Dropouts from trials

One concern with equivalence and non-inferiority trials is that a positive conclusion of equivalence/non-inferiority could result from an insensitive trial by default. If, for example, equivalence is established then this could mean either that the two treatments are equally effective, or indeed equally ineffective. If chosen endpoints are insensitive, dosages of the drugs too low, patients recruited who are not really ill and the trial conducted in a sloppy fashion with lots of protocol deviators and dropouts, then the treatments will inevitably look very similar Clearly we must ensure that a conclusion of equivalence/non-inferiority from a trial is a reflection of the true properties of the treatments. The regulatory guidelines (ICH ElO) talk in terms of assay sensitivity as a requirement of a clinical trial that ensures this. 

That SSRIs are more acceptable from the patient s perspective than traditional TCAs is largely accepted by clinicians but has recently been questioned in an article by F. Song et al. [1993]. These researchers conducted a metaanalysis of 63 randomized controlled trials comparing the efficacy and acceptability of SSRIs with those of tricyclic and related antidepressants. The article has been cited widely, especially in the United Kingdom. The authors suggest that the dropout rate in patients receiving SSRIs is similar to that of patients receiving TCAs in controlled studies. They concluded that the routine use of SSRIs as the first-line treatment of depressive illness was not... 

Munjack et al. [1991] conducted a pilot study of buspirone in the treatment of social phobia. Subjects meeting DSM-lll-R criteria for social phobia were entered into an 8-week, open-label trial. Buspirone was started at 5 mg twice a day and increased by 5 mg every 2-3 days to a maximum dosage of 60 mg/day, or until side effects prevented further dose escalation. Of the 17 subjects entered in this study, 11 completed it. The 6 dropouts resulted from lack of responsiveness, adverse effects, inability to attend appointments, and a loss to follow-up. At week 6, of the 11 subjects completing the trial, 5 reported a little and 6 endorsed moderate change in their symptomatology. At the end of week 8, two subjects reported a little, 5 noted moderate, and 4 endorsed marked improvement. Although the global measures demonstrated the above results, instruments used to measure the features specific to social phobia demonstrated mixed results. 

From the perspective of clinical trial methodology, concurrent medications can create a dilemma for the investigator by complicating the interpretation of results. Intermediate rescue medications are often required, however, because mood stabilizers are relatively slow in their onset of action. Further, if rescue medications are avoided, this usually introduces the confound of dropouts before the experimental drug can be fully effective. When feasible, a reasonable compromise is the use of modest amounts of a benzodiazepine (BZD), such as lorazepam, only when necessary for a limited time (e.g., 7 to 10 days) into the active phase of treatment. This can reduce the number of nonresponding, highly agitated patients who may otherwise drop out of treatment and in a trial of several weeks, the initial lorazepam effect should have dissipated by the final assessments. 

A combination of olanzapine and fluoxetine was used in two randomized, double-blind simultaneous 8-week trials in 249 patients with major depression with psychotic features (trial 1 n = 124, mean age 41 years, 52% women trial 2 n = 125, mean age, 41 years, 50% women), which have been jointly published (69). This multicenter study was completed by 51 subjects in trial 1 (41%) and 59 subjects in trial 2 (47%). Altogether, there were no significant differences in the rates of discontinuation due to adverse events among the different treatment groups placebo (n = 100), monotherapy with olanzapine 5-20 mg/day (n = 101), and olanzapine 5-20 mg/day plus fluoxetine 20-80 mg/day (n = 48). Dropout percentages were 59% in trial 1 (similarly distributed in the three groups) and 53% in trial 2 (ranging from 40% of dropouts... 

Topiramate reduces glutamate release from neurons and antagonizes activation of the alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptor, a glu-tamatergic excitatory amino acid receptor. A double-blind, placebo-controlled trial conducted in 296 ALS patients over 12 months showed a faster rate of decline in arm strength and no effect on survival from topiramate (Cudkowicz et al., 2003). This trial had a large dropout rate and proceeded to a phase III trial without first exploring the effects of topiramate in the SOD model. 

A marker validated as Type 0 or Type I may next be validated as a Type II marker or surrogate endpoint. The validation of surrogate endpoints is best done with data from Phase 2 or 3 studies, where dropout rates have been low, the treatment has continued unchanged over the duration of the study, the biomarker was measured early in treatment, patients have been followed for a sufficiently long time, and a difference between control and treatment (especially for placebo-controlled trials) was demonstrated. Note in the case of no difference placebo-controlled trials the biomarker then would be a Type 0 marker. It is desirable to have this type of data from several studies to establish the validity of the marker across studies, drugs, and various patient populations. 

One of the most potent applications of pharmacometrics is the informative construction of clinical trials by using clinical trial simulation (CTS). Population PM models are of great value when used in CTS because estimates of typical parameters along with parameter variability can be incorporated. There are three basic types of models needed to execute a CTS an input-output model, a covariate model, and an execution model. These are described in detail in Chapter 34 of this book. Clinical trial simulation can improve pediatric study structure by examining the impact of many important factors such as dropouts, choosing varying endpoints, and deviations from protocol. Pediatric PM models find great utility when applied to CTS. 

The first portion of this section is an identification of those trials that fulfill the requirements for adequate and well-controlled trials showing that the drug has its intended effect. This is followed by a discussion of those trials as well as the data from all controlled clinical trials. Any differences in outcome between trials of similar design should be explained where possible. Tables showing major trial design features, numbers of subjects, numbers of dropouts, and major outcomes are sometimes useful. 

Subject Sample. Provide a statement describing the total number of subjects expected to complete the trial this number can include an estimate of treatment failures but should not include administrative dropouts. (If subjects are transferred from inpatient clinics, if they relocate, or if they do not complete the trial for any other reason [not drug related], they should not be included in the total.) Establish the number of subjects that must complete the trial and be part of the final statistical analysis. Additional subjects should always be enrolled in a clinical trial to make certain that, minus trial dropouts, the required number of subjects needed for meaningful statistical analyses is still available, and that administrative dropouts will not jeopardize numbers needed for the final statistical analysis. 

A systematic review of adverse events reported in clinical trials of St. John s wort indicated that data from 35 doubleblind randomized trials showed that dropout and adverse event rates in patients receiving St. John s wort extracts were similar to placebo, lower than with older antidepressants, and somewhat lower than with SSRI antidepressants. Dropout rates due to adverse events ranged from 0 to 5.7% in 17 observational studies that included 35,562 patients. No serious adverse events were reported in any of the studies (Knuppel and Linde 2004). 

---