1 Center for Clinical and Genetic Economics, Duke Clinical Research Institute, Duke University Medical Center, Durham, NC, USA; 2 Centre Hospitalier de l'Université de Montréal, Hôpital Notre-Dame, Montréal, Québec, Canada; 3 Novartis Pharmaceuticals Corporation, East Hanover, NJ, USA
* Correspondence to: Dr K. P. Weinfurt, Center for Clinical and Genetic Economics, Duke Clinical Research Institute, PO Box 17969, Durham, NC 27715, USA. Tel: +1-919-668-8101; Fax: +1-919-668-7124; Email: kevin.weinfurt{at}duke.edu
![]() |
Abstract |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Patients and methods: Data were from a clinical trial of zoledronic acid versus placebo in the treatment of SREs associated with advanced prostate cancer metastatic to bone. Patients (n=248) were included if they experienced an SRE during the study. Outcome measures were assessed at fixed intervals. We used mixed-effects models to estimate changes in outcomes after each patient's first SRE.
Results: There were clinically meaningful and statistically significant declines in physical well-being after: radiation and pathologic fractures; functional well-being after radiation; and emotional well-being after radiation and pathologic fractures. There also were meaningful and significant declines in preference and utility scores after radiation and fracture. Pain intensity declined after radiation, but not after other SREs; no other pain measure changed substantively.
Conclusions: SREs have important and significant effects on measures of health-related quality of life in men with prostate cancer. Treatments that prevent SREs may not demonstrate corresponding effects on outcomes if the effects of SREs occur between scheduled outcome assessments. Implications for trial design are discussed.
Key words: antineoplastic agents, bone neoplasms, diphosphonates, prostatic neoplasms, quality of life, research design
![]() |
Introduction |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Despite the findings by Saad et al. [4], concerns have been raised about the clinical significance of SREs, because the investigators found no corresponding treatment differences in health-related quality of life (HRQOL) [5
, 6
]. This lack of correspondence between differences in clinical events and patient-reported outcomes may have resulted from a lack of clinical effect of the therapy on those outcomes. Alternatively, patients who experienced SREs might have experienced the predicted declines in outcomes, but these results were not observed in the primary analysis of the intent-to-treat population. Although patient-reported outcomes were included in the trial [4
], the investigators did not attempt to determine how the outcomes may have been related to SREs.
Clinical experience suggests a potentially significant effect of SREs on HRQOL [7, 8
]. SREs may affect patients general physical well-being and ability to perform basic functions of daily living. The extent to which SREs affect other domains of HRQOL is less clear. Using data from the trial reported by Saad et al. [4
], we sought to characterize the clinical relevance of SREs in terms of their impact on patients trajectories of HRQOL, pain and health state preferences.
![]() |
Methods |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
Demographic and clinical variables were collected at randomization. Occurrence of SREsdefined as radiation to bone, pathologic fracture (vertebral or nonvertebral), spinal cord compression, surgery to bone or change in antineoplastic therapy to treat bone painwas recorded at each clinic visit. For the analysis, we condensed these types into three categories (i) radiation to bone, (ii) pathologic fractures and (iii) other SREs. Date of disease progression was recorded based on clinical observation of one of the following during a scheduled trial visit: (i) marked increase in serum prostate-specific antigen level; (ii) new and/or progression of existing osteoblastic or osteolytic skeletal lesions; or (iii) new and/or progression of existing nonskeletal malignant lesions.
Quality-of-life measures
HRQOL was measured during the clinical trial with the Functional Assessment of Cancer TherapyGeneral (FACT-G), the Brief Pain Inventory (BPI) and the EuroQol instrument. Assessments were scheduled every 90 days, so we coded the first assessment made 100 days after an SRE as the post-SRE assessment. Patients without an assessment during that interval were considered to have missing post-SRE assessments. All measures were administered before the patient was interviewed by the physician or received study medication.
The FACT-G is a 27-item questionnaire that addresses physical well-being, social/family well-being, emotional well-being and functional well-being, as well as overall HRQOL using a total score [9]. It has demonstrated acceptable psychometric reliability, validity and sensitivity to change in populations of mixed cancer types [9
, 10
]. Patients completed the questionnaire at baseline and every third month up to month 24.
To measure pain intensity and the interference of pain with daily activities, we used the pain intensity and interference items from the BPI [1114
]. Patients completed the questionnaire at baseline, months 1 and 2, and every other month thereafter up to month 24.
The EuroQol instrument [15], composed of a five-item health state profile and a visual analog scale (VAS), allows for assessment of health state preferences from the patient's (VAS) and society's (profile) perspectives. Health state preferences refer to the desirability or worth of the health state. The VAS asks the patient to rate his current health state on a scale from 0 (worst imaginable health state) to 100 (best imaginable health state). The five-question profile responses are transformed to utilities using published community weights to reflect the societal perspective [16
]. The EuroQol was administered at baseline and every third month thereafter up to month 24.
Statistical analysis
We computed descriptive statistics at baseline for all clinical and demographic characteristics and for the FACT-G and BPI. We estimated the effect of each SRE category on each measure of patient-reported outcomes separately. For each SRE category, only patients who experienced that category as the first SRE were included. Nine patients experienced more than one type of SRE as the first SRE and so were excluded from the analysis. To determine the immediate impact of an SRE on patients HRQOL, we included all outcome assessments up to the date of the SRE of interest, as well as the first post-SRE assessment.
We used linear mixed-effects models to model each patient's longitudinal trajectory of patient-reported outcomes before the SRE of interest and to estimate how far the post-SRE value of HRQOL deviated from that trajectory [17]. All models included fixed effects of time (in months) and an SRE indicator variable denoting the post-SRE data point for each patient. We estimated random intercepts and slopes for time to allow the growth curves to vary across patients. The trajectory-adjusted mean change (TAMC) was defined as the mean deviation of the post-SRE value from the expected value based on patients pre-SRE trajectory. We used the PROC MIXED command in SAS (version 8.2, SAS Institute Inc, Cary, NC) to estimate all models, employing a likelihood-based approach to handling missing data points for patient-reported outcomes.
For purposes of comparison, we calculated the standardized effect size () by dividing each adjusted mean change by the corresponding overall standard deviation (SD) from the entire trial sample (n=643) at baseline (available from the authors). Thus,
translates the change in patient-reported outcome scores into SDs. Following Cohen's benchmarks [18
, 19
], we characterized effect sizes of 0.2, 0.5 and 0.8 as small, medium and large, respectively. The benchmark effect sizes correspond to changes in post-SRE functioning that move the average patient in the original trial from the 50th percentile to approximately the 40th, 30th and 20th percentiles, respectively.
We also conducted a sensitivity analysis to determine whether any observed changes in patient-reported outcomes after an SRE could be due to overall disease progression. We applied the modelling strategy described above to a refined sample that excluded patients whose date of disease progression was between the last pre-SRE assessment and the post-SRE assessment. If changes in outcomes after the SREs are due to disease progression rather than SREs, the TAMC estimates should be smaller for patients who did experience disease progression around the same time as their first SRE.
![]() |
Results |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
|
|
|
|
|
![]() |
Discussion |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
All categories of SREs affected at least some outcomes to a clinically significant extent. There were differences, however, among types of SREs with regard to which aspects of patients experiences were affected. Radiation to bone had the broadest impact (affecting four out of five FACT-G scores) and affected both measures of patient preferences. This finding highlights the pervasive negative effects of radiation therapy on HRQOL, even when radiation might alleviate bone pain. These negative effects might reflect side effects of radiotherapy, the repeated hospital visits required for radiotherapy, and the psychological effect of recognizing that the disease is severe enough to require radiation. Pathologic fractures were associated with changes in two of five FACT-G scores and in both measures of patient preferences. Finally, the effect sizes for other SREs, though not statistically significant (due to a markedly smaller sample), appear to reflect deficits across multiple domains of HRQOL as assessed by the FACT-G.
We found that patients suffered declines in emotional well-being after radiation to bone, pathologic fractures and perhaps other types of SREs. An SRE signifies a permanent downturn in the course of disease, and experiencing a fracture or requiring radiation indicates to patients that their situation is worsening. Feelings of depression and anxiety that result from the realization of deteriorating physical health may be the most important cause of the declining emotional well-being scores we observed.
The lack of large changes in either pain intensity or interference in functioning due to pain is understandable given the definitions of SREs used in this study and the design of the assessment schedule. With regard to the first issue, SREs that involved a significant increase in bone pain were identified by the treatment initiated to reduce the pain. The events themselves constitute evidence of increased pain from skeletal complications. For these events, if there is additional change detected, it should be a decrease in pain after the reported SRE. Thus, the BPI intensity score decreased slightly (=0.19) after initiation of radiation treatment. This explanation is consistent with results from a recent reanalysis [20
] of the pain data reported by Saad et al. [4
], in which significant differences in pain were observed between the zoledronic acid 4 mg and placebo groups. Specifically, the zoledronic acid group was 8.3% less likely than the placebo group to experience clinically significant pain, which fits well with the finding that patients in the zoledronic acid group were 11% less likely to experience an SRE [4
].
Issues related to study design and patient population may have led us to underestimate changes in outcomes caused by SREs. First, because outcome assessments were scheduled to occur every 90 days, it was necessary to define a broad post-SRE window in which to observe responses. A patient who suffered a moderate fracture may not have had an outcome assessment until 2 to 3 months after the event, by which time many of the fracture's effects might have subsided. The relatively brief recall period for the outcome measures (e.g. 7 days for the FACT-G) likely exacerbated this issue. Also, the effect of some SREs (i.e. spinal cord compression) is gradual and would, therefore, affect the outcome before the discrete event we defined as an SRE. Our analysis measured the effect as a deviation from the patient's pre-event trajectory, but the effect of the underlying skeletal morbidity might already be present in the pre-event trajectory.
Despite these limitations, we observed relatively large effects, with most in the range of 0.350.48 SDs. Our data indicate that when someone with an average patient-reported outcome score experiences an SRE, their post-SRE functioning puts them at about the 33rd percentile of the trial population, rather than the 50th percentile. Furthermore, the effects we report are for each patient's first SRE and do not reflect the cumulative effects of multiple SREs.
Our findings demonstrate that SREs lead to significant functional declines in patients daily lives. Therefore, any therapy that reduces the occurrence of SREs should also reduce the effects of SREs on HRQOL. However, Saad et al. [4] reported no such effect. Now that we have ascertained that outcome measures were generally sensitive to the effects of SREs, it is clear that other factors in the trial design were responsible for the absence of observed treatment differences. One aspect of the design that merits attention is the schedule of outcome assessments, which was similar to most other trials of new cancer therapies. That is, a fixed number of assessments were made at fixed study visits. This schedule seems optimal when the clinical course of the disease involves changes in health that result in relatively prolonged periods of the same health state or in gradual worsening or improvements in health. In such cases, sampling the patient's HRQOL at fixed time points is assumed to provide good estimates of patients underlying trajectories of health.
In this clinical trial [4], however, the underlying trajectories were presumably influenced by repeated acute events, the effects of which could be resolved prior to the next outcome assessment. Using fixed quarterly assessments, especially when the interval between visits is longer than the time it takes for the acute event to resolve, may not provide good estimates of the patient's underlying trajectory. To the extent that the assessment schedule misses the full impact of acute events, it is unlikely that a sensitive treatment group comparison can be made.
In contrast, it was possible to construct an accurate estimate of the clinical trajectory (i.e. the status of the skeletal system), because the exact dates of SREs were known. These data allowed for a sensitive test of treatment group differences in SREs. The structure of the assessment schedule helps to explain why treatment group differences were found for the occurrence of SREs but not for the outcomes we have shown to be sensitive to SREs. Investigators designing trials that involve repeated acute events should consider alternative measurement strategies for patient-reported outcomes, such as event-triggered data collection or more intensive, diary-type assessments.
![]() |
Acknowledgements |
---|
Received for publication September 9, 2004. Accepted for publication November 23, 2004.
![]() |
References |
---|
![]() ![]() ![]() ![]() ![]() ![]() ![]() |
---|
2. Carlin BI, Andriole GL. The natural history, skeletal complications, and management of bone metastases in patients with prostate carcinoma. Cancer 2000; 88: 29892994.[CrossRef][Medline]
3. Pelger RC, Soerdjbalie-Maikoe V, Hamdy NA. Strategies for management of prostate cancer-related bone pain. Drugs Aging 2001; 18: 899911.[ISI][Medline]
4. Saad F, Gleason DM, Murray R et al. A randomized, placebo-controlled trial of zoledronic acid in patients with hormone-refractory metastatic prostate carcinoma. J Natl Cancer Inst 2002; 94: 14581468.
5. Atkins CD. Correspondence Re: A randomized, placebo-controlled trial of zoledronic acid in patients with hormone-refractory metastatic prostate carcinoma. J Natl Cancer Inst 2003; 95: 332.
6. Rosenthal M. Correspondence Re: A randomized, placebo-controlled trial of zoledronic acid in patients with hormone-refractory metastatic prostate carcinoma. J Natl Cancer Inst 2003; 95: 332.
7. Litwin MS, Hays RD, Fink A et al. Quality-of-life outcomes in men treated for localized prostate cancer. J Am Med Assoc 1995; 273: 129135.[Abstract]
8. Villavicencio H. Quality of life of patients with advanced and metastatic prostatic carcinoma. Eur Urol 1993; 24(Suppl 2): 118121.
9. Cella DF, Tulsky DS, Gray G et al. The Functional Assessment of Cancer Therapy scale: development and validation of the general measure. J Clin Oncol 1993; 11: 570579.[Abstract]
10. Esper P, Mo F, Chodak G et al. Measuring quality of life in men with prostate cancer using the functional assessment of cancer therapy-prostate instrument. Urology 1997; 50: 920928.[CrossRef][ISI][Medline]
11. Cleeland CS, Ryan KM. Pain assessment: global use of the Brief Pain Inventory. Ann Acad Med Singapore 1994; 23: 129138.[Medline]
12. Janjan NA, Payne R, Gillis T et al. Presenting symptoms in patients referred to a multidisciplinary clinic for bone metastases. J Pain Symptom Manage 1998; 16: 171178.[CrossRef][ISI][Medline]
13. Jensen MP, Karoly P. Pain-specific beliefs, perceived symptom severity, and adjustment to chronic pain. Clin J Pain 1992; 8: 123130.[ISI][Medline]
14. Radbruch L, Loick G, Kiencke P et al. Validation of the German version of the Brief Pain Inventory. J Pain Symptom Manage 1999; 18: 180187.[CrossRef][ISI][Medline]
15. Euroqol Group. EuroQola new facility for the measurement of health-related quality of life. The EuroQol Group. Health Policy 1990; 16: 199208.[CrossRef][ISI][Medline]
16. Dolan P. Modeling valuations for EuroQol health states. Med Care 1997; 35: 10951108.[CrossRef][ISI][Medline]
17. Verbeke G, Molenberghs G. Linear Mixed Models for Longitudinal Data. New York: Springer-Verlag 2000.
18. Cohen J. Statistical Power Analysis for the Behavioral Sciences, 2nd edition. Hillsdale: Lawrence Erlbaum Assoc 1988.
19. Samsa G, Edelman D, Rothman ML et al. Determining clinically important differences in health status measures: a general approach with illustration to the Health Utilities Index Mark II. Pharmacoeconomics 1999; 15: 141155.[ISI][Medline]
20. Weinfurt KP, Anstrom KJ, Castel LD et al. Effect of zoledronic acid on clinically meaningful changes in pain associated with metastatic prostate cancer. J Clin Oncol 2004; 22 (Suppl): 4680.