Health-related quality of life as an outcome variable in Parkinson’s disease

Abstract

In the past three decades, health-related quality of life (HRQoL) has become an outcome variable in Parkinson’s disease clinical trials. This review considers the measuring tools that have been developed, suitability of data reporting, complexity of outcome interpretation, and clinical application to provide evidence regarding available therapeutic interventions to date. In the introduction, different terms regarding quality of life are clearly defined. The methodology section offers an overview of generic, disease specific, and recommended HRQoL scales in Parkinson’s disease and the most important psychometric attributes a scale should meet. The interpretation of HRQoL outcomes is complex and not intuitive. Thus, appropriate reporting of data is crucial in order to calculate relative change, a result that facilitates understanding to what extent an intervention is beneficial. The concept of minimally important change/difference is explained as well as the different approaches to its calculation (anchor-based and distribution-based methods). In the results section, a brief overview of the impact on HRQoL of currently available treatments in Parkinson’s disease is provided. Special emphasis is given to data assessment, highlighting reports that helped understanding of the clinical significance of the intervention and therefore aided in making therapeutic decisions. The discussion section emphasizes the need for more clinical trials with HRQoL as a primary outcome and standardized reporting in order to further our understanding of the complexity of treatment effects and make evidence-based clinical decisions regarding HRQoL in patients with Parkinson’s disease.

Introduction

Currently, the expression ‘quality of life’ is frequently used in conversation and by mass media to evoke feelings of well being and satisfaction in important aspects for the individual that are related to personal judgment and social standards. More than three decades ago, the term was introduced in medicine as the objective of care, in contrast to the indiscriminate ‘quantity of life’ (‘to live better rather than to live more’). Since then, terminology regarding quality of life in medical settings and literature has expanded and can be rather confusing since there are no universally accepted definitions and differing terms are erroneously used interchangeably. In our understanding, quality of life encompasses ‘individuals’ perception of their position in life in the context of the culture and value systems in which they live and in relation to their goals, expectations, standards and concerns’ [The WHOQOL Group, 1995], whereas health-related quality of life (HRQoL) is ‘the perception and evaluation, by patients themselves, of the impact caused on their lives by the disease and its consequences’ [Martinez-Martin, 1998]. Health status influences and predicts HRQoL, but should not be considered equivalent to HRQoL [Den Oudsten et al. 2007] since it assesses physical and mental symptoms, disability, and social dysfunction related to a health situation in a mere descriptive way, lacking judgments about their impact on the individual’s well being and expectations.

The arrival of the current biopsychosocial model to medicine has lead to participation of patients in their own care, thus their judgments about perceived health status, preferences and satisfaction with treatment are now considered essential part of health care. Subsequently, instruments for measurement of these judgments have been developed and are integrated in assessment procedures as ‘patient-reported outcomes’ (PROs), defined as ‘information provided by patients on their health status and treatment’. PROs are scientific measures for evaluation of health status and change and they should be developed, validated, and analyzed as any other variable in a study [Acquadro et al. 2003]. Together with clinician-reported evaluation, ancillary tests (biological, physiological, imaging), and caregiver-reported assessment, PROs are useful indicators of disease activity and treatment efficacy and thus should be included in clinical trials and taken into account for evidence-based practice [European Medicines Agency, 2005; Food and Drug Administration, 2009]. HRQoL is a kind of PRO which refers to quality of life aspects affected by health status and health care.

Methods

HRQoL scales

The most frequently included domains in HRQoL measures are symptoms, physical function, mental status (emotion, cognition), social (well being, activities), role function, and global judgment of health [Berzon et al. 1993; Schipper et al. 1996; Fitzpatrick et al. 1998; Fitzpatrick and Alonso, 1999]. Therefore, HRQoL measures combine personal judgments and reflections about well being and satisfaction with health. HRQoL scales may be classified as generic or specific. Generic scales, in turn, may be health profiles or utility measures. Profiles adopt the format of questionnaires and seek to cover the most important domains of health. Typical examples of this category are the Sickness Impact Profile (SIP) [Bergner et al. 1976], the Nottingham Health Profile (NHP) [Hunt et al. 1981] and the Medical Outcomes Study-Short form (SF-36) [Ware and Sherbourne, 1992]. Utility measures provide a unique number (‘utility’) that represents quality of life status in relation to best health status (1.0) and death (0.0). They are used for econometric analysis and decision making in health policy. The EQ-5D [EuroQol Group, 1990] is a prototypic example of these kind of measures. Their possible disadvantages are related with potential inability to assess areas of specific interest and lack of sensitivity to detect changes.

Specific instruments address areas of particular interest for a disease, population, function, or dimension. Their advantages are the ability to evaluate health problems with high prevalence in the target population for which they were developed and their responsiveness. Therefore, specific instruments are more appropriate than generic measures for clinical trials, although ideally they should be combined to obtain relevant data on general health.

Table 1 shows a list of the generic and specific HRQoL scales used in Parkinson’s disease (PD) with their respective acronym and some of their basic characteristics. These scales have been recently reviewed by an ad hoc Movement Disorder Society Task Force concerning their psychometric attributes and validity evidence. Four generic (SIP, NHP, SF-36, EQ-5D) and five specific (PDQ-39, PDQ-8, PIMS, PDQL, SCOPA-PS) scales gained the qualification of ‘recommended’ for application in patients with PD. That designation means a scale fulfills ‘criteria of successful clinimetric testing, established validation, reliability, and sensitivity assessments, and prior application in PD’ [Goetz et al. 2007; Martinez-Martin et al. 2011].

Table 1.

Health-related quality of life scales applied in Parkinson’s disease studies.

Scale	Acronym	Number of		Time frame	Reference
		Items	Domains
Generic Scales
Sickness Impact Profile	SIP	136	12	Today	Bergner et al. [1976]
Nottingham Health Profile	NHP	38	6	At present time	Hunt et al. [1981]
EuroQoL 5 Dimensions	EQ-5D	5	5	Today	EuroQol Group [1990]
Medical Outcomes Study-Short Form 36 items	SF-36	36	8	Last 4 weeks	Ware and Sherbourne [1992]
Schedule for the Evaluation of Individual Quality of Life – Direct Weighting	SEIQOL-DW	-	5	At present time	Hickey et al. [1996]
Quality of life questionnaire 15 Dimensions	15-D	15	15	At present time	Sintonen [2001]
WHO Quality of Life Assessment Short Version	WHOQOL- BREF	26	4	Last 4 weeks	Saxena et al. [2001]
Questions on Life Satisfaction
– Movement Disorders	QLS -MD	12	1	Last 4 weeks	Kuehler et al. [2003]
– Deep Brain Stimulation	QLS-DBS	5	1
Specific Scales
Belastungsfragebogen Parkinson Kurzversion	BELA-p-k	19	4	At present time	Ellgring et al. [1993]
Parkinson’s Disease Questionnaire
– 39 items	PDQ-39	39	8	Last month	Peto et al. [1995], Jenkinson et al. [1997]
– 8 items	PDQ-8	8	8
Parkinson’s Impact Scale	PIMS	10	4	At present time	Calne et al. [1996]
Parkinson’s Disease Quality of Life Questionnaire	PDQL	37	4	Last 3 months	de Boer et al. [1996]
Fragebogen Parkinson Lebensqualität	Fragebogens PLQ	44	9	Past week	Van den Berg [1998]
Parkinson’s Disease Quality of Life Scale	PDQUALIF	32	7	Flexible	Welsh et al. [2003]
Parkinson’s Problem Schedule	PPS	39	3	(unknown)	Brod et al. [1998]
Scales for Outcomes in Parkinson’s Disease-Psychosocial	SCOPA-PS	11	1	Last month	Marinus et al. [2003]

Scale attributes that influence outcomes

The outcome analysis of a clinical trial including HRQoL as an endpoint of interest will be affected by the characteristics of the scale applied [Sprangers et al. 2002; Marquis et al. 2004; European Medicines Agency, 2005; Food and Drug Administration, 2009]. The most relevant psychometric properties when testing a scale are summarized in Table 2. A study can be unsuccessful if any of these problems occur: missing data in a considerable proportion of the sample; accumulation of patients scoring in the high or low end of the scale; indexes reflecting high random error due to poor item cohesion or measure stability; lack of association with other measures applied for assessing the same or related constructs or, in contrast, nonnegligible correlation with measures evaluating constructs nonrelated with that of interest; large distance between correlative scores; and lack of sensitivity to capture a real change.

Table 2.

Psychometric properties especially relevant for outcome analysis.

Attribute	Meaning
Data quality	The completeness of item- and scale-level data
Acceptability	Extent to which the scores are well distributed in the sample
Reliability	Extent to which the scale is free from random error
A - Internal consistency	- intercorrelations of the items at one point in time
B - Reproducibility	- stability over time (test–retest) and interrater agreement
Construct validity	Degree of coherence in the behavior of scores according to theoretical assumptions related with the concept measured
A - Convergent and divergent validity	- degree of association with other measures used for evaluating the same construct or a nonrelated construct
B - Known-groups validity	- ability of the scale to detect differences between (extreme) groups at a point in time
Precision (sensitivity)	Ability of the instrument to detect small differences
Responsiveness	Ability of the scale to detect change over time
Interpretability	Degree to which an easily understandable meaning can be assigned to the quantitative scores

Interpretation of outcomes

Validation procedures in the realm of HRQoL instruments are quite uniform, but interpretation of outcomes is not, and this fact leads to uncertainty and confusion. Difficulty in establishing the meaningfulness of change is related to the complexity of the outcome itself (difference by severity level; multiple items and domains; influence of nonrelated aspects to the disease/treatment) [Crosby et al. 2003; European Medicines Agency, 2005] and to the lack of intuitive association between the magnitude of change and significance.

Reporting of HRQoL is often incomplete since most studies simply report the mean and standard deviation at baseline and at the end of follow up, with a p value that determines statistical significance (calculated, for example, by a t-test or repeated measures analysis of variance [ANOVA]). However, it is important to keep in mind that p value ‘is the probability of having observed our data when the null hypothesis is true’ [Altman, 1991]. The usual cut-off, p = 0.05, is an arbitrary level and nothing will be very different if a p value is 0.045 or 0.055. Really, a ‘p’ value depends on the standard error of the observed value (related, in turn, to the sample size) and informs about the null hypothesis, but says nothing about the real meaningfulness of the observed difference [Osoba et al. 1998; Hurley, 2010]. Confidence interval and power of the study add relevant information for interpreting the certainty of results [Altman, 1991; Sprangers et al. 2002; Charter and Feldt, 2009], but give no information on the importance of change. A better way of approaching the significance of change is to provide the percentage that the change represents of the baseline value (relative change = (mean₂ – mean₁)/ mean₁) [Deyo and Centor, 1986]. The higher the magnitude of relative change, the more likely that it will be meaningful.

For interpretation of clinical significance when analyzing HRQoL measures, the minimally important change (MIC) or difference (MID) [Jaeschke et al. 1989, 1991; Barrett et al. 2005; Sloan et al. 2005] is a relevant criterion [European Medicines Agency, 2005]. A definition of MID is ‘the smallest difference in score in the outcome of interest that informed patients or informed proxies perceive as important, either beneficial or harmful, and that would lead the patient or clinician to consider a change in the management’ [Brozek et al. 2006]. The MID works as a threshold value, marking the border between a negligible and a significant benefit (or harm).

Two strategies have been used most frequently to determine a MID:

1. The anchor-based approach considers the relationship between change in the HRQoL measure and an independent clinical or patient-based measure (the anchor). The aim is to assign patients to groups showing no change, mild, moderate, or large (positive or negative) change in health status. The anchor should be easily interpretable and closely related with the target [Guyatt et al. 2002]. It may be an external measure of symptoms, severity level, functional status, etc. or a within-patient rating, typically in the format of a transition question on, for example, a scale from 0 (no change) to 7 (a very large change) (Table 3). The mean change in score observed for the group of patients referring small change is considered the MID [Jaeschke et al. 1989; Juniper et al. 1994; Guyatt et al. 2002].

2. Distribution-based methods seek to determine MID on the basis of the observed scores and, therefore, they simply express the change in a standardized way. Different formulas have been proposed for this purpose, such as the effect size [(mean₂ – mean₁)/SD₁)] [Cohen, 1977; Terwee et al. 2003], the standardized response mean [(mean₂ – mean₁)/SD_difference)] [Crosby et al. 2003; Terwee et al. 2003], half a standard deviation at baseline [Norman et al. 2003; Sloan et al. 2005], the standard error of measurement [SEM = SD × √(1 – r_xx)] [Wyrwich et al. 1999, 2005], and the reliable change index [1.96 × S_diff] [Jacobson and Truax, 1991; Fitzpatrick et al. 2004].

Table 3.

Transition questions usable as within-patient anchor measures.

Rating	Change
0 – No change	No
1 – Almost the same; hardly any better (worse) at all	Small
2 – A little bit better (worse)
3 – Somewhat better (worse)
4 – Moderately better (worse)	Moderate
5 - A good deal better (worse)
6 – A great deal better (worse)	Large
7 – A very great deal better (worse)

Based on Guyatt et al. [2002].

There is no consensus about which of the proposed methods is most suitable. Anchor-based methods are conceptually more sound for determining MID, but they have some problems: transition scales are single-item measures, whereas HRQoL is a complex construct; the anchor choice may be inappropriate; minimal, moderate and large change levels are arbitrary, dependent on the patients’ subjectivity and lacking clinical control; and the assessment is based on a retrospective judgment, which may be defective. On the other hand, distribution-based methods do not directly explore the importance of change and the relationship between values obtained by distribution-based methods and clinical significance is not intuitive; some methods depend heavily on the sample’s characteristics (although this may be appropriate); and some (e.g. SEM) really determine the minimal detectable or real change rather than an important change [Crosby et al. 2003; de Vet et al. 2006; Martinez-Martin et al. 2008; Revicki et al. 2008; Turner et al. 2010].

Finally, the MID may be challenged in some aspects: is the same magnitude of effect valid for another sample? Is it valid for mild and severe patients? Is the change equally significant across the range of scores of a scale? Determining the MID for each sample and context through estimations from several methods (triangulation) to offer a range of values presumably containing the real MID value [Wyrwich et al. 2007; Revicki et al. 2008] has been proposed. Another approach is based on dichotomization of patients according to whether they reach or do not reach, for example, a certain score threshold (e.g. 65/100 points), a change by a certain score (e.g. increase of 3 points or more) or a percentage of score of the baseline (e.g. improve by 10% or more), or an improvement over the established MID. Using proportions, this data may be transformed into the number needed to treat (i.e. NNT = 1/proportion of patients achieving benefit) [Guyatt et al. 1998, 2002; Norman et al. 2001; Brozek et al. 2006; Schünemann et al. 2006], a well-understood concept that helps the interpretation of outcomes at a glance.

Results

Application of HRQoL measures to PD treatments

The discovery of the beneficial effects of treatment with levodopa [Barbeau, 1962; Cotzias et al. 1967] as a precursor and source of lacking dopamine in PD, started the era of rational medical treatment in the late 1960s. Since then, other strategies have been developed and currently PD is the only neurodegenerative disease with efficacious treatment since both medical and surgical therapies have demonstrated successful symptomatic control. However, treatment evaluation is complex as the spectrum of PD symptoms has broadened to include many nonmotor symptoms. Furthermore, other considerations such as social and professional limitations and treatment adverse events must be taken into account. When evaluating intricate treatment effects in clinical trials, a multidimensional construct such as HRQoL is useful, since it is a comprehensive outcome that informs about the final effect from the patient’s perspective. In this results section, the impact of currently available treatments in PD on HRQoL is reviewed. Special emphasis is given to appropriate reporting of data, highlighting reports that helped to understand the clinical significance of the intervention and therefore aided in making therapeutic decisions.

L-dopa

To the best of the authors’ knowledge there are no clinical trials evaluating the effect on HRQoL of standard levodopa (L-dopa) therapy versus placebo. This is probably because the beneficial effect of dopamine therapy was well established before the concept of HRQoL was developed in the late 1970s. There are data from four clinical trials testing controlled release versus standard L-dopa or transfer from standard to controlled-release formulations. According to the most recent review on the effect of medical and surgical interventions on HRQoL in PD [Martinez-Martin and Deuschl, 2007], these trials have shown significant improvement in HRQoL in favor of controlled release formulation, in total scores [Grandas et al. 1998; Martinez-Martin et al. 1998, 1999; Martinez-Martin and Koller, 1999], in all domains [Martinez-Martin, 1998; Martinez-Martin and Koller, 1999], or for some dimensions (home management, emotions, pain, and physical mobility) [Pahwa et al. 1997; Grandas et al. 1998; Martinez-Martin et al. 1999]. The studies in which the magnitude of effect on HRQoL was reported showed 25% improvement with a generic measure and 30% with a specific HRQoL scale. The authors conclude that there is level I and III evidence showing that conversion from standard L-dopa to controlled release-L-dopa is efficacious in improving HRQoL, even at long term.

Duodenal L-dopa infusion

The first clinical trial with HRQoL as an outcome testing the effect of a duodenal L-dopa infusion suggested that this intervention may be more efficacious than oral L-dopa to improve patient HRQoL in the short term [Nyholm et al. 2005]. The trial was randomized, controlled and used a disease-specific scale (PDQ-39) and a generic scale (15D) to measure HRQoL but only included 24 patients with 1.5 months follow up and is therefore considered level II evidence. PDQ-39 summary index (SI) scores at the end of the trial were lower for the infusion group and median scores of seven of the eight dimensions of the PDQ-39 were significantly decreased with infusion. No data on baseline PDQ SI scores were reported, therefore RC and ES could not be calculated.

A more recent prospective, open-label clinical trial evaluating the effect of L-dopa infusions provided more data regarding HRQoL which was established as a secondary outcome (nonmotor symptoms measured by the NMSS were the primary outcome) [Honig et al. 2009]. Based on their sample of 22 patients, this group reported mean baseline (44.2 ± 18.4) and follow up (20.7 ± 12.0) PDQ-8 scores, and calculated mean change in score of 23.4 (p = 0.0003), relative change = –53%, and ES = 1.3. This article is an example of good data reporting, reporting outcomes based on magnitude of change and distribution-based methods.

Dopamine agonists

The effect of dopamine agonists (DAs) on management of PD symptoms has been tested by a wide number of clinical trials. To the best of the authors’ knowledge, the first article reporting the effect of a DA on HRQoL applied generic measures (FSQ and EQ-5D) and was published in 1997 [Guttman, 1997]. Since then, other reported data have included HRQoL results, but according to the most recent review [Martinez-Martin and Kurtis, 2009] no study has included this construct as a primary outcome. Only one clinical trial [Noyes et al. 2006] reported a post hoc study specifically designed to analyze HRQoL. However, it was based on the extension of a study comparing the development of motor complications with pramipexole versus L-dopa as initial treatment [Parkinson Study Group, 2000; Holloway et al. 2004]. The 4-year extension study provided ambiguous [Holloway et al. 2004] and conflicting post hoc results [Noyes et al. 2006].

Using a procedure similar to that used by the Movement Disorders task force [Task Force of the Movement Disorder Society, 2002], this review classified DAs according to their efficacy in improving HRQoL. The authors conclude that there is level I evidence demonstrating that pramipexole, extended release ropinirole, rotigotine, cabergoline and pergolide are efficacious in improving HRQoL as add-on therapy to L-dopa in patients with wearing-off and motor fluctuations. The most widely studied agonist was pramipexole where there was level I evidence demonstrating that it is efficacious in improving HRQoL (SF-36), at 3 months follow up in PD patients with depression on stable L-dopa treatment, without motor fluctuations or dyskinesia [Barone et al. 2006]. Recently, a study on mild to moderate PD patients confirmed a modestly beneficial effect of pramipexole on depression [Barone et al. 2010].

However, as initial treatment, there is level I evidence showing that pramipexole is nonefficacious in improving HRQoL (EQ-5D, PDQUALIF) at 2 years follow up [Parkinson Study Group, 2000]. This is also the case for rotigotine, where a class I study demonstrated that this agonist is nonefficacious in improving HRQoL (as measured by the EQ-5D) of early PD patients not receiving levodopa or other DA, at 6 months follow up [Jankovic et al. 2007]. To date, there is insufficient evidence regarding subcutaneous apomorphine injections and infusions as to their effect on HRQoL.

One of the highest quality agonist trials was an evidence level I study which favored tolcapone as an add on to L-dopa therapy over pergolide in patients with fluctuations. The authors reported that at 3 months follow up, tolcapone produced significantly greater improvement than pergolide in HRQoL as measured by SIP (p = 0.005 RC = 22.6% versus 20.3%; ES = 0.32 versus 0.28) and PDQ-39 scores (p = 0.005; RC = 15.7 versus 10.3; ES = 0.59 versus 0.35) [Koller et al. 2001].

Other medical therapies

There are two randomized, prospective, double-blind, placebo-controlled clinical trials which show improvement of HRQoL with the addition of entacapone to L-dopa in nonfluctuating PD patients [Olanow et al. 2004; Fung et al. 2009]. The first study included 750 patients with 26 weeks of follow up and considered quality of life as a secondary outcome, measured by a disease-specific scale (PDQ-39) and two recommended generic scales (SF 36 and SIP). The second more recent study included 184 patients and investigated HRQoL with the PDQ-8. Both of them report change between baseline and final scores for each group, showing significant differences in favor of the patients on entacapone. However, this way of reporting data does not allow for calculations of relative change and effect size. A recent review that included medical treatment with catechol-O-methyltransferase (COMT) and monoamine transferase (MAO) inhibitors [Gallagher and Schrag, 2008] found that the addition of entacapone to patients with motor fluctuations is unclear [Fenelon et al. 2003; Reichmann et al. 2005]. Tolcapone (only tested in patients with motor fluctuations) resulted in significant improvement in HRQoL measured by generic scales (SIP) in two of four studies [Baas et al. 1997; Welsh et al. 2000]. These inconsistent results lead to uncertainty about the effect of tolcapone and entacapone as adjuncts to L-dopa therapy to improve patients HRQoL in PD patients with motor fluctuations. Rasagiline has improved HRQoL as monotherapy in early PD [Parkinson Study Group, 2002; Biglan et al. 2006] but not clearly in more advanced disease [Parkinson Study Group, 2005].

Surgery

Historically, the first study to use HRQoL measures in PD considered a surgical intervention [Wilson and Goetz, 1990]. Traditionally more surgery studies than medical studies have approached the issue of HRQoL. With respect to lesional surgery, most clinical data on quality of life considered unilateral pallidotomy. Published studies through December 2010, included two, high-quality, class Ib comparative studies, one contrasting this surgery with best medical therapy [de Bie et al. 1999] and another comparing it with bilateral subthalamic nucleus (STN) deep brain stimulation (DBS) [Esselink et al. 2004]. Considering these studies, including a total of 71 patients, followed for an average of 6 months, the average improvement in HRQoL using the PDQL for unilateral pallidotomy was around 16%.

As far as deep brain stimulation, most clinical trials considering HRQoL to date, regard bilateral STN DBS. Considering four class I studies on bilateral STN published to date [Esselink et al. 2004; Deuschl et al. 2006; Follet et al. 2010; Williams et al. 2010] with a total of 855 patients (mean: 214; 34–366) and mean follow up of 12 months, the average improvement in HRQoL using the PDQ-39 was 15.4%. These results are lower than those reported in reviews dating 4 or 5 years ago, where published trials on bilateral subthalamic DBS showed a HRQoL improvement of 60.5% in generic measures [Martinez Martin and Deuschl, 2007] and 33.8–34.5% in disease-specific scales [Kleiner-Fisman et al. 2006; Martinez Martin and Deuschl, 2007]. These reviews concluded that the intervention obtaining maximal improvement in global HRQoL was bilateral subthalamic DBS.

However, recent data challenge this traditionally established view, in favor of unilateral globus pallidum interna (GPI) DBS. In a recent large class I clinical trial considering bilateral GPI DBS, this intervention provided an average improvement in HRQoL that was similar to most recently reported results with bilateral STN DBS (11.2%, as measured by the PDQ-39, at 24 months follow up) [Follet et al. 2010]. However, in a class I study considering unilateral GPI DBS in contrast to unilateral STN (included 42 patients with a 6-month follow up), the average improvement in HRQoL as measured by the PDQ-39 after unilateral STN was 14.6%, with ES of 0.38, while results for unilateral pallidal DBS showed a magnitude of change of 38% and an ES of 0.88 [Zahodne et al. 2009]. This study is a good example of how to report data since baseline and follow-up scores were reported so that RC and ES could be easily calculated for both interventions.

Exercise interventions

In a recent systematic review and meta-analysis on the effectiveness of exercise for people with PD [Goodwin et al. 2008], four studies were found that included HRQoL as an outcome variable [Ellis et al. 2005; Schmitz-Hubsch et al. 2005; Burini et al. 2006; Ashburn et al. 2007]. Two of these studies were considered to be of high quality [Ellis et al. 2005; Burini et al. 2006]. Reported findings included 292 participants in total, across three quality of life outcomes, including the Sickness Impact Profile (SIP-68), the Parkinson’s disease questionnaire (PDQ-39), and the EuroQOL (EQ-5D). Only one reported a statistically significant benefit in favor of the exercise intervention group [Ashburn et al. 2007]. The authors extracted relevant data from the four papers and calculated the standardized mean difference on the ES (SMD = 0.27, 95% CI 0.04–0.51). Their results suggest that exercise interventions are likely to result in improvements in HRQoL, although more good quality studies are needed.

Discussion

HRQoL measures are currently essential for the evaluation of patients in clinical research. A very recent meta-analysis on HRQoL in later PD found that only three studies included patient-rated quality of life measures [Stowe et al. 2011]. These results highlight the need for including (generic and specific) HRQoL assessments in all clinical trials. For trials not focusing specifically on this construct, short versions or brief scales (e.g. EQ-5D and PDQ-8) may be enough, although more precise information will be needed when HRQoL is the main endpoint (e.g. SF-36 and PDQ-39). We would like to insist on the need to use HRQoL measurement as a primary outcome variable in adequately powered clinical trials [Martinez-Martin and Deuschl, 2007]. The lack of studies focusing mainly on HRQoL is paradoxical since this construct has been proposed as a comprehensive and rational outcome for evaluating the complex effects potentially related to PD therapy [Clarke and Guttman, 2002; Martinez-Martin and Kurtis, 2009].

This review emphasizes the need to standardize reporting of clinical trial outcomes regarding HRQoL, providing enough data (at least, mean ± SD at baseline and follow up) for the calculation of informative parameters of change, such as RC and ES. Many reports only give p values that merely provide statistical probability for rejection of the null hypothesis; they do not give information about the magnitude and clinical importance of change. A small sample size can impede reaching statistical significance whereas other outcome parameters are indicative of a potentially important change. For example, in a study with duodenal L-dopa infusion (n = 12) there was a –10 point difference from follow up to baseline in the PDQ-39 summary index (p = 0.076) [Isacson et al. 2008]. The difference was not ‘statistically significant’, but our calculations show that RC was −27.4% and the ES = 0.88, indicative of a potentially important effect. Inclusion in the protocol of a transition question allowing to assign the MID directly, details of scoring at baseline at follow up, and/or a priori settled thresholds for calculation of NNT are necessary for an appropriate interpretation of outcomes. Obviously, outcomes should be reported for all groups in the study (e.g. placebo and active treatment, surgical and best medical therapy, etc.), both within and between groups. A consensus on a common minimal body of explicit and standardized elements for reporting HRQoL outcomes, similar to the FDA recommendations, would be welcome [Food and Drug Administration, 2009].

Another relevant point is that some therapies have been tested in diverse situations (for example, as add-on therapy in advanced phases of disease and as initial treatment in early stages). This variability requires specific evaluation of intervention impact in each scenario, and obviously results have to be interpreted accordingly. In this sense, it seems necessary to determine the MID for each study, adjusted to the characteristics of the sample and circumstances of the study, rather than considering a fixed value. We propose obtaining MID values from several anchor- and distribution-based methods applied simultaneously, a strategy producing a range of values that should contain the true MID [Wyrwich et al. 2007; Revicki et al. 2008; Honig et al. 2009]. Determining the importance of change is a fundamental step for making decisions in clinical research, pharmacoeconomics, assignment of resources, and health policy.

The application of HRQoL scales in the investigation of PD treatment has demonstrated efficacy of certain interventions. Traditionally, surgical trials have included HRQoL as an outcome more often than clinical drug trials, a trend that is rapidly changing as HRQoL progressively gains attention from the scientific community. Summarizing the literature on medical therapy and HRQoL in PD: controlled release L-dopa has consistently shown improvement; DAs are efficacious as add-on therapy to L-dopa in patients with wearing-off and motor fluctuations; addition of entacapone to L-dopa in stable PD patients has demonstrated improvement; and rasagiline has demonstrated beneficial effects in monotherapy in early PD patients. There is insufficient evidence to conclude on the effects of COMT inhibitors on fluctuating PD patients or on exercise, although they are likely efficacious. In patients with advanced disease, there is level II evidence that intraduodenal infusion of L-dopa improves HRQoL, while there is insufficient evidence to conclude on the effect of apomorphine infusions. Regarding PD surgery, recent results looking specifically at HRQoL, challenge the traditional supremacy of bilateral STN target in favor of unilateral pallidal DBS [Zahodne et al. 2009], although both interventions are efficacious.

More clinical trials with HRQoL as a primary outcome and standardized reporting are necessary to further our understanding of the complexity of treatment effects in order to make evidence based clinical decisions regarding HRQoL in patients with PD.