Association between changes in quality of life and mortality in hemodialysis patients: results from the DOPPS

Abstract

Background: Cross-sectional health-related quality of life (HR-QOL) measures are associated with mortality in hemodialysis (HD) patients. The impact of changes in HR-QOL on outcomes remains unclear. We describe the association of prior changes in HR-QOL with subsequent mortality among HD patients.

Methods: A total of 13 784 patients in the Dialysis Outcomes and Practice Patterns Study had more than one measurement of HR-QOL. The impact of changes between two measurements of the physical (PCS) and mental (MCS) component summary scores of the SF-12 on mortality was estimated with Cox regression.

Results: Mean age was 62 years (standard deviation: 14 years); 59% were male and 32% diabetic. Median time between HR-QOL measurements was 12 months [interquartile range (IQR): 11, 14]. Median initial PCS and MCS scores were 37.5 (IQR: 29.4, 46.2) and 46.4 (IQR: 37.2, 54.9); median changes in PCS and MCS scores were −0.2 (IQR: −5.5, 4.7) and −0.1 (IQR: −6.8, 5.9), respectively. The adjusted hazard ratio (HR) for a 5-point decline in HR-QOL score was 1.09 [95% confidence interval (CI): 1.06–1.12] for PCS and 1.05 (95% CI: 1.03–1.08) for MCS. Adjusting for the second QOL score, the change was not associated with mortality: HR = 1.01 (95% CI: 0.98–1.05) for delta PCS and 1.01 (95% CI: 0.98–1.03) for delta MCS. Categorizing the first and second scores as predictors, only the second PCS or MCS score was associated with mortality.

Conclusions: In our study, only the most recent HR-QOL score was associated with mortality. Hence, the predictive power of a measurement of HR-QOL is not affected by changes in HR-QOL prior to that measurement; more frequent HR-QOL measurements are needed to improve the prediction of outcomes in HD. Further studies are needed to determine the optimal frequency and appropriate instrument to be used for serial measurements.

INTRODUCTION

Patients with end-stage renal disease (ESRD) treated with maintenance hemodialysis (HD) suffer from disproportionately higher rates of depression, impairments in sleep quality, nutrition and sexual function, and reduced vocational abilities [1–8]. Consequently, cross-sectional, validated measures of health-related quality of life (HR-QOL) among HD patients have consistently demonstrated dramatically reduced QOL compared with the general population and with other chronic medical conditions [9–13]. Furthermore, cross-sectional evaluation of HR-QOL among ESRD patients has also consistently demonstrated an independent association between reduced HR-QOL and adverse events, particularly hospitalization and mortality [10–13].

The Kidney Disease Outcomes Quality Initiative (KDOQI) guidelines emphasize the importance of HR-QOL as a key outcome and recommend repeated assessment of HR-QOL measures to serve as a parameter for monitoring quality of care delivered to maintenance dialysis patients [14]. The KDOQI recommendations highlight the central role of patient-reported outcomes, including baseline measurement and changes in HR-QOL, in the evaluation of the patients' well-being, as well as the effectiveness of novel therapies and interventions. It is therefore important for providers and policymakers to have an understanding of the natural history of changes in QOL among maintenance HD patients. Such an understanding would include: how changes in HR-QOL affect outcomes compared with those derived from cross-sectional assessment, what proportion of HD patients experience either improvement or decline in HR-QOL and how these changes affect long-term clinical outcomes. We hypothesized that a decline in HR-QOL would be positively associated with mortality, controlling for confounders. In the present study of the international Dialysis Outcomes and Practice Patterns Study (DOPPS) cohort, our primary aim was to relate longitudinal changes in HR-QOL among HD patients with mortality and QOL.

MATERIALS AND METHODS

Data source

The DOPPS is an international prospective cohort study of in-center HD patients aged ≥18 years. The patients were randomly selected from a representative sample of dialysis facilities in each country [15–17]. In this analysis, data from participants in DOPPS Phase 1 (1996–2001), Phase 2 (2002–2004), Phase 3 (2005–2008) and Phase 4 (2009–2011) in Australia, Belgium, Canada, France, Germany, Italy, New Zealand, Japan, Spain, Sweden, the UK and the USA were included. Data collection in Australia, Belgium, Canada, New Zealand and Sweden did not begin until Phase 2. Demographics, comorbid conditions and laboratory values were abstracted from medical records using uniform and standardized data collection tools in all countries. Mortality data were collected during study follow-up.

The DOPPS self-administered patient questionnaire (PQ) was typically completed shortly after study entry [median: 1.4 months, interquartile range (IQR): 0.7, 2.7] and was administered annually thereafter, approximately three times per phase. The PQ contained the Kidney Disease Quality of Life (KDQOL-36) survey. The scores of the physical component summary (PCS) and mental component summary (MCS) were determined using the patients' response to questions from the SF-12, a subset of the KDQOL-36 [18]. Among ESRD patients, the SF-12 has been validated against the SF-36, which has been widely studied in the ESRD population [19–21]. Higher PCS and MCS scores indicate better QOL. Data from one question on the DOPPS Phase 4 Unit Practice Survey (UPS) were used for descriptive analyses: ‘Is patient quality of life routinely (at least once yearly) assessed?’ The question was completed in 321 facilities by the unit's social worker or the facility nurse manager. Study approval and patient consent were obtained as required by national and local ethics committee regulations.

Statistical analysis

Patient characteristics were summarized descriptively by the number of PQs completed by the patient. Among patients who completed more than one PQ, boxplots were used to summarize absolute PCS and MCS scores (at the time of the first PQ) by region; within-patient changes in HR-QOL were summarized overall and by vintage (<1 versus >1 year).

Cox regression was used to assess the association between the change in HR-QOL and mortality. Models were stratified by country and phase, accounted for facility clustering using robust sandwich covariance estimators and adjusted for case mix. Two sets of models were used to evaluate the association between change in HR-QOL and mortality: (i) PCS score was categorized into three groups (<30, 30–45 and >45) at both the first PQ (PQ1) and second PQ (PQ2). Categories were chosen to include ∼50% of patients in the intermediate (30–45) group. The nine distinct categories with a single reference group, defined as patients who had a PCS of 30–45 at both time points, provided the opportunity to examine the association between mortality and various patterns of change in PCS over time. (ii) Change in PCS, defined as PCS at PQ2 minus PCS at PQ1, was analyzed as a continuous variable and also categorized into five groups: decreased by >10, decreased by 5–10, no change ± 5, increased by 5–10, increased by >10 based on both the clinical significance of a change (at least ±3) and the observed distribution [21]. We examined the relationship between mortality and change in PCS, both without and with adjustment for PCS at PQ2. The latter model tests whether the trajectory of PCS is associated with mortality after accounting for the most recent value. To further understand the relationship between PCS and mortality, we included PCS as a continuous variable at each time point in the model, with and without adjustment for PCS at the other time point. For both sets of models, analyses were repeated using MCS in place of PCS score.

Because our cohort of interest only included those patients with two assessments of HR-QOL, time at risk began at PQ2. Time at risk ended at the time of death, 7 days after leaving the facility due to transfer or change in renal replacement therapy modality, loss to follow-up, transplantation, end of study phase or the most recent date of data availability (whichever event occurred first). Median follow-up time was 11.0 months (IQR: 6.4, 17.7) from PQ2. Model adjustments for vascular access and comorbid conditions were based on baseline data. Other covariates included in the models were age, vintage, body mass index (BMI), serum albumin and hemoglobin—all measured at the start of follow-up.

Missing covariate values were multiply imputed using the chained equation method [22] by IVEWARE [23]. Missing values were sequentially updated using the bootstrap or the Markov Chain Monte Carlo method, based on multiple regression models with other variables as covariates. This procedure was carried out for 10 cycles, thereby constructing an imputed data set. Results from 20 such imputed data sets were combined for the final analysis using Rubin's formula [24]. The proportion of missing data was above 1% for the following covariates included in models: albumin (18%), BMI (15%), hemoglobin (10%), vascular access (4%) and vintage (1%). All analyses used SAS software, version 9.3 (SAS Institute, Cary, NC, USA).

RESULTS

Study population

HR-QOL data available on at least two PQs, a minimum of 30 days apart, were available for 13 784 patients. For patients with more than two PQs, only the first two were considered for analysis. Median time between the first two PQs was 12.5 months (IQR: 11.2, 14.3). There were 17 169 patients with only one PQ completed during follow-up; among these patients, 23% died prior to completing a second PQ, 52% were censored due to study end or loss to follow-up and 24% remained in the study for a minimum of 13 months but did not complete a second PQ. There were an additional 11 019 patients who did not complete any PQs during follow-up. Patient characteristics varied depending on the number of PQs completed by the patient (Table 1). Patients completing more PQs were younger, had longer dialysis vintage, were less likely to dialyze with a catheter, had higher albumin and hemoglobin, and had generally fewer comorbid conditions.

Table 1.

Patient characteristics by the number of PQs completed

Characteristic (mean ± SD or %)	Number of PQs completed
	0 PQs	1 PQ	≥2 PQs
Number of patients, n	11 019	17 169	13 784
Age (years)	63.9 ± 15.7	62.8 ± 15.1	62.0 ± 14.1
Gender (% male)	57	59	59
Black race (%)	18	11	7
BMI (kg/m2)	25.2 ± 6.0	25.2 ± 5.9	24.4 ± 5.6
Vintage (years)	2.6 ± 4.2	3.9 ± 5.1	6.4 ± 6.3
Catheter use (%)	41	27	17
Albumin (g/dL)	3.5 ± 0.6	3.7 ± 0.5	3.8 ± 0.4
nPCR (g/kg/day)	0.96 ± 0.27	1.00 ± 0.25	1.04 ± 0.23
Phosphorus (mg/dL)	5.4 ± 1.9	5.4 ± 1.8	5.4 ± 1.6
Hemoglobin (g/dL)	10.8 ± 1.7	11.0 ± 1.6	11.2 ± 1.5
Comorbid conditions (%)
Coronary artery disease	50	44	37
Cancer (non-skin)	13	13	11
Other cardiovascular disease	36	34	31
Cerebrovascular disease	22	17	13
Congestive heart failure	42	33	25
Diabetes	48	40	32
Gastrointestinal bleeding	8	6	5
Hypertension	83	82	79
Lung disease	15	12	9
Neurologic disease	17	10	7
Psychiatric disorder	25	19	14
Peripheral vascular disease	32	27	22
Recurrent cellulitis	12	9	6

For 0 PQ patients, all values were collected at baseline. For patients with ≥1 PQ, age, BMI, vintage and labs were updated at the time of the most recent PQ; others were collected at baseline.

PQ, patient questionnaire; SD, standard deviation; BMI, body mass index.

HR-QOL: descriptive statistics

Mean (± standard deviation) PCS score at PQ1 was much higher in Japan (43.4 ± 9.1) than Europe–Australia/New Zealand (35.6 ± 10.4) and North America (35.0 ± 10.2) (Figure 1). MCS was more comparable across regions with North America having the highest mean score (47.4 ± 11.3), followed by Japan (45.8 ± 10.1) and Europe (44.8 ± 12.0). Mean ± standard deviation change in PCS was −0.4 ± 8.8 and change in MCS was −0.4 ± 11.1. There was slightly more variability in change in MCS than change in PCS (Figure 2). Mean change in HR-QOL varied minimally by region; mean delta PCS and mean delta MCS were −0.5 and −0.2 in Japan, −0.2 and −0.6 in Europe–Australia/New Zealand, and −0.5 and −0.2 in North America, respectively.

FIGURE 1.

Distribution of PCS and MCS by region. Higher scores indicate better QOL. Diamond symbol represents the mean. PCS and MCS, physical and mental component summary of the SF-12; HR-QOL, health-related quality of life; A/NZ, Australia/New Zealand.

FIGURE 2.

Distribution of change in PCS and MCS, overall and by vintage. PCS and MCS, physical and mental component summary of the SF-12; HR-QOL, health-related quality of life.

Change in HR-QOL: association with mortality

Approximately 11% of patients (1478/13 784) died during follow-up. Table 2 shows the association between mortality and various patterns of change in PCS over time. When compared with a common reference group, patients with a low PCS (<30) at PQ2 had similarly high hazard ratios (HRs) regardless of whether the patient had a low PCS (<30) at PQ1 [HR = 1.53, 95% confidence interval (CI): 1.30–1.79], a slightly higher PCS (30–45) at PQ1 (HR = 1.46, 95% CI: 1.22–1.74) or a much higher PCS (>45) at PQ1 (HR = 1.54, 95% CI: 1.08–2.21). This demonstrates that the trajectory toward a lower PCS score from the previous year—whether the patient had consistently low HR-QOL or a sharp decline in HR-QOL—was not predictive of mortality. We observed a consistently strong trend toward higher mortality with lower PCS score at PQ2 across each of the three columns of PQ1, indicating little evidence of an interaction effect. Results were qualitatively consistent when analyzing MCS score (Table 3). Only a small proportion of patients (<3%) switched from either the highest to lowest, or lowest to highest category of PCS or MCS, as illustrated by the wide confidence intervals.

Table 2.

Association between mortality and PCS from PQ1 and PQ2

		PCS at PQ1
		<30	30–45	>45
	<30	1.53 (1.30–1.79)	1.46 (1.22–1.74)	1.54 (1.08–2.21)
		2326 (17%)	1442 (10%)	180 (1%)
PCS at PQ2	30–45	1.20 (1.00–1.45)	1 (Ref.)	1.03 (0.81–1.32)
		1224 (9%)	3656 (27%)	1213 (9%)
	>45	0.86 (0.49–1.52)	0.50 (0.36–0.68)	0.67 (0.52–0.86)
		164 (1%)	1127 (8%)	2452 (18%)

A total of 13 784 patients and 1478 deaths (11%). HRs (95% CI) of death compared with a single reference group are shown; n patients in each cell shown along with % of total study population.

Higher scores indicate better QOL. Follow-up began after PQ2. Model stratified by phase and country, adjusted for gender, black race, 13 comorbidities, baseline vascular access, age and vintage at PQ2, and BMI, albumin and hemoglobin at both PQ1 and PQ2. PQ1, first patient questionnaire; PQ2, second patient questionnaire ∼1 year later; PCS, physical component summary of the SF-12; HR, hazard ratio; CI, confidence interval; BMI, body mass index; QOL, quality of life.

Table 3.

Association between mortality and MCS from PQ1 and PQ2

		MCS at PQ1
		<35	35–55	>55
	<35	1.45 (1.22–1.73)	1.42 (1.20–1.68)	1.58 (1.09–2.29)
		1292 (9%)	1281 (9%)	197 (1%)
MCS at PQ2	35–55	1.03 (0.84–1.26)	1 (Ref.)	1.05 (0.88–1.25)
		1174 (9%)	5021 (36%)	1458 (11%)
	>55	0.97 (0.62–1.52)	0.88 (0.72–1.07)	0.75 (0.61–0.91)
		183 (1%)	1426 (10%)	1752 (13%)

A total of 13 784 patients and 1478 deaths (11%). HRs (95% CI) of death compared with single reference group shown; n patients in each cell shown along with % of total study population.

Higher scores indicate better QOL. Follow-up began after PQ2. Model stratified by phase and country, adjusted for gender, black race, 13 comorbidities, baseline vascular access, age and vintage at PQ2, and BMI, albumin and hemoglobin at both PQ1 and PQ2. PQ1, first patient questionnaire; PQ2, second patient questionnaire ∼1 year later; MCS, mental component summary of the SF-12; HR, hazard ratio; CI, confidence interval; QOL, quality of life; BMI, body mass index.

When analyzing the absolute change in HR-QOL from PQ1 to PQ2, we observed that negative change in PCS was strongly associated with mortality: HR = 1.09 per 5-point decrease in PCS (P < 0.0001) and HR = 1.05 per 5-point decrease in MCS (P < 0.0001). However, after adjustment for HR-QOL score at PQ2, these associations were markedly attenuated: HR = 1.01 for PCS (P = 0.41) and HR = 1.01 for MCS (P = 0.64). In treating change in PCS and MCS scores as categorical variables in Figure 3, we observed that the type of trajectory toward a lower HR-QOL score from the previous year was not predictive of mortality. We also included both the first and second scores as predictors, and only the second PCS or MCS score predicted mortality (data are not shown).

FIGURE 3.

Association between mortality and change in HR-QOL. A total of 13 784 patients and 1478 deaths (11%). Follow-up began after PQ2. Two separate models stratified by phase and country, adjusted for gender, black race, 13 comorbidities, baseline vascular access, age and vintage at PQ2, and BMI, albumin and hemoglobin at both PQ1 and PQ2. P-values correspond to testing change in HR-QOL as a continuous variable. PQ1, first patient questionnaire; PQ2, second patient questionnaire ∼1 year later; PCS and MCS, physical and mental component summary of the SF-12; HR-QOL, health-related quality of life; CI, confidence interval; BMI, body mass index.

Facility frequency of HR-QOL measurement

From a survey of 321 DOPPS facilities in Phase 4 (2009–2011), we found that almost all facilities in the USA (97%) measure HR-QOL in patients at least once per year, and 13% of facilities measure HR-QOL at least once every 6 months (Figure 4). In the other DOPPS countries, at least half of the facilities responding to the UPS do not measure HR-QOL routinely in their patients. Facilities in Japan were least likely to routinely measure HR-QOL (11%).

FIGURE 4.

Frequency of facility HR-QOL measurement by country. A total of 321 facilities responding to the DOPPS Phase 4 (2009–2011) UPS question: Is patient QOL routinely (at least once yearly) assessed? HR-QOL, health-related quality of life; DOPPS, Dialysis Outcomes and Practice Patterns Study; UPS, Unit Practice Survey; US, United States; A/NZ, Australia and New Zealand; Can, Canada; Spa, Spain; Bel, Belgium; Ger, Germany; Ita, Italy; UK, United Kingdom; Swe, Sweden; Fra, France; Jpn, Japan.

DISCUSSION

In this sample of over 13 000 international HD patients with at least two sequential measures of HR-QOL, we observed that the most recent HR-QOL score was strongly associated with mortality, while changes in HR-QOL score from 1 year prior had no additional predictive value.

Previous studies demonstrating an association between HR-QOL and mortality in maintenance HD cohorts largely employed the use of cross-sectional measures of HR-QOL [10–13]. In a prior study using DOPPS data, Mapes et al. found a strong inverse association between MCS and PCS scores and mortality and hospitalization [10]. With respect to international variation in HR-QOL, and consistent with the present findings, Mapes et al. found higher PCS scores among Japanese HD patients than among participants from other countries, including the USA, whereas MCS scores were highest among North American patients [10].

The present study extends these findings by examining longitudinal measures of HR-QOL and the impact that changes in MCS and PCS scores had on mortality. We found a time-dependent relation between HR-QOL and mortality: the most recent measures of HR-QOL were most predictive of mortality, controlling for prior measures or changes from prior measures of HR-QOL. The relationship between HR-QOL scores and mortality is similar to the one between depression scores and mortality demonstrated by Kimmel et al. using the Beck Depression Inventory; depression was associated with mortality only when considered as a time-varying covariate [25]. These findings suggest that HR-QOL scores may be tightly linked to acute events and deteriorations in clinical condition. These findings would suggest that if HR-QOL is to be incorporated as a predictive tool in routine clinical practice, more frequent and routine assessments of HR-QOL might be necessary to identify a group of at-risk patients.

Our findings are particularly salient given that we also found that among the majority of facilities in the present study, routine measurements of HR-QOL are not performed (Figure 4). The exception is in the USA where 97% of facilities reported performing an annual assessment of HR-QOL, while a minority of facilities (13%) performs even more frequent measures of HR-QOL. This likely reflects the fact that as of 1 April 2008, the Centers for Medicare & Medicaid Services (CMS) mandated that an annual measurement of HR-QOL be performed on the majority of dialysis patients as 1 of 26 clinical performance measures, which were introduced to monitor the quality of care being delivered to ESRD patients in the USA (http://www.cms.hhs.gov/cpmproject). The present study would support and endorse the need for routine serial assessments of HR-QOL to maximize its prognostic significance.

While CMS mandates annual measures of HR-QOL, what remains unclear is whether or not a formal assessment of HR-QOL needs to be performed more frequently than annually. Indeed, we found that HR-QOL measures more than 1 year prior to the most recent measurement were not predictive of mortality, after accounting for the most recent measurement, suggesting that more frequent measurements (at least annually or more frequently) may be able to further delineate and identify high-risk patients. A recent study demonstrated that the functional status of HD patients remained stable up until 1 month prior to death, further bolstering the need for more frequent self-reported measures of physical function to identify at-risk patients [26]. Since economic and practical considerations, including patient adherence and burnout in completing lengthy and frequent assessments of HR-QOL, may be limiting more frequent administration of the KDQOL-SF, more frequent and abbreviated assessment tools may be needed. A previous analysis of DOPPS data from Phase 1 used two simple questions from the Medical Outcomes Study SF-36 Health survey to assess symptoms of depression and found that lower scores on these two questions alone were associated with adverse outcomes [27]. Moreover, both an analysis based on data collected in the USA and a recent analysis of a Dutch database (NECOSAD) revealed a similar performance of the shorter SF-12 compared with the SF-36 in predicting the risk of death and adverse events in a cohort of ESRD patients [17, 19]. Taken together, further research is needed to assess the optimal frequency with which to measure HR-QOL and to validate additional abbreviated measures of HR-QOL.

The mean MCS (45.7 ± 11.4) and PCS (37.8 ± 10.7) scores observed are comparable to previous studies in dialysis patients, and are much lower compared with the general population [28, 29]. In our cohort, we observed no difference, on average, in HR-QOL scores compared with 1 year prior. However, because a substantial number of patients died or were lost to follow-up before completing a second HR-QOL assessment, we would expect that HD patients' condition deteriorated to a greater degree than observed in our study, but this remains unknown. Given these sobering findings, efforts need to focus on strategies to improve the HR-QOL among these patients. Interventions that have shown promise among the ESRD population, although controversial, include the treatment of depression, the use of structured exercise programs, the management of anemia and the use of alternative HD schedules including short daily and nocturnal HD delivered either in-center or within the home [30–37]. Other strategies including the role of increasing social support; addressing symptoms such as assessment and treatment of sleep disturbances, pruritus, pain, stress and anxiety; and sexual dysfunction likely also play important roles but have not been rigorously examined [38–40]. Improving HR-QOL among ESRD patients will require a multifaceted approach—one that engages a wider net of patient resources including the support and engagement of the entire multidisciplinary renal care team including nurses, pharmacists and social workers [39]. In this regard, formal assessments of HR-QOL such as those employed in the present study may only serve as a screen, and declining scores need to be followed-through with more detailed, directed assessments that are culturally sensitive, engage patients, their families and their caregivers, and allow for a root-cause analysis of declining HR-QOL.

There are several strengths and limitations of the present study. This is the largest study to date to assess longitudinal measurements of HR-QOL among an internationally representative cohort of HD patients, employing extensive case-mix adjustment of comorbidities, biochemistry and treatment-related variables. While the association between a single HR-QOL measurement and mortality is well established [10], this study aims to assess whether the change in HR-QOL score from 1 year prior has any additional predictive value; this study does not prospectively evaluate or predict patients' change in HR-QOL. Nonetheless, there are some limitations. The nonresponse rate to one assessment of HR-QOL was substantial in this study and even more so for two longitudinal measures of HR-QOL. We were only able to assess HR-QOL once per year, on average, and were thus unable to assess the impact of any adverse events that may have caused HR-QOL to fluctuate in the interim. To address our question of interest, our study population was restricted to patients who had a HR-QOL assessment at the beginning of follow-up (PQ2) and 1 year prior (PQ1). Patients who did not complete any PQs or patients who only completed one PQ were thus not eligible for the study. Requiring an HR-QOL score from 1 year prior limits generalizability of our results to patients who (i) have been on dialysis at least 1 year and (ii) were healthy enough and willing to complete a PQ at both time points.

We identified a strong association between HR-QOL and survival, yet causal inference of this relationship cannot be verified. It is possible that impaired HR-QOL may itself increase the risk of death. Equally plausible is that the relative importance of a recent HR-QOL measure in predicting death may be a direct consequence of recent and/or acute changes in health status and/or other clinical changes events, which may be taking place in the immediate premorbid period. Indeed, in the present study, HR-QOL scores were more predictive of more short-term (3-month) versus longer-term (1-year) mortality (Supplementary data, Table S1). Further studies are required to better characterize the nature of the association between HR-QOL and mortality.

Notwithstanding these limitations, the present study demonstrates that repeated measurements of HR-QOL might be important for healthcare providers to utilize in the routine care of dialysis patients. Earlier recognition of low HR-QOL may allow for a simplified method of identification of potentially vulnerable patients who are at an increased risk of death. However, independent of its strong association with death, perhaps the most compelling reason to frequently measure HR-QOL is due to the inherit value of HR-QOL itself as a pivotal outcome measure in ESRD care. The challenge is how to address the patient whose HR-QOL scores are low. Careful assessments of those factors responsible for the low HR-QOL scores need to then be elucidated and appropriate interventions discussed with the patient. Whether or not implementation of the subsequently selected interventions can lead to improvements in both HR-QOL and survival needs to be tested via rigorous future studies.

SUPPLEMENTARY MATERIAL

Supplementary data are available online at http://ndt.oxfordjournals.org.