Nutritional supplementation for hip fracture aftercare in older people

Alison Avenell; Toby O Smith; James P Curtain; Jenson CS Mak; Phyo K Myint

doi:10.1002/14651858.CD001880.pub6

. 2016 Nov 30;2016(11):CD001880. doi: 10.1002/14651858.CD001880.pub6

Nutritional supplementation for hip fracture aftercare in older people

Alison Avenell ^1,^✉, Toby O Smith ², James P Curtain ³, Jenson CS Mak ⁴, Phyo K Myint ⁵

Editor: Cochrane Bone, Joint and Muscle Trauma Group

PMCID: PMC6464805 PMID: 27898998

Abstract

Background

Older people with hip fractures are often malnourished at the time of fracture, and subsequently have poor food intake. This is an update of a Cochrane review first published in 2000, and previously updated in 2010.

Objectives

To review the effects (benefits and harms) of nutritional interventions in older people recovering from hip fracture.

Search methods

We searched the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register, CENTRAL, MEDLINE, MEDLINE In‐Process & Other Non‐Indexed Citations, Embase, CAB Abstracts, CINAHL, trial registers and reference lists. The search was last run in November 2015.

Selection criteria

Randomised and quasi‐randomised controlled trials of nutritional interventions for people aged over 65 years with hip fracture where the interventions were started within the first month after hip fracture.

Data collection and analysis

Two review authors independently selected trials, extracted data and assessed risk of bias. Where possible, we pooled data for primary outcomes which were: all cause mortality; morbidity; postoperative complications (e.g. wound infections, pressure sores, deep venous thromboses, respiratory and urinary infections, cardiovascular events); and 'unfavourable outcome' defined as the number of trial participants who died plus the number of survivors with complications. We also pooled data for adverse events such as diarrhoea.

Main results

We included 41 trials involving 3881 participants. Outcome data were limited and risk of bias assessment showed that trials were often methodologically flawed, with less than half of trials at low risk of bias for allocation concealment, incomplete outcome data, or selective reporting of outcomes. The available evidence was judged of either low or very low quality indicating that we were uncertain or very uncertain about the estimates.

Eighteen trials evaluated oral multinutrient feeds that provided non‐protein energy, protein, vitamins and minerals. There was low‐quality evidence that oral feeds had little effect on mortality (24/486 versus 31/481; risk ratio (RR) 0.81 favouring supplementation, 95% confidence interval (CI) 0.49 to 1.32; 15 trials). Thirteen trials evaluated the effect of oral multinutrient feeds on complications (e.g. pressure sore, infection, venous thrombosis, pulmonary embolism, confusion). There was low‐quality evidence that the number of participants with complications may be reduced with oral multinutrient feeds (123/370 versus 157/367; RR 0.71, 95% CI 0.59 to 0.86; 11 trials). Based on very low‐quality evidence from six studies (334 participants), oral supplements may result in lower numbers with 'unfavourable outcome' (death or complications): RR 0.67, 95% CI 0.51 to 0.89. There was very low‐quality evidence for six studies (442 participants) that oral supplementation did not result in an increased incidence of vomiting and diarrhoea (RR 0.99, 95% CI 0.47 to 2.05).

Only very low‐quality evidence was available from the four trials examining nasogastric multinutrient feeding. Pooled data from three heterogeneous trials showed no evidence of an effect of supplementation on mortality (14/142 versus 14/138; RR 0.99, 95% CI 0.50 to 1.97). One trial (18 participants) found no difference in complications. None reported on unfavourable outcome. Nasogastric feeding was poorly tolerated. One study reported no cases of aspiration pneumonia.

There is very low‐quality evidence from one trial (57 participants, mainly men) of no evidence for an effect of tube feeding followed by oral supplementation on mortality or complications. Tube feeding, however, was poorly tolerated.

There is very low‐quality evidence from one trial (80 participants) that a combination of intravenous feeding and oral supplements may not affect mortality but could reduce complications. However, this expensive intervention is usually reserved for people with non‐functioning gastrointestinal tracts, which is unlikely in this trial.

Four trials tested increasing protein intake in an oral feed. These provided low‐quality evidence for no clear effect of increased protein intake on mortality (30/181 versus 21/180; RR 1.42, 95% CI 0.85 to 2.37; 4 trials) or number of participants with complications but very low‐quality and contradictory evidence of a reduction in unfavourable outcomes (66/113 versus 82/110; RR 0.78, 95% CI 0.65 to 0.95; 2 trials). There was no evidence of an effect on adverse events such as diarrhoea.

Trials testing intravenous vitamin B1 and other water soluble vitamins, oral 1‐alpha‐hydroxycholecalciferol (vitamin D), high dose bolus vitamin D, different oral doses or sources of vitamin D, intravenous or oral iron, ornithine alpha‐ketoglutarate versus an isonitrogenous peptide supplement, taurine versus placebo, and a supplement with vitamins, minerals and amino acids, provided low‐ or very low‐quality evidence of no clear effect on mortality or complications, where reported.

Based on low‐quality evidence, one trial evaluating the use of dietetic assistants to help with feeding indicated that this intervention may reduce mortality (19/145 versus 36/157; RR 0.57, 95% CI 0.34 to 0.95) but not the number of participants with complications (79/130 versus 84/125).

Authors' conclusions

There is low‐quality evidence that oral multinutrient supplements started before or soon after surgery may prevent complications within the first 12 months after hip fracture, but that they have no clear effect on mortality. There is very low‐quality evidence that oral supplements may reduce 'unfavourable outcome' (death or complications) and that they do not result in an increased incidence of vomiting and diarrhoea. Adequately sized randomised trials with robust methodology are required. In particular, the role of dietetic assistants, and peripheral venous feeding or nasogastric feeding in very malnourished people require further evaluation.

Plain language summary

Nutritional supplementation for older people after hip fracture

Background and aim

Older people with hip fractures are often malnourished at the time of their fracture and many have poor food intake while in hospital. Malnutrition may hinder recovery after hip fracture. We reviewed the effects of nutritional interventions in older people recovering from hip fracture.

Search results

We searched the scientific literature up to November 2015 and include 41 studies including 3881 participants. All nutritional interventions were started within one month of hip fracture. The studies had flaws in their methods that may affect the validity of their results. Some evidence was very low quality which means we are very unsure of the results.

Key results

Eighteen studies examined the use of additional oral feeds that provided energy from sources other than protein, protein, some vitamins and minerals. There was low‐quality evidence that these multinutrient oral feeds may not reduce mortality but that they may reduce the number of people with complications (e.g. pressure sore, infection, venous thrombosis, pulmonary embolism, confusion). There was very low‐quality evidence that oral multinutrient feeds may reduce unfavourable outcome (death or complications) and that they did not result in increased vomiting and diarrhoea.

Four studies examined nasogastric tube feeding, where liquid food is delivered via a tube inserted into the nose and passed down into the stomach, with non‐protein energy, protein, some vitamins and minerals. These studies provided very low‐quality evidence that tube feeding, which was poorly tolerated, did not seem to make a difference to mortality or complications. Unfavourable outcome was not recorded and there was insufficient evidence on adverse events.

One study provided very low‐quality evidence that nasogastric tube feeding followed by oral feeds may not affect mortality or complications. It reported that tube feeding was poorly tolerated.

One study provided very low‐quality evidence that giving feed into a vein initially and then by mouth may not affect mortality but may reduce complications. However, we were surprised that this intervention was being used in people who seemed to be able to take nutrition orally.

Increasing protein intake in an oral feed was tested in four studies. These provided low‐quality evidence of no clear effect on mortality or complications and very low‐quality evidence for a reduction in unfavourable outcome.

Studies testing intravenous vitamin B1 and other water soluble vitamins, oral 1‐alpha‐hydroxycholecalciferol (vitamin D), high dose bolus vitamin D, different oral doses or sources of vitamin D, intravenous or oral iron, ornithine alpha‐ketoglutarate versus an isonitrogenous peptide supplement, taurine versus placebo, and a supplement with vitamins, minerals and amino acids, provided low‐ or very low‐quality evidence of no clear effect on mortality or complications, where reported.

One study, evaluating the use of dietetic assistants to help with feeding, provided low‐quality evidence that this may reduce mortality but not the numbers of people with complications.

Conclusions

Oral supplements with non‐protein energy, protein, vitamins and minerals started before or soon after surgery may prevent complications after hip fracture in older people but may not affect mortality. Adequately sized randomised studies with better design are required. We suggest that the role of dietetic assistants, and of peripheral venous feeding or nasogastric feeding in very malnourished patients, require further evaluation.

Summary of findings

Background

Description of the condition

Hip (proximal femur) fractures are a common cause of substantial morbidity and mortality in older people living in industrialised societies (Kanis 2012). Nine months after their hip fracture people still have poorer quality of life than age and sex matched controls (Cranney 2005). Many people fail to return to their own homes and previous state of mobility after a hip fracture. In industrialised societies, mortality in the year after hip fracture may be as high as 30% (Bentler 2009; Mariconda 2015), and averages 11% during the first few months after fracture (Lyons 1997). Mortality in the first four months after hip fracture surgery is age dependent. For instance, mortality was reported as 5% in people aged 50 to 69 years, compared with 28% in those people aged 90 years or over in the Scottish Hip Fracture Audit Report (Holt 2008). A meta‐analysis of prospective studies found the relative hazard for mortality during the first three months following hip fracture to be 5.75 (95% CI 4.94 to 6.67) for women, and higher in men at 7.95 (95% CI 6.13 to 10.30) (Haentjens 2010). Excess mortality was also found to persist for as long as 10 years after hip fracture for both men and women. For those who survive, acute hospital costs are substantial, but long‐term costs in rehabilitation and extra care in the community are even greater (Dolan 1998; Haentjens 2005; Johnell 1997).

People with hip fractures, who are more likely to be older and frailer, are often malnourished or at risk of becoming malnourished at the time of the fracture (Bachrach 2001; Bastow 1983a; Koren‐Hakim 2012; Lumbers 2001). Social, psychological, physical, economic, medical and cognitive influences may all contribute to the risk of malnutrition. Dietary intake in people recovering from hip fracture in hospital is frequently suboptimal (Bell 2014; Jallut 1990; Lumbers 2001; Nematy 2004; Patterson 1992; Stableforth 1986).

Under‐nutrition leads to depressed mood, muscle wasting and reduced muscle power, and impaired cardiac function (Keys 1950). All of these will impair mobility and increase the tendency to develop postoperative medical complications (e.g. pneumonia, pressure sores, deep venous thrombosis) and hinder recovery, both in hospital and subsequently (Lennard‐Jones 1992), increasing health and social care costs. Malnutrition also impairs the immune response, which will enhance the risk of postoperative infection (Lesourd 1997). Poor nutritional status is associated with an increased risk of pressure ulcers after hip fracture (Lindholm 2008).

Description of the intervention

Examined in this review are nutritional interventions started within the first month after a hip fracture that are aimed at improving recovery from hip fracture by increasing the intake of energy, protein, vitamins and minerals, alone or in combination. Nutrition interventions can be provided by various routes: oral (by mouth), enteral (tube feeding into the stomach or small bowel, including percutaneous endoscopic gastrostomy) or parenteral (intravenous and intramuscular), also alone or in combination. Also considered are interventions that revolve round the administration of nutrition, such as the use of dietetic assistants in hospital.

How the intervention might work

There is an association between frailty, including that related to nutrition, and unfavourable outcomes following a hip fracture. Modification of nutritional status in the rehabilitation period, particularly early on, could be beneficial in reducing functional decline and reducing complications.

Making links between nutritional status and fracture recovery is complicated by the fact that markers of dietary protein depletion measured in blood, such as albumin, prealbumin, and transferrin are partly affected by fluid shifts and responses to injury and infection. Nevertheless, associations have been shown between low serum albumin and increased postoperative complications and poorer survival (Foster 1990; Patterson 1992). Another factor that has been implicated is vitamin C, which is required for an effective immune response and collagen formation required for wound healing. Low leucocyte vitamin C levels have been associated with the development of pressure sores in people with hip fracture (Brown 1992a; Goode 1992).

More direct markers of nutritional status are anthropometric indices, such as weight in relation to height, triceps skinfold for body fat, and mid‐upper arm circumference for muscle and fat mass. People with hip fracture have lower triceps skinfold and mid‐upper arm circumference than healthy people in the same age category (Mansell 1990; Nematy 2004). In a study of 744 people with hip fracture, Bastow 1983a found that low triceps skinfold and arm muscle circumference was associated with lower calorie intake on the ward and predicted poorer survival after hip fracture.

Why it is important to do this review

As described above, people with hip fracture are sometimes undernourished, and poor food intake may occur during routine care, hindering recovery. There is therefore an argument for nutritional supplementation in this group, and consequently a need to evaluate the use of nutrition interventions in this group of people by examining the evidence from relevant randomised controlled trials. This is the seventh update of our Cochrane review first published in 2000, and previously updated in 2010. The previous update (Avenell 2010) continued to point to the insufficiency of the available evidence to draw robust conclusions.

Objectives

To review the effects (benefits and harms) of nutritional interventions in older people recovering from hip fracture.

We considered comparisons where people with hip fracture, who were randomly allocated a nutritional intervention, including supplements, were compared with those allocated to no intervention or placebo. Where possible, effects were examined according to pre‐existing nutritional status: malnourished or not malnourished.

We also considered comparisons between nutritional interventions if these were compared in a randomised controlled trial.

Methods

Criteria for considering studies for this review

Types of studies

We included all randomised controlled trials (RCTs) and quasi‐randomised (e.g. allocation by date of birth or hospital record number) controlled trials of nutritional supplements post hip fracture. We also included trials that could not be analysed on an intention‐to‐treat basis, and those that lacked blinding or use of placebo treatment.

Studies of nutritional interventions that examined the secondary prevention of osteoporotic fractures after hip fracture were not considered in this review.

Types of participants

We included trials of older people recovering from any type of hip fracture. It was anticipated that most participants would be over 65 years of age. If the number of younger participants was relatively small, and provided there was adequate randomisation with unbiased distribution of this age group between the intervention and control groups, we retained them. Trials that focused specifically or mainly on younger people, people with multiple trauma or people with pathological fractures (e.g. cancer‐related fractures) were excluded. We also excluded trials published before 1980 with undefined geriatric populations or with mixed populations with fewer than five participants with hip fracture in each intervention group.

Studies reporting results on mixed populations of orthopaedic or other geriatric patients were only included, either if separate data were available from the participants with fracture of the hip, or when contact with the study authors resulted in the provision of such data.

The participants studied may have resided in a hospital or in a rehabilitation unit or any location after discharge from either of these facilities.

Types of interventions

We included trials of nutritional interventions aimed to improve the recovery from hip fracture by increasing the intake of energy, protein, vitamins and minerals, alone or in combination. Nutritional interventions were provided by oral (by mouth), enteral (tube feeding into the stomach or small bowel, including percutaneous endoscopic gastrostomy) or parenteral (intravenous and intramuscular) routes, alone or in combination. Interventions included those evaluating the administration of nutrition, such as the use of dietetic assistants. The interventions examined were started within the first month after hip fracture, and given for up to one year. Trials evaluating intravenous fluid administration in the immediate postoperative period for hydration purposes were excluded.

Interventions included multinutrient supplements (providing non‐protein energy, protein, vitamins and minerals) given orally, enterally or intravenously, compared with supplements containing less or none of these components, or no treatment. We included interventions of vitamins, minerals, amino acids or related compounds compared with lower doses, placebo or no treatment. We also included trials examining different policies to provide nutrition, for example, additional assistance from dietetic assistants.

Types of outcome measures

We sought information on the following outcomes, which we split into main outcomes (and further categorised into primary and secondary outcomes) and other outcomes. Additionally, we made the collection of 'unfavourable outcome' explicit.

Main outcomes

Primary outcomes

All cause mortality
Morbidity, postoperative complications (e.g. wound infections, pressure sores, deep venous thromboses, respiratory and urinary infections, cardiovascular events)
'Unfavourable outcome'. This was defined as the number of trial participants who died plus the number of survivors with complications. Alternatively, where these data were unavailable, we accepted a slightly different definition (mortality or survivors with a major complication or two or more minor complications) originally presented in Delmi 1990.

Secondary outcomes

Length of hospital and rehabilitation unit stay
Postoperative functional status (cognitive functioning, mobility and ability to perform activities of daily living)
The level of care and extent of support required after discharge
Patient perceived quality of life after discharge
Fracture healing
Putative side effects of treatment (e.g. diarrhoea, aspiration pneumonia, specific intravenous line complications)

Other outcomes

Patient tolerance of/compliance with nutrition interventions
Carer burden and stress
Economic outcomes

For this update, we shortened the list of 'other outcomes' that appeared in previous versions of this review (Avenell 2010). The removed outcomes are listed in Differences between protocol and review.

Search methods for identification of studies

Electronic searches

We searched the Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (9 October 2014), the Cochrane Central Register of Controlled Trials (CENTRAL; 2015 issue 12) in The Cochrane Library, MEDLINE (1966 to October Week 5 2015), MEDLINE In‐Process & Other Non‐Indexed Citations (10 November 2015), Embase (1980 to 2015 Week 45), CAB Abstracts (1973 to 2015 Week 44), and the Cumulative Index to Nursing and Allied Health Literature (CINAHL) (1982 to 10 November 2015). For this update, the search results were limited from 2008 onwards.

In MEDLINE (Ovid), we combined the sensitivity‐maximizing version of the Cochrane Highly Sensitive Search Strategy for identifying randomised trials (Lefebvre 2011) with subject‐specific terms. We modified this strategy for use in the other databases (see Appendix 1 for search strategies).

We also searched the ISRCTN registry (17 February 2015), the WHO International Clinical Trials Registry Platform (17 February 2015), the UK Clinical Research Network Study Portfolio (17 February 2015), to identify ongoing trials.

We did not apply any language restrictions. We have given details of the search methods used for the previous version of the review in Appendix 2.

Searching other resources

We checked reference lists of articles, searched books related to orthopaedics, geriatric medicine and nutrition, and corresponded with colleagues and investigators.

Data collection and analysis

Selection of studies

Two review authors (from AA, TS, JC) independently assessed reports of potentially eligible studies and resolved any differences by discussion.

Data extraction and management

Two review authors (from AA, TS, JC, JM) independently extracted data. We resolved all differences by discussion. We extracted data using a pre‐derived data extraction form and entered the agreed results into Review Manager (RevMan) (RevMan 2014). If necessary, we contacted trialists for further information on methodology and data.

Assessment of risk of bias in included studies

In this update, two review authors (from AA, TS, JC, JM) independently assessed risk of bias in all included trials using the Cochrane 'Risk of bias' tool (Higgins 2011). This assesses sequence generation, allocation concealment, blinding of participants or personnel, blinding of outcome assessment, completeness of outcome data, selective outcome reporting and other potential sources of bias. We considered primary and secondary outcomes separately in our assessment of blinding of outcome assessment and completeness of outcome data. We resolved any differences of opinion by consensus or by consulting a third party.

Our risk of bias assessment superceded our assessment of methodological quality in previous versions of this review (Avenell 2010); see Differences between protocol and review.

Measures of treatment effect

For each study, risk ratios and 99% confidence intervals (CI) were calculated for dichotomous outcomes and mean differences and 99% CIs for continuous outcomes. The choice of 99% CIs reflects the extra burden of proof we considered appropriate for individual trials, in view of their generally poor quality. Summary estimates for meta‐analysis are provided as 95% CIs.

Unit of analysis issues

Although we would have included cluster‐randomised trials, the unit of randomisation in all included trials was the individual participant.

Dealing with missing data

We have presented mortality results using denominators based on the numbers of participants at randomisation (intention‐to‐treat analysis), where available. Generally, we presented the results for other outcomes using denominators based on the numbers of participants available at follow‐up. In some cases, we investigated the effect of drop outs and exclusions by conducting worst scenario analyses for the primary outcomes, where those who were missing to follow‐up in the intervention group were assumed to have the poorer outcome but not those who were missing in the control group. We were alert to the potential mislabelling or non identification of standard errors and standard deviations. Unless missing standard deviations could be derived from confidence intervals or standard errors, we did not assume values in order to present these in the analyses.

Assessment of heterogeneity

Heterogeneity was assessed by visual inspection of the forest plot (analysis) along with consideration of the Chi² test for heterogeneity and the I² statistic (Higgins 2003).

Assessment of reporting biases

We considered that there were sufficient data available to present funnel plots to explore the potential for publication bias for multinutrient supplements and the outcomes of mortality and complications. Our search of 'grey literature', dogged pursuit of trials listed in clinical trial registers and contact with trial authors should have helped to avoid some publication bias.

Data synthesis

Where appropriate, we combined the results of comparable groups of trials using both fixed‐effect, as the main analysis, and random‐effects models. We have presented all data for individual trials in meta‐analyses as 99% CIs, and pooled data with 95% CIs.

Subgroup analysis and investigation of heterogeneity

Our primary subgroup analysis was based on pre‐existing nutritional status (malnourished targeted versus malnourished not targeted). We also presented data on multinutrient supplements stratified by route of delivery: oral supplements, nasogastric tube feeding, nasogastric tube feeding and oral supplements, and intravenous feeding and oral supplements. To test whether the subgroups were statistically significantly different from one another, we tested the interaction using the inbuilt facility in RevMan 2014 that is based on methods outlined by Deeks 2011 in the Cochrane Handbook for Systematic Reviews of Interventions (Chapter 9).

Sensitivity analysis

We planned sensitivity analyses based on aspects of trial and review methodology. We have explored the risk of bias associated with inadequate concealment of allocation (unclear or high risk of selection bias) and the choice of statistical model for pooling data (fixed‐effect versus random‐effects).

'Summary of findings' tables and assessment of the quality of the evidence

We have presented the results for mortality, participants with complications and unfavourable outcomes (our primary outcome measures) and adverse events (e.g. vomiting and diarrhoea) in separate 'Summary of findings' tables for the comparisons of oral multinutrient supplements versus control and nasogastric multinutrient supplements versus control. For each outcome for each comparison, we graded the evidence as 'very low', 'low', 'moderate' or 'high' in accordance with the GRADE working group criteria (Guyatt 2008).

Results

Description of studies

Results of the search

We updated the search from 2008 to November 2015. We screened a total of 2459 records from the following databases: Cochrane Bone, Joint and Muscle Trauma Group Specialised Register (7), CENTRAL (340), MEDLINE (483), Embase (847), CAB Abstracts (234), and CINAHL (548). We did not identify any additional new trials from Current Controlled Trials, the WHO International Clinical Trials Registry Platform or the UK clinical research network study portfolio. We also identified one potentially eligible study from contact with the author (Luo 2015).

The search update resulted in the identification of 32 new studies (many published in multiple articles) for potential inclusion, for which we obtained reports. Upon study selection, we found 17 trials eligible for inclusion (Anbar 2014; Bischoff‐Ferrari 2010; Botella‐Carretero 2010; Chevalley 2010; Fabian 2011; Flodin 2014; Glendenning 2009; Kang 2012; Luo 2015; Myint 2013; Papaioannou 2011; Parker 2010; Prasad 2009; Scivoletto 2010; Serrano‐Trenas 2011; Van Stijn 2015; Wyers 2013), we excluded six studies (Bell 2014; Gunnarsson 2009; Hitz 2007; Hoekstra 2011; Holst 2012; Li 2012), we placed five in ongoing trials (ACTRN12609000241235; ACTRN12612000448842; NCT01404195; NCT01505985; Rowlands) and four await classification (Benati 2011; Bernabeu‐Wittel 2016; Ekinci 2015; Ish‐Shalom 2008).

We excluded one previously ongoing study (Cameron 2011). A second (NCT00523575) was published and is now an included study (Wyers 2013).

Overall, there are now 41 included studies, 43 excluded studies, seven ongoing trials and six studies awaiting classification.

Further details of the process of screening and selecting studies for inclusion in the review are illustrated in Figure 1. The results of the search reported in the previous version of the review (Avenell 2010) can be found in Appendix 3.

Included studies

Details of study methods, population, interventions and outcomes of individual trials are provided in the Characteristics of included studies.

We obtained further details (including clarifications) on methodology, trial participants and outcomes, from trialists of 23 studies (Bastow 1983b; Botella‐Carretero 2008; Botella‐Carretero 2010; Brown 1992b; Bruce 2003; Chevalley 2010; Day 1988; Duncan 2006; Eneroth 2006; Espaulella 2000; Flodin 2014; Hankins 1996; Hartgrink 1998; Houwing 2003; Luo 2015; Miller 2006; Myint 2013; Neumann 2004; Parker 2010; Prasad 2009; Sullivan 1998; Sullivan 2004; Tidermark 2004) and other sources for two trials (Ronald Koretz for Gallagher 1992; Jane Robertson for Hoikka 1980).

Design

Thirty‐seven trials were RCTs. The other four trials were quasi‐randomised trials (Bastow 1983b; Brown 1992b; Bruce 2003; Hoikka 1980). There were no cluster or cross‐over randomised trials.

Sample sizes

The 41 included studies involved a total of 3881 participants. Sample size ranged from 10 participants in Brown 1992b to 318 participants in Duncan 2006.

Setting

The publication dates of the trials span 35 years, Hoikka 1980 being the earliest. Most of the trials were based in a single centre. Trials were conducted in 15 countries (Australia, Austria, Canada, China, Finland, Israel, Italy, Korea, the Netherlands, Russia, Spain, Sweden, Switzerland, UK, USA), with eight trials being conducted in the UK, five each in Australia and Switzerland, four in the USA, three in Spain, and three each in the Netherlands and Sweden.

Participants

The majority of participants were female and in 10 studies all participants were female (Bastow 1983b; Bean 1994; Brown 1992b; Bruce 2003; Chevalley 2010; Duncan 2006; Fabian 2011; Serrano‐Trenas 2011; Stableforth 1986; Tidermark 2004). Sullivan 1998 and Sullivan 2004 were the only studies where male participants formed the majority. Where reported, the mean age of participants was usually over 80 years. Luo 2015 and Papaioannou 2011 had younger participants with a mean age of 69 years. Gallagher 1992 gave no details on age, but the rest of the details provided in the abstract were compatible with an older population. Only Bean 1994 applied an upper age limit, this being 85 years.

All studies (except Miller 2006, which included participants with lower limb fractures) included only participants with hip fracture. We obtained separate data for participants with hip fracture for Miller 2006. Nineteen studies provided information on the types of hip fractures suffered by the participants (Anbar 2014; Day 1988; Delmi 1990; Eneroth 2006; Espaulella 2000; Flodin 2014; Hartgrink 1998; Myint 2013; Parker 2010; Prasad 2009; Schürch 1998; Scivoletto 2010; Serrano‐Trenas 2011; Stableforth 1986; Sullivan 1998; Sullivan 2004; Tidermark 2004; Tkatch 1992; Wyers 2013). Seventeen studies excluded people with dementia or severe cognitive dysfunction. Many studies excluded people with a wide range of medical conditions (Anbar 2014; Bastow 1983b; Bean 1994; Bischoff‐Ferrari 2010; Brown 1992b; Chevalley 2010; Delmi 1990; Eneroth 2006; Espaulella 2000; Flodin 2014; Luo 2015; Myint 2013; Schürch 1998; Scivoletto 2010; Tidermark 2004; Tkatch 1992; Van Stijn 2015; Wyers 2013). Eight studies indicated that consent (assent) was acceptable if given by a relative or guardian (Day 1988; Duncan 2006; Espaulella 2000; Hankins 1996; Houwing 2003; Parker 2010; Sullivan 1998; Sullivan 2004).

Eight studies, involving 616 participants, examined the effect of supplementation on malnourished participants (Bastow 1983b; Bean 1994; Brown 1992b; Gallagher 1992; Hankins 1996; Luo 2015; Miller 2006; Myint 2013). Gallagher 1992 and Luo 2015 defined participants as malnourished on the basis of serum albumin; other studies used anthropometric measurements, such as mid‐upper arm circumference. Myint 2013 recruited participants if BMI was < 25 kg/m²and mean BMI was actually 21.7 kg/m². We discussed this with consultant geriatrician colleagues, who advised that participants in this trial be considered under 'malnourished targeted' category of subgroup analysis.

Interventions

The 41 included trials evaluated a variety of nutritional supplements, mostly in comparison with a control group. We have provided details of these and the method of delivery in individual studies in the Characteristics of included studies. The comparisons under test fell into five categories (as detailed below).

Four studies had three treatment groups each. Madigan 1994 had three groups: the two supplemented groups (one with a multivitamin and mineral supplement) were subsequently combined in the report, owing to small numbers at follow‐up. Since these two groups both fit the criterion in this review for a 'multinutrient' supplement group, the combined results for these two groups, compared with the control, are also presented here. Botella‐Carretero 2008 also had three groups: oral protein and energy, oral protein, and control; both supplemented groups have been combined for this review, also owing to small numbers. Papaioannou 2011 compared an initial oral bolus dose of 100,000 IU vitamin D2 versus 50,000 IU vitamin D2 versus placebo; followed by 1,000 IU vitamin D3 for 90 days in all groups. Finally, Chevalley 2010 compared three different protein sources: oral casein protein versus oral whey protein versus oral whey protein plus essential amino acids.

Miller 2006 had four groups: a nutrition supplementation group, a physical activity intervention group, a combined intervention group, and an attention control group. We have only used data from the nutritional supplementation only and control groups here. Bischoff‐Ferrari 2010 had a factorial design with randomisation to two different doses of vitamin D3 and standard or extended physiotherapy.

We made the following comparisons:

Multinutrient supplements (oral, nasogastric, intravenous) versus control

The multinutrient supplements under investigation usually provided non‐protein energy, protein, some vitamins and minerals. These were delivered either orally, via a nasogastric tube, intravenously, or combinations of these.

Oral supplements

The 18 studies testing oral supplements involved 1190 participants (Anbar 2014; Botella‐Carretero 2008; Botella‐Carretero 2010; Brown 1992b; Bruce 2003; Delmi 1990; Fabian 2011; Flodin 2014; Hankins 1996; Houwing 2003; Kang 2012; Luo 2015; Madigan 1994; Miller 2006; Myint 2013; Stableforth 1986; Tidermark 2004; Wyers 2013). Anbar 2014 undertook three measurements of resting energy expenditure to estimate requirements for the intervention group. Wyers 2013 included five dietetic visits and five follow‐up phone calls for the intervention group. Interventions were usually started pre‐operatively or in two days postoperatively and most continued for at least a month.

Nasogastric tube feeding

Four studies involving 377 participants (Bastow 1983b; Gallagher 1992; Hartgrink 1998; Sullivan 1998), examined supplementation starting within five days of surgery and continued usually until oral intake was sufficient, or hospital discharge.

Nasogastric tube feeding and oral supplements

One study involving 57 participants (Sullivan 2004), examined supplementation postoperatively until oral intake was sufficient.

Intravenous feeding and oral supplements

One study involving 80 participants (Eneroth 2006) examined supplementation for the first 10 days in hospital.

High protein‐containing supplements versus low‐protein or non‐protein‐containing supplements

Protein supplementation was delivered within oral feeds, usually starting within a week of fracture and given for one to six months. This was assessed in four studies involving 371 participants (Espaulella 2000; Neumann 2004; Schürch 1998; Tkatch 1992). Whereas the protein supplement resulted in extra calories in the intervention group in Tkatch 1992, the energy content of both intervention and placebo groups were equivalent in Espaulella 2000 and Schürch 1998. Moderate quantities of minerals and vitamins were also provided with the protein supplement in Espaulella 2000 and Schürch 1998; none were in sufficient doses to detract from these being predominantly protein supplements. In Neumann 2004 there were differences in vitamin and mineral intakes between the high‐ and lower‐protein supplements, and the carbohydrate intake in the lower‐protein supplement resulting in similar energy contents of the two supplements.

Comparison of different protein sources

One study with 45 participants compared 20 g daily of oral casein protein versus 20 g of oral whey protein versus 15 g of oral whey protein and 5 g of essential amino acids in a ratio identical to casein, given from a mean of 10 days post‐fracture for a month (Chevalley 2010).

Vitamin supplementation versus control or lower dose supplementation

This comparison was based on four studies involving 335 participants (Bischoff‐Ferrari 2010; Day 1988; Hoikka 1980; Papaioannou 2011). Day 1988 investigated intravenous thiamin (vitamin B1) and water soluble vitamins versus control. Hoikka 1980 investigated the use of oral 1 mcg 1‐alpha‐hydroxycholecalciferol and 1 g calcium as calcium carbonate daily versus placebo and 1 g calcium as calcium carbonate daily for four months. Bischoff‐Ferrari 2010 investigated daily 2000 IU vitamin D3 compared with daily 800 IU vitamin D3; all participants also received 1 g of calcium as calcium carbonate daily over one year. Papaioannou 2011 compared an initial oral bolus dose of 100,000 IU vitamin D2 versus 50,000 IU vitamin D2 versus placebo; followed by 1000 IU vitamin D3 for 90 days in all groups. Interventions were commenced pre‐operatively or up to a mean of four days postoperatively and continued for between five days (Day 1988) and 3 to 12 months (Bischoff‐Ferrari 2010; Hoikka 1980; Papaioannou 2011).

Comparison of different vitamin D sources

One study with 95 participants compared oral vitamin D3 1000 IU/d and calcium carbonate equivalent to 600 mg/d to vitamin D2 1000 IU/d and calcium carbonate equivalent to 600 mg/d for three months from the inpatient stay (Glendenning 2009).

Iron supplementation versus control

Three studies with 568 participants investigated oral or intravenous iron supplementation compared with no intervention or placebo, started pre‐operatively or early postoperatively, for the first month after hip fracture (Parker 2010; Prasad 2009; Serrano‐Trenas 2011).

Vitamin, mineral and amino acid supplementation versus control

One study with 107 participants investigated six weeks of an oral Restorfast supplement daily (L‐carnitine, calcium, magnesium, vitamin D3, L‐leucine) followed by 10 weeks of an oral Riabylex supplement daily (creatine, L‐carnitine, coenzyme Q10, nicotinamide, pantothenic acid, riboflavin) (Scivoletto 2010). The start time for the intervention was unclear. Creatine, coenzyme Q and L‐carnitine were also included but are not vitamins, minerals or amino acids, and can be manufactured by the body.

Isonitrogenous ornithine alpha‐ketoglutarate versus peptide supplement

One study with 59 participants made this comparison (Bean 1994). Ornithine alpha‐ketogluturate is metabolised in part to the amino acid glutamine, and is used to improve nitrogen conservation. The interventions were probably delivered orally, and were given for two months, start time unclear.

Taurine versus placebo

One trial with 236 participants compared taurine with a placebo control (Van Stijn 2015). Taurine or placebo capsules were started pre‐operatively and then supplied for six days postoperatively. Taurine, which has been described as a conditionally essential amino acid, is a semi‐essential amino acid with antioxidant action to theoretically reduce oxidative stress which can be induced by surgical tissue injury. Taurine was provided three times a day with a scheme of 2‐1‐2 capsules of 1.2 g taurine to reach 6 g per day daily dose.

Dietetic assistants versus usual care

One study, involving 318 participants, tested the provision of extra assistance in the form of dietetic assistants, above that of dietitians and nurses, to help improve people's dietary intake (Duncan 2006). The dietetic assistants gave support for a median of 16 to 17 days; the start time for this assistance was unclear.

Excluded studies

We have given reasons for excluding 43 studies in the Characteristics of excluded studies. Six excluded studies were published in languages other than English, sufficient translation having been obtained to establish non‐eligibility. The major reasons for exclusion included studies not being RCTs (Bachrach 2001; Bell 2014; Bradley 1995; Giaccaglia 1986; Groth 1988; Gunnarsson 2009; Harju 1989; Hoekstra 2011; Holst 2012; Kacmaz 2007; Lawson 2003; Ravetz 1959; Tassler 1981); studies not recruiting (or presenting separate data for) people who had sustained a hip fracture (Brocker 1994; Cameron 2011; Goldsmith 1967; Hitz 2007; Larsson 1990; Lauque 2000; Lawson 2003; Pedersen 1999; Volkert 1996); and studies not presenting the outcomes of interest (Beringer 1986; Boudville 2002; Gegerle 1986; Stumm 2001; Wong 2004; Zauber 1992).

Ongoing studies

We have given details of seven ongoing trials in the Characteristics of ongoing studies. Of the ongoing studies, two (ACTRN12612000448842; Rowlands) with a total of 350 participants will examine the use of intravenous iron supplementation. The size of the trial by NCT00497978, which will examine the use of taurine supplementation, is unclear. ACTRN12609000241235 will recruit 150 participants to examine the effect of fish oils compared with other oils. In a trial with 340 participants, ACTRN12610000392066 will examine the use of oral 250,000 IU vitamin D3 compared with placebo. NCT01505985 and NCT01404195 will examine the use of oral multinutrient supplements compared with placebo or usual care in a total of 124 participants.

Studies awaiting classification

We have given details of the six studies in this category in the Characteristics of studies awaiting classification. We have requested further details from the trial investigators, where required.

Risk of bias in included studies

The quality of trial methodology, as reported, was disappointing and we could not rule out risk of bias associated with poor trial methods. Many of the trials failed to report trial methodology in sufficient detail. We obtained additional information on methods for nine trials (Brown 1992b; Bruce 2003; Day 1988; Espaulella 2000; Hankins 1996; Hartgrink 1998; Houwing 2003; Luo 2015; Sullivan 1998). We have summarised the risk of bias judgements in Figure 2 and Figure 3.

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies

Risk of bias summary: review authors' judgements about each risk of bias item for each included study

Allocation

Sequence generation

We judged random sequence generation as adequate resulting in low risk of bias in 13 trials (32%) (Anbar 2014; Bischoff‐Ferrari 2010; Chevalley 2010; Day 1988; Espaulella 2000; Houwing 2003; Luo 2015; Miller 2006; Papaioannou 2011; Prasad 2009; Sullivan 2004; Van Stijn 2015; Wyers 2013). Four trials (10%) were quasi‐randomised and thus at high risk of bias (Bastow 1983b; Brown 1992b; Bruce 2003; Hoikka 1980). The remaining trials we judged to be at unclear risk of risk because of insufficient details.

Allocation concealment

Concealment of allocation was judged to be adequate resulting in low risk of bias in 19 (46%) trials (Anbar 2014; Botella‐Carretero 2008; Botella‐Carretero 2010; Chevalley 2010; Duncan 2006; Espaulella 2000; Flodin 2014; Glendenning 2009; Hankins 1996; Houwing 2003; Miller 2006; Myint 2013; Papaioannou 2011; Parker 2010; Serrano‐Trenas 2011; Sullivan 1998; Sullivan 2004; Van Stijn 2015; Wyers 2013). Allocation was unlikely to be concealed in three of the quasi‐randomised studies (Bastow 1983b; Brown 1992b; Bruce 2003), which were judged to be at high risk of bias. The remaining trials were judged to be at unclear risk of risk because of insufficient details.

Blinding

We judged eight (20%) trials to be at low risk of performance bias (blinding of participants and personnel) (Bean 1994; Bischoff‐Ferrari 2010; Espaulella 2000; Glendenning 2009; Houwing 2003; Papaioannou 2011; Schürch 1998; Van Stijn 2015). These trials generally had placebo interventions, or were comparisons of different kinds of supplement. We judged 29 trials at high risk of performance bias and four at unclear risk.

We judged almost all trials (95%) to be at low risk of detection bias relating to blinding of outcome assessment for primary outcomes, with the exception of two trials reporting putative side effects of interventions (Chevalley 2010; Prasad 2009). Blinding of secondary or other outcomes was less likely to be judged low risk, with only six trials (15%) judged as being low risk of detection bias (Bischoff‐Ferrari 2010; Espaulella 2000; Houwing 2003; Myint 2013; Papaioannou 2011; Van Stijn 2015). The remaining trials we judged to be at unclear risk of detection bias for both domains.

Incomplete outcome data

We judged 15 trials to be at low risk of bias for incomplete outcome data (attrition bias) for primary outcomes. Fourteen trials were judged to be at high risk of bias in this category (Botella‐Carretero 2010; Chevalley 2010; Delmi 1990; Flodin 2014; Glendenning 2009; Luo 2015; Madigan 1994; Myint 2013; Papaioannou 2011; Prasad 2009; Scivoletto 2010; Serrano‐Trenas 2011; Tkatch 1992; Van Stijn 2015). The remainder were judged at unclear risk of attrition bias for primary outcomes, where reported.

Incomplete outcome data were more problematic for secondary outcome data, and we judged only 10 trials to be at low risk of attrition bias. Thirteen trials were judged to be at high risk of bias and the remainder, where secondary outcomes were reported, at unclear risk of attrition bias.

Selective reporting

We judged 14 trials (34%) to be at low risk of bias for selective reporting of outcomes. However, we judged seven trials to be at high risk of bias (Bischoff‐Ferrari 2010; Bruce 2003; Chevalley 2010; Day 1988; Fabian 2011; Gallagher 1992; Luo 2015), usually as a result of data not presented that would be expected from their methods, or data that were provided not mentioned in methods, for example, length of stay, mortality, functional status. The remainder were at unclear risk of selective reporting bias.

Other potential sources of bias

For other potential sources of bias, we assessed adequacy of the length of follow‐up, adequacy of information on nutritional status, whether there were major between‐group imbalances in key baseline characteristics, and whether there was drug company involvement.

Recovery from hip fracture in older people takes time, with long‐term implications for morbidity and functional status. Sixteen studies followed up participants for six months or over; with six of these extending follow‐up to one year (Bischoff‐Ferrari 2010; Flodin 2014; Miller 2006; Schürch 1998; Van Stijn 2015; Wyers 2013).

Details of the nutritional status of the groups were often missing. Related to this was the lack of information on anthropometric parameters. While it is difficult to measure height and weight in people with hip fracture, 11 trials (27%) failed to provide any information on baseline anthropometry (e.g. mid‐upper arm circumference or weight) or an anthropometry‐derived nutrition risk score.

An appraisal of the trials for baseline imbalances found important differences between the two groups for age in two trials (Papaioannou 2011; Sullivan 2004), for type of hip fracture in Tidermark 2004; and for body weight in Stableforth 1986.

Twenty trials reported receiving some drug company sponsorship or provision of supplements, and were judged to be at high risk of bias. One trial (Anbar 2014) was judged to be at high risk of bias as a result of stopping early due to poor recruitment, when the interim analysis showed a 'positive result'. Another trial (Van Stijn 2015) was judged to be at high risk of bias because the power calculation was based on a very unlikely 50% reduction in mortality.

Effects of interventions

See: Table 1; Table 2

Summary of findings for the main comparison. Multinutrient supplements (oral) versus control for hip fracture aftercare in older people.

Multinutrient supplements (oral) versus control for hip fracture aftercare in older people
Patient or population: Older people undergoing hip fracture aftercare  Settings: Acute hospital  Intervention: Multinutrient supplements (oral route) in addition to standard care. (Typically, supplements were started either pre‐operatively or within 2 days postoperatively and continued for at least a month) Comparison: Standard postoperative nutritional support and care in control groups
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect  (95% CI)	No of Participants  (studies)	Quality of the evidence  (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Multinutrient supplements (oral) versus control
Mortality by end of study Follow‐up: 1‐12 months	Study population		RR 0.81   (0.49 to 1.31)	968  (15 studies)	⊕⊕⊝⊝  low³	The statistical test for subgroup differences between the results for the 5 trials targeting malnourished participants and those 10 trials not targeting malnourished participants did not confirm a difference between the two subgroups for mortality
	72 per 1000¹	59 per 1000   (36 to 95)
	High risk²
	250 per 1000	203 per 1000   (123 to 328)
Participants with complications (e.g. pressure sore, chest infection) at end of study   Follow‐up: 1‐12 months	Study population		RR 0.71   (0.59 to 0.86)	727  (11 studies)	⊕⊕⊝⊝  low⁶	Only 2 trials targeting malnourished people reported these data
	443 per 1000⁴	315 per 1000   (262 to 381)
	Moderate risk⁵
	290 per 1000	206 per 1000   (171 to 250)
Unfavourable outcome ⁷ by end of study Follow‐up: 1‐12 months	Study population		RR 0.67 (0.51 to 0.89)	334 (6 studies)	⊕⊝⊝⊝  very low⁸	Only 1 trial targeting malnourished people reported these data
	500 per 1000⁴	335 per 1000 (255 to 445)
Putative side effects of treatment (e.g. vomiting and diarrhoea) Follow‐up: during supplementation period	Study population		RR 0.99 (0.47 to 2.05)	442 (6 studies)	⊕⊝⊝⊝  very low⁹	Three of the 6 trials reported no adverse effects
	50 per 1000⁴	50 per 1000 (24 to 103)
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).  CI: Confidence interval; RR: Risk ratio;
GRADE Working Group grades of evidence  High quality: Further research is very unlikely to change our confidence in the estimate of effect.  Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.  Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.  Very low quality: We are very uncertain about the estimate.

Multinutrient supplements (nasogastric) versus control for hip fracture aftercare in older people⁷
Patient or population: Older people undergoing hip fracture aftercare  Settings: Acute hospitals  Intervention: Multinutrient supplements (nasogastric). (Started within 5 days of surgery and continued usually until oral intake was sufficient or hospital discharge.)¹ Comparison: Standard postoperative nutritional support and care in control groups
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect  (95% CI)	No of Participants  (studies)	Quality of the evidence  (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Multinutrient supplements (nasogastric) versus control
Mortality by end of study   Follow‐up: 1‐12 months	Study Population		RR: 0.99 (0.50 to 1.97)	280 (3 studies)	⊕⊝⊝⊝  very low³	Only 1 trial targeting malnourished participants reported these data
	156 per 1000²	155 per 1000 (78 to 308)
Participants with complications (e.g. pressure sore, aspiration pneumonia) at end of study   Follow‐up: 6 months	Study Population		RR: 1.09 (0.73 to 1.64)	18 (1 study)	⊕⊝⊝⊝  very low⁵	For consistency we have presented 95% CI here but have used 99% CI for single trial data in the main text: 99% CI 0.64 to 1.86.⁶
	800 per 1000⁴	872 per 1000 (584 to 1000)
Unfavourable outcome Follow‐up: 1‐12 months	See comment		See comment			Outcome not reported
Putative side effects of treatment (e.g. aspiration pneumonia) Follow‐up: during supplementation period	See comment		See comment			Insufficient data to draw any conclusions. However, poor toleration of tube feeding was noted.¹ There was no report of aspiration pneumonia (1 study; 140 participants). One study reported 18 (28% of 64) participants in the intervention group developed diarrhoea ‐ this was ascribed to antibiotics in 16 ‐ but did not report on the control group. One study (18 participants) reported 3 cases of "bloating" in the intervention group; it found no feed‐induced diarrhoea
The basis for the assumed risk* (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).  CI: Confidence interval; RR: Risk Ratio.
GRADE Working Group grades of evidence  High quality: Further research is very unlikely to change our confidence in the estimate of effect.  Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.  Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.  Very low quality: We are very uncertain about the estimate.

Study ID	Intervention  (n, mean days, SD)			Control  (n, mean days, SD)			Mean difference (99% confidence intervaI)
Multinutritional oral supplements
Anbar 2014	22	10.1	3.2	28	12.5	5.5	‐2.40 days (‐5.60 to 0.80)
Botella‐Carretero 2010	30	13.3	4.3	30	12.8	4.0	0.50 days (‐2.26 to 3.26)
Brown 1992b	5	27.00	10.00	5	48.00	37.00	‐21.00 days (‐65.15 to 23.15)
Bruce 2003	50	17.70	9.40	58	16.60	9.20	1.10 days (‐3.53 to 5.73)
Madigan 1994	18	16.00	8.00	12	15.00	11.00	1.00 day (‐8.51 to 10.51)
Myint 2013	61	26.2	8.2	60	29.9	11.2	‐3.70 days (‐8.30 to 0.90)
Nasogastric tube feeding
Sullivan 1998	8	38.20	36.90	7	23.70	20.00	14.50 days (‐24.34 to 53.34)
High protein supplements
Espaulella 2000	85	16.40	6.60	86	17.20	7.70	‐0.80 days (‐3.62 to 2.02)
Neumann 2004	18	23.20	5.52	20	28.00	11.63	‐4.80 days (‐12.29 to 2.69)
Iron supplementation versus control
Parker 2010	150	18.8	17.4	150	21.3	20.6	‐2.50 days (‐8.17 to 3.17)
Serrano‐Trenas 2011	99	13.5	7.1	97	13.1	6.9	0.40 days (‐2.18 to 2.98)
Vitamin B1
Day 1988	28	35.00	34.00	30	29.00	30.00	6.00 days (‐15.75 to 27.75)
Vitamin, mineral and amino acid supplementation versus control
Scivoletto 2010	49	15.4	6.8	47	17.9	7.3	‐2.50 days (‐6.21 to 1.21)
Semi‐essential amino acid
Van Stijn 2015	111	13	10	123	13	11	0.00 days (‐3.54 to 3.54)

Date	Event	Description
30 October 2016	New search has been performed	In this seventh update of the review, we updated our trial search to November 2015. Of the newly identified studies for this update, 17 trials were selected for inclusion (Anbar 2014; Bischoff‐Ferrari 2010; Botella‐Carretero 2010; Chevalley 2010; Fabian 2011; Flodin 2014; Glendenning 2009; Kang 2012; Luo 2015; Myint 2013; Papaioannou 2011; Parker 2010; Prasad 2009; Scivoletto 2010; Serrano‐Trenas 2011; Van Stijn 2015; Wyers 2013), one of which (Wyers 2013) was previously an ongoing study. Six new studies were excluded (Bell 2014; Gunnarsson 2009; Hitz 2007; Hoekstra 2011; Holst 2012; Li 2012) and one previously ongoing study was excluded (Cameron 2011). Five were placed in ongoing trials (ACTRN12610000392066; ACTRN12612000448842; NCT01404195; NCT01505985; Rowlands) and five await classification (Benati 2011; Bernabeu‐Wittel 2016; Ekinci 2015; Ish‐Shalom 2008; Stratton 2006). New interventions examined were: high dose bolus vitamin D; different oral doses or sources of vitamin D; intravenous or oral iron; types of protein supplement; a supplement with vitamins, minerals and amino acids; and taurine (an amino acid). We have assessed the risk of bias for all new trials and all previously included trials with the Cochrane 'Risk of bias' tool. This replaces our former assessment of methodological quality. We have assessed the quality of the evidence using GRADE. We have constructed and presented 'Summary of findings' tables.
30 October 2016	New citation required and conclusions have changed	Conclusions were changed for oral multinutrient supplements, which now have low‐quality evidence for prevention of complications. There have been changes to the byline.

Date	Event	Description
12 November 2009	New citation required and conclusions have changed	In this sixth update, published in Issue 1, 2010 of The Cochrane Library, we updated our trial search to September 2008. Of the 10 newly identified studies for this update, one trial is included (Botella‐Carretero 2008), five trials are excluded (Boudville 2002; Hommel 2007; Kacmaz 2007; Olofsson 2007; Thomas 2008) and one trial awaits classification (Gerstorfer 2008a). Three new trials are ongoing (Dagneliea; NCT00497978; ACTRN12609000241235). Of previously identified trials: one former ongoing trial is now included (Eneroth 2006), and one trial formerly awaiting classification (Miller 2006) is now included. A new category (intravenous feeding and oral supplements) was set up for one new trial. There was slight modification to the conclusions that reflected reappraisal of the available evidence.
15 August 2008	Amended	Converted to new review format.
4 August 2006	New search has been performed	In the fifth update, published in The Cochrane Library Issue 4, 2006, we updated our trial search to January 2006. Of the six newly identified studies for this update, one trial is included (Sullivan 2004), three trials are excluded (Ashworth 2006; Carlsson 2005; Wong 2004) and two trials await assessment (Eneroth 2005; Stratton 2005). Of two former ongoing trials, one is now included (Duncan 2006, formerly Johansen 2002) and the other awaits assessment (Miller 2006, formerly Crotty 2003). One trial formerly awaiting assessment is now included (Neumann 2004). Two existing categories were modified to accommodate two newly included trials. A new category (dietetic assistants versus usual care) was set up for the third new trial.
3 November 2003	New search has been performed	In the fourth update, published in The Cochrane Library Issue 1, 2004, we updated our trial search to August 2003. Two new trials were included (Bruce 2003; Houwing 2003). Two newly identified trials were excluded (Hedström 2002; Stumm 2001). One newly identified trial is awaiting assessment (Tidermark 2003). Updates to all three ongoing trials were provided (Cameron 2000; Crotty 2003; Johansen 2002). The review conclusions were unchanged.
1 May 2002	New search has been performed	In the third update, published in Issue 3, 2002 of The Cochrane Library, we updated our trial search to April 2002. No new trials were included. Two newly identified trials were excluded (Bachrach 2001; Lauque 2000). Four trials previously awaiting assessment were now excluded. Two newly identified trials (Crotty 2003; Johansen 2002) were included as ongoing trials. The review conclusions were unchanged.
1 May 2001	New search has been performed	In the second update, published in Issue 3, 2001 of The Cochrane Library, the trial search was updated to April 2001. No new trials were included. Two more trials were excluded: one previously awaiting assessment (Doshi 1998) on the basis of a full journal publication (Lawson 2000) and the other (Bachrach 2000) was newly identified. One newly identified trial, only available as a conference abstract, was placed in Studies awaiting assessment (Moller‐Madsen 1988) and further details sought. The review conclusions were unchanged.
1 August 2000	New search has been performed	In the first update, published in Issue 4, 2000 of The Cochrane Library, we extended our trial search to January 2000. We identified one new ongoing trial (Cameron 2000), and obtained new information on four included trials and two studies placed in the awaiting assessment category in the first version of this review. This extra information resulted in one included trial (Williams 1989) being excluded, and one of the two studies pending assessment being included (Espaulella 2000) and the other excluded (Pedersen 1999). The inclusion of the new trial, which evaluated the effect of protein in an oral feed, and the other new information did not substantially alter the conclusions of the original review. Relative risks instead of Peto odds ratios were presented for dichotomous outcomes. Again, this did not affect the conclusions of the review.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality by end of study	20	1385	Risk Ratio (M‐H, Fixed, 99% CI)	0.79 [0.55, 1.15]
1.1 Oral supplements	15	968	Risk Ratio (M‐H, Fixed, 99% CI)	0.81 [0.49, 1.32]
1.2 Nasogastric tube feeding	3	280	Risk Ratio (M‐H, Fixed, 99% CI)	0.99 [0.50, 1.97]
1.3 Nasogastric tube feeding and oral supplements	1	57	Risk Ratio (M‐H, Fixed, 99% CI)	0.74 [0.23, 2.35]
1.4 Intravenous feeding and oral supplements	1	80	Risk Ratio (M‐H, Fixed, 99% CI)	0.11 [0.01, 2.00]
2 Participants with complications at end of study	14	882	Risk Ratio (M‐H, Fixed, 99% CI)	0.69 [0.59, 0.81]
2.1 Oral supplements	11	727	Risk Ratio (M‐H, Fixed, 99% CI)	0.71 [0.59, 0.86]
2.2 Nasogastric tube feeding	1	18	Risk Ratio (M‐H, Fixed, 99% CI)	1.09 [0.73, 1.64]
2.3 Nasogastric tube feeding and oral supplements	1	57	Risk Ratio (M‐H, Fixed, 99% CI)	1.11 [0.75, 1.65]
2.4 Intravenous feeding and oral supplements	1	80	Risk Ratio (M‐H, Fixed, 99% CI)	0.21 [0.10, 0.46]
3 Participants with complications at end of study: random‐effects model	14	882	Risk Ratio (M‐H, Random, 99% CI)	0.70 [0.53, 0.91]
3.1 Oral supplements	11	727	Risk Ratio (M‐H, Random, 99% CI)	0.72 [0.58, 0.89]
3.2 Nasogastric tube feeding	1	18	Risk Ratio (M‐H, Random, 99% CI)	1.09 [0.73, 1.64]
3.3 Nasogastric tube feeding and oral supplements	1	57	Risk Ratio (M‐H, Random, 99% CI)	1.11 [0.75, 1.65]
3.4 Intravenous feeding and oral supplements	1	80	Risk Ratio (M‐H, Random, 99% CI)	0.21 [0.10, 0.46]
4 Unfavourable outcome (death or complications) at end of study	6	334	Risk Ratio (M‐H, Fixed, 99% CI)	0.67 [0.51, 0.89]
4.1 Oral supplements	6	334	Risk Ratio (M‐H, Fixed, 99% CI)	0.67 [0.51, 0.89]
4.2 Nasogastric tube feeding	0	0	Risk Ratio (M‐H, Fixed, 99% CI)	0.0 [0.0, 0.0]
4.3 Nasogastric tube feeding and oral supplements	0	0	Risk Ratio (M‐H, Fixed, 99% CI)	0.0 [0.0, 0.0]
4.4 Intravenous feeding and oral supplements	0	0	Risk Ratio (M‐H, Fixed, 99% CI)	0.0 [0.0, 0.0]
5 Unfavourable outcome (death or complications) ‐ oral supplements extra analyses	6		Risk Ratio (M‐H, Fixed, 99% CI)	Subtotals only
5.1 Oral supplements: worst case scenario	6	353	Risk Ratio (M‐H, Fixed, 99% CI)	0.81 [0.62, 1.04]
5.2 Oral supplements: Hankins 1996 acute hospital data	1	31	Risk Ratio (M‐H, Fixed, 99% CI)	0.96 [0.71, 1.31]
5.3 Oral supplements: Hankins 1996 post discharge	1	31	Risk Ratio (M‐H, Fixed, 99% CI)	1.10 [0.50, 2.41]
6 Adverse effects (putatively related to treatment)	8		Risk Ratio (M‐H, Fixed, 99% CI)	Subtotals only
6.1 Oral supplements (mainly diarrhoea or/and vomiting)	6	442	Risk Ratio (M‐H, Fixed, 99% CI)	0.99 [0.47, 2.05]
6.2 Nasogatric tube feeding	1	18	Risk Ratio (M‐H, Fixed, 99% CI)	8.56 [0.51, 144.86]
6.3 Intravenous feeding and oral supplements	1	57	Risk Ratio (M‐H, Fixed, 99% CI)	1.85 [0.49, 7.03]
6.4 Nasogastric tube feeding and oral supplements	0	0	Risk Ratio (M‐H, Fixed, 99% CI)	0.0 [0.0, 0.0]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality by end of study by risk of bias for allocation concealment	20	1385	Risk Ratio (M‐H, Fixed, 99% CI)	0.79 [0.55, 1.15]
1.1 Low risk of bias	10	682	Risk Ratio (M‐H, Fixed, 99% CI)	0.57 [0.32, 1.01]
1.2 Unclear risk of bias	7	462	Risk Ratio (M‐H, Fixed, 99% CI)	1.22 [0.65, 2.28]
1.3 High risk of bias	3	241	Risk Ratio (M‐H, Fixed, 99% CI)	0.78 [0.34, 1.79]
2 Participants with complications at end of study by risk of bias for allocation concealment	14	882	Risk Ratio (M‐H, Fixed, 99% CI)	0.69 [0.59, 0.81]
2.1 Low risk of bias	9	622	Risk Ratio (M‐H, Fixed, 99% CI)	0.78 [0.66, 0.92]
2.2 Unclear risk of bias	5	260	Risk Ratio (M‐H, Fixed, 99% CI)	0.38 [0.24, 0.61]
2.3 High risk of bias	0	0	Risk Ratio (M‐H, Fixed, 99% CI)	0.0 [0.0, 0.0]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality by end of study	4	361	Risk Ratio (M‐H, Fixed, 99% CI)	1.42 [0.85, 2.37]
1.1 Protein‐containing supplement v non‐protein‐containing supplement	3	315	Risk Ratio (M‐H, Fixed, 99% CI)	1.38 [0.82, 2.34]
1.2 High protein‐containing supplement v low protein‐containing supplement	1	46	Risk Ratio (M‐H, Fixed, 99% CI)	2.18 [0.21, 22.42]
2 Unfavourable outcome (death or complications) at end of study	2	223	Risk Ratio (M‐H, Fixed, 99% CI)	0.78 [0.65, 0.95]
2.1 Protein‐containing supplement v non‐protein‐containing supplement	2	223	Risk Ratio (M‐H, Fixed, 99% CI)	0.78 [0.65, 0.95]

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality by end of study	1		Risk Ratio (M‐H, Fixed, 99% CI)	Totals not selected
2 Participants with complications at end of study	1		Risk Ratio (M‐H, Fixed, 99% CI)	Totals not selected

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Participants with complications at end of study	1		Risk Ratio (M‐H, Fixed, 99% CI)	Totals not selected
2 Mortality by end of study	1		Risk Ratio (M‐H, Fixed, 95% CI)	Totals not selected

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality by end of study	3	566	Risk Ratio (M‐H, Fixed, 99% CI)	0.98 [0.65, 1.46]
2 Participants with complications at end of study	2	266	Risk Ratio (M‐H, Fixed, 99% CI)	1.23 [0.63, 2.42]

Methods	Method of randomisation: use of a computer programme, balanced in blocks of four, by independent person  Intention‐to‐treat analysis: probably ‐ appears so  Lost to follow‐up: probably none
Participants	Location: three centres, Arnhem, Deventer and Nieuwegein, in The Netherlands  Period of study: April 1998 to December 1999  103 participants  Inclusion criteria: hip fracture, pressure ulcer score >8 (Dutch Consensus Meeting scoring system), consent from patient or legal representative  Exclusion criteria: terminal care, metastatic hip fracture, insulin‐dependent diabetes, renal disease, hepatic disease, morbid obesity (BMI > 40), therapeutic diet incompatible with supplementation, pregnancy, lactation  Sex: 84 female, 19 male  Age: mean age 81 years  Fracture type: not given (48 internal fixation presumed extracapsular fractures, 44 hemi‐arthroplasty presumed intracapsular fractures)
Interventions	Timing of intervention: supplemented from immediately postoperative period for four weeks or until discharge, given between regular meals  (a) 400 ml/day oral supplement (600 kcal or 2.51 MJ, 40 g protein, 6 mg arginine, 20 mg zinc, 500 mg vitamin C, 200 mg vitamin E as alpha‐tocopherol, 4 mg carotenoids (Cubitan, NV Nutricia, The Netherlands)); and regular diet  (b) Placebo supplement was a non‐caloric, water‐based drink with sweeteners, colourants and flavourings in similar packaging, look and taste not identical to active supplement; and regular diet  Allocated: 51/52  Assessed: 51/52
Outcomes	Length of follow‐up: 28 d or earlier if discharged  Main outcomes:  Morbidity and complications: pressure sores  Other outcomes: Patient compliance: mean percentage intake/day, days supplemented
Notes	Request for further details (method of randomisation, other complications, adverse events, length of stay, further details of supplement) sent 13/10/03.  Further details of randomisation method received 29/10/03.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Use of a computer programme, balanced in blocks of four, by an independent person. Information from trialists. Comment: probably low risk.
Allocation concealment (selection bias)	Low risk	Use of a computer programme, balanced in blocks of four, by an independent person. Information from trialists.
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	States "double‐blind" but also states " look and taste of both supplements were not exactly identical, but supplements were given in similar, blinded packages to mask the differences". Comment: probably done.
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	States "double‐blind" but also states " look and taste of both supplements were not exactly identical, but supplements were given in similar, blinded packages to mask the differences". Assessed by nurses and unlikely to have been influenced by unblinding. Comment: probably done.
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Low risk	States "double‐blind" but also states " look and taste of both supplements were not exactly identical, but supplements were given in similar, blinded packages to mask the differences". Assessed by nurses. Comment: probably done.
Incomplete outcome data (attrition bias)   Primary outcomes	Low risk	All participants accounted for in data.
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Low risk	All participants accounted for in data.
Selective reporting (reporting bias)	Unclear risk	Pressure ulcer reporting agrees with methods, but would expect reporting of other complications
Other bias	High risk	Funded by Numico Research BV, nutrition company.

Methods	Method of randomisation: concealed, computer‐generated programme Intention‐to‐treat analysis: carried out Lost to follow‐up: all participants followed‐up
Participants	Location: ortho‐geriatric unit, Department of Geriatrics, Rabin Medical Center, Petah Tikva, Israel Period of study: May 2010–December 2011 50 participants Inclusion criteria: > 65 years, admitted following hip fracture within 48 h of the injury and orthopaedic surgery was the treatment of choice Exclusion criteria: presented to hospital > 48 h after the injury, receiving steroids and/or immunosuppression therapy; active oncologic disease, multiple fractures, diagnosed dementia, required supplemental nasal oxygen which precluded the measurement of resting energy expenditure (REE) Sex: 33 female, 17 male Age: mean 83 years Fracture type: 40% pertrochanteric, 20% subcapital, 6% subtrochanteric, 6% base of femoral neck, 28% other
Interventions	Timing of intervention: 24 h after surgery for 14 d (a) Calories with an energy goal determined by three REE measurements in first 7 d using indirect calorimetry (IC) (Fitmate, Cosmed, Italy) which was based on hospital‐prepared diets (standard or texture‐adapted). Oral nutritional supplements (ONS) amount adjusted to make up the difference between energy received from hospital food and measured energy expenditure. These ONS were provided in the form of Ensure plus (Abbott Laboratories) containing 355 kcal/237 ml and 13.5 g protein or Glucerna (Abbott Laboratories) containing 237 kcal/237 ml and 9.9 g protein/237 ml. The participant, family and caregivers educated regarding importance of nutritional support and more attention was given to personal food preferences. 24‐h food diaries were filled in by the medical staff, family and caregivers. (b) Usual hospital food (standard or texture‐adapted) and a fixed dose of ONS if already prescribed prior to hospitalisation. Hospital‐prepared diets provided a mean of 1800 kcal and 80 g of protein if meals completely eaten by the participants Allocated: 22/28 Assessed: 22/28
Outcomes	Length of follow‐up: length of hospital stay Main outcomes: Mortality Length of hospital stay Total complications Infectious complications Pressure ulcers Other outcomes: Protein and energy intakes
Notes	Power calculation indicated needed 66 participants. In view of the slow rate of expected recruitment an interim analysis was planned after 50 participants. In the presence of a positive result, the study was discontinued. No funder reported
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	States "Randomization was performed using a concealed, computer generated program."
Allocation concealment (selection bias)	Low risk	States "Randomization was performed using a concealed, computer generated program. RA enrolled participants and assigned them to interventions while YB enrolled patients but was blinded to the intervention." Comment: probably done
Blinding of participants and personnel (performance bias)   All outcomes	High risk	No placebo group
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	No placebo group. Comment: unlikely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No placebo group. Comment: may have been influenced by lack of blinding
Incomplete outcome data (attrition bias)   Primary outcomes	Low risk	All participants accounted for, with no drop‐outs.
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Low risk	All participants accounted for, with no drop‐outs.
Selective reporting (reporting bias)	Low risk	No protocol available, but expected outcomes reported
Other bias	High risk	Power calculation indicated needed 66 participants. In view of the slow rate of expected recruitment an interim analysis was planned after 50 participants. In the presence of a positive result, the study was discontinued. No funder reported

Methods	Method of randomisation: quasi‐randomised  Intention‐to‐treat analysis: appears so  Lost to follow‐up: appears none
Participants	Location: hospital, Nottingham, UK  Period of study: over 18 months, probably prior to 1983  122 participants  Inclusion criteria: hip fracture, mid‐arm circumference or triceps skinfold, or both, 1 to 2 SD below the mean (thin group) or over 2 SD below the mean (very thin group)  Exclusion criteria: incapable of understanding study, severe dementia, serious concomitant physical disorder e.g. stroke  Sex: all female  Age: range 68‐92 years  Fracture type: further details not given
Interventions	Timing of intervention: nasogastric feeding started within 5 d of surgery, 8 h overnight with tube disconnected during the day, until discharge or death. Feeding stopped if participant did not tolerate tube or removed tube on 3 occasions  (a) 1 L Clinifeed Iso (4.2 MJ or 1000 kcal, 28 g protein, 270 mosmol/L) via fine bore nasogastric tube using peristaltic pump, and normal ward diet, with free access to snacks and drinks  (b) Normal ward diet, with free access to snacks and drinks  Allocated: 64/58  Assessed: 60/49 for independent mobility
Outcomes	Length of follow‐up: until discharge or death  Main outcomes:  Mortality  Morbidity and complications: infection  Length of stay: hospital stay  Postoperative functional status: days to weight bearing with support, days to independent mobility  Putative side effects of treatment: aspiration, diarrhoea  Other outcomes:  Voluntary food intake  Patient compliance: tolerance of tube, duration of feeding
Notes	There was an administrative limit imposed of a maximum of 6 participants being fed at one time. Data presented from 1983 paper for numbers of participants are correct, error in number of participants in 1985 paper. Slight discrepancy with days to reach independent mobility presented in 1984 abstract. Reply from trialists (15 February 2000) gave details of randomisation (on recall: either by date of admission or birth), outcome assessment, inclusion criteria, denominators and baseline comparability
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Quasi‐randomised. On recall by trialists: "either on the basis of odd and even dates of birth or of admission".
Allocation concealment (selection bias)	High risk	Quasi‐randomised. On recall by trialists: "either on the basis of odd and even dates of birth or of admission".
Blinding of participants and personnel (performance bias)   All outcomes	High risk	No placebo group. Comment: likely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	No placebo group. Comment: unlikely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No placebo group. Comment: may have been influenced by lack of blinding
Incomplete outcome data (attrition bias)   Primary outcomes	Low risk	All participants accounted for, with no drop‐outs.
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Low risk	All participants accounted for, with no drop‐outs.
Selective reporting (reporting bias)	Low risk	Protocol not available, but study report includes all outcomes reported in methods and those that would be expected. Comment: probably done
Other bias	High risk	States that Bastow was "supported by a grant from Rousell Laboratories Ltd", manufacturers of Clinifeed nasogastric feed used in trial

Methods	Method of randomisation: states double‐blind, but no other details  Intention‐to‐treat analysis: claimed by authors, but no details to support  Lost to follow‐up: details not given
Participants	Location: hospitals; Nottingham, Leeds and Doncaster, UK  Period of study: recruitment over 2.5 years  59 participants  Inclusion criteria: fractured femur, 70‐85 years, mean arm circumference < 25 cm, triceps skinfold < 18 mm  Exclusion criteria: other major medical disorder, failure to gain consent, demented (Cape score less than 9/12)  Sex: all female  Age: not given  Fracture type: further details not given
Interventions	Timing of interventions: start time unclear, twice daily for 2 months,  (a) Cetornan (ornithine alpha‐ketoglutarate) 20 g/d (0.293 MJ or 70 kcal, 2.73 g N), presumed orally  (b) Pro‐up (defined formula peptide supplement, 0.293 MJ or 70 kcal, 2.73 g N), presumed orally  Allocated: ?/?  Assessed: ?/?
Outcomes	Length of follow‐up: 6 months  Main outcomes:  Mortality  Morbidity and complications: all complications and delay in major complications (nr)  Length of stay: duration of treatment or hospitalisation (nr)  Postoperative functional status: fatigue score (nr)  Other outcomes: Food intake (nr)  Patient compliance: proportion completing 2 months' treatment (nr)
Notes	Conference abstract only. No denominators for intention‐to‐treat analysis, so cannot use data in analysis. Data on arm muscle circumference, fatigue score and food intake presented for 35 participants completing 2 months of treatment. Request for further details (including denominators) sent 19 May 1999, re‐sent 4 February 2000
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Abstract only. No details provided
Allocation concealment (selection bias)	Unclear risk	Abstract only. States "randomized in a double‐blind fashion", no other details provided
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	Abstract only. States "double‐blind" and "unlabelled identical sachets"
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	Abstract only. Comment: unlikely to have been influenced by unblinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	Abstract only. States "double‐blind" and "unlabelled identical sachets". Comment: unclear if done
Incomplete outcome data (attrition bias)   Primary outcomes	Unclear risk	Abstract only. Insufficient details on attrition and exclusions provided
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Unclear risk	Abstract only. Insufficient details on attrition and exclusions provided
Selective reporting (reporting bias)	Unclear risk	Abstract only. Insufficient details provided
Other bias	Unclear risk	Abstract only. Insufficient details provided. No details on sponsor

Methods	Method of randomisation: Factorial design computer‐based randomisation performed by study statistician. Randomisation for the dosage of cholecalciferol was double‐blinded Intention‐to‐treat analysis: carried out Lost to follow‐up: 14% lost to follow‐up
Participants	Location: Triemli City Hospital, Zurich, Switzerland Period of study: screening for recruitment 2005‐2007 173 participants Inclusion criteria: age 65 years or older, surgical repair of acute hip fracture, Folstein Mini‐Mental State Examination score of 15 or more, understand German, able to walk at least 3 m before fracture Exclusion criteria: prior hip fracture at the newly fractured hip, metastatic cancer or chemotherapy in last year, severe visual or hearing impairment, creatinine clearance of 15 mL/min or less, kidney stone in the past 5 years, hypercalcaemia, primary hyperparathyroidism or sarcoidosis Sex: 137 female, 36 male Age: mean 84 years Fracture type: further details not given
Interventions	Timing of intervention: from mean of 4.2 d after hip fracture surgery for 12 months (a) With breakfast, participants took a study capsule containing 1200 IU of cholecalciferol. For breakfast and at bedtime, participants took a tablet containing 400 IU of cholecalciferol and 500 mg of elemental calcium as calcium carbonate (Nycomed, Wädenswil, Switzerland). (b) With breakfast, participants took a placebo capsule (identical in appearance and taste to active tablet). For breakfast and at bedtime, participants took a tablet containing 400 IU of cholecalciferol and 500 mg of elemental calcium as calcium carbonate (Nycomed, Wädenswil, Switzerland). Groups a and b were also randomised to standard or extended physiotherapy Allocated: 86/87 Assessed: 73/75
Outcomes	Length of follow‐up: 12 months Main outcomes: Mortality Complications Functional status Level of care Putative side effects Other outcomes: Compliance
Notes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	States "computer‐based randomization"
Allocation concealment (selection bias)	Unclear risk	States "Randomization for the dosage of cholecalciferol was double‐blinded, whereas randomization for PT (physiotherapy) was single‐blinded (all study staff except the treating physiotherapist who instructed the home program were blinded to the PT treatment allocation). Comment: allocation concealment unclear
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	States double‐blind and vitamin D placebo identical in appearance and taste
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	States double‐blind and vitamin D placebo identical in appearance and taste
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Low risk	States double‐blind and vitamin D placebo identical in appearance and taste
Incomplete outcome data (attrition bias)   Primary outcomes	Low risk	Reasons for missing data provided and missing data balanced across groups
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Low risk	Reasons for missing data provided and missing data balanced across groups
Selective reporting (reporting bias)	High risk	Trial registration on clinicaltrials.gov gives outcomes of numbers of people who fell, disability, health care utilisation and quality of life (EuroQol); not provided in published paper
Other bias	Low risk	Funded by Swiss National Foundations, Vontobel Foundation (charitable foundation),  Baugarten Foundation

Methods	Method of randomisation: sealed opaque envelopes, prepared independently from recruitment  Intention‐to‐treat analysis: unclear  Lost to follow‐up: details given
Participants	Location: Hospital Ramon y Cajal, Madrid, Spain  Period of study: February 2006‐February 2007  90 participants  Inclusion criteria: > 65 years, surgery for hip fracture, written informed consent  Exclusion criteria: weight loss > 5% in previous month or > 10% in previous 6 months, and/or albumin < 27 g/dL. Acute or chronic renal failure, hepatic insufficiency or cirrhosis (Child B or C), severe heart failure (New York heart classification III or IV), respiratory failure, gastrointestinal condition precluding adequate oral intake. Also: previous oral nutrition supplements or nutrition support in previous 6 months.  Sex: 71 female, 19 male  Age: mean age 84 years  Fracture type: 58% gamma nail surgery (presumed extracapsular fractures), 42% total hip replacement (presumed intracapsular fractures)
Interventions	Timing of intervention: started 48 h after operation, until hospital discharge  (a) Four 10 g packets a day of Vegenat‐med Proteina (Vegenat SA, Badajoz, Spain) each providing 9 g protein and 38 kcal, dissolved in water, milk or soup from diet  (b) Two 200 ml bricks a day (Resource Hiperproteico, Novartis Medical Nutrition, Barcelona) providing total of 37.6 g protein and 500 kcal  (c) no oral nutrition supplements  Allocated: 30/30/30  Assessed: 28/30/27
Outcomes	Length of follow‐up: up to hospital discharge  Main outcomes:  Mortality  Complications: urinary, respiratory, wound infection; pressure ulcer, dysphagia, ischaemic heart disease; severe hyponatraemia; anaphylaxis; vomiting and/or diarrhoea  Length of acute hospital stay  Level of care: time to mobilisation  Other outcomes:  Energy and protein intake
Notes	Emailed 22 January 2009 requesting mortality information. Author replied 23 January confirming no participants had died during the trial.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	States "randomized" only. No further details provided
Allocation concealment (selection bias)	Low risk	States used of "sealed opaque envelopes". Independent preparation of envelopes: "The investigator recruiting the patients ....had no role in the randomisation process"
Blinding of participants and personnel (performance bias)   All outcomes	High risk	No placebo provided
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	No details provided on blinding of outcome assessment, but outcome assessment unlikely to have been influenced by unblinding.
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No details provided on blinding of outcome assessment, and outcome assessment may have been influenced by unblinding
Incomplete outcome data (attrition bias)   Primary outcomes	Low risk	All participants accounted for in analysis
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Unclear risk	Denominators unclear for length of hospital stay, length of immobilisation and supplement intake
Selective reporting (reporting bias)	Unclear risk	Insufficient details provided
Other bias	Unclear risk	Funding source (Fundacion para la Investigacion Biomedica, Hospital Ramon y Cajal, Madrid, Spain) and source of supplemental nutrition (Hospital Ramon y Cajal) do not appear related to manufacturer of the supplements.

Methods	Method of randomisation: randomised, open two‐arm trial, using sealed opaque envelopes Intention‐to‐treat analysis: in acute hospital; complications, length of stay, mobilisation not collected after moved to another centre for rehabilitation Lost to follow‐up: 53% lost to complete follow‐up (moved to another centre for rehabilitation)
Participants	Location: Hospital Universitario Ramon y Cajal, Madrid, Spain Period of study: recruitment May 2007–September 2008 60 participants Inclusion criteria: age > 65 years, hip fracture where orthopaedic surgery considered treatment of choice Exclusion criteria: moderate–severe malnutrition (weight loss of > 5% in the previous month or > 10% in the previous 6 months, and/or serum albumin concentrations < 2.7 g/dL), acute and/or chronic renal failure, hepatic insufficiency or cirrhosis (Child B or C), severe heart failure with class III or IV of the New York Heart Association, respiratory failure, gastrointestinal condition precluding adequate oral nutritional intake. Sex: 44 female, 16 male Age: mean 84 years Fracture type: fracture type not given
Interventions	Timing of intervention: from admission (including pre‐operative) until discharge (a) Energy and protein supplements by means of commercial enteral nutrition for oral intake (Fortimel, 200 mL bricks, each provides 20 g protein and 200 kcal, Nutricia Advanced Medical Nutrition – Danone Group) to aim at 40 g of protein and 400 kcal per day (2 bricks a day) and every participant was prescribed a standard or texture‐adapted diet to meet their calculated metabolic rate. The Harris–Benedict equation was employed to calculate the basal metabolic rate and a coefficient of 1.3 was employed to estimate the total metabolic rate. In‐hospital diets provided a mean of 100 g of protein per day (range 80–120 g). (b) Every participant was prescribed a standard or texture‐adapted diet to meet their calculated metabolic rate. The Harris–Benedict equation was employed to calculate the basal metabolic rate and a coefficient of 1.3 was employed to estimate the total metabolic rate. In‐hospital diets provide a mean of 100 g of protein per day (range 80–120 g). Allocated: 30/30 Assessed: 18/14
Outcomes	Length of follow‐up: until discharge from hospital Main outcomes: Mortality Postoperative hospital stay, Postoperative hospital complications Requiring rehabilitation Other outcomes: Compliance
Notes	Emailed jbotella.hrc@salud.madrid.org 25 November 2014 to enquire about numbers in intervention and control groups going to rehabilitation hospital (text differs from flow chart) and whether data were collected in rehabilitation hospital for complications, mobilisation and length of stay. Replied with further information 26 November 2014 for all these queries
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	States "patients were randomized using sealed opaque envelopes to yield two groups with 30 patients each."
Allocation concealment (selection bias)	Low risk	States "patients were randomized using sealed opaque envelopes to yield two groups with 30 patients each... The investigators who designed the study prepared the envelopes and assigned participants to their groups, but had no contact with the patients throughout the study. The investigator recruiting the patients, administering the interventions and evaluating the outcomes had no role on the randomization process."
Blinding of participants and personnel (performance bias)   All outcomes	High risk	No placebo group. Comment: likely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	No placebo group. States also "The investigator recruiting the patients, administering the interventions and evaluating the outcomes had no role on the randomization process." Comment: unlikely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No placebo group. States "The investigator recruiting the patients, administering the interventions and evaluating the outcomes had no role on the randomization process." Comment: may have been influenced by lack of blinding
Incomplete outcome data (attrition bias)   Primary outcomes	High risk	Missing outcome data balanced in numbers across intervention (18) and control groups (14), but proportion high enough to likely induce a clinically relevant bias in observed effect size
Incomplete outcome data (attrition bias)   Secondary and other outcomes	High risk	Missing outcome data balanced in numbers across intervention (18) and control groups (14), but proportion high enough to likely induce a clinically relevant bias in observed effect size
Selective reporting (reporting bias)	Unclear risk	Insufficient details provided to judge
Other bias	Low risk	States "The funding source, Fundacion para la Investigacion Biomedica, Hospital Ramon y Cajal (FIBio‐RyC), Madrid, Spain, had no role in the study design, the collection, analysis, and interpretation of data, the writing of the report, or the decision to submit the paper for publication. The ONS employed in this study were provided by the Hospital Ramo´n y Cajal, Madrid, Spain."

Methods	Method of randomisation: alternating numbers  Intention‐to‐treat analysis: carried out  Lost to follow‐up: no losses to follow‐up
Participants	Location: hospital, Ipswich, UK  Period of study: 6 months, probably prior to 1992  10 participants  Inclusion criteria: thin (based on weight for height, triceps skinfold, mid‐arm circumference ‐ 2 out of 3 more than 1 SD below reference mean), elderly, women 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"  Fracture type: trochanteric or subcapital hip fracture
Interventions	Timing of intervention: from second day of admission until discharge (including rehabilitation hospital)  (a) Participant 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 1 kcal/ml, as 15% protein energy, 30% fat energy and 55% carbohydrate energy  (b) Normal hospital diet  Allocated: 5/5  Assessed: 5/5
Outcomes	Length of follow‐up: no details (21+ days)  Main outcomes:  Mortality  Morbidity and complications: pressure sore (nr)  Length of stay: days to discharge from orthopaedic surgeon  Postoperative functional status: 2‐stage walking goals  Other outcomes:  Dietary intake (nr)
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 May 1999, re‐sent 3 February 2000
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Alternating numbers (information from trial author)
Allocation concealment (selection bias)	High risk	Alternating numbers (information from trial author), states randomly assigned with no further details
Blinding of participants and personnel (performance bias)   All outcomes	High risk	No placebo group. Comment: likely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	No placebo group. Comment: unlikely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No placebo group. Comment: may have been influenced by lack of blinding
Incomplete outcome data (attrition bias)   Primary outcomes	Unclear risk	Insufficient details provided on pressure sores
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Unclear risk	Insufficient details provided on 2‐stage walking goals
Selective reporting (reporting bias)	Low risk	Protocol available and all outcomes provided
Other bias	Unclear risk	Source of funding for study unclear

Methods	Method of randomisation: block randomisation of 15. Table of randomisation by statistician not involved in study Intention‐to‐treat analysis: insufficient details provided Lost to follow‐up: insufficient details provided
Participants	Location: orthopaedic ward of Geneva University Hospital, Switzerland Period of study: recruited March 1999‐June 2000 45 participants Inclusion criteria: women older than 60 years with a recent hip fracture, i.e. within two weeks, that was attributable to osteoporosis such as occurring on a fall from standing height, and with the ability to give a written informed consent. Exclusion criteria: pathologic fracture; fracture caused by severe trauma; cardiac or pulmonary failure; advanced renal insufficiency with plasma creatinine concentration 200 mmol/L or more; hepatic failure; severe mental impairment; acute illness before the fracture that could interfere with the study protocol; active metabolic bone disease; consumption of protein supplement or of anti‐osteoporotic active drugs or medication known to alter bone metabolism, such as sex hormones or corticosteroids; severe malnutrition (serum albumin level < 15 g/L); life expectancy of less than one year Sex: all female Age: mean 81.3 (SD 7.4) years Fracture type: not given
Interventions	Timing of intervention: from a mean of 10 d post fracture for 28 d a) 20 g milk protein (casein) in 200 ml water, including 550 mg calcium and 500 IU vitamin D3, daily for 28 d b) 20 g whey protein in 200 ml water, including 550 mg calcium and 500 IU vitamin D3, daily for 28 d c) 15 g whey protein and 5 g of essential amino acids in ratio identical to casein in 200 ml water, including 550 mg calcium and 500 IU vitamin D3, daily for 28 d Allocated: 15/15/15 Assessed: unclear
Outcomes	Length of follow‐up: 28 d Main outcomes: Putative adverse events from supplements Other outcomes: Compliance
Notes	Emailed thierry.chevalley@hcuge.ch 9 October 2014 to ask for further information on outcomes, reply received 14 October 2014 with details of putative side effects and compliance
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	States " randomization was performed in blocks of 15 patients...table of randomization was established by a statistician who was not directly involved in the study"
Allocation concealment (selection bias)	Low risk	States " randomization was performed in blocks of 15 patients...table of randomization was established by a statistician who was not directly involved in the study"
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	States "dietician as well as both the medical staff and subjects involved in the study were blinded to the experimental groups" but no further details on how this was achieved
Blinding of outcome assessment (detection bias)   Primary outcomes	Unclear risk	No details provided and putative adverse events from supplements may have been influenced by unblinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No details provided and compliance may have been influenced by unblinding
Incomplete outcome data (attrition bias)   Primary outcomes	High risk	Numbers in email differ from publication: give 11 dropouts (5 casein, 4 whey, 2 whey and amino acids), with 12 mentioned in publication
Incomplete outcome data (attrition bias)   Secondary and other outcomes	High risk	Numbers in email differ from publication: give 11 dropouts (5 casein, 4 whey, 2 whey and amino acids), with 12 mentioned in publication
Selective reporting (reporting bias)	High risk	No details on outcome activities of daily living provided
Other bias	High risk	Supported by Novartis Cosumer Health (Berne, Switzerland)

Methods	Method of randomisation: computer‐generated random sequence, insufficient indication of adequate safeguards  Assessor blinding: blinded assessment of mental state, other outcomes not stated  Intention‐to‐treat analysis: analysis performed  Lost to follow‐up: details given
Participants	Location: hospital, Cardiff, UK  Period of study: recruitment over 6 months, probably prior to 1988  60 participants  Inclusion criteria: people with acute proximal femur fracture, age > 60 years  Exclusion criteria: unable to be assessed preoperatively, not seen within 24 h of admission, pathological fracture, difficulty obtaining consent from participant or relative  Sex: 44 female, 16 male  Age: 60 years and older (inclusion criterion)  Fracture type: 17 cervical, 9 trochanteric, 2 other/16 cervical, 14 trochanteric, 2 other
Interventions	Timing of intervention: 2 doses of vitamin preparation given preoperatively, and then 1 dose daily for 5 d postoperatively  (a) Intravenous Parentrovite IVHP (containing 250 mg thiamine hydrochloride, 4 mg riboflavine, 50 mg pyridoxine, 160 mg nicotinamide, 500 mg ascorbic acid, 1 g anhydrous dextrose)  (b) No supplement  Allocated: 28/32  Assessed: 28/32 for abbreviated mental test at day 2
Outcomes	Length of follow‐up: 3 months  Main outcomes:  Mortality  Morbidity and complications: total number of complications, numbers of participants with complications  Length of stay: hospital  Postoperative functional status: acute confusional state, acute on chronic confusional state, abbreviated mental test, objective learning test, Ishihara Colour Plates  Care required after discharge: final placement  Putative side effects of treatment: serious and other adverse events
Notes	Request for further details (method of randomisation, constituents of Parentrovite IVHP, other outcomes) sent. Reply from trialists (27 May 1999) gave details of the intervention, randomisation, and information on fracture type, baseline albumin levels, complications and hospital stay
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Allocation of patients was based on randomly generated numbers (0 or 1)"
Allocation concealment (selection bias)	Unclear risk	States "randomly allocated", no further details provided
Blinding of participants and personnel (performance bias)   All outcomes	High risk	No placebo
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	No placebo group. Comment: unlikely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No placebo group. Comment: may have been influenced by lack of blinding, apart from mental health status which was "assessed by a psychology technician who remained blind as to the treatment group of each patient"
Incomplete outcome data (attrition bias)   Primary outcomes	Unclear risk	Data provided for all participants, apart from putative adverse events (no data provided for control group)
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Unclear risk	Data provided for all participants, apart from putative adverse events (no data provided for control group)
Selective reporting (reporting bias)	High risk	Data on outcome final placement not available
Other bias	High risk	Bencard provided Parenterovite

Methods	Method of randomisation: not stated  Intention‐to‐treat analysis: appears intention‐to‐treat, but denominators unclear  Lost to follow‐up: mortality reported, but unclear if other losses to follow‐up
Participants	Location: orthopaedic unit in hospital and recovery hospital, Geneva, Switzerland  Period of study: 1 March‐15 May 1985  59 participants  Inclusion criteria: femoral neck fracture after an accidental fall, aged over 60 years  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 82 years  Fracture type: 26 femoral neck, 33 inter‐trochanteric
Interventions	Timing of intervention: from admission to orthopaedic unit to end of stay in second (recovery) hospital, supplement given once daily at 20:00 hours for a mean period of 32 d  (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: ?25/?27 at 6 months
Outcomes	Length of follow‐up: 6 months  Main outcomes:  Mortality  Morbidity and complications: complications (total, bedsore, severe anaemia, cardiac failure, infection, gastrointestinal ulcer, other), favourable clinical course (excludes death, major complication, or two or more minor complications)  Length of stay: orthopaedic unit and recovery hospital  Other outcomes:  Energy, protein and calcium intake
Notes	Numbers of complications unclear, request for further details sent 24 May 1999, re‐sent 7 February 2000
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	States "randomised", no other details provided
Allocation concealment (selection bias)	Unclear risk	States "randomised", no other details provided
Blinding of participants and personnel (performance bias)   All outcomes	High risk	No placebo group. Comment: likely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	No placebo group. Comment: unlikely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No placebo group. Comment: may have been influenced by lack of blinding
Incomplete outcome data (attrition bias)   Primary outcomes	High risk	Data provided for only 25/27 intervention group and 27/32 control group
Incomplete outcome data (attrition bias)   Secondary and other outcomes	High risk	Length of stay data not provided for 6/27 intervention group and 4/32 control group, i.e. length of stay for survivors presented
Selective reporting (reporting bias)	Unclear risk	Insufficient details provided
Other bias	High risk	Sandoz‐Wander supplied the supplement, but do not appear to have funded the study

PERMALINK

Nutritional supplementation for hip fracture aftercare in older people

Alison Avenell

Toby O Smith

James P Curtain

Jenson CS Mak

Phyo K Myint

Abstract

Background

Objectives

Search methods

Selection criteria

Data collection and analysis

Main results

Authors' conclusions

Plain language summary

Summary of findings

Background

Description of the condition

Description of the intervention

How the intervention might work

Why it is important to do this review

Objectives

Methods

Criteria for considering studies for this review

Types of studies

Types of participants

Types of interventions

Types of outcome measures

Main outcomes

Primary outcomes

Secondary outcomes

Other outcomes

Search methods for identification of studies

Electronic searches

Searching other resources

Data collection and analysis

Selection of studies

Data extraction and management

Assessment of risk of bias in included studies

Measures of treatment effect

Unit of analysis issues

Dealing with missing data

Assessment of heterogeneity

Assessment of reporting biases

Data synthesis

Subgroup analysis and investigation of heterogeneity

Sensitivity analysis

'Summary of findings' tables and assessment of the quality of the evidence

Results

Description of studies

Results of the search

1.

Included studies

Design

Sample sizes

Setting

Participants

Interventions

Multinutrient supplements (oral, nasogastric, intravenous) versus control

Oral supplements

Nasogastric tube feeding

Nasogastric tube feeding and oral supplements

Intravenous feeding and oral supplements

High protein‐containing supplements versus low‐protein or non‐protein‐containing supplements

Comparison of different protein sources

Vitamin supplementation versus control or lower dose supplementation

Comparison of different vitamin D sources

Iron supplementation versus control

Vitamin, mineral and amino acid supplementation versus control

Isonitrogenous ornithine alpha‐ketoglutarate versus peptide supplement

Dietetic assistants versus usual care

Excluded studies

Ongoing studies

Studies awaiting classification

Risk of bias in included studies

2.

3.

Allocation

Sequence generation

Methods	Method of randomisation: sequentially numbered opaque sealed envelopes, initially in blocks of 20, later reduced to blocks of 10, prepared by member of staff outside trial, opened sequentially  Intention‐to‐treat analysis: post‐randomisation exclusion of people for conservative care of hip fracture  Lost to follow‐up: details given
Participants	Location: single trauma ward, University Hospital of Wales, Cardiff, UK  Period of study: recruitment May 2000‐August 2003  318 participants  Inclusion criteria: women aged over 65 years presenting to trauma ward with acute non‐pathological hip fracture, consent or assent to trial  Exclusion criteria: none  Sex: all female  Age: mean age 84 years  Fracture type: further details not given
Interventions	Timing of intervention: unclear when commenced, during stay in acute trauma ward, median 16‐17 d. Dietetic assistant present on ward 6 h/d for 7 d/week  (a) Additional attention of dietetic assistant (previous NHS experience, given 14‐d period of orientation and training), working closely with specialist dietitian. Asked to ensure participants met nutritional needs, including by: checking personal and cultural food preferences; co‐ordinating appropriate meal orders with catering staff; ordering nutritional supplements; provision of feeding aids; assisting with food choice, portion size and positioning at mealtimes; sitting with, encouraging and feeding; collecting information to aid nutritional assessment by dietitian  (b) Nurse‐ and dietitian‐led care, including routine provision of oral nutritional supplements to all participants  Allocated: 153/165  Assessed: 145/157 for mortality
Outcomes	Length of follow‐up: 4 months  Main outcomes:  Mortality  Morbidity and complications: on trauma ward in survivors  Length of trauma ward and hospital stay  Other outcomes: energy intake
Notes	Request for further details on participants with complications sent 15 March 2006. Reply from trialists (15 March 2006) provided number and per cent of live participants having had complications on trauma ward. A letter to the editor in Age and Ageing Advance Access (24 June 2006) by Hewitt and Torgerson pointed out the numerical difference between the two groups was higher than expected given the reported block size of 10. The reply from Duncan indicated that they initially started the study with a block size of 20.
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Randomisation was by sequentially numbered, opaque, sealed envelope method in blocks of 10, prepared by a member of staff not directly involved in the trial." No further details
Allocation concealment (selection bias)	Low risk	"Randomisation was by sequentially numbered, opaque, sealed envelope method in blocks of 10, prepared by a member of staff not directly involved in the trial."
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Not blinded
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	No placebo group. Comment: unlikely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No placebo group. Comment: may have been influenced by lack of blinding
Incomplete outcome data (attrition bias)   Primary outcomes	Low risk	Missing outcome data balanced in numbers across intervention groups, with similar reasons for missing data across groups and unlikely to relate to outcome
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Low risk	Missing outcome data balanced in numbers across intervention groups, with similar reasons for missing data across groups and unlikely to relate to outcome
Selective reporting (reporting bias)	Unclear risk	Appears Waterlow score of pressure sore risk and Abbreviated Mental Test score collected as outcomes, but not provided
Other bias	Unclear risk	Funding from Women's Royal Voluntary Service, British Dietetic Assocation, Innovations in Care, Wales Office of Research and Development, Shire Pharmaceuticals (funded nutritional assessments, research assessments)

Methods	Method of randomisation: states "randomly divided" only Intention‐to‐treat: unclear Lost to follow‐up: unclear
Participants	Location: Trauma Center Meidling, Vienna, Austria Period of study: before September 2010 23 participants Inclusion criteria: aged > 65 years with hip fractures (femoral neck, intertrochanteric and subtrochanteric) Exclusion criteria: acute or chronic renal disease, liver failure, severe congestive heart failure, severe pulmonary disease, and any gastrointestinal condition that might preclude the participant from adequate oral nutritional intake Sex: all female Age: mean age 84 years Fracture type: further details not given
Interventions	Timing of intervention: after operation whilst hospitalised a) Oral supplements administered individually when energy and/or protein intake calculated by dietary records did not exceed a level of 20–25 kcal and/or 1–1.5 g protein/kg body weight/ day as recommended by the European Society for Clinical Nutrition and Metabolism per 1000 ml – 4.2 MJ (40% energy as protein), 1.88 mg vitamin A, 13 mcg vitamin D, 23 mg vitamin E, 0.1 mg vitamin K, 190 mg vitamin C, 2.8 mg thiamine, 3.1 mg riboflavin, 34 mg niacin, 3.3 mg pyridoxine, 0.5 mg folate, 10 mg pantothenic acid, 7 mcg vitamin B12, 75 mcg biotin, 500 mg sodium, 2 g potassium, 420 mg magnesium, 2.8 g calcium, 2 g phosphorus, 900 mg chloride, 23 mg zinc, 30 mg iron, 3.4 mg copper, 0.25 mg iodine, 0.13 mg chromate, 1.9 mg fluoride, 6.3 mg manganese, 0.19 mg molybdenum, 0.11 mg selenium b) Usual care Allocated: 14/9 Assessed: 14/9 (numbers not certain)
Outcomes	Length of follow‐up: length of hospitalisation Main outcomes: Length of hospital stay
Notes	Emailed ibrahim.elmadfa@univie.ac.at 31 December 2014 to request more details of denominators
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	States "randomly divided" only
Allocation concealment (selection bias)	Unclear risk	States "randomly divided" only
Blinding of participants and personnel (performance bias)   All outcomes	High risk	No placebo group. Comment: likely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No placebo group. Comment: may have been influenced by lack of blinding
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Unclear risk	Denominators not given for length of stay
Selective reporting (reporting bias)	High risk	Length of stay only provided, with no other details of clinical outcomes. Length of stay not included in methods
Other bias	Low risk	Funded by Trauma Center, Meidling, Vienna

Methods	Method of randomisation: randomised into 3 groups in blocks of 12, using a sealed envelope technique Intention‐to‐treat analysis: appears undertaken Lost to follow‐up: 20% of groups examined here
Participants	Location: 4 university hospitals in Stockholm, Sweden Period of study: before 2014 54 participants Inclusion criteria: age 60 years or older, no severe cognitive impairment (Short Portable Mental Questionnaire score ≥ 3), ambulatory before fracture, body mass index 28 kg/m2 or lower Exclusion criteria: pathological fractures and bisphosphonate treatment within the last year; alcohol/drug abuse or overt psychiatric disorders; abnormal hepatic or renal laboratory parameters such as serum‐alanine aminotransferase or serum‐aspartate‐aminotransferase twice the normal reference range or higher, respectively; serum‐creatinine levels higher than 130 μmol/L or glomerular filtration rate lower than 30 mL/minute; bone metabolic disorders such as primary hyperparathyroidism, osteogenesis imperfecta, Paget’s disease, or myeloma; lactose intolerance, dysphagia, oesophagitis, gastric ulcer, or malignancy; diabetes mellitus associated with nephropathy or retinopathy; active iritis or uveitis Sex: 37 female, 17 male Age: mean 81 years Fracture type: 41% femoral neck fracture, 59% trochanteric fracture
Interventions	Timing of intervention: as soon as participants were stable from a cardiovascular standpoint, able to take food by mouth, and able to sit in an upright position for 1 h after taking their tablets for 6months (a) Fresubin (Fresenius Kabi, Bad Homburg, Germany) protein energy drink, 200 mL twice daily, totaling 600 kcal with 40 g protein and 35 mg risedronate once weekly for 12 months (b) 35 mg risedronate once weekly for 12 months Allocated: 26/28 Assessed: 18/25
Outcomes	Length of follow‐up: 1 year Main outcomes: Mortality Complications Putative side effects Other outcomes: Compliance
Notes	Emailed lena.flodin@karolinska.se on 9 December 2014 to enquire if more data on outcomes available. Author provided more details 15 December 2014 A third group ('control') was not included in this review
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	States "randomized into three groups in blocks of twelve, using a sealed envelope technique", no details of sequence generation
Allocation concealment (selection bias)	Low risk	States "randomized into three groups in blocks of twelve, using a sealed envelope technique"
Blinding of participants and personnel (performance bias)   All outcomes	High risk	No placebo intervention
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	No placebo group. Comment: unlikely to have been influenced by lack of blinding.
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No placebo group. Comment: may have been influenced by lack of blinding
Incomplete outcome data (attrition bias)   Primary outcomes	High risk	8/26 nutrition group lost to follow‐up versus 3/28 in control group
Incomplete outcome data (attrition bias)   Secondary and other outcomes	High risk	8/26 nutrition group lost to follow‐up versus 3/28 in control group
Selective reporting (reporting bias)	Unclear risk	Insufficient details provided
Other bias	Unclear risk	About 10% difference in weight between groups, although BMI only differs by 1.3 kg/m2 Fresenius Kabi provided supplement, but states not involved in the planning or implementation of the study, nor in the analyses, conclusions, or manuscript writing

Methods	Method of randomisation: block randomised, double‐blind. Randomisation was performed by the Royal Perth Hospital Pharmacy Department, and those involved in this process had no other study involvement. Intention‐to‐treat analysis: not undertaken Lost to follow‐up: 26% did not complete study
Participants	Location: 2 teaching hospitals, Perth, Australia Period of study: before November 2008 95 participants Inclusion criteria: vitamin D‐deficient (serum 25O HD b50 nmol/L) by DiaSorin radioimmunoassay Exclusion criteria: ionised hypercalcaemia, chronic kidney disease (serum creatinine > 150 μmol/L), history of thyrotoxicosis or Cushing's syndrome, concomitant anticonvulsant drug therapy, and use of other medications affecting bone metabolism (including oestrogen, raloxifene, calcitriol, anabolic steroids, bisphosphates, sodium fluoride, oral glucocorticoids > 7.5 mg/day or inhaled glucocorticoids > 1000 μg/day) within the preceding 3 months; poor prognosis or who were unlikely to comply with therapy Sex: not given Age: mean 83 years Fracture type: further details not given
Interventions	Timing of intervention: 3 months from inpatient stay (a) Vitamin D3 1000 IU/d and 1 placebo daily and calcium carbonate equivalent to 600 mg/d (b) Vitamin D2 1000 IU/d and 1 placebo daily and calcium carbonate equivalent to 600 mg/d Allocated: 47/48 Assessed: 36/34 for compliance
Outcomes	Length of follow‐up: 3 months Main outcomes: Mortality, Hypercalcaemia Other outcomes: Compliance
Notes	Boots Health Care provided vitamin D2 and matching placebo. Study funded by Royal Perth Hospital Medical Research Foundation
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	States "Randomization was performed by the Royal Perth Hospital Pharmacy Department, and those involved in this process had no other study involvement", no further details
Allocation concealment (selection bias)	Low risk	States "Randomization was performed by the Royal Perth Hospital Pharmacy Department, and those involved in this process had no other study involvement"
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	States double‐blind
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	States double‐blind and unlikely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	States double‐blind but blinding of outcome assessment not described and may have been influenced by lack of blinding
Incomplete outcome data (attrition bias)   Primary outcomes	High risk	8/47 on vitamin D3 and 7/48 on vitamin D2 appear not to have been included in follow‐up
Incomplete outcome data (attrition bias)   Secondary and other outcomes	High risk	8/47 on vitamin D3 and 7/48 on vitamin D2 appear not to have been included in follow‐up
Selective reporting (reporting bias)	Unclear risk	Insufficient details provided
Other bias	Unclear risk	Boots Health Care provided vitamin D2 and matching placebo. Study funded by Royal Perth Hospital Medical Research Foundation

Methods	Method of randomisation: sealed, opaque envelopes in blocks of 10, appears stratified by place of residence  Assessor blinding: not done  Intention‐to‐treat analysis: carried out  Lost to follow‐up: details given
Participants	Location: acute care in Hornsby‐Kuringai Hospital and rehabilitation hospitals, Sydney, Australia  Period of study: admissions from 16 May‐8 August 1996  32 participants  Inclusion criteria: fractured neck of femur after accidental fall; admitted from home, hostel or nursing home; age 65 years 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 86 years  Fracture type: further details not given
Interventions	Timing of intervention: started within 5 d of surgery, given once in the morning and once in the evening for 30 d, served on meal tray in hospital by nurses, given by family or self‐administered out of hospital  (a) Oral supplement of 250 ml Sustagen twice daily (total daily intake 22.5 g protein, 10 g fat, 60 g carbohydrate, 1.712 MJ or 409 kcal energy, 500 mcg vitamin A, 6.6 mcg vitamin D, 50.8 mg vitamin C, 1.2 mg thiamin, 1.15 mg riboflavin, 13 mg niacin, 1.3 mcg vitamin B12, 825 mg calcium, 670 mg phosphorus, 8 mg iron, 66 mcg iodine, 1.2 g potassium, 370 mg sodium) plus standard hospital diet  (b) Standard hospital diet  Allocated: 17/15  Assessed: 17/14
Outcomes	Length of follow‐up: 2 months  Main outcomes:  Mortality  Morbidity and complications: complications (total, infection, pressure sores, pulmonary embolism, delirium, anaemia, cardiac failure, acute renal failure), favourable clinical course (excludes death, major complication, or 2 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 2 months  Other outcomes:  Energy, protein intakes from food and supplement; calcium, iron and vitamin C intakes from food  Patient compliance: numbers completing full 30 d of supplement
Notes	Request for further details (blinding of outcome assessors, details of supplement administration, further information on outcomes) sent. Reply from trialists (11 June 1999) gave details of outcome assessor blinding, supplement administration and outcomes
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Sealed, numbered opaque envelopes in blocks of 10". Information from Ian Cameron
Allocation concealment (selection bias)	Low risk	"Sealed, numbered opaque envelopes in blocks of 10". Information from Ian Cameron
Blinding of participants and personnel (performance bias)   All outcomes	High risk	No placebo group. Comment: likely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	No placebo group. Comment: unlikely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No placebo group. Comment: may have been influenced by lack of blinding
Incomplete outcome data (attrition bias)   Primary outcomes	Low risk	Only one participant withdrew in control group, data provided by Ian Cameron for all other participants
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Low risk	Only one participant withdrew in control group, data provided by Ian Cameron for all other participants
Selective reporting (reporting bias)	Low risk	Thesis provides details that all outcomes reported
Other bias	High risk	Mead Johnson pharmaceutical company provided Sustagen supplement

Methods	Method of randomisation: states randomised controlled trial, no further details Intention‐to‐treat analysis: no details Lost to follow‐up: no details
Participants	Location: Daejin Medical Center, Bundang Jesaeng General Hospital, Korea Period of study: before September 2012 60 participants Inclusion criteria: aged over 65 years admitted to hospital for hip fracture surgery Exclusion criteria: none provided Sex: not given Age: mean age 81 years Fracture type: further details not given
Interventions	Timing of intervention: 2 weeks postoperatively (a) Oral nutritional supplements, trace element supplements and dietetic counselling (b) Usual care Allocated: 30/30 Assessed: unclear
Outcomes	Length of follow‐up: mean of 120 days Main outcomes: Mortality Complications
Notes	Abstract only. Letter to Dr Kang requesting more details sent 3 October 2014
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Abstract only. States randomized controlled trial, no further details
Allocation concealment (selection bias)	Unclear risk	Abstract only. States randomized controlled trial, no further details
Blinding of participants and personnel (performance bias)   All outcomes	High risk	Abstract only. No placebo group. Comment: probably not done
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	Abstract only. No placebo group. Comment: unlikely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	Abstract only. No placebo group. Comment: unclear if done
Incomplete outcome data (attrition bias)   Primary outcomes	Unclear risk	Abstract only. No details provided
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Unclear risk	Abstract only. No details provided
Selective reporting (reporting bias)	Unclear risk	Abstract only. Insufficient details provided
Other bias	Unclear risk	Abstract only. Insufficient details provided. No details on sponsor

Methods	Method of randomisation: not stated, stratified by type of hip fracture  Intention‐to‐treat analysis: unclear  Lost to follow‐up: details given
Participants	Location: 3 rehabilitation hospitals, USA  Period of study: unclear  46 participants  Inclusion criteria: within 3 weeks of surgical repair of hip fracture (intertrochanteric or femoral neck), expected to stay 1‐3 weeks in rehabilitation, aged 60 years or over, BMI < 30 kg/m2, informed consent, able to be reached by phone after discharge  Exclusion criteria: fracture due to non‐osteoporotic disease, e.g. pathological fracture; significant trauma to other organ systems or medical conditions significantly affecting outcome (severe hepatic dysfunction bilirubin > 3 mg/dL, severe renal dysfunction creatinine at least 3 mg/dL or dialysis, uncontrolled diabetes: 2 random blood glucose values > 200 mg/dL or > 140 mg/dL fasting)  Sex: 33 female, 13 male  Age: mean age 83 years  Fracture type: further details not given
Interventions	Timing of intervention: consecutive 28‐d period at least two 8 oz cans/d  (a) Boost HP high protein liquid supplement (Mead Johnson, Evansville, Indiana, USA) providing per 8 oz can: 240 kcal, 15 g protein, 33 g carbohydrate, 6 g fat, 1110 IU vitamin A, 89 IU vitamin D, 6.7 IU vitamin E, 27 mcg vitamin K, 13.3 mg vitamin C, 89 mcg folic acid, 0.33 mg thiamin, 0.4 mg riboflavin, 0.47 mg vitamin B6, 1.33 mcg vitamin B12, 4.7 mg niacin, 56 mg choline, 67 mcg biotin, 2.3 mg pantothenic acid, 220 mg sodium, 490 mg potassium, 350 mg chloride, 240 mg calcium, 220 mg phosphorus, 90 mg magnesium, 33mg iodine, 0.67 mg manganese, 0.47 mg copper, 3.3 mg zinc, 4 mg iron, 15.8 mcg selenium, 27 mcg chromium, 16.9 mcg molybdenum  (b) Ensure liquid supplement (Ross Labs, Columbus, Ohio, USA) providing per 8 oz can: 250 kcal, 8.8 g protein, 40 g carbohydrate, 6.1 g fat, 1250 IU vitamin A, 100 IU vitamin D, 7.5 IU vitamin E, 20 mcg vitamin K, 30 mg vitamin C, 100 mcg folic acid, 0.38 mg thiamin, 0.43 mg riboflavin, 0.50 mg vitamin B6, 1.50 mcg vitamin B12, 5.0 mg niacin, 100 mg choline, 75 mcg biotin, 2.5 mg pantothenic acid, 200 mg sodium, 370 mg potassium, 310 mg chloride, 300 mg calcium, 300 mg phosphorus, 100 mg magnesium, 38 mcg iodine, 1.3 mg manganese, 0.50 mg copper, 3.8 mg zinc, 4.5 mg iron, 18 mcg selenium, 30 mcg chromium, 38 mcg molybdenum  Allocated: 22/24  Assessed: 18/20 for length of stay
Outcomes	Length of follow‐up: 3 months  Main outcomes:  Mortality  Morbidity: complications (nr), adverse events (nr)  Length of rehabilitation hospital stay  Location for discharge  Postoperative functional status: mobility subscale of FIM instrument (Uniform Data System for Medical Rehabilitation)  Other outcomes:  Days of supplement consumption
Notes	Request for further details (mortality, denominators for length of stay, complications) sent 13 October 2004. Details of mortality and denominators received 06 January 2005
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No information other than: "randomized, double‐blind, parallel‐group study"
Allocation concealment (selection bias)	Unclear risk	No information other than: "randomized, double‐blind, parallel‐group study"
Blinding of participants and personnel (performance bias)   All outcomes	Unclear risk	States double‐blind but no further details
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	States double‐blind and unlikely to have been influenced by lack of blinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	States double‐blind but no further details, and may have been influenced by lack of blinding
Incomplete outcome data (attrition bias)   Primary outcomes	Unclear risk	No details on denominators for complications provided
Incomplete outcome data (attrition bias)   Secondary and other outcomes	High risk	Length of stay data for 4 participants on Boost, and 4 on Ensure not provided. Numbers for purported adverse events, mobility and discharge destination not provided
Selective reporting (reporting bias)	Unclear risk	Insufficient details provided
Other bias	High risk	Part funded by Mead Johnson, manufacturer of Boost HP

Methods	Method of randomisation: randomised in blocks according to computer‐generated randomisation, in‐patient pharmacy co‐ordinated the randomisation and drug distribution Intention‐to‐treat analysis: not carried out Lost to follow‐up: 18/65 lost to follow‐up
Participants	Location: two academic hospital sites, Hamilton, Ontario, Canada Period of study: October 2007‐April 2009 65 participants Inclusion criteria: over age 50 with an acute fragility hip fracture (defined as femoral neck, trochanteric, subtrochanteric or subcapital) which was the result of a minimal trauma accident, defined as a fall from standing height or less Exclusion criteria: pelvic fractures; pathological fractures secondary to malignancy or intrinsic bone disease (e.g. Paget’s disease); pre‐existing bone abnormality; cancer in the past 10 years likely to metastasize to bone; renal insufficiency (creatinine < 30 mls/min); renal stones in past 10 years; hypercalcaemia (primary hyperparathyroidism; granulomatous diseases); hypocalcaemia; stroke within the last 3 months; or had taken hormone replacement therapy, calcitonin, bisphosphates, raloxifene, or parathyroid hormone during the previous 24 months; admitted from long‐term care facilities/nursing homes Sex: 36 female, 25 male Age: mean 69 years Fracture type: further details not given
Interventions	Timing of intervention: day 1 for 90 d (a) Oral placebo bolus day 1, then a daily tablet of 1000 IU vitamin D3 for 90 d (b) 50,000 IU vitamin D2 oral bolus day 1, then a daily tablet of 1000 IU vitamin D3 for 90 d (c) 100,000 IU vitamin D2 oral bolus day 1, then a daily tablet of 1000 IU vitamin D3 for 90 d Allocated: 22/22/21 Assessed: 12/18/17 at 90 d
Outcomes	Length of follow‐up: 90 d Main outcomes: Mortality Adverse events Other outcomes: Compliance
Notes	Emailed PAPAIOANNOU@HHSC.CA 6 November 2014 for details of allocation of participants who died or had adverse events. No details received
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	States "Patients were randomized in blocks according to a computer‐generated randomization list"
Allocation concealment (selection bias)	Low risk	States "The central in‐patient pharmacy at McMaster University Medical Centre coordinated the randomization procedure and the distribution of study drugs"
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	Blinded, placebo‐controlled trial and states "The medication treatment group was concealed and all participants, study coordinators, physicians, staff, and caregivers were blinded to treatment group allocation"
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	Blinded, placebo‐controlled trial and states "The medication treatment group was concealed and all participants, study coordinators, physicians, staff, and caregivers were blinded to treatment group allocation"
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Low risk	Blinded, placebo‐controlled trial and states "The medication treatment group was concealed and all participants, study coordinators, physicians, staff, and caregivers were blinded to treatment group allocation"
Incomplete outcome data (attrition bias)   Primary outcomes	High risk	18 participants from 65 lost to follow‐up by 90‐d final follow‐up
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Unclear risk	18 participants from 65 lost to follow‐up by 90‐day final follow‐up
Selective reporting (reporting bias)	Low risk	Principally a study of vitamin D dose responses and adverse events
Other bias	High risk	Signficant imbalance in age between two intervention groups (reported P = 0.024). Study supported by Merck Frosst Canada Ltd

Methods	Method of randomisation: sealed opaque numbered envelopes Intention‐to‐treat analysis: undertaken Lost to follow‐up: no participants lost to follow‐up
Participants	Location: Peterborough District Hospital, UK Period of study: recruitment January 2003–July 2007 300 participants Inclusion criteria: postoperative haemoglobin level of < 110 g/L within 5 d after hip fracture surgery. Exclusion criteria: participant unwilling to give written informed consent or for whom the relative or next of kin was unavailable or declined to give assent, postoperative haemoglobin level of ‡110 g/L, multiple trauma (defined as either > 2 other fractures or any other fracture requiring surgery other than simple manipulation), participant unable to take oral iron medication because of adverse effects, participant taking iron therapy at time of admission, haemoglobin level of < 110 g/L at time of admission, participant unable to attend routine follow‐up in the hip fracture clinic, age of < 60 years Sex: 245 female, 55 male Age: mean age 82 years Fracture type: 45% intracapsular fracture, 21% intramedullary nail and 34% extramedullary fixation (presumed not intracapsular fractures)
Interventions	Timing of intervention: immediately post‐randomisation for 28 d (a) Oral iron therapy (ferrous sulphate, 200 mg twice daily) (b) No iron supplement Allocated: 150/150 Assessed: 150/150 at 12 months
Outcomes	Length of follow‐up: 12 months Main outcomes: Mortality, Hospital length of stay Putative side effects of treatment
Notes	Emailed Dr Martyn Parker (Martyn.Parker@pbh‐tr.nhs.uk) 16 October 2014 about further details on length of hospital stay data, reply received 16 October 2014
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	No details provided on sequence generation
Allocation concealment (selection bias)	Low risk	States "randomization was accomplished by opening a sealed opaque numbered envelope for each patient"
Blinding of participants and personnel (performance bias)   All outcomes	High risk	No placebo group
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	No placebo group. Comment: unlikely to have been influenced by unblinding
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	No placebo group. Comment: may have been influenced by unblinding
Incomplete outcome data (attrition bias)   Primary outcomes	Low risk	Data for all participants randomised provided
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Unclear risk	13/150 discontinued iron therapy in intervention group and 5 in control group commenced iron therapy. 7/150 in intervention group unable to attend outpatient follow‐up and 16/150 in control group likely to have influenced putative side effects of treatment
Selective reporting (reporting bias)	Unclear risk	Study protocol not available and unclear if all expected outcomes provided
Other bias	Low risk	Non‐pharmaceutical funding (funded by Peterborough Hospital Hip Fracture Fund)

Methods	Method of randomisation: states random number table and double‐blind study, but unclear if those who assigned were blinded  Intention‐to‐treat analysis: unclear  Lost to follow‐up: incomplete report of drop outs
Participants	Location: orthopaedic ward in hospital and recovery hospital, Geneva, Switzerland  Period of study: April 1992‐February 1994  82 participants  Inclusion criteria: hip fracture within 2 weeks attributable to osteoporosis (minor trauma), aged over 60 years, able to give written consent  Exclusion criteria: pathological fracture; fracture caused by severe trauma; history of contralateral hip fracture; severe mental impairment; active metabolic bone disease; renal failure (plasma creatinine equal to or greater than 200 mcmol/L); acute illness that could interfere with study protocol; severe malnutrition (serum albumin less than 15 g/L); on drugs known to alter bone metabolism, e.g. calcitonin, fluoride, sex hormones, corticosteroids, bisphosphates; life expectancy less than 1 year  Sex: 74 female, 8 male  Age: mean 80.7 years  Fracture type: 31 cervical, 51 trochanteric
Interventions	Timing of intervention: mean randomisation time 6.5 (SD 1.9) d after fracture, supplemented 5 d a week for 6 months  (a) Oral protein supplement (1.05 MJ or 250 kcal, 20 g protein, 3.1 g fat, 35.7 g carbohydrate, 1000 IU vitamin A, 30 mcg vitamin K1, 20 mg vitamin C, 550 mg calcium, 91 mg magnesium, 429 mg phosphorus, 228 mg sodium) plus oral 200,000 IU vitamin D3 once at baseline during study  (b) Placebo without protein made isocaloric by addition of maltodextrins, plus oral 200,000 IU vitamin D3 once at baseline during study  Allocated: 41/41  Assessed: ?/?
Outcomes	Length of follow‐up: 12 months  Main outcomes:  Mortality  Length of stay: orthopaedic ward, rehabilitation stay  Postoperative functional status: activities of daily living score  Putative side effects: drop outs due to nausea and diarrhoea  Other outcomes:  Patient compliance: refusals
Notes	Composition of placebo unclear, denominators not clear. Request for further details sent 27 May 1999, re‐sent 7 February 2000
*Risk of bias*
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Using a random number table", no further details provided
Allocation concealment (selection bias)	Unclear risk	"Using a random number table, we assigned ..."  Although "double‐blind", it is unclear whether allocation was concealed
Blinding of participants and personnel (performance bias)   All outcomes	Low risk	Oral protein supplement and placebo made isocaloric, states "double‐blind". Comment: probably done
Blinding of outcome assessment (detection bias)   Primary outcomes	Low risk	Oral protein supplement and placebo made isocaloric, states "double‐blind" and unlikely to be influenced by unblinding. Comment: probably done
Blinding of outcome assessment (detection bias)   Secondary and other outcomes	Unclear risk	Oral protein supplement and placebo made isocaloric, states "double‐blind" and may have been influenced by unblinding as no details on who assessed outcomes
Incomplete outcome data (attrition bias)   Primary outcomes	Low risk	All participants accounted for and drop‐outs do not appear to differ between groups
Incomplete outcome data (attrition bias)   Secondary and other outcomes	Unclear risk	No denominators for lengths of stay, activities of daily living unclear
Selective reporting (reporting bias)	Unclear risk	Insufficient details provided
Other bias	High risk	Study supported by Sandoz Nutrition Ltd