Post-marketing studies of new insulins: sales or science?

Post-marketing studies

Post-marketing research has received little scientific scrutiny. Current procedures to license a drug represent a compromise between the need to make potentially valuable new drugs available without undue delay and the need to assure an acceptable level of safety. This means that relatively few patients have been exposed to a given agent by the time it is licensed. The drug in question may then require further investigation in special patient subgroups, in special situations, or for other indications that might give useful pointers to its future use. Such studies may result in an application for a change in the prescribing information or “label.” Phase IV studies, defined by ClinicalTrials.gov as “post-marketing studies [which] delineate additional information including the drug’s risks, benefits and optimal use,”5 are intended to fulfil this function. A further and more recent development has been the observational study, defined as “a study that provides estimates and examines associations of events in their natural settings without recourse to experimental intervention.”6 The European Medicines Agency currently uses the terms interventional and non-interventional studies to clarify the distinction.4

From the regulatory standpoint, pharmacovigilance is the main justification for non-interventional studies. The US Food and Drug Administration has advised that “postmarketing safety data collection and risk assessment based on observational data are critical for evaluating and characterising a product’s risk profile” and has produced non-binding recommendations as to how this might be achieved.7 In practice, and reasonably enough, drug companies often seek added value from post-marketing studies. This may (for example) be obtained by investigations designed to show clinical effectiveness as well as safety. Since such studies typically introduce the new drug to many clinical investigators in different environments, they also serve a tacitly acknowledged marketing function. Post-marketing studies thus represent a fine balance between the marketing advantage sought by the companies and the safety data sought by the regulators, a balance which, as the following analysis reveals, may require careful scrutiny.

Insulin analogues

In 2007 insulin accounted for 48.4% of the total UK drug bill for diabetes related prescriptions,8 and in 2010 global sales for insulin exceeded $12bn (£8bn; €9.5bn). Insulin was obtained from animal pancreas until the 1980s, when biosynthetic human insulin began to replace animal insulins, with important supply implications but no clinical advantage.9 The result was the first of two worldwide switch campaigns, in the course of which smaller competitors were eliminated by the cost of producing genetically engineered insulins. Three manufacturers now dominated the market.

A second switch campaign began when human insulin was replaced by genetically engineered insulin analogues at 2–4 times the cost. Company reports for 2010 show that insulin glargine, the most successful analogue, helped to give Sanofi-Aventis insulin sales of around $5.1bn, compared with $4.7bn for Novo-Nordisk and $3.1bn for Eli Lilly. Evidence based reviews have consistently concluded that that the insulin analogues offer little benefit to most people with type 2 diabetes, except those who experience troublesome hypoglycaemia.10 11 12 13 Nor are they cost effective as first line treatment for type 2 diabetes.14 Skilled marketing has been needed to overcome these limitations.

This analysis reports the use of post-marketing studies by the three main insulin manufacturers since trial registration was introduced in 2005. It argues that marketing and clinical functions of post-marketing studies need to be kept in proper balance and queries whether better regulation might ensure that they are. The search strategy and findings are outlined in box 1.

Summary of findings

Nearly 400 000 people have been recruited into post-marketing studies of the insulin analogues, and one company alone has sponsored studies involving nearly 360 000 individuals. Not all participants were exposed to analogue insulins, however, estimated enrolment may differ from actual recruitment, and not all relevant studies were registered.

The novel feature of the Novo Nordisk studies is their enormous scale. To take one example, A1chieve, a study that set out to “remedy the deficit of data on the efficacy and safety of insulin analogues in less well resourced/newly developed countries” recruited 66 726 people from 3166 centres in 28 nations in 18 months (see box 2).17 The scientific limitations of a study with no control arm and which lacked a well defined question or hypothesis are readily apparent. Furthermore, the “deficit of data” seems puzzling, given that more than 100 000 patients had already been studied by the company in non-Western environments, including 21 of the 28 countries involved in A1chieve. The study was extravagantly powered to detect an adverse effect such as hypoglycaemia. The European Medicines Agency has pointed out that a sample size calculation implies a question—for example, a comparison between therapies or evaluation of a risk reduction strategy.4 In the event, severe hypoglycaemia fell from a remembered rate of 0.81 events a year in previous users of insulin to an almost undetectable level of <0.01 events a year by the end of the study. This is an implausible outcome in individuals who increased their insulin dose by an average of 10–20% while experiencing a 2% fall in HbA_1c, and the authors acknowledge that there may have been a problem in the way in which hypoglycaemic events were recorded. The protocol required a severe hypoglycaemic event to be confirmed by blood glucose measurement, but, since patients who are used to hypoglycaemic episodes and their helpers rarely see the need to measure blood glucose at the time of an episode, the apparent reduction in the number of events might simply have been because episodes were not confirmed by blood glucose measurement.

The European Medicines Agency notes that prescription of a medicine should be “clearly separated from the decision to include the patient in the study.”4 The extent to which this was possible in A1chieve is unclear; the patients seem to have been actively recruited, provided signed consent, and completed questionnaires on two occasions—more like study participants than a population under observation. The authors acknowledge that increased attention and greater engagement by the patients may have contributed to many outcomes unrelated to the use of analogue insulins, such as the fall in systolic blood pressure of 6 mm Hg. In the event, this massive study found only minor differences between the three insulin regimens. Patient outcomes did improve, but in predictable ways (such as improved wellbeing and glucose control in poorly controlled patients converted to insulin). Last but not least, the role of the insulin analogues in bringing about these benefits—the main focus of the study—remains undefined, and the relevant comparison with cheaper human insulins was not undertaken.17

The scientific value of such data might be doubtful, but the commercial rationale is good. The regulatory environment is undemanding (Good Clinical Practice regulations do not have to be observed in non-interventional studies), the outcomes are more positive, there is no requirement to publish negative findings, and the studies pay for themselves. Let us take a hypothetical example. Insulin prices vary enormously from one country to another,18 but, for illustrative purposes, I will extrapolate from costs in the UK. The PREDICTIVE trial19 was sponsored by Novo Nordisk, ran from 2004 to 2010 in 26 countries (15 in Europe), and recruited 47 565 individuals already taking or newly transferred to insulin detemir for the six month study; the cost of the prescription was absorbed by the patient or healthcare system. Participating physicians in the UK received £125 per patient, to a maximum of £2500. Rates of remuneration differed from one country to another, but at UK rates nearly £6m would have been paid directly to participating physicians (this may be an overestimate because UK rates are likely to be higher than in less developed countries). The administrative costs must have been considerable, but patients are likely to have remained on the new insulin at the end of the study, leaving up to 47 000 people on long term treatment with insulin detemir. The estimated average cost for a year’s treatment in the UK is £408, as compared with £109 for human NPH insulin supplied in vials and made by the same company.20 Extrapolating to the whole study, this would yield ongoing annual sales of £19.2m, as against £5.1m for human insulin. The company was thus well placed to recover its costs, to acquire more than 40 000 long term users, to influence the prescribing habits of thousands of physicians,21 and (potentially) to claim tax exemption for this research activity.

Why has one company invested so heavily in post-marketing studies when the others have not? The answer can only be speculative, but the other companies have self evidently not perceived the same scientific or clinical need for additional information about their products. The legislation controlling such studies is considered less stringent in the EU than in the US, which might explain why Eli Lilly has undertaken notably fewer studies of this type, and why none of the Novo Nordisk observational studies was done in the US. Commercial pressure on Novo Nordisk to regain market share lost to insulin glargine, the first and most successful long acting analogue on the market, must have been high. There would have been the further incentive to establish a strong market position in less well resourced but developing economies.

What is the role of post-marketing studies?

A wider question relates to the role and legitimacy of post-marketing studies, a topic that has attracted little academic interest. Well designed observational studies can produce estimates of treatment effect similar to those found in randomised controlled clinical trials, but many of the post-marketing studies considered here did not ask a clear or novel research question, addressed the same issues, lacked any comparator group, and seem larger than needed to answer the questions proposed. Safety is a key issue, but studies that are excessively large, short term, uncontrolled, and loosely supervised (A1chieve recruited 66 000 patients from 3166 clinical centres in 18 months, which must have posed a logistical nightmare for the company) are unlikely to offer much information of value in this respect. Novo Nordisk files reports from all its studies on its website,15 and many publications are listed. Some are well rated, but most have appeared in journals catering for this type of report, typically with coauthors or writing support from the company, and describe selected subanalyses with positive outcomes. For example, a PubMed search for terms “insulin detemir” and “PREDICTIVE study” identified around 25 published subanalyses of data from PREDICTIVE, but no main study report.

It might be argued that even poor quality studies could have useful spin-off effects, such as by setting higher standards for clinical care or by introducing the potential benefits of analogue insulins to new environments. The rejoinder is that healthcare improvement can be done much more cheaply and effectively outside the context of poor quality research and that there is little or no evidence to show that the benefits of the analogues outweigh their costs in routine use, especially in less wealthy environments. Poor quality studies merely impede progress.

Since marketing can be guaranteed to expand in any direction that is not expressly forbidden, this type of study will continue unless tighter regulation is introduced. This should provide a much clearer definition of what constitutes a legitimate non-interventional study. One proposed definition is “a study where the medicinal product(s) is (are) prescribed independent to [sic] inclusion of the patient in the study and as part of a therapeutic strategy, including diagnostic and monitoring procedures, which is not decided in advance by a study protocol.”22 Further criteria should include a scientifically valuable question, prospective design, a comparator group wherever possible, a representative population, and adequate (but not excessive) power to answer the research question.23 Eli Lilly has proposed similar criteria for observational studies.24 Last but not least, a company’s activities should be judged by the regulatory and legal standards of the country from which it originates, rather than those in which its research is undertaken.

Conclusions

It seems reasonable to ask whether a company would invest in such large scale activities without a good commercial rationale. Participating physicians, particularly those with no previous research experience, may participate in good faith, unaware that one aim of the study is to influence their prescribing habits. Their judgment is nonetheless open to question, particularly when their own participation is remunerated and their patients are exposed to extra cost for little benefit. Patients may volunteer on altruistic grounds or with inflated expectations of personal benefit. The patient or healthcare system pays for a more expensive agent instead of one that is cheaper and equally effective, and the public is offered misleading claims of comparative merit based on studies of limited scientific value.