Key to the practice of evidence- and outcome-based medicine is data that are trustworthy. Sponsors of novel medical therapies are required to provide data from well-controlled studies to support their claims of safety and efficacy.

The demands on scientific researchers and clinicians are increasing with more competition for grants from the National Institutes of Health and other agencies, demand for profitability by investors, and growing clinical costs in an environment of decreasing relative clinical resources. Clearly, a for-profit firm has a known conflict of interest—a financial one. However, as I pointed out previously, so does everyone else, and these potential conflicts must be disclosed. Recently, the editors of the leading American ophthalmology journals upgraded the disclosure requirements for their journals, as well as adding checks for plagiarism. There is a call for increasing transparency in clinical trials, especially those sponsored by pharmaceutical firms. While transparency is a seemingly simple and unobjectionable goal, I want to share with readers some aspects of the current status of clinical research conducted by for-profit pharmaceutical firms.

Colleagues have expressed to me concerns about the selection of biased analyses that favor a particular hypothesis in trials sponsored by pharmaceutical firms. Of particular concern are post hoc analyses, especially on subsets of the population—that is, “painting the target after the arrows are shot.” I highlight an aspect of Good Clinical Practices, a set of international guidances for the conduct of quality clinical research (International Conference on Harmonisation). There are also companion guidances for conducting preclinical safety studies, Good Laboratory Practices, and manufacturing, Good Manufacturing Practices. These guidances are part of the regulations in most countries, and are followed by pharmaceutical firms developing new medications. I refer to the guidance E9, Statistical Principles for Clinical Trials. This guidance states that the Sponsor must identify a priori the primary outcome and analysis population. Of course, subsequent analyses may be performed, but these must be identified as such. Such secondary analyses may inadvertently provide spurious and misleading results. That said, sometimes secondary or post hoc analyses may be of great importance (eg, the relationship between control of intraocular pressure and visual field deterioration in the Advanced Glaucoma Intervention Study and central corneal thickness as a risk factor in glaucomatous progression in patients with ocular hypertension in the Ocular Hypertension Treatment Study).

Reproducibility and independent substantiation of experimental results is key in scientific research. Based on the 1962 Kefauver-Harris amendment to the Federal Food, Drug, and Cosmetic Act, for approval of a new drug, the US Food and Drug Administration (FDA) requires at least 2 controlled studies providing substantial evidence of safety and efficacy. These studies are typically planned at 80% power or greater. Based upon the Food and Drug Administration Modernization Act of 1997, the FDA will consider a single trial as the basis for approval, pending several assumptions, including a very clear clinically and statistically significant treatment effect. Typically, pharmaceutical firms conduct at least 2 pivotal studies at 80% power to support registration, whereas National Institutes of Health extramural programs conduct 1 study, albeit usually of 90% power.

Is the FDA requirement of 2 clinical studies justified? In the 1980s, I was involved in the development of a novel ocular hypotensive agent. As part of the development, and based on the pharmacology and pharmacokinetics of the molecule, we decided to evaluate the compound once daily. We conducted a single trial against a positive control. In that trial, we found a statistically and clinically significant benefit of the new agent over the positive control. Given that this result was somewhat unexpected, we repeated that study twice and were unable to replicate that finding. The simplest explanation is that even at a P value of .05, 1 in 20 times, one may obtain an incorrect answer.

Are the controls (including record keeping) as prescribed in Good Practices that important? I am reminded of a multi-year study conducted in nonhuman primates at a major research university. The authors initially reported the major finding that chronic use of (±)3,4-methylenedioxymethamphetamine (“ecstasy”) caused toxicity to dopaminergic neurons. Subsequently, they found that owing to a simple labeling issue, the primates actually received methamphetamine, a drug whose neurotoxicity was already well accepted. Although errors may happen to anyone, if this study was conducted in compliance with the guidances, I suggest that such an error would have been highly unlikely.

Recently, the US FDA approved ranibizumab (Lucentis) for the additional indication of treatment of diabetic retinopathy in patients with diabetic macular edema. This was apparently NOT an a priori primary outcome measure. However, I assume that the FDA used its guidance regarding efficacy evidence that this was a highly clinically and statistically significant finding, especially in a condition for which there was no approved treatment.

One might question whether the Consolidated Standards of Reporting Trials statement, adopted by the top ophthalmology journals, would cover the above statistical issues. Indeed, the statement does speak to some of the issues regarding statistical planning. However, it is a brief checklist, typically reviewed and completed at the end of a clinical trial during the manuscript preparation process. Good Practices, including guidance E9, are regulatory guidelines to be followed from the initial planning through to the reporting of clinical trials.

It is well recognized that there is inherent bias against publishing negative data in a host of research areas. In particular, this issue was raised in clinical trials, wherein if only positive trials were published, scientists would not know the “denominator”—that is, how many trials were at risk. The scientific community responded to this issue over 10 years ago with journal editors requiring up-front registration of clinical trials for all sponsors (industrial, academic, and government). This later became law in the United States in 2007, and elsewhere. More recently, and as anticipated in the 2007 law, the NIH and the FDA have asked for more rigorous registration and reporting of results.

Not all researchers follow these guidances. The FDA has a public listing of clinical investigators against whom disqualification proceedings have been conducted, as well as a list of pharmaceutical professionals who have been debarred. Good medical practice and the best interests of the patient require that physicians use legally available drugs, biologics, and devices according to their best knowledge and judgment. However, pharmaceutical firms are allowed to promote only for the indications on the approved package insert. The US Justice Department has brought charges against several companies for off-label promotion.

I recommend that ophthalmologists continue to carefully read the literature—primary study results as well as meta-analyses and reviews. They should continue to be physicians in how they use such findings in deciding on the best course of treatment for each patient.