Important Group Differences on the Functional Assessment of Cancer Therapy–Kidney Symptom Index Disease-Related Symptoms (FKSI-DRS) in Patients with Metastatic Renal Cell Carcinoma

Abstract

Background:

The Functional Assessment of Cancer Therapy–Kidney Symptom Index Disease-Related Symptoms (FKSI-DRS) is important to gauge clinical benefit in metastatic renal cell carcinoma (mRCC).

Objective:

To estimate important difference (ID) in FKSI DRS scores that is considered to be meaningful when comparing treatment effect between groups, using mRCC trial data.

Methods:

Data were derived from two pivotal phase III mRCC trials comparing sunitinib versus interferon-alfa (N=750) in first-line mRCC, and axitinib versus sorafenib (N=723) in second-line mRCC. The change from baseline in FKSI-DRS score was examined as a function of set of anchors using the repeated measures model (RMM). Several anchors were evaluated: FKSI item “I am bothered by side effects of treatment”, EuroQoL Group questionnaire (EQ-5D) utility score, and adverse events (AEs).

Results:

When “I am bothered by side effects of treatment” score was used as an anchor, ID ranged between 1.2 and 1.3 points. When change in EQ-5D utility score was used as an anchor the FKSI-DRS ID ranged between 0.62 and 0.63 point. Selecting the AEs that corresponded to a maximum worsening in the FKSI-DRS score in either trial, the ID ranged between 0.62 and 0.74.

Conclusions:

Among patients undergoing treatment for mRCC, between-group differences in FKSI-DRS scores as low as 1 point might be meaningful.

Introduction

In the past decade, a number of targeted anti-angiogenic therapies have been approved for the treatment of advanced renal cell carcinoma (RCC) (1). These therapies have improved the response rate, progression-free survival (PFS), and overall survival (OS) in many patients with advanced RCC (2). However, anti-angiogenic therapies are not curative, and have multiple side effects that may have an impact on health-related quality of life (HRQoL) (3–5).

Comparison of HRQoL among patients treated with different targeted therapies for metastatic RCC (mRCC) is an important exercise to guide treatment selection (3). However, use of cancer-specific global quality of life (QoL) instruments, such as the Functional Assessment Cancer Therapy (FACT-G) (6), may obscure important and significant changes in disease-related and treatment-related symptoms. To address this, an RCC-specific 15-item instrument, the Functional Assessment of Cancer Therapy–Kidney Symptom Index (FKSI-15), was developed and validated by Cella et al. (7), by weighting the relative importance of each of the 15 items based on ratings from both RCC patients and clinicians who treat RCC. The FKSI-15 evaluates both RCC-related symptoms and treatment-related side effects. Retaining nine items from the FKSI, which comprise the Disease-Related Symptom (DRS) subscale, the FKSI-15 was expanded to 19 items in the National Comprehensive Cancer Network FKSI-19, which created a more complete list related specifically to kidney cancer (8). The nine-item FKSI-DRS was subsequently validated by Cella et al. (9).

Accurate interpretation of the HRQoL scores — specifically, determination of what constitutes an “important difference” (10–12) — is required to assess overall clinical benefit of a therapy and also to determine the optimal sample size in clinical trials (13, 14). Although the clinically important difference for the FKSI-DRS has been assessed previously, the number of patients studied was small, not all patients received active treatment, and the follow-up time was relatively brief (9).

The goal of this study was to provide an estimation of the important difference (ID) in FKSI-DRS score that is considered to be meaningful when comparing treatment groups (as opposed to difference on the individual level, i.e., within-subject change), especially under conditions of active treatment in mRCC patients. The defined unit (the estimated group difference) can be used in a study to benchmark the between-group FKSI-DRS difference.

Analyzing data independently from two pivotal clinical trials of targeted agents in the treatment of mRCC (15, 16), we used a series of models to understand the relationship between change in the FKSI-DRS score and the FKSI item “I am bothered by side effects of treatment” score, and change from baseline in the EuroQoL Group questionnaire (EQ-5D™) utility score, as well as the treatment-emergent adverse events (AEs). These estimates were then triangulated and evaluated relative to distribution-based estimates, in order to determine the ID for the FKSI-DRS scale.

Methods

Patient populations and trial design

Patient-reported outcome data from two pivotal phase III clinical trials were used in this analysis. Trial 1, by Motzer et al. (15), compared sunitinib (50 mg/day; 4-weeks-on/2-weeks-off schedule) with interferon-alfa (3–9 MU/dose thrice-weekly) in a population of treatment-naive patients with clear-cell mRCC. Trial 2, by Rini et al. (16), compared axitinib (5 mg twice daily) with sorafenib (400 mg twice daily) in patients with clear-cell mRCC who had failed one previous first-line therapy. Inclusion criteria in both trials included Eastern Cooperative Oncology Group performance status 0 or 1 and adequate renal, hepatic, hematologic, and cardiac function. Both trials have published final efficacy and safety results, including patient-reported outcome data (15–20). The data from each trial were analyzed separately and the results were compared.

Patient-reported outcomes and safety assessments

Patients in both trials were administered the FKSI questionnaire. FKSI-DRS, a nine-item subdomain of FKSI, assessed patient-reporting of the following symptoms: lack of energy, pain, weight loss, bone pain, fatigue, shortness of breath, coughing, bothered by fevers, and hematuria. Each question is rated on a five-point Likert-type scale ranging from 0 = “not at all” to 4 = “very much” (9). However, the answers are reverse coded: a score of 0 indicates the most symptoms and a score of 4 is associated with no symptoms. The final score, which can range from 0 (most severe symptoms) to 36 (no symptoms), is calculated as the sum of the nine individual item scores.

Patients in both trials were also administered the EQ-5D. The EQ-5D is a preference-based generic health status measure that comprises a descriptive system with five dimensions — mobility, self-care, usual activities, pain/discomfort, and anxiety/depression — each with three levels (no problems, moderate problems, extreme problems), and a visual analogue scale score for the overall health state (21). In both trials, utility scores, which ranged from 1 (best possible health) to 0 (dead) and negative scores representing “worse than dead” condition, were calculated based on societal reference developed from general population-based valuation studies in the United Kingdom (22).

Questionnaires were provided at screening (baseline), during treatment (on day 1 and day 28 of every cycle in Trial 1, and every 4 weeks in Trial 2), and at treatment end (18, 23, 24). Questionnaires were also provided at follow-up (28 days after treatment end) in Trial 2 (18).

Safety assessments

Safety assessments included documentation of AEs, which were graded in both studies (15, 16) according to Common Terminology Criteria for Adverse Events of the National Cancer Institute, version 3.0. For the current analyses, we selected AEs with a high number of events and occurring in both studies.

Models for ID estimation

Four different approaches were investigated to estimate ID in FKSI DRS scores that is considered to be meaningful when comparing treatment effect between groups: the FKSI item GP5 (“I am bothered by side effects of treatment”), EQ-5D utility score, AEs and Effect Size. We chose FKSI item GP5 because, like AEs, it is associated with the patient’s overall perception of how they are feeling and functioning (14).

The repeated measures model (RMM) with FKSI item “I am bothered by side effects of treatment” as a continuous predictor variable describes the change in FKSI-DRS score corresponding to a one-category difference in FKSI item “I am bothered by side effects of treatment”. Each response category of FKSI item GP5 represents a distinct state and, therefore, can serve as a useful anchor for the corresponding difference on the FKSI-DRS. The RMM model with FKSI item “I am bothered by side effects of treatment” as a categorical predictor does not impose any functional relationship between change from baseline FKSI-DRS and FKSI item “I am bothered by side effects of treatment” and was used to examine the appropriateness of the linearity assumptions. The RMM included all available data for a patient (provided that there is at least one measurement available). This model assumes that missing data are missing at random.

The same type of models was also used to examine the relationship between the change from baseline in the FKSI-DRS score (outcome) and the change from baseline in the EQ-5D utility score (predictor). In all cases, we examined the change in FKSI-DRS score that corresponded to what would generally be considered a clinically important difference (the value of 0.10 points) in the EQ-5D utility score (25).

In addition, a series of models were used to assess the relationship between the change in the FKSI-DRS score from baseline (as outcome) and AE grade (as predictor) for frequently occurring AEs.

The overall mean change from baseline in FKSI-DRS for a patient and the overall mean of a predictor for the same patient were calculated effectively creating a dataset with only one observation per patient. This dataset was used to estimate Pearson correlation coefficients.

RESULTS

Patients

Trial 1 enrolled 750 patients (375 randomized to each treatment arm), of whom 375 patients received treatment with sunitinib, 360 patients received treatment with interferon-alfa, and 15 patients withdrew consent before the start of treatment. Trial 2 enrolled 723 patients (361 randomized to axitinib and 362 to sorafenib), of whom 359 patients received treatment with axitinib, 355 received treatment with sorafenib, and 9 patients withdrew before the start of treatment. Baseline demographic and clinical characteristics were similar between the two trials (Table 1).

Table 1 –

Patient demographic and baseline clinical characteristics

Variable	Trial 1		Trial 2
	Sunitinib n=375	IFN-α n=375	Axitinib n=361	Sorafenib n=362
Age, median, years	62	59	61	61
Sex (%)
Male	71	72	73	71
Female	29	28	27	29
ECOG PS, %
0	62	61	54	55
1	38	38	45	44
>1	0	1	<1	0
MKSCC risk groups, %
0	38	34	28	28
1–2	56	59	37	36
2– −3	6	7	33	33
n/a	0	0	2.5	3.0
Baseline QoL scores, mean (SD)
FKSI-DRS	29.74 (5.24)	29.55 (5.03)	28.87 (5.19)	28.98 (5.19)
EQ-5D	0.76 (0.23)	0.76 (0.23)	0.73 (0.28)	0.73 (0.26)

ECOG PS, Eastern Cooperative Oncology Group Performance Status; EQ-5D, EuroQoL 5 Dimensions; FKSI-DRS, Functional Assessment of Cancer Therapy–Kidney Symptom Index Disease-Related Symptoms; IFN-α, interferon alfa; MSKCC, Memorial Sloan Kettering Cancer Center; n/a, not available; QoL, quality of life; SD, standard deviation.

Relationship between change from baseline in FKSI-DRS score and the FKSI item “I am bothered by side effects of treatment” score

The relationship between changes from baseline in scores for FKSI-DRS and “I am bothered by side effects of treatment,” determined by the different models, is shown in Figure 1. This graph also indicates that the linearity assumption for the relationship between changes in FKSI-DRS and “I am bothered by side effects of treatment” item is appropriate. Data from 681 of 750 (91%) randomized patients from Trial 1, and from 674 of 723 (93%) randomized patients from Trial 2 were available for RMM. The correlations between changes in these scores were 0.31 and 0.28 for Trials 1 and 2, respectively (P < 0.05 for both). Based on the RMM model with “I am bothered by side effects of treatment” score as a continuous predictor, slopes were 1.26 for Trial 1 and 1.20 for Trial 2 (P < 0.05 for both).

Fig. 1 – Relationship between change from baseline in FKSI-DRS score and the FKSI item “I am bothered by side effects of treatment (gp5)” score.

FKSI-DRS, Functional Assessment of Cancer Therapy–Kidney Symptom Index Disease-Related Symptoms.

Relationship between changes from baseline in the EQ-5D utility score and change in FKSI-DRS

The relationship between change from baseline in the EQ-5D utility score and change from baseline in the FKSI-DRS score, as determined by the different models, is shown in Figure 2. This Figure also indicates that the linearity assumption for the relationship between changes in FKSI-DRS and changes in EQ-5D utility scores is appropriate. Visually notable differences at the edges corresponded to the small number of observations with those large changes in EQ-5D. Data from 689 of 750 (92%) randomized patients from Trial 1, and from 659 of 723 (91%) randomized patients from Trial 2 were available for RMM. Generally, correlations between changes in scores were moderate: 0.44 and 0.49 for Trial 1 and 2, respectively (P < 0.05). According to the RMM model with change in EQ-5D utility score as a continuous predictor/predicting variable, slopes for both trials were statistically significant (P < 0.05) and practically identical (0.62 and 0.63 for Trials 1 and 2, respectively; Figure 2). Based on this model, a 0.1-point decrease in the EQ-5D utility score corresponded to a 0.6-point worsening in the FKSI-DRS score.

Fig. 2 – Relationship between change in FKSI-DRS and EQ-5D scores using the RMM models with EQ-5D as a continuous or categorical predictor variable.

EQ-5D, EuroQoL Group questionnaire, FKSI-DRS, Functional Assessment of Cancer Therapy–Kidney Symptom Index Disease-Related Symptoms, RMM, repeated measures model.

Relationship between individual AEs and change from baseline in FKSI-DRS

From all the models studied to assess relationships between individual most frequent AEs (fatigue, diarrhea, nausea, dysgeusia, palmar-plantar erythrodysesthesia syndrome, anorexia, hypertension, decreased appetite, asthenia, weight decreased, rash, vomiting, arthralgia, and dyspepsia) and change in FKSI-DRS score from baseline using AE grade as a continuous predictor, a one-grade increase in AE severity for fatigue (slope, −0.62, p<0.0001) in Trial 1 corresponded to a largest 0.62-point decrease (worsening) in the FKSI-DRS score (Supplementary Figure 1 and Supplementary Table 1). Data from 707 of 750 (94%) randomized patients from Trial 1were available for RMM. In Trial 2, a one-grade increase in AE asthenia (slope, −0.74, p<0.0001) corresponded to a largest 0.74-point decrease (worsening) in the FKSI-DRS score (Supplementary Table 1). Data from 674 of 723 (93%) randomized patients from Trial 2 were available for RMM. For both above-mentioned AEs, the correlations were weak (below 0.30) and generally were in the anticipated direction, indicating a worsening in AE grade associated with negative change in the FKSI-DRS score.

DISCUSSION

The new agents available for the treatment of mRCC increased the complexity of treatment patterns. In addition to safety and efficacy data, meaningful changes in QoL scales would inform decision-making by patients and providers. The change in HRQoL represents an important criterion to compare overall clinical benefit. However, the interpretation of what constitutes a clinically meaningful difference has been challenging. Mixed qualitative and quantitative methods, including clinical standards setting (26), combined anchor and distribution approaches (11, 27, 28), and search for convergence across anchors (29, 30) have emerged to address this.

The current study used multiple anchor-based models to determine the estimation of the ID for the FKSI-DRS instrument in patients undergoing treatment in the first- or second-line treatment of mRCC. Each of the different methods using different anchors, along with the distribution-based approach, gave supportive and consistent estimates of the ID, that is, all slopes were pointing in the same direction.

Results using the FKSI “I am bothered by side effects of treatment” score as an anchor were in agreement with previous results (ID ranged between 1.2 and 1.39 points). Additionally, when comparing changes in FKSI-DRS score with change in the EQ-5D utility score, for which an ID has been previously defined (25), and using RMM with change in EQ-5D as a continuous predictor, a 0.1-point change in EQ-5D was found to correspond to a 0.63-point change in the FKSI-DRS score. As the 0.1-point change in EQ-5D utility score is considered clinically relevant (25), the estimated FKSI-DRS ID can be considered to be 0.63 points based on this model. Based on RMM with AE grade as a continuous predictor, and using the most conservative estimate, the FKSI-DRS ID can be estimated as 0.74 points.

Previous research has suggested moderate level agreement between quality of life and AEs (31). The FKSI “disease-related symptom” scale, despite its name, also taps treatment-related AEs in part; it neither asks nor expects patients to attribute the queried symptom to disease activity. the reason this scale is called “disease-related symptoms” is because they were identified by clinical experts as predominantly (but not exclusively) disease-related rather than treatment-related (9). In our study, the relationship between FKSI-DRS and individual AEs was lower than expected, possibly due to the range of AEs. Nevertheless, for anchors other than AEs, where the correlations were above 0.3, results were comparable to those when AEs were used as an anchor.

The standard deviation in baseline FKSI-DRS scores in Trial 1 and Trial 2 were almost the same, and corresponded to approximately 5 points. In a study by Pickard et al. (25), the EQ-5D ID corresponded to a small effect size ≤0.2 points. In the current analysis, the effect size of 0.2 would be equivalent to 1 point on the FKSI-DRS scale (32). When considering data from the different models in combination with the value obtained by a distribution-based approach, a 1-point group difference in FKSI-DRS score can be considered to be potentially important. This value can help gauge the clinical relevance of the treatment effect of a drug between different treatment groups; it should not be used on an individual level to define a responder (33).

A multi-pronged approach that utilizes both anchor-based and distribution-based methods is required for accurate estimation of the clinically important difference (34). Here, we used multiple anchor-based models and also included a distribution-based assessment and triangulated the estimations from these different approaches to determine ID for the FKSI-DRS scale.

A limitation of this analysis is that most of the correlations between anchors and FKSI-DRS scores were relatively low, sometimes falling below 0.30. Low correlations with anchors can tend to under-estimate the ID for differences in scores. To account for that, we conducted a series of analyses that supports our findings and conclusion. Specifically, despite the weak correlation between AE anchor and FKSI-DRS the estimated ID (based on AE anchor) was consistent with the two other anchors, EQ-5D, and the FKSI item “I am bothered by side effects of treatment” and distribution based estimation.

We recognize the apparent incongruity in the use of a side-effect bother question as an anchor for a scale labeled as “disease-related symptoms.” However, in reality, disease symptoms and treatment side effects are difficult for patients to distinguish, and more research is needed to better understand the patient perspective. Finally, in both trials, only sparse data were available for higher, grade 4 AEs.

Conclusions

This analysis of two randomized trials in mRCC provides an estimate of as low as 1 point for the ID of the FKSI-DRS scale in group comparisons, which is lower than the previously suggested range of 2–3 points (9). Determination of the ID for the FKSI-DRS will allow clinicians to determine what constitutes a relevant change in disease-specific symptom severity among treatment groups of patients with mRCC, whereas application at the individual level will require greater changes. Estimation of the FKSI-DRS ID will allow accurate interpretation of patient-reported outcomes data, a potentially important endpoint in clinical trials in mRCC.