Factors Associated with Variation in Utility Scores among Patients with Prostate Cancer
ABSTRACT
Objective: We aimed to assess the effects of age, comorbidity, and disease-specific functions on utility scores derived from three methods on prostate cancer.
Methods: A total of 330 Japanese prostate cancer patients were asked to answer self-administered questionnaires. Community-weighted utility scores were derived from the EuroQoL-5D (EQ-5D) and the Short Form-36 (SF-36), while the patient's directly elicited utility score was derived from time trade-off technique. Univariate and multivariate analyses were performed to examine the relation between covariates and utility scores. We assigned age, the Index of Co-existent Disease, and disease-specific functions including sexual, urinary, bowel, and hormonal function as covariates.
Results: Bowel and hormonal function were related to utility scores, while age and sexual function were not. Comorbidities were more closely related to utility scores derived from EQ-5D and SF-36.
Conclusions: These results contribute to an understanding of which factor has an impact on utility scores in patients with prostate cancer.
Introduction
Recently, early detection of clinically localized prostate cancer has been achieved by the prevalence of prostate-specific antigen (PSA) screening [1]. Patients newly diagnosed have to make decisions on treatment. By receiving any treatments, they may experience treatment complications including sexual, urinary, bowel, and hormonal dysfunctions. To enable a comparison of treatments taking quality into consideration in addition to survival, assigning utility scores derived from the general public or patients on each health state is necessary. Utility scores are numbers that represent the strength of an individual's preference for different health outcomes under conditions of uncertainty. We aim to evaluate the effects of age, comorbidity, and treatment complications on utility scores derived from the general public or patients before conducting decision analyses for localized prostate cancer in the PSA era.
Methods
Ethic Approval and Informed Consent
The research protocol was approved by the ethics review committees of Juntendo University, Juntendo Urayasu Hospital, and the International Medical Center. All participating patients gave informed consent to be interviewed for the study.
Participants
Between October 2004 and September 2005, prostate cancer outpatients were recruited at these three institutions in Japan. Patients receiving radical prostatectomy, external beam radiotherapy, brachytherapy, primary hormonal therapy, watchful waiting, or a combination of these for localized prostate cancer, and patients with hormone refractory prostate cancer participated in this study. Patients who had difficulty communicating verbally were excluded.
Measures
All patients were asked to answer the self-administered questionnaires. The EuroQoL-5D (EQ-5D) [2] and the Short Form-36 version 2.0 (SF-36) [3] were used to assess the generic quality of life (QOL). The UCLA Prostate Cancer Index (PCI) version 1.2 [4] which measured sexual function, urinary function, and bowel problem, the hormonal function domain of Expanded Prostate Cancer Index Composite (EPIC) [5], and the International Prostate Symptom Score [6] which measured lower urinary tract symptoms (LUTS), were used to assess disease-specific QOL. Patient age and the Index of Co-existent Disease (ICED) [7] were also asked in the questionnaires. ICED includes two subscales: 1) the severity of each set of categories of coexistent medical conditions, the Index of Disease Severity; and 2) the degree of physical impairment, and the Index of Physical Impairment. The final index is an ordinal scale in which the two subscales are combined to form four levels of severity. All questionnaires except EPIC were made through the process of translation, back-translation, refinement by the developer of the original, and establishment of the modified Japanese version. Because the Japanese version of EPIC was not developed, we translated hormonal function domain of EPIC from English to Japanese and examined its test–retest reliability in 21 Japanese patients with prostate cancer. The weighted kappa coefficient of each item ranged from 0.41 to 0.86.
The interview of time trade-off (TTO) technique using props was performed to directly derive utility scores. The props were specially designed boards and a sliding scale on the board showed the number of years in perfect health compared to t years in the poor health state [8].
For a current health state better than dead, a “10-year” TTO elicits the number of years, t (<10), where the respondent is indifferent to the following two prospects:
- •
to live in full health for t years;
- •
to live in the current health state for 10 years.
For a state worse than dead, it elicits the value of t (<10) where the respondent is indifferent to the two prospects:
- •
to live in the current health state for t years;
- •
then in full health for (10-t) years;
- •
immediate death.
The responses thus derived need to be adjusted so that they lie within the boundary of −1 and +1, with 0 equivalent to dead. Conventionally, this is represented by h = t/10 for states better than dead, and h = (t/10)−1 for states worse than dead, where t represents the obtained response and h represents the adjusted TTO value [9]. This study used 10 years as the reference duration, and 1 year as the smallest unit of measurement.
Time trade-off technique is the scaling method most widely used [10] because TTO using props causes few misunderstandings and the interview can be conducted within 5 min per person. We examined its test–retest reliability in 21 Japanese patients with prostate cancer. Intraclass correlation coefficient was 0.83 (95% CI, 0.68–0.92).
Community-weighted utility scores were derived from EQ-5D and SF-36. Indirectly elicited utility scores of EQ-5D were derived from the general public using TTO [2]. We adapted utility scores from SF-36 using the approach proposed by Brazier et al. [11].
Statistical Analyses
We assessed the relation between covariates and utility scores on prostate cancer. Age, ICED, and disease-specific functions were assigned as covariates. Pearson correlation coefficients were calculated to confirm multicollinearity in the disease-specific functions. After stratifying age, ICED, and disease-specific functions into quartiles or clinical classification, trend tests were performed to examine the linearity between these covariates and the utility scores. To assess these relations simultaneously, we applied the general linear model by the least-squares means method in consideration of the intrapatient correlation of utility scores. Statistical analyses were performed with SAS software. All statistical tests were two-sided.
Results
A total of 330 patients were recruited at three institutions. Seven patients were excluded because of poor interview quality. The remaining 323 patients were included in this study. Mean age was 71.5 (SD 6.0) years. Mean utility score derived from EQ-5D, SF-36, and TTO was 0.90 (SD 0.15), 0.74 (SD 0.08), and 0.89 (SD 0.15), respectively (Table 1). In the univariate analyses, all the variables except age and sexual function had a linear trend with TTO, EQ-5D, and SF-36 (ICED, P = 0.048, P < 0.001, and P < 0.001; urinary function, P = 0.013, P < 0.001, and P = 0.006; bowel problem, P < 0.001, P < 0.001, and P < 0.001; hormonal function, P < 0.001, P < 0.001, and P < 0.049; LUTS, P = 0.007, P < 0.001, and P < 0.002, respectively, Table 1). The absolute value of Pearson correlation coefficients for disease-specific function scores ranged from 0.034 to 0.323 (Table 2), and multicollinearity in the disease-specific functions was not confirmed. In the multivariate analyses using general linear model, bowel problems and hormonal function were related to the utility scores derived from TTO, EQ-5D, and SF-36 (bowel problems, P < 0.001, P = 0.015, and P = 0.007; hormonal function, P < 0.001, P < 0.001, and P = 0.075, respectively, Table 3), while age and sexual function were not related in TTO, EQ-5D, and SF-36 (age, P = 0.085, P = 0.631, and P = 0.721; sexual function, P = 0.526, P = 0.941, and P = 0.202, respectively, Table 3). Urinary function and LUTS were only related to those derived from EQ-5D (P = 0.005 and P < 0.001, respectively, Table 3). Comorbidity was more closely related to those derived from EQ-5D and SF-36 (P < 0.001 and P = 0.002, respectively, Table 3).
| Covariates | Group | n | TTO | EQ-5D | SF-36 | |||
|---|---|---|---|---|---|---|---|---|
| Utility score (SD) | P | Utility score (SD) | P | Utility score (SD) | P | |||
| All patients | 323 | 0.89 (0.15) | 0.90 (0.15) | 0.74 (0.08) | ||||
| Age | 51–59 | 7 | 0.81 (0.40) | 0.182 | 0.77 (0.22) | 0.212 | 0.72 (0.19) | 0.958 |
| 60–69 | 112 | 0.90 (0.13) | 0.92 (0.13) | 0.74 (0.07) | ||||
| 70–79 | 172 | 0.88 (0.15) | 0.91 (0.14) | 0.74 (0.06) | ||||
| 80–89 | 32 | 0.90 (0.12) | 0.85 (0.21) | 0.72 (0.12) | ||||
| ICED* | 0 | 60 | 0.90 (0.19) | 0.048 | 0.93 (0.14) | <0.001 | 0.75 (0.09) | <0.001 |
| 1 | 173 | 0.91 (0.12) | 0.93 (0.12) | 0.75 (0.06) | ||||
| 2 | 74 | 0.84 (0.17) | 0.85 (0.15) | 0.72 (0.06) | ||||
| 3 | 16 | 0.83 (0.21) | 0.71 (0.23) | 0.64 (0.14) | ||||
| Sexual function† | 16–85 | 80 | 0.93 (0.10) | 0.163 | 0.93 (0.11) | 0.128 | 0.76 (0.05) | 0.016 |
| 4–16 | 57 | 0.84 (0.19) | 0.90 (0.15) | 0.73 (0.07) | ||||
| 3 | 35 | 0.89 (0.17) | 0.89 (0.14) | 0.72 (0.07) | ||||
| 0 | 149 | 0.88 (0.16) | 0.90 (0.16) | 0.73 (0.09) | ||||
| Urinary function† | 100 | 170 | 0.91 (0.12) | 0.013 | 0.94 (0.11) | <0.001 | 0.75 (0.07) | 0.006 |
| 75–95 | 73 | 0.87 (0.18) | 0.88 (0.17) | 0.71 (0.10) | ||||
| 11–74 | 80 | 0.86 (0.18) | 0.84 (0.17) | 0.72 (0.09) | ||||
| Bowel problem† | 100 | 134 | 0.92 (0.11) | <0.001 | 0.94 (0.12) | <0.001 | 0.75 (0.07) | <0.001 |
| 93–95 | 47 | 0.89 (0.12) | 0.91 (0.15) | 0.73 (0.08) | ||||
| 81–92 | 62 | 0.92 (0.09) | 0.91 (0.13) | 0.74 (0.06) | ||||
| 0–80 | 80 | 0.81 (0.23) | 0.84 (0.18) | 0.71 (0.09) | ||||
| Hormonal function† | 100 | 155 | 0.92 (0.10) | <0.001 | 0.93 (0.14) | <0.001 | 0.74 (0.08) | 0.049 |
| 95 | 26 | 0.88 (0.17) | 0.91 (0.12) | 0.73 (0.06) | ||||
| 85–90 | 71 | 0.90 (0.14) | 0.92 (0.12) | 0.75 (0.06) | ||||
| 35–80 | 71 | 0.82 (0.23) | 0.84 (0.17) | 0.71 (0.09) | ||||
| LUTS‡ | 0–7 | 168 | 0.92 (0.10) | 0.007 | 0.93 (0.12) | <0.001 | 0.74 (0.06) | 0.002 |
| 8–19 | 116 | 0.86 (0.17) | 0.89 (0.16) | 0.74 (0.09) | ||||
| 20–35 | 38 | 0.84 (0.24) | 0.83 (0.19) | 0.70 (0.10) | ||||
- * ICED has four grades of severity from 0 (none) to 3 (severe).
- † Lower scores on sexual function, urinary function, bowel problem, and hormonal function indicate higher degrees of dysfunction.
- ‡ Lower score on LUTS indicates lower degrees of symptoms.
- EQ-5D, EuroQoL-5D; ICED, Index of Co-existent Disease; LUTS, lower urinary tract symptoms; SD, standard deviation; SF-36, Short Form-36; TTO, time trade-off.
| Methods | Covariates | Estimates | Standard error | P |
|---|---|---|---|---|
| TTO | Intercept | 0.3170 | 0.1350 | |
| Age | 0.0002 | 0.0015 | 0.885 | |
| ICED | −0.0036 | 0.0115 | 0.753 | |
| Sexual function | 0.0003 | 0.0005 | 0.526 | |
| Urinary function | 0.0005 | 0.0004 | 0.239 | |
| Bowel problem | 0.0027 | 0.0006 | <0.001 | |
| Hormonal function | 0.0032 | 0.0008 | <0.001 | |
| LUTS | −0.0015 | 0.0013 | 0.254 | |
| EQ-5D | Intercept | 0.5651 | 0.1261 | |
| Age | −0.0007 | 0.0014 | 0.631 | |
| ICED | −0.0390 | 0.0108 | <0.001 | |
| Sexual function | 0.0000 | 0.0005 | 0.941 | |
| Urinary function | 0.0012 | 0.0004 | 0.005 | |
| Bowel problem | 0.0014 | 0.0006 | 0.015 | |
| Hormonal function | 0.0025 | 0.0007 | <0.001 | |
| LUTS | −0.0024 | 0.0012 | 0.055 | |
| SF-36 | Intercept | 0.5644 | 0.0721 | |
| Age | 0.0003 | 0.0008 | 0.721 | |
| ICED | −0.0189 | 0.0061 | 0.002 | |
| Sexual function | 0.0004 | 0.0003 | 0.202 | |
| Urinary function | 0.0003 | 0.0002 | 0.162 | |
| Bowel problem | 0.0009 | 0.0003 | 0.007 | |
| Hormonal function | 0.0007 | 0.0004 | 0.074 | |
| LUTS | −0.0006 | 0.0007 | 0.418 |
- EQ-5D, EuroQoL-5D; ICED, Index of Co-existent Disease; LUTS, lower urinary tract symptoms; SF-36, Short Form-36; TTO, time trade-off.
Discussion
Evaluating the effects of disease-specific functions on utility scores is important because the disease-control benefits of curative therapy may be offset by treatment complications. This study also evaluates the effects of age and comorbidity on utility scores, because little has been published [12]. When discussing treatment options with prostate cancer patients, the effects of age, comorbidity, and treatment complications may be a central point in addition to clinical stage, pretreatment PSA, and Gleason score.
In three methods, although the utility scores derived from EQ-5D showed a relation to disease-specific function, the influences were much less than those from comorbidities. In conducting decision analysis using utility scores derived from EQ-5D, adjusting with comorbidity may be required.
The utility score from SF-36 was influenced by comorbidity, similar to EQ-5D, and was less related to the disease-specific function.
Our results showed that bowel problem and hormonal function were closely related to the utility score derived from TTO, while age and comorbidities were not correlated. These findings suggest that utility scores derived from Japanese men may differ from those elicited in foreign population. In the study of Stewart et al. [13], they reported the bowel problem was rated as significantly more bothersome than impotence and urinary incontinence, and age was reported as a significant predictor of higher utility ratings. Krahn et al. evaluated the disease-specific functions with UCLA PCI and utility scores with different methods in prostate cancer [14]. They reported that sexual, urinary, and bowel dysfunctions had less impact on utility scores than reported in previous studies when adjustments were made for concomitant changes in other QOL domains. Saigal et al. reported the utility score derived from TTO for current sexual function decreased by 0.13 and for current urinary function decreased by 0.09 at 6 months after undergoing treatment including radical prostatectomy, external beam radiotherapy, brachytherapy, or watchful waiting. [15] Nevertheless, they reported that the utility score for current bowel problems showed no change between pre- and post-treatment, despite the dropping of bowel problem scores. The use of scores derived from Japanese may be justified if cultural differences determine large differences in utility scores. The limitation of our study is that the comparison between the different races was not performed.
In conclusion, bowel problem and hormonal function in the disease-specific functions might have impact on utility scores derived from all three methods, while age and sexual function had less impact. Comorbidity was related to utility scores derived from EQ-5D and SF-36. A clear understanding of the effects of age, comorbidity, and treatment complications should be gained in conducting decision analyses.
Source of financial support: None.
Supplementary material for this article can be found at: http://www.ispor.org/publications/value/ViHsupplementary.asp
