients Inclusion criteria: Adult men and women between 55 and 85 years and in good health BMI between 16 and 36, 5 years post menopause, consumption of 1.5 servings or fewer per day of dairy foods, willingness to consume additional 3x8oz milk/day Exclusion criteria: Unstable hypertension or dyslipidemia within last month, unstable hormone therapy use within last year, chronic or life threatening diseases, serious abnormality that would interfere with study participation, substance or alcohol abuse, no calcium supplementation for more than 4 weeks before enrolment in the study, diabetes, blood pressure over systolic or 95 mm Hg diastolic, total blood cholesterol greater than 6.75mmol/L, fasting blood glucose level greater than 7.8mmol/L Sex: 35‐36% male, 64‐65% female Age: mean age 65y Living at home
Interventions	a: Participants added 3 x 8 oz servings of fluid milk daily (low fat or fat free) to their usual consumption of dairy products b: Participants maintained their usual diets Allocated: 101/103 Assessed: 98/102 for all outcomes
Outcomes	Main outcomes: Additional outcomes: Anthropometric indices ‐ weight change Functional status ‐ Barthel index, mental health inventory, general perceived health scale, work activity scale Measures of dietary intake ‐ change in energy and protein intake
Notes	No further information required
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Benati 2001.

Methods	Method of randomisation: not stated ‐ allocation concealment: B Assessor blinding: not mentioned Intention to treat: unclear Lost to follow‐up: unclear Timing of the intervention: 2 weeks Length of follow‐up: 2 weeks
Participants	Location: Department of Geriatric Medicine, Forli, Italy 10 patients Inclusion criteria: Inpatients with severe cognitive impairment. MMSE 15 or less (maximum 30), and pressure ulcers Exclusion criteria: Unlikely to benefit from nutritional supplementation Sex: 3 female 7 male Age: 71‐91 years
Interventions	a: Normal hospital diet and 2 x 200ml/day of high protein and calorie supplementary feeding (500 kcal, 37g protein approx) b: Normal hospital diet Allocated: 5?/5? Assessed: 5/5
Outcomes	Main outcomes: pressure sore status Additional outcomes: Measures of nutritional status
Notes	Third arm received supplement enriched with arginine
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Bonnefoy 2003.

Methods	Method of randomisation: remote allocation ‐ allocation concealment: A Assessor blinding: not mentioned Intention to treat: unclear for one participant Lost to follow‐up: unclear which arm 5 clinical events, 6 retracted consent, 3 dropped out Timing of the intervention: Twice daily between meals Length of follow‐up: 9 months
Participants	Location: 16 retirement homes in Lyon, France 57 patients Inclusion criteria: Multiple diagnoses, length of stay at least 3 years in retirement homes, over 72 years, defined as frail by GP Exclusion criteria: Uncontrolled or rapidly evolving diseases, dementia, type 1 diabetes, severe renal insufficiency, functional handicap preventing exercising, long‐term corticosteroid therapy with receipt of vitamins before the study Sex: 50 female 7 male Age: 83 (?SEM 0.91), 83 (1.24)
Interventions	a: 2 x 200 ml, total intake 1686 kJ (400 kcal) 30% Prot, 50% CHO, 20% Fat, providing approximately 50% of the RDA for vitamins and minerals, four different flavours in unmarked containers, twice daily at 10.00 and 16.00 b: 4 different flavours in unmarked containers, neither energy, protein, vitamins or minerals Allocated: 28/29 Assessed: 22/20
Outcomes	Main outcomes: mobility and muscle power Additional outcomes: compliance to supplements Measures of nutritional status‐ BMI, Fat free mass
Notes	Factorial design half participants also had exercise programme and half memory training. One death unclear from which group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Bourdel 2000.

Methods	Method of randomisation: Cluster randomised Assessor blinding: no Intention to treat: carried out Lost to follow‐up: intervention 0; control 0 Timing of the intervention: 15 days or less if discharged Length of follow‐up: 15 days
Participants	Location: Bordeaux, France 672 patients Inclusion criteria: Ward had >40% of patients >65years, patients in an acute phase of critical illness, unable to move by themselves, unable to eat independently at admission Exclusion criteria: patients with pressure ulcers on admission Sex: 437 female, 235 male Age: 83.6 (SD7.3), 83.0 (7.1)
Interventions	a: 2 x 200 kcal commercial supplements with breakfast and mid afternoon (400 kcal, 30g protein) and assistance with meals b: usual nutritional care Allocated: 295/377 (Cluster randomised) Assessed: 295/377
Outcomes	Main outcomes: mortality, pressure ulcers Additional outcomes: Measures of dietary intake ‐energy and protein intake
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Broqvist 1994.

Methods	Method of randomisation: not stated Assessor blinding: not reported Intention to treat: carried out, Lost to follow‐up: one unable to continue because of illness, 2 died Timing of intervention: 8 weeks
Participants	Location: not given, presumably University Hospital, Linkoping, Sweden 19 patients Inclusion criteria: Patients with severe congestive heart failure (NewYork Heart Association functional class III‐IV) Exclusion criteria: Diabetes mellitus, severe liver or renal insufficiency Sex: 3 female, 19 male Age: mean age 72y
Interventions	a: 500 ml oral nutritional supplement (Biosorb 1500 (Pharmacia, Germany) 750 kcal, 30 g protein) b: 1:10 diluted version of intervention as placebo (7.5 kcal, 3 g protein) Allocated: 9/13 Assessed: 7/12
Outcomes	Main outcomes: Mortality Morbidity and complications ‐NYHA (New York Heart Association) functional class, complications (renal failure and diabetic coma), malnourished after 8 weeks Additional outcomes: Measures of nutritional status ‐ weight, triceps skinfold (mm), arm muscle circumference (cm) Measures of dietary intake ‐ energy and protein intake
Notes	Request for further details sent 11/5/01 on inpatient or outpatient status, location of trial and further details regarding composition of supplement and how it was provided
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Brown 1992.

Methods	Method of randomisation: alternating number Assessor blinding: blinded Intention to treat: carried out Lost to follow‐up: no losses to follow‐up Timing of intervention: from second day of admission until discharge
Participants	Location: hospital, Ipswich, UK 10 patients Inclusion criteria: thin (based on weight for height, triceps skinfold, mid‐arm circumference ‐ two out of three more than one standard deviation below reference mean), elderly, females with hip fracture Exclusion criteria: malignant disease, mental illness, renal or hepatic failure, neurological disorder, stroke, diabetes Sex: all female Age: not given, but "elderly"
Interventions	a: Patient offered oral nutritional supplement Fresubin (Fresenius) calculated to make up deficit between intake from normal hospital diet and requirement. Fresubin provides 4.2 kJ or 1kcal/ml, as 15% protein energy, 30% fat energy and 55% carbohydrate energy b: Normal hospital diet Allocated: 5/5 Assessed: 5/5
Outcomes	Main outcomes: Mortality Morbidity and complications ‐ pressure sore Length of stay ‐ days to discharge from orthopaedic surgeon Postoperative functional status ‐ two stage walking goals Additional outcomes: Measures of nutritional status ‐ percentage losses in weight, triceps skinfold, midarm circumference, arm muscle circumference Measures of dietary intake
Notes	Author provided protocol of trial and information on method of randomisation and outcome assessment. Request for further details (other outcomes, period of follow‐up) sent 19/5/99, resent 3/2/00
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Bruce 2003.

Methods	Method of randomisation: quasi‐randomised by year of birth Assessor blinding: not reported Intention to treat: unclear Lost to follow‐up: intervention 3; control 1 Timing of intervention: started within 2 to 3 days after surgery, for 28 days Length of follow‐up: 6 months
Participants	Location: Royal Perth hospital, Freemantle, Australia 109 patients Inclusion criteria: female patients with hip fracture, consent given Exclusion criteria: BMI <20 or >30 kg/m2, nursing home resident, resident outside metropolitan Perth (preventing follow‐up), diseases expected to influence nutritional intake (malignancy, severe organ failure), diabetes (to avoid potential hyperglycaemia), fracture due to major trauma Sex: 109 female Age: mean 84 years
Interventions	a: One 235 ml can of Sustagen Plus daily (Mead Johnston), providing 352 kcal or 1.47 MJ, 17.6 g protein, 11.8 g fat, 44.2 g carbohydrate, vitamins and minerals; chocolate and vanilla flavours. Dietitian carried out preliminary taste test and offered encouragement and strategies to help with compliance, e.g. ways to alter taste and timing of supplement. And routine care b: Routine care Allocated: 50/59 Assessed: 47/58 (mortality)
Outcomes	Main outcomes: mortality Length of stay ‐ hospital Postoperative functional status ‐ % with fall in Katz score, level of care and extent of support required after discharge ‐ % discharged home, % home at 6 months Additional outcomes: Anthropometric indices ‐ weight Nutritional indicators in blood ‐ albumin Patient compliance ‐ consumption of cans of supplement
Notes	More information requested and received August 2003 Classified as nourished, acute admission and hospitalised.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Carver 1995.

Methods	Method of randomisation: not stated Assessor blinding: unclear whether all or some outcomes Intention to treat: not carried out, those unable to consume supplements not included in the analysis Lost to follow‐up: intervention 2 withdrawn due to reluctance to take extra drinks, 1 withdrawn due to objection to being weighed and measured; control 1 withdrawn due to reluctance to take extra drinks, 2 withdrawn due to objection to being weighed and measured Timing of intervention: 12 weeks, twice daily
Participants	Location: All residents of care of the elderly wards at a large psychiatric teaching hospital, Edinburgh 46 patients Inclusion criteria: diagnosed as having some degree of senile dementia, BMI 15.1‐19.9 Exclusion criteria: physical pathology, likely to be discharged during the study period, no consent from relatives Sex: 36 female, 10 male Age: men: mean age 68‐69y, women: 79‐80y Health Status: malnourished, senile dementia
Interventions	a: 200 ml oral supplement Fortisip twice daily providing 600 kcal/day in addition to normal meals b: 200 ml oral vitamin preparation twice daily providing the same vitamins as Fortisip, but virtually no macronutrients, in addition to normal meals Allocated: 23/23 Assessed: 20/20
Outcomes	Main outcomes: Mortality Additional outcomes: Measures of nutritional status ‐ weight change, body mass index, mid upper arm circumference, triceps skinfold Patient compliance ‐ numbers consuming all drinks offered
Notes	Request for further details sent 18/9/01 on method of randomisation, blinding of outcome assessors and treatment providers and further details of anthropometry outcomes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Collins 2005.

Methods	Method of randomisation: not stated Assessor blinding: only study nurse and participants blinded not research dietitian or investigator Lost to follow‐up: unclear Intention to treat: very poor detail Lost to follow‐up: unclear Timing of the intervention: given for 4 weeks 3 times/day with meals Length of follow‐up: 4 weeks
Participants	Location: elderly home nursed patients referred for wound management, Australia 38 patients Inclusion criteria: over 60 years old, informed consent, all types of wounds‐ skin grafts, lacerations, skin tears, ulcers, pressure ulcers, post surgical wounds Exclusion criteria: allergy or intolerance to milk based products Sex: 55.3% male Age: 79.2 (SD 6.3), 81.0 (SD 9.5)
Interventions	a: 237 ml of 8 kJ/ml oral nutritional supplement b: 237 ml of 4 kJ/ml oral nutritional supplement Allocated: 18/20 Assessed: unclear
Outcomes	Main outcomes: Dietary intake, cognitive function, quality of life, wound healing Additional outcomes:
Notes	Insufficient description of allocation concealment
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Daniels 2003.

Methods	Method of randomisation: pharmacy maintained, computer generated, opaque envelopes concealment: A Assessor blinding: yes Intention to treat: yes Lost to follow‐up: intervention 2 withdrawn; control 1 withdrawn Timing of the intervention: from day 7 post surgery. From discharge to 6 weeks Length of follow‐up: 12 weeks
Participants	Location: Elderly people post lower limb fracture 100 patients Inclusion criteria: Nutritionally at risk (MAC 25th percentile or less) Exclusion criteria: did not reside within southern Adelaide, unable to comprehend instructions relating to positioning of upper arm for eligibility assessment, unable to fully weight bear on the side of injury for more than 7 days post admission, not independently mobile post fracture, medically unstable more than 7 days post admission, suffering from cancer, chronic renal failure, unstable angina or unstable diabetes Sex: 38/41 female Age:83y/84y
Interventions	a: 1.5 kcal/ml sip feed for 6 weeks, individually prescribed, home visits 3 x week from discharge to 6 weeks b: home visits 3 x week from discharge to 6 weeks Allocated: 49/51 Assessed: 45/48
Outcomes	Main outcomes: mortality. mobility and muscle power Additional outcomes: compliance to supplements Measures of nutritional status‐ % weight change
Notes	Allocated to nutrition alone or nutrition plus resistance training exercise, resistance training exercise alone, or neither
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Deletter 1991.

Methods	Method of randomisation: not stated Assessor blinding: not mentioned Intention to treat: carried out Lost to follow‐up: intervention 0; control 0 Timing of intervention: 9 weeks
Participants	Location:Veterans Affairs Medical Center‐ Lexington 37 patients Inclusion criteria: outpatients, over 55 years old, irreversible airway obstruction, no COPD exacerbation within 4 weeks of enrolment, less than 90% ideal body weight, ambulatory, English speaking, able to communicate verbally and in writing Exclusion criteria: muscular discomfort of chest wall, pain on inspiration, diabetes, thyroid disease, neoplastic disease, serious heart disease, alcoholism, hepatic failure, renal failure, malabsorption, surgery within 3 months of the study. Sex: all male Age: mean age 67y
Interventions	a: high fat, low carbohydrate formula (Pulmocare, Ross laboratories, Columbus, OH) 1 can/day, 16.7% protein, 28% carbohydrate, 55% fat b: normal dietary routines Allocated: 37 Assessed: 18/17
Outcomes	Main outcomes: Lung function, 6 and 12 minute walking distance Additional outcomes: Measures of nutritional status‐weight, changes in skinfold thickness from baseline to the end of the supplemented period Measures of dietary intake‐ energy and protein intake
Notes	Further details: Request for further details (energy content and volume of supplement, blinding) sent 18/10/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Delmi 1990.

Methods	Method of randomisation: not stated Assessor blinding: not reported Intention to treat: appears intention to treat, but denominators unclear Lost to follow‐up: deaths reported, but unclear if other losses to follow‐up Timing of intervention: from admission to orthopaedic unit to end of stay in second (recovery) hospital,given once daily for a mean period of 32 days Length of follow‐up: 6 months
Participants	Location: orthopaedic unit in hospital and recovery hospital, Geneva, Switzerland 59 patients Inclusion criteria: femoral neck fracture after an accidental fall, aged over 60y Exclusion criteria: fracture from violent external trauma, pathological fracture due to tumour or non‐osteoporotic osteopathy; overt dementia; renal, hepatic, or endocrine disease; gastrectomy or malabsorption; taking phenytoin, steroids, barbiturates, fluoride or calcitonin Sex: 53 female, 6 male Age: mean age 82y
Interventions	a: 250 ml oral nutritional supplement (1.06 MJ or 254 kcal, 20.4 g protein, 29.5 g carbohydrate, 5.8 g lipid, 525 mg calcium, 750 IU vitamin A, 25 IU vitamin D3, nicotinamide, folate, calcium pantothenate, biotin, minerals; and vitamins E, B1, B2, B6, B12, C) and standard hospital diet b: Standard hospital diet Allocated: 27/32 Assessed: unclear/unclear at 6 months
Outcomes	Main outcomes: Mortality Morbidity and complications ‐ complications (total, bedsore, severe anaemia, cardiac failure, infection, gastrointestinal ulcer, other) Side effects of treatment ‐ favourable clinical course (excludes death, major complication, or two or more minor complications) Length of stay ‐ orthopaedic unit and recovery hospital Additional outcomes: Measures of dietary intake ‐ energy and protein intake
Notes	Numbers of complications unclear, request for further details sent 24/5/99, resent 7/2/00
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Edington 2004.

Methods	Method of randomisation: envelope prepared by statistician‐ allocation concealment: A Assessor blinding: not reported Intention to treat: for some outcomes Lost to follow‐up: 42 withdrawn (unclear which arm) Timing of intervention: 8 weeks from baseline visit within 7 days after discharge from hospital Length of follow‐up: 6 months
Participants	Location: North Staffordshire Hospital NHS trust, Hammersmith NHS Trust and Newcastle upon Tyne Hospital NHS Trust (4 Hospitals) 100 patients Inclusion criteria: female elderly malnourished patients newly discharged from hospital, aged 65 years or older, BMI<20, or <25 with documented evidence of weight loss of at least 10% of body weight in the last 6 months prior to the study period or 5% of more in the last 3 months prior to the study period. Score of 7 or more on the abbreviated mental test. Exclusion criteria: Incapable of taking supplements to provide a minimum of 600 kcal/day, not able to be weighed standing, intolerant of any of the ingredients in the study supplements, history of diabetes, hyperglycaemia or chronic renal failure, requiring Parenteral Nutrition or Enteral feeds as a sole source of nutrition or had been prescribed supplements during the last week of their hospital stay. Informed consent. Sex: 55 female, 25 male
Interventions	a: The choice of one or more nutritional supplements: Ensure Plus tetrapak, Enlive tetrapak, Formance pudding or Ensure Bar, Abbott Laboratories Ltd. Supplements 600‐1000 kcal aim to increase to ideal body weight 0.5kg/ week based on Schofield equations b: Routine care Allocated: 51/49 Assessed: 32/26 (weight)
Outcomes	Main outcomes: mortality Functional status ‐ Handgrip, health related quality of life, requirement for health and social care services, health care costs Additional outcomes: Anthropometric indices ‐ weight change, %AMC change, dietary intake change
Notes	Classified as malnourished, unwell, acute admission and in the community Commercial trial
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Eneroth 2004.

Methods	Method of randomisation: states randomisation but no description Assessor blinding: yes Intention to treat: withdrew patients if hospitalised or non‐compliant Lost to follow‐up: intervention 8 (withdrawn excluding 1 death); control 3 (withdrawn excluding 1 death) Timing of the intervention: daily for 6 months taken between meals Length of follow‐up: 2 years
Participants	Location: Dept of Internal Medicine, Lund University Hospital, Sweden 53 patients Inclusion criteria: patients over 60 years with diabetes mellitus and a Wagner grade or 2 foot ulcer over 4 weeks duration, distal blood pressure measured in the last 3 months
Interventions	a: 400 ml Fortimel (1 kcal/ml Nutricia AB, Netherlands) b: 400 ml placebo Allocated: 26/27 Assessed: 17/23
Outcomes	Main outcomes: Mortality, wound healing, amputations, nutritional status, compliance Additional outcomes: wound healed at 6 months, amputations
Notes	No description of allocation concealment
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Fiatarone 1994.

Methods	Method of randomisation: not stated Assessor blinding: not reported Intention to treat: carried out Lost to follow‐up: intervention 1 dropped out due to lack of interest, 1 lost to follow‐up (excluding death); control 1 dropped out due to lack of interest, 3 lost to follow‐up (excluding death) Timing of interventions: once per day in the evening for 10 weeks, administered in blinded containers
Participants	Location: Hebrew Rehabilitation Center for the Aged, Boston 50 patients no exercise, 50 patients with exercise Inclusion criteria: residents in long‐term care, over 70y, ability to walk 6 metres unaided Exclusion criteria: severe cognitive impairment, rapidly progressive terminal illness, acute illness, unstable chronic illness, myocardial infarction or fracture within 6 months, IDDM, weight loss diet, undergoing resistance training, musculoskeletal cardiovascular abnormality, BMI over 32 Sex: 31 female, 19 male (no exercise), 32 female, 18 male (exercise) Age: 86/89 (no exercise) Health Status: 'healthy' residents of long term care, undernourished analysed separately
Interventions	a: usual diet plus Exceed (Ross laboratories) 240 ml: 360 kcal, 60% carbohydrate, 23% fat, 17% protein b: usual diet plus Crystal light (4 kcal) Allocated: no exercise 24/26, exercise 25/25 Assessed: unclear
Outcomes	Main outcomes: Mortality Functional status ‐ stair climbing, physical activity, gait Side effects of treatment ‐ diarrhoea Additional outcomes: Measures of nutritional status ‐ weight change, BMI, skinfold thickness, change in lean body mass Measures of dietary intake ‐ energy and protein intake
Notes	Data from non exercise group only, request for further details regarding numbers of side effects, weight denominators and intake denominators for all groups, blinding of outcome assessors sent 18/9/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

FOOD trial 2005.

Methods	Method of randomisation: remote phone call to gain computer randomisation Assessor blinding: 6 months follow‐up may have been blinded Intention to treat: carried out Lost to follow‐up: intervention 7; control 4 Timing of the intervention: from randomisation until discharge Length of follow‐up: 6 months
Participants	Location: 125 hospitals in 15 countries 4023 patients Inclusion criteria: able to swallow, uncertainty of clinician about need to use supplements, recent stroke (first or recurrent no more than 7 days before admission), patient or relative consent within 30 days of admission or within 30 days of a stroke occurring in hospital Exclusion criteria: subarachnoid haemorrhage Sex: 2149 male, 1874 female Age: 71 (SD 12), 71 (SD 13)
Interventions	a: protein energy supplement 360 ml (2.26MJ, 22.5g Pr), suitable commercial supplements used in most centres e.g. liquid, yoghurt, pudding. Prescribed on drug charts b: normal hospital diet Allocated: 2016/2007 Assessed: 2012/2000
Outcomes	Main outcomes: Mortality or poor outcome, length of stay, quality of life, complications, residence at follow‐up Additional outcomes:
Notes	Adequate description of allocation concealment
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Gariballa 1998.

Methods	Method of randomisation: telephone assignment, balanced in blocks of four Assessor blinding: action taken to blind assessors but some outcome measures may have involved unblinded assessors (ADL, dietary intake) Intention to treat: carried out Lost to follow‐up: intervention 1; control 1 Timing of intervention: up to four weeks duration or until death or discharge Length of follow‐up: 3 months
Participants	Location: Leicester General Infirmary, UK 42 patients Inclusion criteria: Acute ischemic stroke patients (WHO criteria), impaired nutritional status (TSF and MAC greater than or equal to 1SD below the mean), no difficulty swallowing one week after stroke, conscious during the first week after stroke onset Exclusion criteria: Difficulty swallowing one week after stroke, cerebral subarachnoid haemorrhage, active gastrointestinal disease, gastric surgery, biochemical evidence of hepatic or renal impairment, uncontrolled heart failure, diagnosed malignancy, sepsis, persistent swallowing difficulty, malignancy, chronic renal failure, hepatic disease, no consent from patient or next of kin, did not reside locally, not notified of admission, unstable diabetes, failure to gain consent, severe mental impairment, refusal to participate Sex: 21 female, 21 male Age: mean age 79y
Interventions	a: up to 400 ml oral nutritional supplement (Fortisip) providing 600 kcal/day, 20 g protein/day b: standard hospital diet Allocated: 21 /21 Assessed: 18/13 (at 3 months)
Outcomes	Main outcomes: Mortality Morbidity and complications ‐ number of infective complications requiring systemic antibiotics (chest infections, urinary tract infections, septicaemias) Length of acute hospital stay Discharge destination within 3 months Functional status ‐ Barthel Index Additional outcomes: Measures of nutritional status ‐ percentage losses in weight, mid upper arm circumference, triceps skinfold Measures of dietary intake ‐ energy and protein intake
Notes	Request for further details sent 11/5/01 regarding nutritional composition of supplement, how much was provided per day, criteria for evidence of malnutrition, control group access to supplements, whether patients were given assistance with supplements, whether patients who died were included in the calculation for length of stay, infective complications and total complications
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Gariballa 2006.

Methods	Method of randomisation: generated by trial statistician, sequentially numbered opaque envelopes kept in different city, phoned for randomisation Assessor blinding: action taken to blind assessors through use of identical placebo Intention to treat: carried out Lost to follow‐up: intervention 0; control 0 Timing of intervention: up to four weeks duration or until death or discharge Length of follow‐up: 6 months
Participants	Location: university hospital, UK 445 patients Inclusion criteria: acutely ill hospitalised patients aged 65y or more, able to swallow, able to sign consent Exclusion criteria: gastric surgery, malabsorption, BMI > 40kg/m2, coma, severe dementia (Abbreviated Mental Test score < 6), malignancy, living in an institution, already taking supplements Sex: 234 female, 211 male Age: mean age 77y
Interventions	Bottles of supplement given at 08.00 and 12.00 for 6 weeks, including in the community if out of hospital a: 2 bottles of 200ml each, providing in total 995kcal, 50g protein and 100% of reference nutrient intake for vitamins and minerals, and standard hospital diet b: 2 placebo bottles of 200ml each, providing in total 60kcal and no protein or micronutrients, and standard hospital diet Allocated: 223/222 Assessed: 223/222 (at 6 months)
Outcomes	Main outcomes: Mortality Morbidity and complications ‐ infections Length of acute hospital stay Readmissions Functional status ‐ Barthel Index, cognitive function in subgroup Additional outcomes: Measures of nutritional status ‐ weight, mid upper arm circumference, triceps skinfold Measures of dietary intake ‐ energy and protein intake
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Gazzotti 2003.

Methods	Method of randomisation: unmarked envelope Assessor blinding: no mentioned Intention to treat: appears so for mortality, LOS, destination Lost to follow‐up: none Timing of the intervention: for 8 weeks Length of follow‐up: 60 days
Participants	Location: Geriatric ward of Centre Hospitalier de la Citradelle, Liege, Belgium 80 patients Inclusion criteria: all patients aged 75y or over, admitted for acute conditions, short form MNA<11 within 72 hours of admission, followed by full MNA total score 17‐23.5 (at risk of malnutrition) Exclusion criteria: medical condition preventing oral feeding, end‐of‐life patients, severe dementia, presenting clinical signs of dehydration or heart failure, diseases requiring special dietary treatment Sex: 61 female, 19 male Age: 81.5 (SD 7.6), 78.8 (SD 6.1)
Interventions	a: 1 Clinutren soup (1kcal/ml) and 1 Clinutren (1.5 kcal/ml), 500 kcal, 21g protein/day plus standard diet b: standard diet Allocated: 39/41 Assessed: 39/41
Outcomes	Main outcomes: mortality, length of hospital stay, Additional outcomes: side‐effects Measures of nutritional status‐ MNA, weight Measures of dietary intake ‐energy and protein intake
Notes	Classified as acute admission, undernourished, hospitalised
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Gegerle 1986.

Methods	Method of randomisation: not stated Assessor blinding: not mentioned Intention to treat: carried out Lost to follow‐up: intervention 0; control 0 Timing of intervention: as soon as eating post‐op, for duration of hospitalisation, given at 8pm Length of follow‐up: 12 weeks
Participants	Location: hospital, Geneva, Switzerland 16 patients Inclusion criteria: hip fracture post surgery Sex: 13 female, 3 male Age: mean age 77y
Interventions	a: 250 ml drink daily providing 254 kcal, 20 g protein b: normal dietary routines Allocated: 7/9 Assessed: unclear
Outcomes	Main outcomes: Additional outcomes: Measures of nutritional status‐ Measures of dietary intake‐ energy and protein intake
Notes	Request for further details sent 18/10/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Gray‐Donald 1995.

Methods	Method of randomisation: not stated Assessor blinding: carried out apart from nutrition data Intention to treat: carried out Lost to follow‐up: intervention 0; control 0 Timing of intervention: 12 weeks
Participants	Location: Sherbrooke, Quebec, Canada 48 patients Inclusion criteria: over 60y receiving long‐term publicly financed home help services (housework, personal hygiene or food preparation), defined as at nutritional risk according to the following criteria a) involuntary weight loss of greater than 5% of body weight in the last month, 7.5% in the last 3 months or greater than 10% in the last 6 months and BMI less than 27 or b) BMI less than 24 Exclusion criteria: receiving palliative care, alcoholic, active cancer, illness requiring a therapeutic diet incompatible with supplementation Sex: 34 female, 14 male Age: 76/79 Health Status: living at home, nutritionally at risk
Interventions	a: each subject provided with 2x235 mL cans per day of a commercial liquid formula (Ensure, Ensure Plus, or Enrich, Ross Laboratories, Canada) chosen by themselves in order to improve compliance. Providing between 1045‐1480 kJ per can. Ensure 1045 kJ, 8.74 g protein, Enrich with fibre 1085 kJ, 9.4 g protein, Ensure Plus 1480 kJ, 13.0 g protein b: no treatment provided, but visited every week and given encouragement and suggestions to improve the quality of their diets Allocated: 25/25 Assessed: 22/24
Outcomes	Main outcomes: Mortality Functional status ‐ general well being score, number of falls, self‐perceived health, hand grip strength Additional outcomes: Measures of nutritional status ‐ anthropometric indices ‐ change in weight, triceps, supra iliac, subscapular skinfolds Measures of dietary intake ‐ energy intake Patient compliance ‐ 7 or more cans of supplement per week
Notes	No other information required
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Hampson 2003.

Methods	Method of randomisation: unmarked envelope Assessor blinding: not mentioned Intention to treat: unclear Lost to follow‐up: none Timing of the intervention: for 8 weeks Length of follow‐up: 1 year
Participants	Location: London, UK 71 patients Inclusion criteria: community dwelling elderly women, recruited through GPs, aged over 70y, BMI less than or = to 21kg/m2 and osteoporosis at femora; neck and/or total hip Exclusion criteria: progressive wasting disease, severe renal impairment, severe cardiorespiratory disease, endocrine diseases, therapy with drugs interfering with bone metabolism, cognitive impairment Sex: all female Age: 76 (SD 4.2), 76.7 (SD 5.7)
Interventions	a: 1g calcium and 800iu cholecalciferol/day and advice and supplements to increase BMI by 1 kg/m2 or more over 6 months Fortisip and/or Fortijuice one or two cartons (200ml carton provides 300 kcal, 12g protein, 11g fat, 36.8g carbohydrate, vitamins, minerals and trace elements b: 1g calcium and 800iu cholecalciferol and asked not to change standard diet Allocated: 36/35 Assessed: 31/33
Outcomes	Main outcomes: mortality Additional outcomes: physical activity and well‐being Measures of nutritional status‐ weight, Measures of dietary intake ‐energy, protein and fat intake
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Hankey 1993.

Methods	Method of randomisation: not stated Assessor blinding: not reported Intention to treat: not possible Lost to follow‐up: 4 (unclear which arm) Timing of intervention: 8 weeks, offered with routine drug prescription, mid am and mid pm, maxijul (glucose polymer, SHS) put into all drinks at nurse's discretion
Participants	Location: Sanderson Rehabilitation Hospital, Newcastle 14 patients Inclusion criteria: over 75y, continuing care elderly, plasma albumin less than 40g/L Exclusion criteria: non given Sex: 11 female, 3 male Age: 88/87 Health Status: long‐term care, at nutritional risk
Interventions	a: normal hospital food plus Build‐up (Nestlé, Clinitec), unclear kcal/day, unclear g protein/day and unclear kcal/day b: normal hospital diet Allocated: 10/10 Assessed: 7/7
Outcomes	Main outcomes: none Additional outcomes: Measures of nutritional status‐ Anthropometric indices ‐ change in weight, mid upper arm circumference, triceps skinfold Measures of dietary intake ‐ energy and protein intake
Notes	Request for further information sent 18/9/01. Reply received 18/9/01 with information regarding how many had acute illness, whether all outcomes were measured at 8 weeks, method of randomisation, blinding of outcome assessors, whether care programmes were identical, inclusion and exclusion criteria
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Hankins 1996.

Methods	Method of randomisation: sealed, opaque envelopes in blocks of 10, appears stratified by place of residence Assessor blinding: not done Intention to treat: carried out Lost to follow‐up: intervention 0; control 1 Timing of intervention: started within 5 days of surgery, given once in the morning and once in the evening for 30 days, served on meal tray in hospital by nurses, given by family or self‐administered out of hospital Length of follow‐up: 2 months
Participants	Location: acute care in Hornsby Ku‐ring‐gia Hospital and rehabilitation hospitals, Sydney, Australia 32 patients Inclusion criteria: fractured neck of femur after accidental fall; admitted from home, hostel or nursing home; age 65y or older; mid upper arm circumference less than or equal to 25th centile for sex and age Exclusion criteria: malignancy, chronic renal failure, hepatic disease, no consent from patient or next of kin, did not reside locally, not notified of admission, unstable diabetes Sex: 27 female, 5 male Age: mean 86y
Interventions	a: Oral supplement of 250 ml Sustagen twice daily (total daily intake 22.5 g protein, 10 g fat, 60 g carbohydrate, 1.71 MJ or 409 kcal energy, 500 µg vitamin A, 6.6 µg vitamin D, 50.8 mg vitamin C, 1.2 mg thiamin, 1.15 mg riboflavin, 13 mg niacin, 1.3 µg vitamin B12, 825 mg calcium, 670 mg phosphorus, 8 mg iron, 66 µg iodine, 1.2g potassium, 370 mg sodium) plus standard hospital diet b: Standard hospital diet Allocated: 17/15 Assessed: 17/14
Outcomes	Main outcomes: Mortality Morbidity and complications ‐ complications (total, infection, pressure sores, pulmonary embolism, delirium, anaemia, cardiac failure, acute renal failure) Side effects of treatment ‐ favourable clinical course (excludes death, major complication, or two or more minor complications) Length of stay ‐ acute hospital, rehabilitation hospital, and total stay Postoperative functional status ‐ Barthel Index Care required after discharge ‐ place of residence at two months Additional outcomes: Measures of nutritional status ‐ self‐reported weight, mid upper arm circumference, Measures of dietary intake ‐ energy and protein intakes from food and supplement Patient compliance ‐ numbers completing full 30 days of supplement
Notes	Request for further details (blinding of outcome assessors, details of supplement administration, further information on outcomes) sent. Reply from trialist (11/6/99) gave details of outcome assessor blinding, supplement administration and outcomes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Hubsch 1992.

Methods	Method of randomisation: not stated Assessor blinding: not mentioned Intention to treat: carried out Lost to follow‐up: none Timing of intervention: during hospital stay, mean 25 days
Participants	Location: Heidelberg, Germany 32 patients Inclusion criteria: undernourished geriatric patients aged 75y and older Sex: all female Age: over 75y
Interventions	a: 250 ml (238 kcal, 20 g protein) supplement daily in addition to normal hospital diet b: normal hospital diet Allocated: 16/16 Assessed: 16/16
Outcomes	Main outcomes: mortality Additional outcomes: Measures of nutritional status‐weight
Notes	Request for further details sent 20/04/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Jensen 1997.

Methods	Method of randomisation: sealed envelopes Assessor blinding: blinded Intention to treat: appears intention to treat, but denominators not clear Lost to follow‐up: intervention 5 no intake before discharge; control 7 no intake before discharge Timing of intervention: from time of discharge 110 days elective patients (from 10 days post‐operatively), acute patients from discharge to 110 days post‐discharge
Participants	Location: Surgical Department L, University Hospital of Aarhus, Denmark 34 patients over 75y Inclusion criteria: convalescence after hospital discharge following gastrointestinal surgery, elective patients admitted for colorectal surgery randomised before operation and acute patients randomised prior to discharge operated on due to ileus or peritonitis or had gastric or intestinal surgery performed Exclusion criteria: appendicitis, disseminated malignant disease, inflammatory bowel disease, malabsorption, or dementia Sex: 20 female, 14 male Age: elective 78/82, acute 79/77 Health Status: post‐surgical, not necessarily malnourished
Interventions	a: Advice aimed at a protein intake of 1.5g/kg body weight, milk, quark drink, "Top up special" a complete low fat supplement, and "Plus one" a protein and energy supplement containing no fat (Ferrosan, Soborg, Denmark) (latter 2 offered free). Variety of flavours. Choice of supplement based on patient's taste preference and the recommendations of the dietitian based on estimated requirements. Advice also provided during hospital visits and if felt necessary for compliance, during a home visit. b: Discharged without dietetic advice Allocated: unclear Assessed: 14/20
Outcomes	Main outcomes: Additional outcomes: Measures of nutritional status‐ Anthropometric indices ‐ weight change, LBM Measures of dietary intake ‐ energy and protein intake
Notes	Request for further information sent 25/10/01. Reply with further information received 9/11/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Knowles 1988.

Methods	Method of randomisation: crossover study unmarked envelope Assessor blinding: carried out Intention to treat: carried out Lost to follow‐up: intervention 0; control 0 Timing of the intervention: for 8 weeks Length of follow‐up: 8 weeks
Participants	Location: Pulmonary Research Laboratory, British Columbia 25 patients Inclusion criteria: ambulatory patients with severe COPD, FEV1 <50% of predicted, stable phase of disease. Did not have acute exacerbation in 3 months prior to study or during study Exclusion criteria: known eating disorder, infectious process, pulmonary disease other than COPD, lactose intolerance, other medical illness, intolerance of nutritional supplement Sex: 4 female, 21 male Age: 68(SD11), 70(SD11)
Interventions	a: Sustacal 24% protein, 22% fat, 54% carbohydrate, aimed to increase calorie intake by 50% above normal level b: usual nutritional care crossover trial Allocated: 13/12 Assessed: 13/12
Outcomes	Main outcomes: mortality, lung function Additional outcomes: Measures of nutritional status‐ TSF, MAMC weight
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Krondl 1999.

Methods	Method of randomisation: not stated Assessor blinding: not reported Intention to treat: 4 withdrew ? before or after randomisation Lost to follow‐up: 4 withdrew ? before or after randomisation Timing of intervention: 16 weeks, 24 cans delivered every 28 days, 234ml each, no time of day or daily limit specified, encouraged to spread over days of the week
Participants	Location: Southern Ontario, Canada ‐ community living volunteers 71 patients Inclusion criteria: white, North American/European identity, ability to speak English, independent living, free of major illnesses, not requiring a special diet, free from uncontrolled major disease or infection, avoidance of nutritional supplements at least 30 days prior to study, less than 4 servings of fruit or vegetables per day Exclusion criteria: none given Sex: 54 female, 16 male Age: mean age 70y Health Status: healthy elderly volunteers living at home
Interventions	a: self selected diet plus Boost (Mead Johnson Nutritionals), 234 ml oral nutritional supplement 235 kcal, 11.75 g protein, 5.4 g fat, 35.25 g carbohydrate, 263 mg calcium, 3.8 mg, magnesium 106 mg, potassium 491 mg, phosphorus 261 mg, zinc 3.5 mg, copper 0.5 mg, manganese 0.67 mg, vitamin A 376 RE, vitamin D 1.175 ug, vitamin E 3.52 mg,vitamin C 15 mg, thiamin 0.4 mg, riboflavin 0.47 mg, niacin 5.9 NE, panthothenic acid 1.9 mg, vitamin B‐6 0.59 mg, biotin 14 µg, folate 70 µg, vitamin B‐12 0.9 µg b: self selected diet, no supplements Allocated: 35/36 Assessed: 35/36
Outcomes	Main outcomes: Mortality Functional status ‐ SF‐36, general well‐being Measures of nutritional status ‐ BMI Measures of dietary intake ‐ energy and protein intake
Notes	Request for information regarding whether the 4 people withdrew before or after randomisation, supplemented group table 3 why 36 people not 35, sent 18/9/01. Reply received 15/10/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Kwok 2001.

Methods	Method of randomisation: quasi‐randomised Assessor blinding: not mentioned Intention to treat: not possible Lost to follow‐up: intervention 2 dropped out, 1 moved out; control 2 dropped out, 1 moved out Timing of intervention: 7 weeks
Participants	Location: 2 medium sized private nursing homes Hong Kong 52 patients Inclusion criteria: Exclusion criteria: BMI>27kg/m2 Resident less than 6 months, known wasting disease e.g. cancer or thyrotoxicosis, hospital admission or attendance in accident and emergency department in past month, diabetes mellitus requiring regular medication, regular refusal of milk, milk or oral supplement more than once daily, informed consent from subjects or guardians Sex: 28 female, 19 male Age: mean age 80y
Interventions	a: 4 spoonfuls of milk supplement diluted in warm water twice daily b: no supplement Allocated: 28/24 Assessed: 25/20
Outcomes	Main outcomes: mortality, grip strength Additional outcomes: measures of nutritional status‐weight, changes in triceps skinfold and mid‐upper arm circumference measures of dietary intake‐ energy and protein intake, side‐effects
Notes	Further details: Request for further details (energy content and volume of supplement, blinding) sent 18/10/01. Reply received 30/10/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Larsson 1990.

Methods	Method of randomisation: sealed envelopes Assessor blinding: not reported Intention to treat: not done Lost to follow‐up: unclear Timing of intervention: up to 8 weeks, served in the morning and afternoon, when other patients were getting drinks
Participants	Location: 19 wards, University Hospital, Linköping, Sweden 435 patients Inclusion criteria: newly admitted to long‐term medical care, consent to participate, remains hospitalised for over 3 weeks Exclusion criteria: none given Sex: 263 female, 167 male (430) Age: 81y female, 78y male Health Status: long‐term care, 115 with protein‐energy malnutrition, 320 not with protein‐energy malnutrition
Interventions	a: standard hospital diet plus Biosorb drink (Kabi Nutrition, Sweden) 2x200 ml/day providing 400 kcal, 16 g Protein, 16 g Fat, 44.2 g CHO b: standard hospital diet Allocated: 197/238 Assessed: unclear
Outcomes	Main outcomes: Mortality Morbidity and complications ‐ total number pressure sores, number of sores healed Functional status ‐ improvement in modified Norton scale, activity rating
Notes	Request for information sent 21/08/01 regarding vitamin content of supplement, denominators pressure sores, Norton scale, actual patient weights, actual measures of TSF and AMC, resent 14/7/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Lauque 2000.

Methods	Method of randomisation: not stated Assessor blinding: not reported Intention to treat: 6 withdrew from intervention group, intention to treat analysis not carried out Lost to follow‐up: intervention 6 withdrew consent or admitted to hospital; control 0 Timing of intervention: 60 days, strong encouragement was given to consume the entire amount offered
Participants	Location: 8 privately run 80 bed nursing homes in Toulouse, France 35 patients Inclusion criteria: patients aged 65y and over, informed consent from subjects or legal guardian, at risk of malnutrition (MNA 17‐23.5) Exclusion criteria: acute disease, uncertain life expectancy, undergoing chemotherapy, impaired digestion or absorption Sex: 30 female, 5 male Age: mean age 84y Health Status: long‐term care, nutritionally at risk
Interventions	a: nutritional supplements of 300‐500 kcal in addition to regular meals. Four oral supplementation products (Clinutren, Nestle), soup, fruit or dessert each containing 120‐200 kcal, 7.5‐15 g protein and enriched with vitamins and minerals b: no details Allocated: 19/22 Assessed: 13/22
Outcomes	Main outcomes: Mortality Morbidity and complications ‐ illness Functional status ‐ handgrip Additional outcomes: Measures of nutritional status ‐ mini nutritional assessment score Anthropometric indices ‐ weight change, BMI Measures of dietary intake ‐ energy and protein intake
Notes	Data from groups B and C only (RCT part). Request for more information regarding blinding of outcome assessors, content and administration of supplements and mini nutrition assessment scores sent 18/9/01. Reply received 26/9/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Lauque 2004.

Methods	Method of randomisation: randomised by drawing numbers with sealed envelopes Assessor blinding: the dietitian was the only assessor aware of group allocation Intention to treat: no Lost to follow‐up: intervention 7 excluded (did not comply), 1 lost to follow‐up (hospitalised), 1 withdrawn; control 2 withdrawn Timing of the intervention: for 3 months Length of follow‐up: 6 months
Participants	Location: Geriatric wards and day centres, Toulouse area, France 91 patients Inclusion criteria: Patients with Alzheimer's disease aged 65 years and older, and at risk of undernutrition (MNA score = 23.5)
Interventions	a: Clinutren (Nestlé Clinical Nutrition, Noisiel, France) ranging between 300 and 500 kcal/d in addition to the patients' spontaneous food intake. b: usual care Allocated: 46/45 Assessed: unclear
Outcomes	Main outcomes: Mortality, functional status, weight, body composition, dietary intake Additional outcomes: Hospitalization, no denominators provided for fractures or pressure ulcers
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

MacFie 2000.

Methods	Method of randomisation: sealed envelopes Assessor blinding: not blinded Intention to treat: carried out Lost to follow‐up: apparently none Timing of intervention: patients were instructed to drink the supplements in addition to and not in place of their normal diet and continued up until the day before surgery, pre op minimum 10 days mean 15d (5‐59), post op minimum 7 days mean 8d (0‐20), post op supps as soon as oral fluids were permitted up until 4 weeks after discharge. 4 groups; preop supps, postop supps, both, neither
Participants	Location: preoperative outpatient phase and postoperative inpatient phase, combined gastroenterology unit at Scarborough hospital 100 patients Inclusion criteria: patients requiring major elective gastrointestinal surgery. Exclusion criteria: dementia, major concurrent, metabolic problems, such as uncontrolled diabetes, advanced liver disease, or uraemia, those requiring emergency surgery. Patients were withdrawn if any member of the clinical or nutrition team considered that adjuvant parenteral or enteral support was indicated or whether it was deemed clinically appropriate to proceed with the surgery within 10 days. Sex: 54 female, 46 male Age: mean ages 63, 68, 66, 64y Health Status: patients receiving gastrointestinal surgery, most patients well nourished
Interventions	a: 2 x 200 mL cartons (Fortisip, Nutricia Ltd) in a variety of flavours providing 1.5 kcal, 0.05 g protein and 0.18 g carbohydrate per ml. A fruit flavoured supplement (fortijuice, Nutritia Ltd) was available as an alternative, providing 1.25 kcal, 0.025 g protein and 0.285 g carbohydrate per ml b: usual diet Allocated: 24/24/27/25 Assessed: 24/24/27/25
Outcomes	Main outcomes: Mortality Morbidity and complications ‐ complications (total, septic) Length of stay ‐ post operative stay Side effects of treatment ‐ nausea with supplements Functional status ‐ hand grip, postoperative anxiety and depression Additional outcomes: Measures of nutritional status‐ Anthropometric indices ‐ perioperative weight loss, mid upper arm circumference Measures of dietary intake ‐ energy intake
Notes	Analysed 75/25, request for information regarding how many measured for outcomes, why the supplements were stopped, the number of women in group 2 and when the patients died, sent 18/9/01. Reply received 1/10/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Madigan 1994.

Methods	Method of randomisation: not stated Assessor blinding: not reported Intention to treat: not carried out, results presented for 30 patients, 34 randomised, results from the two supplemented groups were combined Lost to follow‐up: unclear Timing of intervention: started on admission for 10 days, once daily after evening meal Length of follow‐up: 3 months post‐discharge
Participants	Location: Illawarra Regional Hospital, Port Kembla Campus, Woolongong, Australia 34 patients Inclusion criteria: femoral neck fracture resulting from an accidental fall, age over 60y, informed consent Exclusion criteria: pathological fracture due to tumour; fracture due to violent external trauma; elective total hip replacement; renal, hepatic, metastatic or endocrine (affecting skeletal metabolism) disease; admitted from nursing home; failure to gain consent; transferred to another hospital for surgery Sex: 22 female, 8 male, of 30 Age: all over 60y
Interventions	a: 250 ml oral supplement prepared by dietitian from ProMod (protein powder) and Polyjoule (glucose polymer) providing 1.30 MJ or 310 kcal; 16 g protein, 41.4g carbohydrate, 9.2 g fat, 0.19 mg riboflavin, 245 mg calcium, phosphorus 171 mg, and standard hospital diet. b: One multivitamin/mineral tablet daily ((ELEVIT RDI, Roche) providing 750 µg vitamin A, 1.1 mg thiamin, 1.7 mg riboflavin, 20 mg nicotinamide, 7 mg pantothenic acid, 1.9 mg pyridoxine, 2 µg vitamin B12, 200 µg biotin, 200 µg folic acid, 30 mg vitamin C, 200 IU vitamin D3, 15 IU vitamin E, 125 mg calcium, 100 mg magnesium, 125 mg phosphorus, 5 mg iron, 1mg copper, 1mg manganese, 7.5 mg zinc 250 ml), plus oral supplement as above, and standard hospital diet c: Standard hospital diet Allocated: unclear Assessed: 18/12 (a+b/c)
Outcomes	Main outcomes: Mortality Morbidity and complications ‐ numbers of complications (urinary infections, wound infections/delayed healing, pressure sores, pneumonia, deep venous thrombosis, sepsis Length of stay ‐ acute hospital Postoperative functional status ‐ number transferred to rehabilitation hospital, days to reach partial or full weight bearing with support, days to reach independent mobility Care required after discharge ‐ discharge to home, hostel, nursing home, number of subjects returning to pre‐morbid mobility Additional outcomes: Measures of nutritional status ‐ mid upper arm circumference (*nr), triceps skinfold Measures of dietary intake ‐ energy and protein intakes from food and supplements Patient compliance ‐ number taking protein supplement for only 7 days
Notes	In the trial report, the two supplemented groups were combined for analysis for comparison with control group. Three subjects eliminated post‐randomisation from analysis because only took protein supplement for 7 days, and one eliminated for developing diabetes. Numbers of patients assigned/assessed not always clear. Request for further details sent 4/2/00
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Manders 2006.

Methods	Method of randomisation: states randomised but no further details Assessor blinding: double‐blind placebo‐controlled trial Intention to treat: not undertaken Lost to follow‐up: 65 (withdrawn, excluded after randomisation or lost to follow‐up)‐ unclear which arm Timing of intervention: 24 weeks Length of follow‐up: 24 weeks
Participants	Location: homes for elderly or sheltered housing, The Netherlands 176 participants Inclusion criteria: 60 years or older, BMI 30kg/m2 or less Exclusion criteria: serious morbidity (cancer, severe infection, parenteral or tube feeding, gastrointestinal disorders, terminal care); interfering co‐interventions (medications or supplements with effect on safe administration of intervention) Sex: 122 female, 54 male Age: mean age 83y
Interventions	a: 2 x 125ml daily oral liquid supplements with total of 250kcal and 8.75g protein, providing 25‐175% of Dutch RDA of micronutrients b: placebo with water, sweetener, cloudifier, thickening, flavouring, colour, non‐caloric sweetener Allocated: 119/57 Assessed: 78/33
Outcomes	Main outcomes: Mortality Morbidity and complications ‐ number developing illness Functional status ‐ Barthel Index, Frail Elderly Functional Capacity, Berkhov feeding scale, cognitive function Additional outcomes: Measures of nutritional status ‐ weight change Measures of dietary intake ‐ energy and protein intake
Notes	Emailed Dr Manders on 10/01/2008 to ask for death and illness information according to allocation. No information received 16 March 2008 therefore not included
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

McEvoy 1982.

Methods	Method of randomisation: not stated Assessor blinding: not reported Intention to treat: apparently no withdrawals Lost to follow‐up: none Timing of intervention: 4 weeks, patients with malabsorption in both groups received any necessary supplements such as calcium, vitamin D and haematinics, not given encouragement to eat extra food
Participants	Location: acute geriatric ward, Freeman Hospital, Newcastle upon Tyne 51 patients Inclusion criteria: elderly patients admitted to an acute geriatric ward. Poorly nourished according to at least two of the following criteria; weight below 85% of ideal weight for height, triceps skinfold thickness below 85% of standard values, serum albumin less than 34g/l. Exclusion criteria: malignant conditions, metabolic disease such as thyrotoxicosis or diabetes Sex: not given Age: not given Health Status: acute geriatric ward, poorly nourished
Interventions	a: in addition to normal hospital diet, two sachets of Build‐up daily providing 36.4 g Protein and 644 kcal, states consumption ensured by the nursing staff b: normal hospital diet Allocated: 26/25 Assessed: 26/25
Outcomes	Main outcomes: Mortality Additional outcomes: Measures of nutritional status‐ Anthropometric indices ‐ weight change, mid upper arm circumference, triceps skinfold, arm muscle circumference
Notes	More information requested regarding baseline characteristics of study population, care programmes, blinding of outcome assessment, details of supplement sent 14/7/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

McWhirter 1996.

Methods	Method of randomisation: not stated Assessor blinding: not mentioned Intention to treat: states ITT but not clear Lost to follow‐up: 3 withdrawals (unclear which arm) Timing of the intervention: minimum of 7 days Length of follow‐up: until discharge or stopped
Participants	Location: Ninewells Hospital, Dundee 61 patients Inclusion criteria: patients admitted to the medical unit identified as malnourished using anthropometry, BMI < 20 TSF and/or MAMC < 5th percentile Exclusion criteria: none given Sex: not given Age: 69 / 74y
Interventions	a: Tonexis, (Clintec Nutrition Ltd)100 kcal, 3.75g protein, 3.33g fat, 13.8g CHO according to energy requirements and corrected for stress and activity b: no supplement Allocated: ?/? Assessed: 35/26
Outcomes	Main outcomes: none Additional outcomes: Measures of nutritional status ‐% change in weight and AMC
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Meredith 1992.

Methods	Method of randomisation: not stated Assessor blinding: not reported Intention to treat: not carried out, 1 excluded for protocol violation Lost to follow‐up: details given Timing of intervention: exercise sessions and diet intervention conducted three times per week over the 12 week period for both groups
Participants	Location: Human Physiology Laboratory, Boston, Massachusetts 11 participants Inclusion criteria: healthy out‐patient volunteers recruited by advertisement, sedentary, non obese men over 60y Exclusion criteria: a wide range of cardiac related conditions, recent embolism, high dose of phenothiazine agents, uncontrolled metabolic disease, clinically severe hypertension (diastolic above 110), severe anaemia, marked obesity (BMI above 30kg/m2), renal hepatic or other metabolic insufficiency, overt psychoneurotic disturbances, moderate to severe pulmonary disease, coagulation defects, neuromuscular disease, connective tissue diseases Sex: all male Age: 68/65y Health Status: very healthy, non obese, living at home and in the research unit
Interventions	a: ad libitum diet plus complete nutrient mixture (Two Cal HN, Ross Laboratories, Ohio) per 100 ml: 200 kcal, 8.3 g protein, 21.9 g carbohydrate, 8.9 g fat, vitamins and minerals, designed to provide an additional 8 kcal and 0.33 g protein per kg of ideal body weight over and above normal ad libitum diet. The vitamin and mineral content was between 25 and 75% of the RDA. The supplement was consumed as two drinks of about 120 ml each, served cold and in a variety of flavours about 10 am and 10 pm every day b: ad libitum diet with no placebo Allocated: 6/6 Assessed: 6/5
Outcomes	Main outcomes: Mortality Functional status ‐ dynamic strength Additional outcomes: Measures of nutritional status‐ Anthropometric indices ‐ weight change, fat‐free mass, sum of six skinfolds Measures of dietary intake ‐ energy and protein intake
Notes	All participants received strength training and have been included Request for information regarding blinding of outcome assessors and mineral and vitamin content of supplement sent 14/7/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Ovesen 1992.

Methods	Method of randomisation: not stated Assessor blinding: blinded Intention to treat: not carried out, 10 refused to complete due to gastro‐intestinal discomfort ascribed to the sip‐feeds Lost to follow‐up: details given Timing of intervention: 10 days, both interventions have same energy distribution of protein fat and carbohydrate, both served in 200ml tetrabriks without labelling, 10 days supplementation, offered a minimum of 4 x 200ml daily between meals
Participants	24 patients Inclusion criteria: elderly patients, non infectious chronic disease, hospitalisation expected for at least 10 days, weight loss (within 3 months) of more than 10% of their usual body weight or a plasma albumin of less than 0.4 mmol/l, daily energy intake of less than 1.5 x basal energy expenditure, or a protein intake of less than 1g/kg during the first 48 h of admission Exclusion criteria: patients on special diets, unable to participate in taste testing Sex: 17 female, 7 male Age: 74/75y Health Status: inpatients, poor appetite and intake
Interventions	a: liquid supplement containing 6.3 kJ and 0.06 g protein per ml (Nutrison Energirig, Nutricia, DK), nutritionally complete and lactose free, patients chose either vanilla or orange flavour b: liquid supplement containing 4.2 kJ and 0.04 g protein per ml (Nutrison Standard (Nutricia, DK), nutritionally complete and lactose free, patients chose either vanilla or orange flavour Allocated: 17/17 Assessed: 10/14
Outcomes	Main outcomes: Mortality Side effects of treatment ‐ gastrointestinal discomfort Additional outcomes: Measures of dietary intake ‐ energy and protein intake Patient compliance ‐ numbers completing 10 days of supplement
Notes	Request for further information regarding method of randomisation, content of supplement, whether outcome assessors were blinded to treatment status, whether other outcomes were measured sent 19/9/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Payette 2002.

Methods	Method of randomisation: not stated Assessor blinding: not blinding for all outcomes Intention to treat: for deaths only Lost to follow‐up: details provided Timing of the intervention: for 16 weeks Length of follow‐up: 16 weeks
Participants	Location: Sherbrooke, Quebec, Canada (17 community service centres) 83 patients Inclusion criteria: receiving long‐term home help services, Older than 65 years, higher nutritional risk a) involuntary weight loss > 5% of weight in past month, or weight loss > 7.5% in past 3 months or > 10% in past 6 months and BMI < 27 or b) BMI < 24 Exclusion criteria: Palliative care, alcoholic, active cancer, illness requiring therapeutic diet incompatible with supplementation Sex: 59 female 24 male Age: 81.6 (SD 7.5), 78.6 (SD 6.1)
Interventions	a: 2x 235ml cans/day Ensure or Ensure plus aim to gain 0.5 kg/ week b: none Allocated: 43/46 Assessed: 41/42
Outcomes	Main outcomes: mortality, functional status, quality of life Additional outcomes: bed disability days, complications, compliance Measures of nutritional status‐ weight, arm muscle circumference Measures of dietary intake ‐energy and protein intake
Notes	Classified as undernourished, at home and well, need confirmation of denominators
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Payette 2004.

Methods	Method of randomisation: sealed envelopes, insufficient detail to confirm concealed allocation concealment: A Assessor blinding: not all outcomes blinded Intention to treat: unclear Lost to follow‐up: unclear Timing of the intervention: 24 weeks Length of follow‐up: 24 weeks
Participants	Location: Sherbrooke, Quebec, Canada Inclusion criteria: Recruited from Meals‐on Wheels programmes offered by 7 community‐based volunteer agencies. Over 65y, received at least two meals per week, orientated to time and place, at risk of malnutrition (MNA score greater than or equal to 17 and less than 24‐ modified MNA) Exclusion criteria: Palliative care, alcoholic, active cancer, illness requiring therapeutic diet incompatible with supplement, wheel‐chair bound, BMI >30 Sex: 77 female, 22 male Age: 79.4 (SD 6.1)
Interventions	a: Professional counselling to consume food or supplement to achieve at least 100% Canadian nutritional recommendations for energy, protein and all nutrients. Encouraged to take at least one 250 ml can of NuBasics Plus Complete Nutrition Drink (Nestle) for 24 weeks. Dietitian adjusted supplement according to tolerance. Monthly visits for 6 months b: Monthly contact be phone or visit Allocated: 54(50)/51(49)? Assessed: 50/49
Outcomes	Main outcomes: muscle strength, functional outcomes, timed up and go, no others yet reported Additional outcomes: Measures of nutritional status‐ Anthropometric indices ‐ triceps skinfold, weight change Measures of dietary intake ‐ change in energy and protein intake
Notes	Reply received regarding details of study 7/04/2004. Further information requested regarding number of participants and randomisation method.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Potter 2001.

Methods	Method of randomisation: block randomisation with sealed envelopes Assessor blinding: anthropometry and clinical outcome blinded, ADL unlikely to be blinded Intention to treat: carried out Lost to follow‐up: none, all patients accounted for Timing of intervention: from within 48 hours of admission until the time of discharge home, death or referral for institutional placement at 0800, 1400 and 1800 hours
Participants	Location: Medicine for the Elderly Unit, Glasgow, UK 381 patients Inclusion criteria: no known malignancy, able to swallow, non obese (BMI <75th percentile), able to gain consent from patients or relatives Sex: not given Age: not given, but elderly
Interventions	a: 120ml oral nutritional supplement (Entera Frusenius) three times daily intended to provide 540 kcal/day, 22.5 g protein/day b: usual practice Allocated:186/195 Assessed:165/162
Outcomes	Main outcomes: Mortality Complications (sepsis) Length of stay ‐ Acute hospital Discharge destination Functional status ‐ Barthel Index Additional outcomes: Measures of nutritional status‐arm muscle circumference, percentage weight change Measures of dietary intake ‐ energy intake (random sample of one in three patients on day 3)
Notes	Further information received from author regarding: Date study began and ended, timing of baseline data collection and outcome assessments (weight and AMC), length of follow‐up, whether length of stay included those who died, whether care programmes were identical, further details on complications (sepsis)
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Price 2005.

Methods	Method of randomisation: sealed opaque envelopes, prepared by another individual from computer generated random number table Assessor blinding: open trial Intention to treat: yes Lost to follow‐up: details provided Timing of the intervention: daily for 8 weeks after discharge from hospital Length of follow‐up: 6 months
Participants	Location: Dundee area, Scotland recruited from Ninewells or Royal Victoria Hospital 76 patients Inclusion criteria: recruited from acute general medical wards and medicine for the elderly assessment and discharged back to the community (own home, with relatives, sheltered housing or residential homes) evidence of undernutrition defined by admission BMI = 24 with MAMC or TSF < 10th centile and or 5% or more loss in body weight during hospital stay. Sex: 101 female, 35 male Age: 83.7 (SD 5.2), 85.4 (SD 5.4)
Interventions	a: 2 cartons (400 ml, 600 kcal, 24g protein, 72g carbohydrate Fortisip or Fortifresh, Nutricia, UK) in a variety of flavours b: no supplement provided Allocated: 66/70 Assessed: 35/41
Outcomes	Main outcomes: Mortality, complications, functional status, anthropometry, weight, dietary intake Additional outcomes: Unplanned readmissions, one or more prescription of antibiotics, new antidepressant prescriptions, introduction of other new medication, out patient or day hospital attendance, respite admission, falls, planned admissions, admissions to residential home
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Rosendahl 2006.

Methods	Method of randomisation: stratified cluster randomised controlled trial with exercise and nutritional intervention and placebo in 2x2 factorial model Assessor blinding: all assessors blinded Intention to treat: yes Lost to follow‐up: recorded Timing of intervention: 3 months Length of follow‐up: 6 months
Participants	Location: residential care facilities in Emea, Sweden Inclusion criteria: >65years, dependent of assistance in one or more personal ADL activities on KATZ index, able to stand from a chair with arms with help of only 1 person or less, MMSE of >10, physician approval Exclusion criteria: not detailed Sex: 139 female, 52male Age: mean 84.5years
Interventions	A: semi protein energy supplement 200ml 7.4g protein and 408kJ /100g B: 200ml drink 0.2g protein and 191kJ per 100g Allocated: 96, 95 Assessed: 82, 81 at 6 months
Outcomes	Main outcomes: Mortality Berg balance scale Gait speed self paced Gait speed maximum 1 RM lower limb strength Others, death recorded
Notes	Mortality data not included in meta‐analysis as cluster randomised and don't have the intra‐class correlation coefficient. Exercise randomised also for each cohort
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Salas‐Salvado 2005.

Methods	Method of randomisation: centralised and both groups stratified by initial BMI Assessor blinding: no information given re attempts to blind assessors as intervention very obvious presume unblinded Intention to treat: no Lost to follow‐up: not recorded but numbers do not add up Timing of intervention: 3 months Length of follow‐up: 3 months
Participants	Location: 6 geriatric institutions Catalonia spain, 53 patients Inclusion criteria: Alzheimers disease DSM IV criteria 3 or above, requiring semi solid or liquid diet, and present weight loss or higher than 5% weight loss in the previous year Exclusion criteria: terminal care, sever acute illness, cancer or history of in last 5years, sever GI disease or any other acute illness able to affect nutritional status during study, significant hepatic or renal disease, enteral or parenteral nutritional support at time of study, chronic treatment with steroids antineoplastic drugs or antibiotics, diabetic patients on insulin, use of nutritional supplements 15days prior to study. Sex: 44 female, 9 male Age: mean 84.7years
Interventions	A: semi solid foods plus dietetic advice, 3 x 451.52kcals per day plus other normal intake B: dietetic advice plus access to above foods as an extra Allocated: 24, 29 Assessed: 15, 23 at 3 months
Outcomes	Weight change Dietetic intake Functional rating scale
Notes	Numbers don't add up in outcomes section
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Saudny 1997.

Methods	Method of randomisation: sealed envelopes Assessor blinding: all strength measurements blinded Intention to treat: not carried out, results presented for 24 patients, 33 randomised Lost to follow‐up: intervention 1 refused to come back or could not be located, 1 refused to continue or data could not be used; control 3 refused to come back or could not be located, 1 refused to continue or data could not be used, 1 too ill to continue Timing of intervention: 14 days
Participants	Location: Montreal Chest Institute 24 patients Inclusion criteria: consecutive patients 40‐85y with diagnosis of COPD and a FEV1 that was less than or equal to 60% of the predicted value Exclusion criteria: patients who required mechanical ventilation, had a gastrointestinal disorder, had active cancer or the condition predisposing to weight loss, were terminally ill, were unable to communicate in English or French, suffered from mental confusion or followed a special diet Sex: 9 female, 15 male Age: mean age 69y Health Status: patients hospitalised with an exacerbation of COPD, not necessarily malnourished
Interventions	a: food and beverages ordered by the participants from the hospital menu, additional oral supplements (Ensure, Ensure plus, or a variety of puddings or extra snacks) to assure a caloric intake of at least 1.5 x REE if their BMI was normal (20‐27) and at least 1.7 X REE if their BMI was below 20 b: food and beverages ordered by the participants from the hospital menu Allocated: 17/16 Assessed: 14/10
Outcomes	Main outcomes: Mortality Morbidity and complications ‐ to ill to continue Length of stay ‐ rehabilitation hospital Functional status ‐ lung function, general well being, grip strength, 6 minute walk test Additional outcomes: Measures of nutritional status‐ Anthropometric indices ‐ weight change Measures of dietary intake ‐ energy and protein intake
Notes	Request for information sent regarding when did deaths occur, were care programmes identical and information about amount of supplement provided sent 14/7/01. Reply with more information received 19/11/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Schols 1995.

Methods	Method of randomisation: not stated Assessor blinding: not reported Intention to treat: carried out, but results not ITT analysis Lost to follow‐up: incomplete report of drop‐outs A group of patients randomised to receive nandrolone decanoate have been excluded from this analysis. All patients included also undertook the standard physical training programme over the period of the study. Length of follow‐up: 8 weeks
Participants	Location: Rehabilitation Centre, University Hospital, Maastricht, the Netherlands 135 patients Inclusion criteria: patients with moderate to severe COPD consecutively admitted to an intensive inpatient pulmonary rehabilitation programme, stable clinical condition, patients who had an acute exacerbation during the study and who were effectively treated with glucocorticosteriods and/or antibiotics remained in the study Exclusion criteria: unstable COPD or greater than 10% increase in FEV of the predicted baseline value after administration of 400ug salbutamol, obesity, malignancies, ischemic heart disease or other cardiac impairment, renal, hepatic, gastrointestinal or endocrine disease, surgery within last 2 months Sex: not given Age: mean age 65y Health Status: patients admitted to an intensive inpatient pulmonary rehabilitation programme with moderate to severe COPD, stratified into two groups : depleted group: less than 90% ideal body weight or fat‐free mass less than 67% (men) 63% women, and non depleted
Interventions	a: meal size and composition were chosen by all the patients themselves and registered daily, patients encouraged to eat their regular meals. In addition, one high calorie drink supplement daily, early evening between 7pm and 9pm for at least 8 weeks: 200 ml, 420 kcal, 51% fat, 35% carbohydrate, 14% protein, containing Nutridrink, Protifar, Fantomalt and oil, 7 flavours b: standard hospital diet, patients encouraged to eat their regular meals Allocated: 72/63 Assessed: 72/63 (33 depleted, 39 non depleted/38 depleted, 25 nondepleted)
Outcomes	Main outcomes: Mortality Functional status ‐ lung function, walking distance Additional outcomes: Measures of nutritional status‐ Anthropometric indices ‐ weight change, mid‐arm muscle circumference, skinfold thickness (4 sites), change in FFM Measures of dietary intake ‐ energy intake Side effects of treatment
Notes	Further information requested regarding randomisation, blinding of outcome assessors, vitamin and mineral content of supplement, denominators for mortality data sent 20/10/01. Reply received 1/11/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Scorer 1990.

Methods	Method of randomisation: not stated ‐ allocation concealment B Assessor blinding: not mentioned Intention to treat: not mentioned Lost to follow‐up: 10 withdrawn (3 insufficient data, 7 weight loss less than 2 kg in run‐in) ‐ unclear which arm Timing of the intervention: 12 weeks Length of follow‐up: 12 weeks
Participants	Location: 53 General Practices, UK carried out by Abbott UK Medical Period of study: 130 patients Inclusion criteria: community dwelling women, 65 years or over, undernourished as defined by GP Exclusion criteria: active neoplastic disease, active metabolic disease, history of malabsorption, inflammatory bowel disease, significant cardiovascular disease, receiving diuretics, corticosteroids or anabolic drugs, institutionalised patients Sex: 85 female 45 male Age: 76/77y
Interventions	a: 3 cans /day Ensure per day between meals and home visits b: carbonated water (330ml can) Allocated: ?/? Assessed: 48/46
Outcomes	Main outcomes: none Additional outcomes: Nottingham Health Profile, compliance Measures of nutritional status‐ weight
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

SG Larsson malnour.

Methods	Subgroup of initially malnourished patients
Participants
Interventions
Outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

SG Larsson nourished.

Methods	Subgroup of well nourished patients
Participants
Interventions
Outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

SG Potter malnourish.

Methods	Subgroup of moderately and severely malnourished patients
Participants
Interventions
Outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

SG Potter moder maln.

Methods	Subgroup of moderately malnourished patients
Participants
Interventions
Outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

SG Potter nourished.

Methods	Subgroup of adequately nourished patients
Participants
Interventions
Outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

SG Potter very maln.

Methods	Subgroup of severely malnourished patients
Participants
Interventions
Outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

SG Volkert comply.

Methods	Subgroup of patients with good acceptance of the supplement (one or nearly one portion per day)
Participants
Interventions
Outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

SG Volkert non compl.

Methods	Subgroup of patients with poor acceptance of the supplement (one portion every two days or less)
Participants
Interventions
Outcomes
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Stableforth 1986.

Methods	Method of randomisation: not stated Assessor blinding: not reported Intention to treat: 3 excluded, intention to treat analysis not possible Lost to follow‐up: none Timing of intervention: started after surgery and 24‐36h of crystalloid intravenous fluids. Intervention provided during waking hours for 10 days Length of follow‐up: 4 weeks
Participants	61 patients randomised Inclusion criteria: hip fracture patients within 12h of fracture, women over 65y Exclusion criteria: none given Sex: all female Age: Mean 81.8y, range 65‐96y
Interventions	a: Encouraged to drink flavoured, Carnation Instant Breakfast in 300ml milk (1.34 MJ or 320 kcal, 18.5 g protein, 11g fat, 40 g carbohydrate, vitamins and minerals) plus ward diet b: Ward diet alone Allocated: unclear Assessed: unclear
Outcomes	Main outcomes: Mortality ‐ all causes Morbidity and complications ‐ anaesthetic, surgical infection, gastrointestinal, urinary Additional outcomes: Measures of nutritional status ‐ weight Measures of dietary intake ‐ energy balance, nitrogen balance
Notes	Limited functional outcomes. Request for further details, especially on longer term follow‐up, sent 13/4/99, resent 7/2/00
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Steiner 2003.

Methods	Method of randomisation: allocated and dispensed by pharmacy‐ allocation concealment: A Assessor blinding: unclear if investigators knew allocation when analysing the results Intention to treat: yes Lost to follow‐up: intervention 15 (plus 2 withdrawn); control 7 (plus 1 withdrawn) Timing of intervention: 7 weeks, dispensed each week when participant attended outpatient rehabilitation clinic each week Length of follow‐up: 7 weeks
Participants	Location: Glenfield Hospital, Leicester, UK 85 patients Inclusion criteria: Referred to pulmonary rehab programme with clinical and spirometric criteria for COPD, stable at recruitment, optimised medical treatment Exclusion criteria: Unsuitable for exercise‐ cardiac, neuropsychiatric or musculoskeletal disorders, diabetes, glucose intolerance, BMI >30 Sex: 32 female, 53 male Age: 66.0 (SD 9.0)/ 68.0 (SD 8.0)
Interventions	a: 125 ml Respifor (Nutricia, Netherlands)‐ 570 kcal/day (CHO 60%, fat 20%, protein 20%), 3 times daily for duration of rehab. b: Non‐nutritive placebo‐ same packaging and flavouring, 3 times daily for duration of rehab. Allocated: 42/43 Assessed: 25/35
Outcomes	Main outcomes: mortality, COPD complications, functional outcomes‐ handgrip strength, muscle strength walking, self reported chronic respiratory questionnaire Additional outcomes: Anthropometric indices ‐ weight change, BMI change, dietary intake change, self reported compliance
Notes	Classified as nourished, unwell, and at home Non commercial trial
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Tidermark 2004.

Methods	Method of randomisation: opaque sealed envelope‐ allocation concealment: A Assessor blinding: unclear for some outcomes Intention to treat: carried out Lost to follow‐up: intervention 1; control 2 Timing of the intervention: 6 months
Participants	Location: Stockholm, Sweden Period of study: Prior to October 2002 Number of patients: 40 Inclusion criteria: Females with acute femoral fracture, > 70 years, BMI less than or equal to 24 kg/m2, absence of severe cognitive disfunction, independent living status and independent walking capability with or without walking aids Exclusion criteria: Patients with fractures not suitable for internal fixation and patients with rheumatoid arthritis or radiographic osteoarthritis Sex: 40 female Age: 83.5 (SD 6.1), 84.1 (SD 4.3)
Interventions	a: Fortimel 200 ml/day, 20 g Protein, plus 1g Calcium and 400 IU Vitamin D b: usual nutritional care plus1g Calcium and 400 IU Vitamin D Allocated: 20/20 Assessed: 18/17
Outcomes	Length of follow‐up: 12 months Main outcomes: mortality, length of hospital stay during first year after surgery, complications, mobility, activities of daily living, hand grip strength, adverse effects, quality of life Additional outcomes: compliance Measures of nutritional status‐ weight, LBM
Notes	Also looked at the effects of nandrolone in a third arm of the trial Classified as undernourished, acute admission and living in the community. Further details required regarding details of intervention
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Vermeeren 2004.

Methods	Method of randomisation: no description Assessor blinding: states double blind but no details Intention to treat: states number and reasons for withdrawal but intention to treat not possible Lost to follow‐up: intervention 3 (nausea); control 1 (nausea) ‐ plus 4 withdrawn due to medical problem and 1 alcohol problem (unclear which arm) Timing of the intervention: from first day of hospitalisation 3 times a day between meals Length of follow‐up: unclear
Participants	Location: University hospital Maastricht and regional hospital 'MaximaMedical Centre' in Veldhaven, Netherlands 56 patients Inclusion criteria: Acutely admitted to hospital for exacerbation of COPD (diagnosis based on GOLD criteria). Recent increase in breathlessness, cough and sputum sufficient for admission judged by independent physician. Sex: 32 male, 65 female Age: 66 (SD 8), 65 (SD 10)
Interventions	a: 125 ml Respifor 3 times day (Nutricia, Zoelemeer, The Netherlands), (2.38 MJ/day) b: 125 ml vanilla flavoured water (0 MJ/day) Allocated: 29/27 Assessed: 23/24
Outcomes	Main outcomes: weight, complications, functional status, dietary intake Additional outcomes: readmissions, side‐effects
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Vlaming 2001.

Methods	Method of randomisation: block randomisation of 100 by pharmacy, sealed envelopes, remote site Assessor blinding: yes Intention to treat: done Lost to follow‐up: intervention 0; control 0 (all patients accounted for) Timing of the intervention: whenever their clinical team allowed Length of follow‐up: mean 14.2 (SD24.9), 11.4 (SD16.4) days
Participants	Location: The Royal London Hospital and St Bartholomew's Hospital, London 549 patients Inclusion criteria: patients admitted though acute services ‐ general medical, surgical or orthopaedic, thin but not seriously undernourished ‐ BMI 18‐22, or unintentional weight loss greater than or equal to 5% Exclusion criteria: planned admissions to medical or orthopaedic wards, <18y, mental illness, already routine treatment with water soluble vitamins, admission clearly for 2 days or less, previously taken part in trial, BMI <18, unintentional weight loss >10%, therapeutic diets, unable to swallow liquids, randomisation clinically unacceptable, unable to gain consent Sex: 314 female, 235 male Age: 67/66
Interventions	a: 2 x 200ml Ensure plus, 600 kcal, 25g protein, 80.8g carbohydrate, 19.6g fat plus vitamins and minerals plus or minus a vitamin supplement b: 400ml placebo drink 100kcal, 25g carbohydrate plus or minus a vitamin supplement Allocated: 275/274 Assessed: 271/274
Outcomes	Main outcomes: mortality, length of hospital stay Additional outcomes: compliance
Notes	Length of follow‐up: 12 months Main outcomes: mortality, length of hospital stay during first year after surgery, complications, mobility, activities of daily living, hand grip strength, adverse effects, quality of life Additional outcomes: compliance Measures of nutritional status‐ weight, LBM
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Volkert 1996.

Methods	Method of randomisation: states randomised by drawing lots but unclear if those who assigned were blinded Assessor blinding: not reported Intention to treat: not carried out, split analysis on the basis of compliance Lost to follow‐up: intervention 11; control 3 Timing of intervention: during hospital stay and after discharge for 6 months
Participants	Location: during hospitalisation in an Acute Care ward of the Geriatric Centre Bethanien in Heidelberg, Germany, and after discharge 46 patients Inclusion criteria: female, aged 75y or older without malignant disease or need for tube feeding or parenteral nutrition, undernourished by clinical judgement of the examining physician, (BMI used to confirm clinical judgement when data on body weight and height were available), expected hospital stay at least 3 weeks, presumed actual life expectancy of more than 6 months, consent from participants or care givers Exclusion criteria: none given Sex: all female Age: mean age 85y Health Status: long term care, nutritionally 'at risk'
Interventions	a: in addition to the standard hospital diet 2 portions x 200ml of a liquid supplement during hospital stay (200ml soup mid‐morning and 200ml sweet drink in the afternoon daily), different brands but with similar composition but different flavours were used in order to increase variety and patient acceptance. On average 2 portions provided 500 kcal and 30 g protein/day, providing 50 ‐ 150% of the recommendation for most micronutrients. After discharge, 1 daily portion of the sweet supplement was provided. Patients were strongly encouraged to consume the entire amount offered. Patients were initially visited at home to provide new supplements every week, and then every two weeks b: usual care without supplements Allocated: 35/37 Assessed: 20/26
Outcomes	Main outcomes: Mortality Functional status ‐ Barthel Index Additional outcomes: Measures of nutritional status‐ Anthropometric indices ‐ weight change, LBM Measures of dietary intake ‐ energy and protein intake Patient acceptance ‐ poor acceptance = one portion every two days or less
Notes	Further information requested regarding method of randomisation, blinding of treatment status, denominators for mortality sent 20/10/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Woo 1994.

Methods	Method of randomisation: computer randomised Assessor blinding: blinded apart from dietary assessment Intention to treat: not carried out Lost to follow‐up: incomplete report of drop‐outs Participants in both groups able to return for medical consultation if they felt unwell in any way. Length of follow‐up: 3 months
Participants	Location: Prince of Wales Hospital, a District General hospital, Hong Kong 81 patients Inclusion criteria: patients consecutively admitted to acute medical wards of a district general hospital not catering for chronic disabled or demented patients, primary diagnosis of chest infection (purulent sputum, increasing shortness of breath, pyrexia, elevated white cell count, with or without radiological changes on chest radiography (75% had underlying chronic lung disease) Exclusion criteria: those with heart failure, renal or hepatic failure, stroke, malignancies or bedridden subjects who could not feed themselves, patients who did not give informed consent Sex: 30 female, 51 male Age: mean age 72/74y Health Status: discharged home following acute hospital stay with chest infection, not necessarily malnourished
Interventions	a: 1 month of supplementation, 500 ml Ensure liquid (Abbott Laboratories Ltd) daily, 500 kcal, 14% protein, 32% fat, 54% carbohydrate, minerals and vitamins. Instructed to take it between meals or before bedtime b: no supplement Allocated: 40/41 Assessed: variable
Outcomes	Main outcomes: Mortality Functional status ‐ Barthel Index, appetite, mental test score Additional outcomes: Measures of nutritional status‐ Anthropometric indices ‐ change in BMI, arm muscle circumference, total body fat, FFM Measures of dietary intake ‐ energy and protein intake
Notes	Request for more information regarding computer randomised assignment, numbers and denominators for mortality, details about the provision of supplements, blinding of treatment providers, recruitment of consecutively admitted patients sent 14/7/01. Reply received 23/10/01, more information awaited
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Wouters 2002.

Methods	Method of randomisation: not stated ‐ allocation concealment: B Assessor blinding: not mentioned Intention to treat: not possible Lost to follow‐up: 3 (2‐taste, 1 consent) ‐ unclear which arm Timing of the intervention: 3 months Length of follow‐up: 3 months
Participants	Location: Two psycho‐geriatric nursing homes in the Netherlands 42 patients Inclusion criteria: 60 years or older, resident at least 2 months in nursing home, diagnosed with dementia syndrome, BMI less than 23 kg/m2 for men and 25 kg/m2 for women Exclusion criteria: Cancer, terminal care, acute illness, severe gastrointestinal disorders, need for parenteral or enteral nutrition, therapeutic diet incompatible with supplementation Sex: 31 female 4 male Age: 85.3 (SD 8.4)/ 78.7 (SD 8.8)
Interventions	a: Two 125ml tetrapaks, twice daily in morning and afternoon between meals 250 kcal/day 8.5g protein, 39.6g carbohydrate, 8.9g fat, vitamins and minerals b: Two 125ml tetrapaks placebo containing water, cloudifier, flavourant, non‐caloric sweetener to resemble supplement in taste and appearance Allocated: 21/21 Assessed: 19/16
Outcomes	Main outcomes: mortality, activities of daily living Additional outcomes: side‐effects of supplement, compliance Measures of nutritional status‐ weight compliance Measures of dietary intake ‐energy and protein intake
Notes	Classified as in long‐term care, malnourished, unwell, in‐patients
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Wouters 2003.

Methods	Method of randomisation: quasi‐randomised Assessor blinding: not mentioned Intention to treat: not possible Lost to follow‐up: intervention 2 ill, 11 did not like taste; 5 too much effort to complete; control 2 ill, 7 did not like taste, 5 too much effort to complete Timing of the intervention: 6 months Length of follow‐up: 6 months
Participants	Location: Residences for the elderly or sheltered housing, Wageningen, The Netherlands 68 patients Inclusion criteria: 65 years or older, BMI less than or = to 25 kg/m2, residence in home for elderly or sheltered housing Exclusion criteria: Cancer, chronic gastrointestinal disorders, diet incompatible with supplementation, mentally unable to answer study questions or to remember taking supplement Sex: 39 female 29 male Age: 81.0 (SD 6.9)
Interventions	a: two flavours in 125ml tetrapaks, twice daily in morning and afternoon between meals 250 kcal/day 8.75g protein, 28.5g carbohydrate, 11.25g fat, vitamins and minerals b: placebo Allocated: 52/49 Assessed: 34/34
Outcomes	Main outcomes: mortality, functional variables, quality of life Additional outcomes: complications (cancer) Measures of nutritional status‐ weight compliance Measures of dietary intake ‐energy and protein intake
Notes	Classified as at home, nourished, well, living in the community
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	High risk	C ‐ Inadequate

Wouters 2005.

Methods	Method of randomisation: states randomised by person not involved in study, in blocks of 4 based on body mass index Assessor blinding: double‐blind placebo‐controlled trial Intention to treat: carried out but no results given Lost to follow‐up: intervention 5 dropped out; control 5 dropped out ‐ plus 13/10 withdrawn or excluded (unclear which arm) Timing of intervention: 6 months Length of follow‐up: 6 months
Participants	Location: homes for elderly or sheltered housing, The Netherlands 101 participants Inclusion criteria: 65 years or older, BMI 25kg/m2 or less Exclusion criteria: cancer, gastrointestinal disease, therapeutic diet incompatible with supplementation, unable to respond to questionnaires or take supplement Sex: 38 female, 29 male (completers) Age: mean age 83y
Interventions	a: 2 x 125ml daily oral liquid supplements with total of 250kcal and 8.75g protein, providing 30‐160% of US RDA of micronutrients b: 2 x 125ml daily placebo with water, sweetener, cloudifier, thickening, flavouring, colour Allocated: 52/49 Assessed: 34/34
Outcomes	Main outcomes: Deaths and illness (combined) Functional status ‐ cognitive function
Notes	Emailed for separate illness and death data by allocation 09/01/2008, no response by 16 March 2008 so illness and death data not included
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Wouters 2006.

Methods	Method of randomisation: states randomised but no further details Assessor blinding: not stated Intention to treat: carried out but no results given Lost to follow‐up: intervention 2; control 3 Timing of intervention: 5 weeks Length of follow‐up: 5 weeks
Participants	Location: nursing home for psychogeriatric residents, The Netherlands 39 patients Inclusion criteria: > 65 years, in nursing home for at least 2 months, acute infection requiring antibiotics Exclusion criteria: cancer, rheumatoid arthritis, insulin dependent diabetes, morbid obesity, need for terminal care, therapeutic diet incompatible with supplementation Sex: 29 female, 5 male (completers) Age: mean age 82.7y
Interventions	a: daily oral liquid supplement with 309kcal and 11.2g protein, vitamins and minerals b: resident referred to dietitian for care, if detected by physician for weight loss, loss of appetite or low intake, usually provided enrichment of food, often energy and protein enriched desserts or drinks Allocated: 20/19 Assessed: 18/16
Outcomes	Main outcomes: Functional status ‐ Zorg Index Geriatrie (ZIG) Additional outcomes: Measures of nutritional status ‐ weight change, triceps skinfold, mid upper arm circumference, arm muscle circumference Measures of dietary intake ‐ energy and protein intake
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Yamaguchi 1998.

Methods	Method of randomisation: Assessor blinding: not mentioned Intention to treat: no Lost to follow‐up: details not given Timing of intervention: 18 months
Participants	Location: community living elderly, USA Inclusion criteria: elderly homebound men and women starting to receive meals‐on‐wheels, at nutritional risk Exclusion criteria: Sex: 35 female, 27 male Age: mean age about 78y
Interventions	a: liquid nutrition supplement aimed to provide per day 600 kcal, 30 g protein, 80 g carbohydrate, 18g fat, 760 µg vitamin A, 30 mg vitamin C, 1mg vitamin B6, 3 mg vitamin B12, 200 µg folate, 700 mg calcium, 9 mg iron, 200 mg magnesium, 10.6 mg zinc b: fruit flavoured beverage aimed to provide 105 kcal, 25.5 g carbohydrate, 15 mg calcium, 9 mg magnesium Allocated: 32/30 Assessed: 11/6
Outcomes	Main outcomes: Additional outcomes: measures of nutritional status‐weight measures of dietary intake‐ energy and protein intake
Notes	Further details requested 15/04/01. Reply received 10/05/01
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Unclear risk	B ‐ Unclear

Young 2004.

Methods	Method of randomisation: cluster randomised by unit or dining room using opaque envelopes, allocation concealment A. Assessor blinding: action taken to blind assessors for some of the cognitive function assessment, but not all. Unblinded assessors (dietary intake) Intention to treat: carried out Lost to follow‐up: intervention 2 withdrawn; control 1 withdrawn Timing of intervention: 21 days Length of follow‐up: 21days for each part of crossover design
Participants	Location: academic geriatric care facility, Toronto, Canada 34 patients Inclusion criteria: residents of Alzheimer's disease units, consent from participants' families or legal guardians. Diagnosis of probable Alzheimer's disease made by qualified clinician, ability to self‐feed or requiring only minimal levels of assistance, e.g. opening containers Exclusion criteria: disease requiring nutritional intervention, e.g. type 1 diabetes, prescription of energy‐restricted diet, swallowing difficulties requiring texture‐modified diet, acute illness Sex: 26 female, 5 male (completers) Age: mean age 88y
Interventions	a: nutritional bar with or without sugar and margarine, Ensure, or food providing 250‐258kcal/d and 9‐11g protein/d, given at 10am about one hour after breakfast b: usual care Allocated: crossover design, 34 allocated Assessed: 31 completed two intervention periods and washout
Outcomes	Main outcomes: Morbidity and complications ‐ acute hospital admission Functional status ‐ Severe Impairment Battery, Global Deterioration Scale, Neuropsychiatric Inventory Nursing Home version, London Psychogeriatric Rating Scale Additional outcomes: Measures of nutritional status ‐ weight change Measures of dietary intake ‐ energy and protein intake
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment?	Low risk	A ‐ Adequate

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Bachrach 2000	Randomised controlled trial of two different surgical methods following femoral neck fracture, oral supplement intervention not randomised
Barateau 1998	Not randomised controlled trial, multicentred randomised survey
Barton 2000	Randomised controlled trial of normal versus fortified menu, intervention had more energy but less protein than control
Bastow 1983	Randomised controlled trial of nasogastric feeding
Bean 1994	Randomised controlled trial of ornithine alpha‐ketoglutarate versus a defined peptide oral supplement
Beattie 2000	Randomised controlled trial of oral protein and energy supplement, controls mean age 62 years, treatment group 54.4 years
Beier 1996	Randomised controlled trial of nasogastric feeding, controls median age 61.5 years, treatment group 66.5 years
Bernstein 2002	Randomised controlled trial of nutrition education versus exercise
Borum 2000	Prospective cohort study of artificial nutritional support in patients over 18 years
Bos 2000	Not randomised controlled trial, oral protein and energy supplement
Breslow 1993	Non randomised controlled trial, high or low protein, nasogastric or meal supplement
Brocker 1994	Randomised controlled trial of oral ornithine oxoglutarate
Brown 1995	Randomised controlled trial of early or late feeding after percutaneous endoscopic gastrostomy insertion
Bunker 1994	Not randomised controlled trial, oral protein and energy supplement, matched control group, aged 70 to 85 years
Bunout 1989	Randomised controlled trial, oral protein and energy supplement, control group mean age 48 years, intervention group mean age 50 years
Bunout 2001	Free‐living elders half received nutrition supplementation, randomised to resistance training or not.
Bunout 2004	Free‐ living elders half received nutritional supplementation not randomised, randomised to receive resistance training or not.
Cabre 1990	Randomised controlled trial of nasogastric feeding, patients with cirrhosis, no age given
Caglar 2002	Not randomised study of oral nutritional supplementation for hemodialysis patients
Calvey 1985	Randomised controlled trial of normal diet versus normal diet plus carbohydrate and protein versus normal diet plus branched chain amino acid protein, mean age all groups 49 years
Campbell 1995	Randomised controlled trial of exercise and two levels of protein intake, mean age 65 years
Carr 1996	Randomised controlled trial of immediate nasogastric feeding versus intravenous fluids, controls mean age 51years, treatment group mean age 60 years
Chee 2003	Randomised controlled trial of the effect of milk supplementation on bone density. Mean age 58.7 and 59.0 years
Creutzberg 2000	Not randomised controlled trial, oral high caloric supplements in patients with COPD
Daly 1992	Randomised controlled trial in cancer patients of nasogastric immunomodulatory diet
Danhof 1982	Not randomised controlled trial, no control group, oral protein and energy supplementation study of nursing home patients
de Jong 2000	Randomised controlled trial of micronutrient supplementation in frail elderly people
Devine 1996	Randomised trial of calcium versus skimmed milk powder, mean age 63 years
Duncan 2006	Randomised controlled trial of additional personal care from dietetic assistants to improve the outcome of hip fracture patients
Efthimou 1988	Randomised controlled trial of oral protein and energy supplementation, mean age 62 years
Elmstahl 1987	Randomised controlled trial of three oral supplements for elderly patients, mean age 85 years with different levels of protein but same energy content
Eneroth 1997	Not randomised controlled trial, controls matched from another hospital
Eneroth 2006	Randomised controlled trial in hip fracture patients, intervention includes both intravenous and oral supplementation
Espaulella 2000	Randomised controlled trial of isocaloric supplement with and without protein in patients with fractured femur
Eyer 1993	Randomised controlled trial of early versus late nasogastric feeding, mean age under 45 years
Forli 2001	Randomised controlled trial of energy rich diet versus normal hospital diet in lung transplantation candidates
Fuenzalida 1990	Randomised controlled trial of oral protein and energy supplement, mean age 62 years with COPD
Gall 1998	Not randomised controlled trial, fortified food and snacks
Gallager 1992	Randomised controlled trial of supplementary overnight nasogastric feeding
Ganzoni 1994	Randomised controlled trial of dietary advice alone
Goris 2003	Randomised controlled trial of oral protein and energy supplements, mean aged 62 years.
Harrington 2004	Randomised controlled trial of high‐protein, high sodium vs adequate protein low sodium, probably too young
Hartgrink 1998	Randomised controlled trial of supplementary overnight nasogastric feeding
Hickson 2004	Randomised controlled trial of intensive feeding support from health care assistants
Hogarth 1996	Randomised controlled trial, oral supplement glucose and vitamins, no protein
Houwing 2003	Randomised controlled trial, oral supplement enriched with arginine versus placebo for hip fracture patients
Hurson 1995	Randomised controlled trial of oral supplement arginine
Jamieson 1997	Not randomised controlled trial, no information on age, wide range of nutrition interventions
Johansen 2004	Randomised controlled trial of specialised nutrition team versus standard regime, average age 62 years.
Johansson 2002	Randomised controlled trial of internal fixation or cemented primary total hip arthroplasty, a non‐randomised subgroup of patients were included in a study of nutrition support
Keane 1998	Randomised controlled trial of vitamin D‐fortified milk versus unfortified milk
Keele 1997	Randomised controlled trial of oral protein and energy supplement versus normal diet (2 phase), control group mean age 60 years, intervention group mean age 65 years
Kemen 1991	Randomised controlled trial of intact versus hydrolysed protein via jejunostomy, no age given
Kerrigan 1996	Not randomised controlled trial of calorifically dense supplement, no control group, nursing home residents, no age given
Kiel 1992	Not randomised controlled trial, fortification and added nutrients at meal times versus supplementation between meals, no age given
Kretser 2000	Not randomised controlled trial, 21 meals per week intensive meals on wheels programme
Kumagai 1999	Not randomised controlled trial, dietary consultation only
Lau 2001	Randomised controlled trial of calcium supplemented milk powder, control group mean age 56.8 years, intervention group mean age 57.1 years.
Lawson 2000	Cluster randomised by ward to protein and energy supplement or control group, post‐operative orthopaedic patients. Too young
Lewis 1987	Randomised controlled trial in patients with chronic pulmonary disease of protein and energy sip feeds, control group mean age 59 years, treatment group mean age 65 years
Neumann 2004	Randomised controlled trial of 2 oral supplements containing protein and energy in patients recovering from hip fracture‐ intervention group additional protein only
Nickerson 1998	Randomised controlled trial of oral amino acid arginine supplementation
Olin 1998	Not randomised controlled trial, of regular meals versus enriched meals in matched groups of elderly nursing home residents
Otte 1989	Randomised controlled trial of oral protein and energy supplementation of patients with pulmonary emphysema, mean age 56 years and 54 years
Pardy 1986	Not randomised controlled trial, all patients given protein and energy oral supplement, no age given
Paton 2004	Randomised controlled trial of protein and energy supplementation, mean age 38.4 years and 39.5 years
Planas 2005	Randomised controlled trial of protein and energy supplementation, mean age 60 years
Price 2006	Non randomised trial of personalized snack based intervention for hip fracture patients
Pupim 2002	Randomised controlled trial of oral nutritional supplementation, mean age 45.6 years and 50.0 years
Rana 1992	Randomised controlled trial of post‐operative oral protein and energy supplements, control group mean age 64 years, treatment group mean age 58 years
Remsburg 2001	Randomised controlled trial of buffet style dining program versus conventional tray style meal in nursing home residents
Rogers 1992	Randomised controlled trial of increased energy and protein feeding, control group mean age 64 years, treatment group mean age 64 years
Schurch 1998	Randomised controlled trial of isocaloric supplement, with or without protein, patients with fractured femur, mean age 80 years
Seri 1984	Randomised controlled trial of early nutritional support via jejunostomy versus conventional care, average age under 34 years
Seven 2003	Randomised controlled trial early versus late feeding in laryngectomized patients with malignant tumour, mean age 55 years and 56 years
Slodkowski 1994	Dietary study, mean age 54 years
Smedley 2004	Randomised controlled trial of the effects of preoperative and postoperative supplements in patients following GI surgery. Too young
Stauffer 1986	Not randomised controlled trial, oral protein and energy supplementation of patients with chronic pulmonary disease, patients used as own controls
Storm 1998	Randomised controlled trial of supplementation with milk versus calcium carbonate versus placebo, no outcomes of interest
Sullivan 1998	Randomised controlled trial of supplementary overnight nasogastric feeding
Tkatch 1992	Randomised control trial of supplements with versus without extra protein for patients with fractured femur
Tschepe 1985	Randomised controlled trial of oral supplementation with branched‐chain amino‐acids, patients with liver cirrhosis, age not given
Vargas 1995	Randomised controlled trial of supplementary nasogastric feeding, mean age 64 years
Vermeeren 2001	Randomised controlled trial of effects of acute dose of nutritional supplement on metabolism and exercise capacity, supplemented for <1day, mean age part 1 65years, part 2 62 years
Von Meyenfeldt 1990	Randomised controlled trial of pre‐operative parenteral nutrition versus nasogastric nutrition, patients with cancer, mean age over 61years
Wachtler 1995	Randomised controlled trial of oral immunomodulatory diet, mean age 65 years
Wara 1985	Not randomised controlled trial, early liquid diet versus conventional treatment and food
Williams 1989	Controlled trial of protein and energy supplementation in elderly orthopaedic patients, problems with randomisation proceedure
Wilson 1986	Not randomised controlled trial, protein and energy supplementation of patients with emphysema, mean age 60 years
Wisten 2005	Patients in geriatric hospital rehabilitation wards randomised into an intervention group (porridge group) and a control group (standard diet without porridge). Unclear if additional protein and energy provided. No outcomes of interest
Wong 2004	A randomised controlled trial of dietary counselling only
Yeh 2000	Randomised controlled trial, megestrol acetate versus placebo in geriatric nursing home patients

Characteristics of ongoing studies [ordered by study ID]

Cameron 2003.

Trial name or title	Effectiveness of oral supplementation for older women with hip and other fractures (EONS)
Methods
Participants	43 older women with hip or other fractures
Interventions	(a) Oral nutritional supplementation: 235ml (1.5kcal/ml) daily for 40 days. (b) Usual care.
Outcomes	follow‐up: 1 and 4 months post fracture. Outcomes: ADL function, nutritional status and medical complications.
Starting date	Started April 2000, follow‐up completed.
Contact information	Prof Ian Cameron Rehabilitation Studies Unit University of Sydney PO Box 6 Ryde New South Wales Australia NSW 1680 ianc@mail.usyd.ed.au
Notes	Study completion confirmed bu Ian Cameron in October 2004.

CENEX 2007.

Trial name or title	CENEX study ‐ cluster randomised trial through 28 community health centres in Santiago, Chile
Methods
Participants	Age 65.0 to 67.9 years
Interventions	Factorial design of protein and energy supplement and/or exercise programme, or neither
Outcomes	follow‐up: 24 months Outcomes: pneumonia, walking capacity, BMI, acute respiratory infection, quality of life (SF36), depression, chronic diseases, physical and functional limitations, productive activity, falls, fracture, blood pressure, anthropometry, timed up‐and‐go, cardiovascular risk factors.
Starting date	2005
Contact information	Alan D Dangour alan.dangour@lshtm.ac.uk
Notes

Elia 2007.

Trial name or title	Community nutrition support trial
Methods
Participants	Elderly individuals in care homes
Interventions	(a) Dietary advice (b) Oral nutritional supplements
Outcomes	follow‐up: 6 months Outcome: quality of life
Starting date	2007‐2010
Contact information	Prof Marinos Elia University of Southampton England elia@soton.ac.uk
Notes

Kvamme 2007.

Trial name or title	Nutritional intervention in malnourished elderly patients
Methods
Participants
Interventions	(a) Fresubin protein energy drink and advice (b) Advice
Outcomes	follow‐up: 6, 12 weeks Outcomes: function, activities of daily living, hand grip, timed up and go, quality of life (SF36), depression, anthropometry
Starting date	2004‐2008
Contact information	Jan M Kvamme University Hospital of Northern Norway jan.magnus.kvamme@unn.no
Notes

Contributions of authors

JAN POTTER: Data from previous review, assistance with selection of studies and data extraction, co‐drafting of the protocol and review.

ANNE MILNE: Literature search, contacting trialists, selection of studies, data extraction, first drafts of protocol and review, data analysis and data presentation.

ANGELA VIVANTI: Assistance with selection of studies and data extraction, co‐drafting of the review.

ALISON AVENELL: Project supervisor, assistance with selection of studies and data extraction, co‐drafting of the protocol and review, assisting with devising data analysis and methodological support.

Sources of support

Internal sources

• University of Aberdeen, UK.

External sources

• Medical Research Council, UK.

• Chief Scientist Office, UK.

• Student Awards Agency for Scotland, UK.

Declarations of interest

One reviewer is also the author of an eligible trial (Potter 2001).

Edited (conclusions changed)