Charles J. Beasley, Jr’s response to Barry Blackwell’s reply
Barry Blackwell: Corporate corruption in the psychopharmaceutical industry
Collated by Olaf Fjetland
First, I thank Barry for what I consider an extreme compliment in his response to my comments on his essay on pharmaceutical industry corporate corruption. To be compared with regard to professional contribution to Alfred Pletscher is a compliment of which I am not sure I am worthy.
While Barry and I probably disagree to some extent on the relative capabilities and efficiencies of free enterprise, for-profit corporate entities, and bureaucratic governmental entities, I believe that there are some points about which we would be in robust agreement. For example, I strongly believe that if legitimate market circumstances (e.g., a small market), tacit collusion or overt collusion (the latter illegal and corrupt) limit competition as a positive influence on price control, then government might well need to intercede with price controls or some influence on price.
As some might be aware from reading the recent “Perspective” in the New England Jounal of Medicine (Green and Padula 2017), a law has recently been passed in the state of Maryland criminalizing price gouging of this nature by generic pharmaceutical manufacturers when competition is limited because of few generic manufacturers. I fully support such legislation as it is not directed at expensive and risky, innovative development and is directed at the absence of competition as an influence on price.
Barry makes an extremely important point that the true value of any treatment is the balance between its benefits and its adverse effects and that we often do not learn of some adverse effects until long after approval. This statement about finding adverse effects long after approval is quite true and is an unfortunate consequence of the statistics of small numbers. Consider a serious event caused by an experimental drug with an incidence of 1 in 1,000 persons in a 6-week observational period where the background incidence approaches 0 (incidence in a placebo- or active-control group and in the drug-treated group when not due to the drug is 0). The required sample size for observing a statistically significant (α<0.05) difference in incidences with only 80% power with equal sample sizes per experimental drug treatment and control groups, using Fisher’s Exact Test, is 9,742 for each treatment group (Power Analysis and Sample Size Software 2017). This sample size does not take into account any dropouts that would adversely impact the ultimate power of the study.
The problem grows worse if the event of interest has a “high” background incidence, compared to the incidence due to drug causation, in the population with the disorder for which the medication is being developed. For example, consider a background incidence of 5% or 1 case in 200 persons (5 cases per 1,000 persons) which is 5-fold the incidence due to drug causation. In this second case, the sample sizes required are 87,851 per treatment group. This second case could be a realistic scenario with an event such as myocardial infarction.
If the event occurs early in treatment (e.g., the majority of cases within six weeks of treatment initiation) due to idiosyncratic, acute toxic effects that is of some advantage in definitive detection. However, if the event is due to a progressive effect over an extended time period, then not only is a large number of subjects required but a lengthy period of observation is required, and due to patient attrition from clinical trials the number of subjects beginning study grows even greater.
An event with a true incidence of 1 in 1,000 while “rare” would occur in 1,000 individuals per 1,000,000 treated during clinical utilization of our hypothetical medication. Commercially successful medications for the treatment of common medical conditions will be used to treat potentially multiple millions of patients, and therefore there would be personal interest for some individuals in understanding risks on the order of magnitude of 1 in 1 million treated.
Development programs that even hint at the possibility of rare (i.e., ≤1 per 1000 incidence) adverse events that are caused by the drug under development (e.g., observe 2-3 events in the experimental treatment group and none in the control group) when the background incidence approaches zero are large and costly. Programs for the definitive identification of events of such incidence are essentially impossible, especially if there is a background incidence of the event exceeds the incidence for the event caused by the drug.
For a demonstration of efficacy when statistical significance reaches only the <0.05 threshold in testing the primary efficacy hypothesis, the conventional standard is a repetition of the finding in two independent experiments. If that same standard of proof was required for definitive demonstration of an adverse safety effect in a development program, then the development program would grow even more impractical. In the safety space, there is rarely a prospective hypothesis of an adverse event and it is unusual to adjust for the multiplicity of adverse events observed, that might or might not be caused by the experimental drug. Therefore, such replication would be even more important for a definitive determination that the experimental drug was causally related to an observed adverse event than for definitive determination of efficacy with its prospective hypothesis and primary outcome.
The statistical examples above deal with events about which there are no prospective hypotheses. Some events might be of prospective interest as a result of perceived class effects. In such instances, studies might be conducted that rely on non-inferiority tests where the incidence with the drug (or some other descriptive statistic) does not exceed the incidence with the control treatment by some pre-specified degree. Two examples of such studies/analyses conducted in pharmaceutical development are: 1) the Thorough QT (TQT) Study required to demonstrate that a non-cardiac drug candidate does not result in delayed cardiac ventricular repolarization (lengthened QTC) of a magnitude that would increase risk for Torsades de Pointes; and 2) an analysis (often a meta-analysis of multiple Phase III studies) to exclude excess risk of major adverse cardiovascular events (MACE) with hypoglycemic agent for treatment of diabetes mellitus. The TQT study can be quite small because it is based on a comparison of mean changes from baseline as a surrogate biomarker rather than incidence of an event, but the required sample sizes in the MACE study/analysis are quite large.
Barry notes that “. . . only two RCT’s required for market approval by FDA.” Two RCTs is the number of positive efficacy studies required under routine circumstances for market approval. In psychiatric indications, these are almost invariably required to be placebo-controlled efficacy studies. Because it is extremely unusual in most psychiatric disorders to achieve positive outcomes in two of two studies, sponsors generally have to conduct three or more efficacy studies (usually more). Furthermore and quite importantly, there are expectations for the minimum number of patients treated for various periods of time with the new drug seeking approval. These requirements are generally met through the conduct of one or more large “safety studies” with an active-comparator or no comparator (although cheaper and faster, a wise sponsor doe does not choose the no comparator option for a variety of reasons). The minimum exposures generally required are 1,500 patients treated with one or more dose of the drug, 300-600 treated for six months, and 100 treated for one year (Guideline for Industry 1995). Most sponsors exceed these minimum requirements in pre-approval development programs by wide margins. For example, in one development program on which I worked with NDA submission in 1995, all patient exposure was 2,500 and a maximum period of treatment was more than four years.
Finally, thanks to Barry for his initial detailed review and stimulus for this important dialogue.
References:
Green JA, Padula WV. Targeting unconscionable prescription-drug prices – Maryland’s anti-price-gouging law. NEJM 2017; published online June 7, 2017; DOI: 10.1056/NEJMp1704907. Accessed June 8, 2017.
Guideline for Industry. The Extent of Population Exposure to Assess Clinical Safety: For Drugs Intended for Long-term Treatment of Non-Life-Threatening Conditions. ICH-E1A March 1995 www.fda.gov/ohrms/dockets/ac/04/briefing/2004-4068B1_09_ICH-E1A-Guidelines.pdf. Accessed 29-Sep-2017.
PASS 15 Power Analysis and Sample Size Software (2017). NCSS, LLC. Kaysville, Utah, USA, ncss.com/software/pass. The computational output provided to INHN.
January 11, 2018