Asking for larger safety databases in clinical trials geared to drug approval could be a cost-effective means of cutting down on adverse reactions once products are on the market, a US analysis suggests.

Researchers at Duke University School of Medicine in Durham, North Carolina developed a model to calculate the magnitude of expected benefit – as measured by avoidance of adverse drug events (ADEs) – as well as the potential cost-effectiveness of requiring larger versus smaller clinical safety databases for regulatory approval of a hypothetical new medicine.

To increase the model’s “real-world applicability”, associate professor of medicine Shelby Reed and colleagues drew on reported rates of cardiovascular and cerebrovascular events among patients enrolled in clinical trials for COX-2 inhibitor painkillers.

As the researchers pointed out in the journal Health Affairs, detection of ADEs is “rarely” considered when sample sizes are calculated for industry-sponsored clinical trials, most of which are powered to measure differences in the primary efficacy endpoint that will drive regulatory approval.

The sample sizes recommended by the International Conference on Harmonisation – e.g., pre-market clinical safety data on 1,500 patients, 300-600 of whom have been treated for at least six months and 100 of whom have been treated for one year – imply “that pre-approval safety evaluations will not characterise rare ADEs”, Reed et al commented, adding: “Yet these databases might not even have adequate power to detect adverse events that occur in as many as 100 patients”.

For extremely rare ADEs, only post-marketing surveillance is practical “or reasonable”, the Duke University researchers acknowledge. But for less rare ADEs – and bearing in mind the associated concerns about pushing up drug development costs – “the methodological benefits of randomisation combined with additional statistical power make a policy that requires larger pre-approval safety databases appealing”, they noted.

Base case

The researchers posited a base-case scenario in which the background annual incidence of the adverse event in the patient population was 0.5% and the odds ratio (OR) to be detected was 2.5. They assumed that data on 2,000 patients per treatment group would be available for the smaller safety database and data on 4,000 patients per treatment group for the larger database.

The researchers also assumed that the US Food and Drug Administration (FDA) would deny approval to the drug when an ADE was detected. If the FDA did approve the drug, the assumption was that the cumulative number of patients treated would be 10 million at increased risk of the adverse event for one year.

In addition, Reed et al conducted sensitivity analyses, evaluating the impact of increasing further the number of patients in the larger safety database, as well as varying the odds ratio to gauge the effect of changing the level of risk associated with the drug, the background incidence of the adverse event, the probability of non-approval by the FDA and the size of the target population.

The researchers then looked at the potential cost-effectiveness of requiring a larger pre-approval safety database, assuming a case fatality rate (based on the COX-2s) of 15% and an average remaining life expectancy of around 22 years. Clinical trial costs were set at US$10,000 per patient.

Statistical power

The analysis found that statistical power for detecting an elevated risk of around 75,000 ADEs was 76% in the safety database with 2,000 patients per treatment group and 96% in the database with 4,000 patients per group.

“Assuming the detection of a statistically significant increase in the risk of ADEs associated with the drug would prevent regulatory approval, approximately 57,000 adverse events on average would be avoided in the target population with the smaller database, and approximately 72,000 would be avoided with the larger database,” the authors commented. “Thus, in the base-case analysis, an additional 15,000 adverse events would be avoided, on average, if larger databases were required to evaluate safety concerns.”

Assuming the FDA would deny approval to a drug with risks corresponding to those factored into the base-case analysis, the incremental cost-effectiveness ratio (i.e., the additional cost per life-year saved in the post-approval patient population were larger rather than smaller safety databases required) was estimated at around US$27,100 per life-year saved in the post-approval patient population when the sample size for safety databases increased from 2,000 to 4,000 patients per treatment group, and around US$70,700 per life-year saved when the sample size increased from 2,000 to 8,000 patients.

When a sample size of 8,000 patients per group was compared with 4,000 patients, however, the incremental cost-effectiveness ratio was about US$361,000 per life-year saved.

The model showed that, in a scenario where smaller databases had adequate statistical power to detect the increased risk of an adverse event, larger safety databases would unnecessarily extend the duration and costs of clinical trials, the researchers concluded. Moreover, where the drug concerned offered therapeutic advantages over existing products, patients would lose out from delayed approval.

In cases where smaller safety databases were inadequately powered to detect ADEs reliably, though, there were potentially substantial gains from requiring bigger safety databases by identifying serious ADEs that would prevent the drug from being approved or prescribed in a large patient population, the researchers noted.

Not detected

It has been estimated that for about half of the drugs on the market, serious adverse events were not detected until after the product had secured regulatory approval, they added. Relying on pre-approval safety evaluation would not be enough to detect increasingly rare adverse events but, as demonstrated in the sensitivity analyses, including high-risk patients in clinical trials “could provide an efficient means of increasing information about drug safety profiles”.

In any case, Reed et al said, post-marketing safety surveillance “is a necessary complement, regardless of changes in requirements for pre-approval safety evaluation. Clinical trials cannot be expected to characterise the risks associated with a drug in real-world clinical use, nor can they reasonably be required to evaluate long-term effects of a drug prior to approval”.

All the same, they argued, regulatory agencies “must come to some agreement about the acceptable level of risk associated with newly approved drug products”, taking into account variations according to the type of adverse event, severity of disease, availability and side-effect profiles of alternative treatments, and levels of risk acceptance from patient to patient. Adverse events “that occur relatively frequently in the target population, particularly those with dire health consequences, should be an obvious focus”.

Summing up, the authors said that in an environment characterised by limited resources, policymakers “must weigh the potential advantages and disadvantages of regulatory requirements for larger safety databases and strengthening safety surveillance efforts after a drug is marketed”. This “complex undertaking” must go beyond considerations of safety and efficacy to include legal, ethical, political and economic factors.