This is an example of how you can play with data to get a positive result: and also that with major amounts of money in a study you better get a positive result. It comes from a large multicentre trial of a novel antidepressant compound that modifies glutamate, basimglurant. The authors used a standard depression clinician rated outcome, the MADRS (Montgomery Asberg Depression Scale), and a patient instrument, the Quick Inventory Depressive symptoms, (QIDS-16).
Both are fairly well established. But they got different results. Interestingly, the patient filled MADRS showed a significant change and the clinician one did not. There is a placebo arm in this study and it may be that the placebo effect influences clinicians more than patients, or that the changes between the arms are of little clinical utility.
But my bias is to follow the patient instruments. Because clinicians have biases.
The trouble is the primary outcome in the protocol is the MADRS, clinician form. This means that you get an abstract, if reported correctly, that reads like this.
Design, Setting, and Participants In this phase 2b, double blind, randomized clinical trial of 333 adult patients with a DSM-IV-TR diagnosis of MDD across 59 research clinics globally, patients were assigned to 1 of 2 doses of basimglurant MR (0.5 or 1.5 mg) or placebo once daily, adjunctive to ongoing antidepressant medication therapy (selective serotonin reuptake inhibitor or serotonin and norepinephrine reuptake inhibitor). Patients were enrolled from October 5, 2011, through July 26, 2013.
Interventions Six-week treatment with 0.5 mg of basimglurant MR, 1.5-mg basimglurant MR, or placebo once daily, adjunctive to ongoing antidepressant medication therapy.
Main Outcomes and Measures The primary end point was the mean change from baseline score on the Montgomery-Åsberg Depression Rating Scale (MADRS), as rated by the clinician at week 6. Other measures included patient-rated MADRS, Quick Inventory of Depressive Symptomatology–Self-Report, Clinical Global Impression–Improvement, Patient Global Impression–Improvement, and Clinical Global Impression–Severity Scales and adverse events.
Results A total of 596 patients were screened, and 333 were randomized into the study (mean [SD] age, 47 [11.2] years; 216 female [65.1%]). The primary end point (mean change in clinician-rated MADRS score from baseline to end of treatment) was not met (effect size [ES]?=?0.16, P?=?.42; intent-to-treat [ITT] mixed-effects model for repeated measures [MMRM] analysis for comparing 1.5-mg basimglurant MR and placebo). Across secondary and exploratory end points, 1.5-mg basimglurant MR revealed larger improvements vs placebo on the patient-rated MADRS (?16.2 vs ?13.3, ES?=?0.28, nominal P?=?.04), Quick Inventory of Depressive Symptomatology–Self-Report (?7.5 vs ?5.8; ES?=?0.37, nominal P?=?.009), Clinical Global Impression–Improvement mean score, and Patient Global Impression–Improvement mean score. Improvements were also seen in the patient-rated MADRS remission rate (36.0% vs 22.0%; nominal P?=?.03) and response rate (50.5% vs 40.4%; nominal P?=?.13), A 0.5-mg dose of basimglurant MR had no benefit over placebo in any of these measures. The most common adverse event was dizziness, which was mostly transient and of mild intensity.
As a practicing clinician, this paper has no relevance. It is not a phase III trial: those are needed. There may be effects in the “Placebo” arm from their concurrent antidepressant, as continuation of the same medication over a longer period of time will show more improvement. This medication is neither being sold nor marketed.
But if I was designing the next study, I would be tempted to make the QIDS, not the MADRS, the primary outcome. None of these instruments are perfect. And it is fairly clear that the instrument you use matters.