Challenges in evaluating patients lost to follow-up in clinical studies of rotator cuff tears

Long-term follow-up studies are necessary to critically evaluate the outcome of a treatment intervention for a specific disorder. However, patients may cease participating in a long-term study and become lost to follow-up; thus, their current condition is unknown. The underlying characteristics that predispose a patient to become lost to follow-up are difficult to identify and control. Patients who are lost to follow-up may be contacted by telephone; however, the effect of administering a functional assessment questionnaire by telephone compared with that of mailing a questionnaire is unknown. The purpose of this study was to compare patients who continued to respond to requests for follow-up with those who did not. A second purpose was to compare responses obtained by mail with those obtained by telephone interview.

Two hundred and twenty-four patients with a rotator cuff tear were enrolled in an ongoing study of shoulder function and general health. Self-assessment questionnaires were mailed to every patient at six-month intervals. Sixty-seven patients (30 percent) regularly responded to mailings (identified as responders in this study), fifty-five patients (25 percent) responded occasionally (these patients were not included in the analysis), and 102 patients (46 percent) ceased to respond and became lost to follow-up (identified as nonresponders in this study). This investigation was performed to determine: (1) the characteristics of nonresponders compared with those of responders, (2) the functional status of nonresponders as assessed with a questionnaire over the telephone, and (3) the effect of administering a self-assessment functional questionnaire by telephone compared with that of sending the same questionnaire by mail.

Nonresponders tended to have lower initial scores for the mental health summary (p = 0.03) and for social function (p = 0.01), were less likely to have had surgery (p = 0.009), and were less likely to consume alcohol (p = 0.03). At the last known time when they completed the mailed questionnaire, nonresponders reported significantly worse shoulder function than responders (p = 0.0001). However, on telephone questioning the mean number of shoulder functions that the nonresponders indicated that they could perform was greater than the mean number documented on their last mailed questionnaire (p < 0.0001). In a random subgroup of responders, the mean number of functions that the patients indicated that they could perform when interviewed by telephone was significantly greater than the number indicated on their most recent mailed questionnaire (p < 0.01). The results obtained by telephone from this random subgroup of responders were similar to those obtained by telephone from the nonresponders.

There are differences between patients who continue to participate in a study and those who become lost to follow-up. Functional assessment questionnaires administered by telephone yield different results than the same questionnaires sent by mail. These considerations are relevant to the design, implementation, and interpretation of clinical studies in which functional questionnaires are used.