Non-inferiority Trial Design in Drug Development

A Primer for Cardiovascular Healthcare Professionals

Fabio Angeli; Paolo Verdecchia; Gianpaolo Reboldi


Am J Cardiovasc Drugs. 2020;20(3):229-238. 

In This Article


Randomized placebo-controlled clinical trials are the gold standard for evaluating a novel therapeutic agent. However, in some instances it may be unethical to plan a placebocontrolled clinical trial. Ethics-based restrictions on the use of placebos as comparators have enhanced the viability of noninferiority trials as alternatives for registration purposes, for risk management once the drug is on the market, and for a marked increase in confidence in the new drug.[16,45]

Major advances in cardiovascular drug discovery generates a seemingly endless supply of new drugs; however, in most cases, these new drugs provide only incremental improvements in efficacy, safety, or the overall risk-benefit ratio; thus, over time, the publication of noninferiority trials using an active comparator has increased.[2,16,45]

Specifically, a noninferiority design is desirable if the new treatment is expected to have a similar efficacy benefit (no worse than) compared with the existing treatment and to have an improved safety profile or some other secondary benefit for the patient. In this case, the new treatment would be considered an improvement over the existing treatment as long as it was noninferior to the existing treatment for key efficacy endpoints.

Some recommendations on how to report the design and conduct and to analyze the results of noninferiority randomized trials in publications should be followed. The following items should be carefully detailed: the rationale for choosing a noninferiority design; the results of trials used to base the active control effect; relevant changes in patient characteristics compared with the previous trials evaluating the active control effect; outcome measures from the previous trials of the active control, including changes in timing of assessment; and interpretation of the results that accounts for sources of potential bias or imprecision.[2,13,14,17]

In this regard, a recent systematic overview[2] determined the number and features causing methodological or reporting limitations related to noninferiority designs among cardiovascular noninferiority trials published in three of the highest-impact medical journals. Bikdeli et al.[2] noted important methodological limitations, including use of wide noninferiority margins and sizeable under-enrollment or post-randomization exclusions. Furthermore, over onethird of trials did not have a published methods paper or study protocol; one in eight of those with a methods paper or protocol had discrepancies related to the noninferiority margin compared with the published manuscript.[2] Finally, several trials suffered from a constellation of factors that could predispose the results towards noninferiority, potentially undermining the confidence in the trial results.[2]

In this review, we provided a comprehensive discussion on the purpose of and issues involved with noninferiority trials. Specifically, key issues and practical considerations to define the noninferiority margin and the methods to analyze noninferiority are summarized.

These aspects are crucial for researchers and clinicians to allow for optimal judgement of noninferiority trials and to assess the quality of noninferiority comparisons conducted in clinical practice.

In this context, it is worth mentioning that the threshold for the maximum loss in efficacy of a new therapy when compared with an established standard treatment (noninferiority margin) is usually chosen using statistical considerations and clinical judgment. Nonetheless, it does not integrate patients' values and preferences.

Recently, patient input incorporating self-reported outcomes (including quality of life) has been suggested as a meaningful measure to be integrated in the evaluation of treatment differences.[46,47] Indeed, it is obvious that clinicians and patients may have different preferences in the choice of a treatment.

We share the opinion elucidated in a recent viewpoint by Acuna et al.[46] suggesting that new techniques incorporating patient-oriented measures and preferences would allow for more meaningful design and interpretation of noninferiority trials.

Nonetheless, in our opinion, such an approach should ideally be used in conjunction with well-established techniques to define potential non-efficacy advantages of a new therapy in cardiovascular medicine.[13] Only the combined use of outcomes and patient-oriented measures of a novel treatment can safely guide both population-based public health policy and individual care in primary care settings.