Sirs, We read the article by Adam et al.1 with great interest. Proper selection of an endpoint is a key determinant for a successful clinical trial. As discussed in the paper, the Gastrointestinal Symptom Score (GIS) appears to be a valuable new tool for the clinical assessment of symptom severity among patients with functional dyspepsia. Methods for the development and initial psychometric evaluation of the GIS as reported are generally consistent with standard approaches. However, we are concerned with the implication that the GIS is superior to global response ratings for assessing treatment response when, in fact, the GIS has not been directly compared with a global rating. Patient-reported global assessments of change in symptom severity measure are both critical for the assessment of treatment response and distinct from clinician-administered measures such as the GIS. In particular, global measures allow patients to weigh the relative importance of their various symptoms when providing a response. The use of a simple sum score for the GIS assumes that each of the 10 items is of equal importance, which is unlikely to be the case. For example, we expect that many patients would be more bothered by ‘vomiting’ than ‘loss of appetite’, but each of these items is given equal weight in the GIS. An additional concern regarding scoring of the GIS is that the authors’ factor analysis revealed ‘four independent factors’ and the number of items per factor was not consistent. As a result, several factors or domains are given greater weight than others in the computation of the total score. Therefore, either further justification or modification of the GI scoring algorithm is warranted, including a comparison of model fit between the one-factor model implied by the current algorithm and the four-factor model described in the paper, as well as an assessment of internal consistency. We also consider it a great benefit to have patients directly assess changes in the severity of their symptoms in response to treatment. Treatment response data collected within the context of a clinical interview, as required for completion of the GIS, are subject to bias in several regards – not only is a patient likely to want to please the clinician by reporting improvements, but clinicians also want to see their patients’ symptoms alleviated. The utilization of self-completed assessments greatly minimizes, if not eliminates, these potential sources of bias. While the GIS could be adapted for self-completion, additional psychometric evaluation would be required. There are several general methods for evaluating responsiveness or the sensitivity of an instrument to change. One is to compare the scores of responders and non-responders as described in Adam et al.;1 the other is to compare the scores of patients who receive an active or superior treatment to those who receive placebo or an inferior treatment. The latter method is preferred when assessing an instrument intended for use in clinical trials. While bias due to interviewer administration of the GIS may have minimal impact on the method utilized by Adam et al.,1 this potential source of measurement error is more likely to limit the ability of the GIS to demonstrate differences among treatment groups by inflating the rate of placebo response. Therefore, we recommend that the authors utilize the clinical trial data described in their study to evaluate whether or not the GIS is sensitive to differences between patients treated with STW 5 and placebo and supplement their previous responsiveness analysis. Although the GIS appears to be a well-developed and useful tool, it is not appropriate to declare it ‘valid’ based on a single evaluation. Instead, instrument validation is considered ‘an ongoing process of conducting various studies to confirm various hypotheses regarding the internal structure of the construct and its relationships with other variables’.2 A patient-reported global assessment of change in symptom severity is a critical component of any evaluation of a new treatment for a condition that relies on patients’ reports of symptoms, such as functional dyspepsia. Indeed, in a condition related to functional dyspepsia, irritable bowel syndrome, the global endpoint of ‘adequate relief’ fulfils many of the validation criteria described in the present study. Adequate relief responses repeatedly produce the same result (test–retest reliability), are able to detect clinically relevant changes (responsiveness), and move in the same direction as other meaningful measures (convergent validity).3-6 The adequate relief endpoint has also been used in a therapeutic trial of functional dyspepsia and was able to show statistically significant benefit over placebo.7
Read full abstract