Assessing patients’ improvement in clinical trials

Abstract

Should the doctor or patient judge improvement, and does it matter?

Doctors need to understand the benefits of new treatments from patients’ perspectives, because they must judge the relative risks and benefits of offering such treatments to patients. The field of health status assessment has evolved to meet this need, and several patient completed instruments have been developed that are valid, reliable, and sensitive to treatment.¹ Such tools can provide invaluable insights into how a treatment affects outcomes such as function and quality of life, which can be more important to patients than survival.^{2 3 4} For some conditions, however, measures of disease specific health status are not available, and global assessments of clinical change are used.

Although valid approaches are available for measuring global change experienced by patients,^{5 6} simple ad hoc measures are often used and can be assessed from the perspective of the patient or the doctor. In the accompanying survey of trials included in systematic reviews, Evangelou and colleagues determine whether patients and doctors differ in their assessments of global change after experimental treatments.⁷

Given the reliance of studies on global assessments of clinical change and the convenience of the doctor doing the assessment, it is important to understand the correlation between patients’ assessments and doctors’ assessments. The studies included in Evangelou and colleagues’ review were mostly from the fields of rheumatology and psychiatry. The authors found a close correlation between the two assessments, which suggests that either should suffice when measuring the benefits of treatment from the patient’s perspective. However, although they performed a thorough literature review, and used the data in a methodologically rigorous fashion, their findings may provide false reassurance.

In the field of cardiology, doctors’ assessments have been shown to be biased and unreproducible. The New York Heart Association (NYHA) classification of functional status in heart failure is meant to measure patients’ symptoms and functional limitations on a four point scale. But interobserver variability is high—agreement between independent assessors is only 55%.^{8 9} Similarly, clinical investigators’ and objective interviewers’ assessments of the severity of angina have been shown to be biased.¹⁰ If reproducible assessments cannot be determined at a single point in time, it is unlikely that doctors will be able accurately to represent changes over time. Yet, as Evangelou and colleagues note, such formal investigation of doctors’ ability to measure change has rarely been performed.

Rather than rely on such crude assessments, clinicians should demand that clinical trials investigate patients’ perspectives of their disease through the serial assessment of valid, reliable, and responsive disease specific health status questionnaires.¹¹ For example, the Kansas City cardiomyopathy questionnaire, a disease specific measure for patients with heart failure, is more sensitive to clinical change than the NYHA,six minute walk test and several generic instruments.¹² Moreover, such measures can identify which domains (for example, symptoms, physical function or social function, etc) are most affected by treatment. This information allows patients to have a more realistic expectation of the benefits from a proposed treatment. When disease specific measures do not exist, generic measures of health status should be considered. Only through the use of psychometrically sound measures can clinicians obtain an accurate assessment of how a treatment affects a patient’s disease process and effectively communicate that information to the patient.

Several possible explanations exist for the discrepancy between Evangelou and colleagues’ findings and the well documented limitations of doctors’ assessments of patients’ health status. Firstly, because the authors used summary data from reviews rather than raw data from clinical observations, they had to rely on group means and may not have appreciated the individual variability of each doctor’s assessment. Secondly, the clinical manifestations of the diseases studied may correlate better with patient’s quality of life than other conditions such as cardiovascular disease. Thirdly, the different measurements of clinical change used in the studies included in the reviews may minimise important differences in perspectives of clinical change that would have been more evident with better assessment measures.^{5 6}

Interestingly, failure to control for variability in the duration of the studies and the potential for recall bias regarding patients’ original health status is more likely to have introduced biases that produce greater differences between patients’ assessments and doctors’ assessments. Finally, because patients are often at a nadir in their disease when enrolled in a clinical trial, a background trend for improvement may have artificially accentuated the similarities between the patients’ assessments and the doctors’ assessments.

Researchers should be encouraged to study the correlations between alternative methods of assessing patients’ clinical status in other diseases and clinical settings. More importantly, however, serial changes in patients’ health status, as assessed by patients themselves, should increasingly be used to document the benefits of treatment. This would obviate the need to rely on global assessments from either the patients’ perspective or the doctors’ perspective.