A combination of distribution- and anchor-based approaches determined minimally important differences (MIDs) for four endpoints in a breast cancer scale

Abstract

Objective

To determine distribution- and anchor-based minimal important difference (MID) estimates for four scores from the Functional Assessment of Cancer Therapy–Breast (FACT-B): the breast cancer subscale (BCS), Trial Outcome Index (TOI), FACT-G (the general version), and FACT-B.

Study design and setting

We used data from a Phase III clinical trial in metastatic breast cancer (ECOG study 1193; n = 739) and a prospective observational study of pain in metastatic breast cancer (n = 129). One third and one half of the standard deviation and 1 standard error of measurement were used as distribution-based criteria. Clinical indicators used to determine anchor-based differences included ECOG performance status, current pain, and response to treatment.

Results

FACT-B scores were responsive to performance status and pain anchors, but not to treatment response. By combining the results of distribution- and anchor-based methods, MID estimates were obtained: BCS = 2–3 points, TOI = 5–6 points, FACT-G = 5–6 points, and FACT-B = 7–8 points.

Conclusion

Distribution- and anchor-based estimates of the MID do show convergence. These estimates can be used in combination with other measures of efficacy to determine meaningful benefit and provide a basis for sample size estimation in clinical trials.

Introduction

Breast cancer is the most commonly diagnosed nondermatologic malignancy in American women. An estimated 211,300 women were diagnosed with invasive breast cancer in 2003 [1]. Available treatment options include surgery (e.g., lumpectomy, mastectomy), radiotherapy, cytotoxic chemotherapy, biologic therapy (e.g., trastuzumab), and hormone therapy (e.g., tamoxifen, aromatase inhibitors) [2], [3], [4], [5]. Most patients receive some combination of these treatments. Treatments for early-stage disease can result in cure; those for metastatic disease typically do not. Systemic treatments for metastatic breast cancer such as chemo- and hormonal therapies can increase survival time, but side effects and toxicities can be severe. Hence, the quality of a patient's survival can be as important as its length [6].

Health-related quality of life (HRQL)—that is, a person's actual or expected physical, emotional, and social well-being resulting from a medical condition or its treatment [7]—can be profoundly impacted by breast cancer treatment [8], [9], [10], [11]. General HRQL concerns of breast cancer patients include pain, fatigue, sexual dysfunction, and disruption in daily activities. Psychosocial concerns including anxiety, depression, and fear of recurrence may also be present. Nausea, vomiting, hair loss, and weight gain are common side-effects of chemotherapy. Beyond these general issues, there are also disease-specific concerns such as breast soreness, arm swelling (lymphedema), body image disturbance, and lost sense of femininity [12]. Assessing all of these issues in breast cancer clinical trials would be ideal [13]. Regulatory agencies such as the U.S. Food and Drug Administration and professional organizations such as the American Society of Clinical Oncology have stated that patient-reported health data (i.e., HRQL) are important endpoints to measure in the assessment of new cancer therapies [14], [15], [16]. The more we know about the importance of changes in scores on these HRQL instruments, the more helpful they can be in decision-making.

Several instruments are now available to measure the HRQL of women with breast cancer. General instruments such as the SF-36 and cancer-specific instruments such as the Functional Assessment of Cancer Therapy (FACT), Cancer Rehabilitation Evaluation System, Functional Living Index Cancer, and the European Organization for the Research and Treatment of Cancer's Core Quality of Life Questionnaire (EORTC QLQ-C30) have been used in studies of breast cancer [11], [17], [18], [19], [20], [21]. Breast cancer–specific modules have also been developed to supplement the EORTC QLQ-C30 [22] and the FACT [12].

While HRQL is now recognized as an important endpoint in cancer clinical trials, aggregated scores on HRQL measures may appear less meaningful to clinicians. One way to enhance the clinical utility of HRQL scores is to assess cross-sectional differences and longitudinal changes by anchoring score differences to clinically familiar events (e.g., response to treatment) and patient-status indicators (e.g., performance status). Sizeable differences in clinically distinct groups are likely to be clinically meaningful. Increasing attention is being paid to clinical significance as HRQL data become more important to the conduct of oncology clinical trials. In October 2000, the Clinical Significance Consensus Meeting Group, a panel of 30 experts, was convened to review the state-of-the-science of clinical significance in quality-of-life measures [23]; their work resulted in six articles devoted to various aspects of this topic [24], [25], [26], [27], [28], [29].

In addition to targeting differences that are clinically significant, identifying those that are minimally important provides a more precise measure of patient-reported treatment effect. Guyatt et al. [24] have recently stipulated that a minimally important difference (MID) on a HRQL measure corresponds to the “smallest difference in score in the domain of interest that patients perceive as important, either beneficial or harmful, and which would lead the clinician to consider a change in the patient's management” (p. 377). Implicit within this definition is that the MID represents the smallest score difference on a HRQL questionnaire that is clinically significant and therefore likely to be meaningful to both patients and clinicians.

MIDs can be determined using both distribution-based and anchor-based methods [24]. Distribution-based methods rely on the statistical distributions of HRQL scores in a given study [30]. Commonly used criteria include proportions of the standard deviation and the standard error of measurement (SEM) [31]. Anchor-based methods anchor or map score differences onto differences in clinical measures. Clinical measures can be objective indicators (e.g., response to treatment) or subjective assessments of patient status (e.g., performance status rating, global ratings of health-status change). Anchor-based differences can be determined either cross-sectionally at a single time point or longitudinally across multiple time points. An anchor-based approach using global ratings of change has identified clinically significant change scores on the QLQ-C30 in advanced breast cancer patients [21]. We have recently used an approach that combines distribution- and anchor-based methods to develop MIDs for the FACT-Lung [32] and the FACT-Anemia and Fatigue scales [33].

Our objective was to establish MIDs on the FACT-Breast cancer scale. We used both distribution- and anchor-based methods to determine MIDs.