Estimating Clinically Meaningful Changes for the Functional Assessment of Cancer Therapy—Prostate: Results from a Clinical Trial of Patients with Metastatic Hormone-Refractory Prostate Cancer
This study has been presented in part at “Determination of Clinically Meaningful Change (CMC) for Functional Assessment of Cancer Therapy—Prostate (FACT–P)”
Carducci MA, Nelson JB, Vogelzang NJ, Mulani P. Journal of Clinical Oncology, 2005 ASCO Annual Meeting Proceedings. Vol 23, No 16S (June 1 Supplement), 2005: 8077
ABSTRACT
Objective: To determine clinically meaningful changes (CMCs) for the Functional Assessment of Cancer Therapy–Prostate (FACT–P).
Methods: We obtained data from a Phase III trial of atrasentan in metastatic hormone-refractory prostate cancer patients (n = 809). We determined anchor-based differences using Karnofsky Performance Status (KPS), bone alkaline phosphatase (BAP), hemoglobin, time to disease progression (TTP), adverse events (AE), and survival. One-third and one-half standard deviation and standard error of measurement (SEM) were used as distribution-based criteria for CMCs. Comparison across baseline FACT–P domains and derived scales [FACT–P total score, Trial Outcome Index (TOI) score, prostate cancer subscale (PCS) score, pain-related score, and FACT Advanced Prostate Symptom Index (FAPSI)] were conducted for KPS, BAP, and hemoglobin using Student's t tests. Twelve-week change scores were compared for TTP, AE, and survival using ANCOVA.
Results: CMCs were estimated as 6 to 10 for FACT–P total score, 5 to 9 for FACT–P TOI score, 2 to 3 for FACT–P PCS, 1 to 2 for the 4 PCS pain-related questions, and 2 to 3 for FAPSI. CMCs were also estimated using distribution-based criteria. Kappa statistics were computed to determine the degree of correspondence between the recommended guideline of 1.0 SEM and empirically derived standards. Most of the kappas for health-related quality of life domains and SEM standards had “substantial” to “almost perfect” concordance.
Conclusions: The significant relationship between clinical and quality of life data provides support for the use of CMCs to increase interpretability of FACT–P scores.
Introduction
For patients faced with advanced illness and little hope for a cure, the medical focus shifts from healing the patient to relieving physical symptoms and maintaining function. This is especially true for advanced prostate cancer, where patients are provided with palliative rather than restorative interventions, and health-related quality of life (HRQoL) assessment is critical in evaluation of such treatments [1]. HRQoL measurements are routinely collected in cancer clinical trials [2–4].
The US Food and Drug Administration recently released draft guidance for inclusion of HRQoL data in product labels [5]. Patient-reported outcomes (PROs) measure patient health status directly from the patient's perspective and can be used to determine the effect of an intervention on one or more aspects of patient health through assessment of simple (e.g., patient symptoms) to complex (e.g., HRQoL) concepts. To support the use of PRO instruments in a clinical trial, researchers must provide information on their reliability and validity, including their ability to detect change. An important component of this is the determination of the minimal important difference or MID [6]. Specifying an MID is a valuable benchmark for interpreting clinical trial results.
One useful method to estimate an instrument's MID entails linking its score differences to meaningful cross-sectional differences and longitudinal change scores. Methods used for this estimation have been classified as anchor-based (comparing score differences or change with established external standards, termed “anchors”). In contrast, distribution-based methods (evaluating the dispersion of scores in the target population as an estimate of the scale's inherent variability) offer a likely range for the MID [6].
In hormone-refractory prostate cancer (HRPC), performance status, bone alkaline phosphatase (BAP), prostate-specific antigen, hemoglobin, time to disease progression (TTP), and pain-related adverse events (AE) can serve as potential anchors [7,8]. Distribution-based measures rely on the statistical distributions of HRQoL scores in a given study, such as standard deviation (SD), standard error of measurement (SEM), or both [9,10]. Distribution-based measures offer a way to identify clinically meaningful changes (CMCs) and can corroborate clinical anchor-based findings. These approaches have been used to determine clinically meaningful differences in HRQoL scores [10,11].
CMCs have been identified for several scales of the Functional Assessment of Chronic Illness Therapy measurement system, including lung cancer [4], fatigue [12], anemia [12], breast cancer [13], biological response modifiers [14], and colorectal cancer [15]. The purpose of this study is to establish CMC ranges for the Functional Assessment of Cancer Therapy—Prostate (FACT–P) using both anchor- and distribution-based methods.
Patients and Methods
Study Population
In the M00-211 trial, 809 patients with metastatic HRPC were randomly assigned to receive 10 mg of a novel endothelin-A antagonist, atrasentan, or placebo. The primary objective was to compare TTP between treatments. A tertiary objective was to evaluate HRQoL. HRQoL was measured at baseline, week 4, week 12, and every 12 weeks thereafter until week 72, or until the patient experienced disease progression. The present research is a secondary analysis to estimate a change in scores likely to be clinically meaningful.
Quality of Life Instruments
The FACT–P is a validated questionnaire [16] used to assess HRQoL in men with prostate cancer. FACT–P consists of FACT–G (general), a 27-item self-report questionnaire that measures general HRQoL in cancer patients, and a 12-item prostate cancer subscale (PCS). The PCS is designed specifically to measure prostate cancer-specific quality of life. The FACT–P Trial Outcome Index (TOI) is based on the physical and functional well-being subscales of the FACT–G and the PCS. The FACT–P total score includes the FACT–G and the PCS. The FACT Advanced Prostate Symptom Index (FAPSI) includes eight items from the FACT–P [17]. A higher overall score indicates better HRQoL. The FACT–P total score, TOI, PCS score, PCS pain-related score (four questions from the FACT–P interrogating pain specifically), and FAPSI were selected as the primary focus of this analysis as they contain the most relevant questions about symptoms and physical functioning.
Demographic and Clinical Data
Demographic information including age, ethnicity, and vital signs were collected through the clinical trial procedures. Physician-rated performance status was determined using the Karnofsky Performance Status (KPS) scale. Clinical variables, including BAP, and hemoglobin were also collected, and pain-related AE were captured in the safety data set. TTP and survival time were calculated.
The most useful clinical anchors are those that correlate at least moderately with HRQoL scores [8,15]. The clinical anchors used in this study were shown to have statistically significant relationships with HRQoL [18,19]. The relationship between changes in HRQoL and TTP (early vs. late progressors), pain-related AE, and survival time was investigated. Patients were not asked to specify a level of change they considered important.
Statistical Analysis
Criterion-related validity was defined as the relationship of test scores to the meaningful anchors KPS score, BAP, and hemoglobin. This information was used to assist in the interpretation of scores based upon group differences. The associated FACT–P total, TOI, FACT–P PCS, PCS pain-related score, and FAPSI differences will likely establish an upper limit for a minimum CMC in the score.
We used the minimum P-value approach [20] to determine the appropriate cut-point for the clinical variables. In the minimum P-value approach, selected values of the clinical variables are examined as candidates for the cut-points. The value is chosen that best separates patient outcomes (i.e., survival) according to maximum chi-square statistics and a minimum P-value or a maximum relative risk. Cut-points were obtained for each clinical variable using survival as an outcome, and patients were divided into two groups around the cut-point.
Using independent samples Student's t tests, we compared baseline HRQoL scores among patients with different baseline clinical characteristics: KPS, BAP, and hemoglobin. Effect sizes were calculated using group-level means and then dividing by the pooled within-group SD [21]. By convention, changes in the range of 0.2 SD units represent a small change, 0.5 a moderate change, and 0.8 a large change [21].
One-way analyses of covariance were used to determine the sensitivity of the HRQoL domains to changes in clinical status, such as TTP, pain-related AE, and survival. Baseline clinical characteristics associated with the HRQoL domain scores were included as covariates. The 12-week follow-up period was selected for these analyses because it proved the optimal balance between having sufficient time to note significant clinical change, while minimizing missing data because of disease progression or other reasons. Change scores were created by subtracting the baseline HRQoL domain scores from the 12-week follow-up scores. The change from baseline to week 12 in the HRQoL domains was analyzed by response to treatment and changes to clinical status. Effect sizes were calculated by dividing the mean change scores by the SD of the baseline HRQoL score.
Finally, several methods were used to identify which magnitudes of change in HRQoL score would be considered clinically meaningful. Empirically derived standards were compared with several established distribution-based standards. Several computations were necessary to accomplish this. First, SD of baseline, week 12, and change scores were computed and divided by 2 and 3 to establish one-half and one-third SD change standards. Next, the SEM was computed: σx(1 − rxx)1/2, where σx = SD of the subscore and rxx = the reliability (internal consistency) of the scale. On the basis of evidence offered by Wyrwich [10], patients were classified on a one SEM standard as having “declined” (change score fell by at least one SEM), “improved” (change score rose by at least one SEM), or “no change” (change score was less than one SEM). Next, patients were further classified into “declined,”“improved,” or “no change” categories based on the empirically derived standards. Cohen's kappa statistic was used to determine the degree of correspondence between the one SEM standard and the empirically derived standard. These methods are based on those used in similar studies estimating CMCs for the FACT within other cancer types [4,13].
Results
Baseline Summary Statistics
Subjects (n = 809) ranged in age from 45 to 93 years (mean = 71.8 years). Subject demographics and baseline clinical variables are summarized in Table 1.
| Variable | Variable | ||
|---|---|---|---|
| Age, years | N = 809 | Hemoglobin, g/dL | N = 793 |
| Mean (SE) | 71.8 (0.29) | Mean (SEa) | 13.2 (0.045) |
| Median | 72.0 | Median | 13.3 |
| Range | 45.0–93.0 | Range | 9.1–18.1 |
| Race | N = 809 | BAP, ng/mL | N = 769 |
| Caucasian | 770 (95%) | Mean (SE) | 59.1 (4.97) |
| Black | 26 (3%) | Median | 25.2 |
| Asian | 8 (1%) | Range | 2.0–1903.8 |
| Other | 5 (1%) | PSA, ng/mL | N = 803 |
| Weight, kg | N = 807 | Mean (SE) | 215.1 (16.1) |
| Mean (SE) | 84.5 (0.53) | Median | 72.9 |
| Median | 82.1 | Range | 1.7–5784.0 |
| Range | 47.0–176.9 | Karnofsky score | N = 809 |
| Mean (SE) | 93.8 (7.7) | ||
| Median | 100 | ||
| Range | 60–100 | ||
- BAP, bone alkaline phosphatase; PSA, prostate-specific antigen.
- Karnofsky Performance Scale Index allows patients to be classified as to their functional impairment. The lower the Karnofsky score, the worse the survival.
Baseline Differences in Clinical Indicators
Hemoglobin. The minimum P-value approach identified 11.6 g/dL as the optimal cut-point for hemoglobin (Table 1). Patients with hemoglobin higher than 11.6 g/dL were found to have higher FACT–P total score (P = 0.0002), TOI score (P < 0.0001), PCS score (P = 0.0004), pain-related score (P = 0.0006), and FAPSI score (P = 0.0002) than patients with hemoglobin lower than 11.6 g/dL. Patients with higher hemoglobin had better HRQoL scores (Table 2).
| Clinical indicator | FACT–P total score | FACT–P TOI score | FACT–P PCS | FACT–P PCS pain | FAPSI |
|---|---|---|---|---|---|
| BAP, ng/mL (N) | |||||
| >52.38 (192) | 111.4 (19.3) | 72.0 (15.1) | 31.0 (6.7) | 9.8 (3.9) | 22.9 (5.4) |
| ≤52.38 (577) | 118.6 (17.2) | 78.0 (13.2) | 33.2 (6.6) | 11.2 (3.8) | 24.6 (4.9) |
| Mean difference | −7.2 | −6.0 | −2.2 | −1.4 | −1.7 |
| P-value | <0.0001 | <0.0001 | <0.0001 | <0.0001 | <0.0001 |
| Effect size | 0.39 | 0.42 | 0.33 | 0.36 | 0.33 |
| Hgb, g/dL (N) | |||||
| >11.6 (709) | 117.8 (17.4) | 77.3 (13.5) | 33.0 (6.5) | 11.0 (3.8) | 24.4 (5.0) |
| ≤11.6 (84) | 107.4 (22.2) | 68.4 (17.6) | 29.7 (7.7) | 9.4 (4.3) | 21.6 (6.1) |
| Mean difference | 10.4 | 8.9 | 3.3 | 1.6 | 2.8 |
| P-value | 0.0002 | <0.0001 | 0.0004 | 0.0006 | 0.0002 |
| Effect size | 0.52 | 0.57 | 0.46 | 0.39 | 0.50 |
| Karnofsky (N) | |||||
| Score = 100 (430) | 121.3 (15.7) | 80.4 (11.8) | 34.2 (6.0) | 11.7 (3.6) | 25.4 (4.4) |
| Score < 100 (379) | 111.1 (19.2) | 71.7 (15.2) | 30.8 (7.0) | 9.8 (4.1) | 22.6 (5.5) |
| Mean difference | 10.2 | 8.7 | 3.4 | 1.9 | 2.8 |
| P-value | <0.0001 | <0.0001 | <0.0001 | <0.0001 | <0.0001 |
| Effect size | 0.58 | 0.64 | 0.52 | 0.49 | 0.56 |
- FACT–P, Functional Assessment of Cancer Therapy—Prostate; TOI, Trial Outcome Index; PCS, prostate cancer subscale; FAPSI, FACT Advanced Prostate Symptom Index; BAP, bone alkaline phosphatase; Hgb, Hemoglobin.
- Karnofsky Performance Scale Index allows patients to be classified as to their functional impairment. The lower the Karnofsky score, the worse the survival.
BAP. Elevated BAP levels have been found to predict for poor outcome in patients with HRPC [8,22,23]. The minimum P-value approach identified 52.38 ng/ mL as the optimal cut-point for BAP. Patients with baseline BAP higher than 52.38 ng/mL had lower FACT–P total score, TOI score, PCS score, pain-related score, and FAPSI score as compared with patients with BAP lower than 52.38 ng/mL (P < 0.0001 for all scales). This suggests that high BAP levels are associated with poor quality of life (Table 2).
KPS Scale. Baseline differences in FACT–P total score, TOI score, PCS, pain-related score, and FAPSI by KPS score are presented in Table 2. Those with a Karnofsky score of 100 had a higher FACT–P total score, TOI score, PCS score, pain-related score, and FAPSI score as compared with those with a Karnofsky score less than 100 (P < 0.0001 for all scales). On these dimensions, patients with higher Karnofsky scores reported better HRQoL (Table 2).
Using the baseline variables as meaningful indicators of patient status, we determined the upper limits for an MID. These limits correspond to differences in FACT–P total, TOI, PCS, pain-related, and FAPSI scores across the baseline variables that represent effects of “small” to “moderate” (approximate range of effect size = 0.20 to 0.60) size, per Cohen. Small to moderate differences are considered minimal estimators of the MID, similar to previous estimations of MID for lung [3] and breast cancer [13]. The absolute differences in FACT–P total score ranged from 7 to 10 (rounding to the nearest decimal). Similarly, for the TOI, the difference ranged from six to nine. PCS score differences ranged from two to three, PCS pain-related score differences ranged from one to two, and FAPSI score differences ranged from two to three.
Changes in HRQoL Over Time
Changes in HRQoL and TTP. The median TTP for the study population was 87 days. A difference in HRQoL change scores was observed between early progressors (EP) and later progressors (LP), with EPs showing a greater decline than the LPs (Table 3). The difference between these groups in FACT–P total score change was 7.2. Differences between the EPs and LPs for the TOI and PCS, and pain-related change scores were 5.9, 2.7, and 1.7, respectively. The EPs and LPs differed on FAPSI by 2.0 (P < 0.0001 for all scales).
| Clinical indicator (N) | FACT–P Total score | FACT–P TOI score | FACT–P PCS | FACT–P PCS pain | FAPSI |
|---|---|---|---|---|---|
| No. of items | 39 | 26 | 12 | 4 | 7 or 8 |
| TTP > median (362) | −2.0 (14.6) | −3.1 (11.6) | −0.6 (6.0) | −0.2 (3.7) | −0.4 (4.4) |
| TTP ≤ median (447) | −9.2 (16.0) | −9.0 (13.0) | −3.3 (6.4) | −1.9 (4.0) | −2.4 (4.8) |
| Difference | 7.2 | 5.9 | 2.7 | 1.7 | 2.0 |
| P-value | <0.0001 | <0.0001 | <0.0001 | <0.0001 | <0.0001 |
| Effect size | 0.47 | 0.48 | 0.44 | 0.44 | 0.43 |
| Pain-related AE (415) | −8.4 (16.9) | −8.8 (13.5) | −3.2 (6.7) | −2.0 (4.0) | −2.4 (5.0) |
| No pain-related AE (394) | −2.1 (13.5) | −2.7 (10.7) | −0.3 (5.5) | 0.05 (3.5) | −0.2 (4.1) |
| Difference | −6.3 | −6.1 | −2.9 | −2.0 | −2.2 |
| P-value | <0.0001 | <0.0001 | <0.0001 | <0.0001 | <0.0001 |
| Effect size | 0.41 | 0.50 | 0.47 | 0.55 | 0.48 |
| Survival time > median (268) | −1.7 (12.3) | −2.5 (9.0) | −0.4 (4.9) | −0.1 (3.1) | −0.4 (3.6) |
| Survival time ≤ median (541) | −7.6 (17.0)–1.7 (12.3) | −7.9 (14.0)–2.5 (9.0) | −2.7 (6.9)–0.4 (4.9) | −1.5 (4.3)–0.1 (3.1) | −1.9 (5.2)–0.4 (3.6) |
| Difference | 5.9–7.6 (17.0) | 5.4–7.9 (14.0) | 2.3–2.7 (6.9) | 1.4–1.5 (4.3) | 1.5–1.9 (5.2) |
| P-value | <0.00015.9 | <0.00015.4 | <0.00012.3 | <0.00011.4 | <0.00011.5 |
| Effect size P Value | <0.0001 | <0.0001 | <0.0001 | <0.0001 | <0.0001 |
| Effect size | 0.40 | 0.46 | 0.38 | 0.37 | 0.34 |
- FACT–P, Functional Assessment of Cancer Therapy—Prostate; TOI, Trial Outcome Index; PCS, prostate cancer subscale; FAPSI, FACT Advanced Prostate Symptom Index; TTP, time to progression; AE, adverse events.
Changes in HRQoL and pain-related AE. A difference in the HRQoL change scores was observed between patients experiencing AE related to pain compared with those who did not, as patients with pain AE showed a greater decline than those without these events (Table 3). The difference between these patient groups in FACT–P total score was 6.3. TOI, PCS, and pain-related change scores differed between these groups by 6.1, 2.9, and 2.0. The two groups also differed on FAPSI by 2.2 (P < 0.0001 for all scales).
Changes in HRQoL and survival. The median survival time for the patient population was 568 days. A difference in HRQoL change scores was observed between patients with survival times less than or equal to the median (short survival: SS) and those with survival times greater than median (long survival: LS), with patients having SS showing a greater decline than those with LS (Table 3). The difference in FACT–P total change score between these groups was 5.9, while the differences were 5.4, 2.3, 1.4, and 1.5, respectively for the TOI, PCS, pain-related, and FAPSI mean change scores (P < 0.0001 for all scales).
From these results, it was possible to estimate more precisely CMCs on the five scales used (Table 4).
| HRQoL domain | Range for clinically distinguishable score | Effect size range |
|---|---|---|
| FACT–P Total | 6 to 10 | 0.39 to 0.58 |
| FACT–P TOI | 5 to 9 | 0.42 to 0.57 |
| FACT–P PCS | 2 to 3 | 0.33 to 0.52 |
| FACT–P PCS pain | 1 to 2 | 0.36 to 0.55 |
| FAPSI | 2 to 3 | 0.33 to 0.56 |
- HRQoL, Health-related quality of life; FACT–P, Functional Assessment of Cancer Therapy—Prostate; TOI, Trial Outcome Index; PCS, prostate cancer subscale; FAPSI, FACT Advanced Prostate Symptom Index.
Comparisons with distribution-based criteria. The hypothesized thresholds for meaningful change were then compared against various distribution-based standards to determine an MID. To specify more precisely a CMC on the HRQoL scales, we determined the correspondence between the empirically derived clinical standards (above) and two established standards of a CMC. The first was one-third to one-half of an SD, which corresponds to the magnitude Cohen suggests is representative of a small to moderate effect size [22], and the second, one SEM, which has been shown to be equivalent to a minimal clinically important intra-individual change [9].
We first calculated one-third and one-half SD for the baseline, week 12, and change scores for the HRQoL domains (Table 5). Baseline and week 12 SEMs were then calculated for all HRQoL domains (Table 5). In comparison with the SD, the SEM is less sample-dependent and accounts for the reliability of the scale [24,25].
| 1/2 SD | 1/3 SD | SEM | |
|---|---|---|---|
| FACT–P Total Score | |||
| Baseline | 9.1 | 6.0 | 6.0 |
| Week 12 | 10 | 6.9 | 6.9 |
| Change from baseline to week 12 | 7.8 | 5.2 | |
| FACT–P TOI Score | |||
| Baseline | 7.1 | 4.7 | 5.5 |
| Week 12 | 8.2 | 5.4 | 6.3 |
| Change from baseline to week 12 | 6.3 | 4.2 | |
| FACT–P PCS | |||
| Baseline | 3.3 | 2.2 | 3.7 |
| Week 12 | 3.7 | 2.5 | 4.1 |
| Change from baseline to week 12 | 3.2 | 2.1 | |
| FACT–P PCS Pain | |||
| Baseline | 2.0 | 1.3 | 2.3 |
| Week 12 | 2.1 | 1.4 | 2.5 |
| Change from baseline to week 12 | 2.0 | 1.3 | |
| FAPSI | |||
| Baseline | 2.6 | 1.7 | 2.3 |
| Week 12 | 2.9 | 1.9 | 2.6 |
| Change from baseline to week 12 | 2.3 | 1.6 | |
- FACT–P, Functional Assessment of Cancer Therapy—Prostate; TOI, Trial Outcome Index; PCS, Prostate Cancer subscale; FAPSI, FACT Advanced Prostate Symptom Index.
We determined the degree of concordance between the hypothesized change standards and the one SEM standard to verify which of our empirically derived, hypothesized standards most closely approximated an MID. This was accomplished as follows: First, we classified patients into “declined,”“improved,” or “no change” using various point change standards for the different scales. 1) 6 to 10 points for the FACT–P total score; 2) 5 to 9 points for the TOI; 3) 2 to 3 points for the PCS; 4) 1 to 2 points for the pain-related score; and 5) 2 to 3 points for the FAPSI. To illustrate, for the 2-point standard for PCS, patients were classified as “declined” if their change score had fallen at least 2 points from baseline to week 12, “improved” if it had risen at least 2 points, and “no change” if it fell between −2 to +2. The same approach was used to classify patients into “declined,”“improved,” or “no change” for the one SEM standard.
Kappa statistics determining concordance between the hypothesized change standards and the one SEM standards were high. For the FACT–P PCS, baseline kappa values were 0.80, 0.98, 0.86, and 0.70 for the 2-, 3-, 4-, and 5-point standards. Week 12 kappa values were 0.66, 0.84, 1.0, and 0.83 for the 2-, 3-, 4-, and 5-point standards. Kappa statistics between 0.61 and 0.80 are “substantial,” while those between 0.81 and 1.00 are “almost perfect. [26]” Most of the kappa values for the HRQoL domains and the SEM standards in this sample were “substantial” to “almost perfect.”
Conclusions
We used a combination of anchor- and distribution-based approaches to identify MIDs for FACT–P end points. Our recommended MID ranges are 6 to 10 for the FACT–P total score, 5 to 9 for the TOI, 2 to 3 for the PCS, 1 to 2 for the pain-related score, and 2 to 3 for the FAPSI. These findings provide indirect evidence for whole number estimates just above the lower bound range values as the MIDs for these HRQoL domains. Because the MID estimates determined here are based on one clinical trial in advanced HRPC patients, future research should focus on verifying these estimates in multiple patient populations, including early-stage disease, to determine their broad applicability.
Effect size ranges for all four HRQoL domains were between 0.26 and 0.72, which is close to the small to moderate range (0.20 to 0.60). Reporting a range as opposed to a single estimate has been recommended because MIDs can vary across patients and groups [9,27]. There is also recent evidence that the one SEM rule of thumb may underestimate the MID for short-term changes in more acute conditions [28]. When interpreting change in response to a specific anchor, one should not expect all domains to change.
On the basis of the rationale that patients who differ on baseline clinical characteristics will also differ on HRQoL domains, baseline indicators were used to determine likely upper limits of the MID. We found that baseline differences on the FACT–P domains in our sample support this notion, as those with better clinical indicators also reported better quality of life.
Week 12 change scores were evaluated for the following clinical indicators: TTP, pain-related AE, and survival. Patients classified as EPs showed greater decline in HRQoL than late progressors, while those with more pain-related AE displayed a more notable decrease than those who did not experience such AE.
Appropriate methods for interpreting changes in HRQoL have not been developed at the same pace as the tools for assessing HRQoL. With a wide range of HRQoL instruments involving varying units of measurement, interpretation is challenging and their meaning may not be evident to clinicians and researchers. Identifying intraindividual change in HRQoL measures is critical in the evaluation of the patient's status. Better interpretation of HRQoL results from clinical trials will enable clinicians to incorporate such information into their practices, including standards of care.
For PRO end point data to be acknowledged as evidence of treatment success, it is essential to document the instrument's psychometrics and MID. CMCs can help clinicians and researchers understand whether HRQoL differences between treatments groups are important or whether changes over time in a group are meaningful, such as when assessing responsiveness to a treatment. One should be cautious when applying these criteria in clinical practice. The MIDs should be used as guide to clinicians when more investigation may be warranted before making therapeutic changes. MIDs can also be helpful as justification for sample size requirements necessary to determine clinical benefit.
Our study is not without limitations. Because these estimates are derived from a single trial, confirmation in other populations is essential. Future research should also incorporate questions that would elucidate meaningful clinical change thresholds for individual patients. These limitations notwithstanding, the MIDs identified in this paper, in conjunction with traditional clinical endpoints, contribute to the enhanced interpretation of FACT–P scores. Clinicians and investigators can apply this information to compare and evaluate treatment efficacy in patients with HRPC.
Source of financial support: This research was partially supported by an unrestricted research grant to the corresponding author.
