Shortcuts in clinical trials are endangering public health

Conor O’Donovan

Deputy Science Editor

Clinical trials – the tests in which new drugs or techniques are trialled on medical patients prior to widespread release – are carried out to answer important questions about how to improve the way patients are treated. Trial results are reported as “end-points”: measures of the outcome, basically the “beef” of what we want to know about what the drug does and how good it is at doing it. The most important, or “primary”, end-points usually dictate whether the trial is considered to have found an overall benefit, harm, or no difference. They are usually of direct relevance to a patient’s health, such as survival time, or incidence of heart attacks etc.

When events such as these are rare, the number of patients recruited into clinical trials must be large, and the duration of the trial considerable, so as to accrue enough events to detect a significant difference between treatment arms in the trial – a rudimentary principle of biostatistics. These trials are hence expensive, and in the case of trials of new medications, the limited patent time for new pharmaceutical entities means there is an onus on drug companies to get their products approved and on the market as quickly as possible.

Surrogate outcomes, as they are known, are a means of dealing with these issues. A surrogate outcome or end-point is a measurement (commonly the result of a blood test or a physical quantity such as blood pressure) that is understood to be an indicator of the underlying disease process. Many are known risk factors for adverse health outcomes. For example, high blood levels of LDL-cholesterol (the “bad” one) predispose to heart disease and general heart-related death. During a clinical trial, it is much quicker and easier to measure a change in LDL-cholesterol levels and find a difference with an investigative drug, than it is to count hard end-points such as deaths or heart attacks.

It is assumed that the disease is well understood scientifically, and that the treatment directly tackles the processes that lead to the manifestations of disease, which matter to patients.

It is likely that the effect of a drug on the measured characteristics (e.g. blood pressure) may be only one of a range of effects, whereas the actual benefit is derived through others, which are not measured. An example is the potential of statins (a widely prescribed class of cholesterol-lowering drugs) to lower blood LDL-cholesterol levels. In cases such as these, it is difficult to confidently ascribe the benefit of the drug to any one of its effects on numerous physiological processes. Other potential biomarkers may even be negatively affected, and not reported merely for the lack of noticeable symptoms arising in patients, which would prompt their measurement.

The major cautions relating to the use of surrogates in clinical trials concern inappropriate extrapolation from limited trial data to broad predictions about a drug’s clinical effectiveness. While all anti-diabetic drugs reduce blood sugar levels (many were designed to do so, and are marketed as such), not all of them provide any meaningful benefit in terms of hard outcomes such as the incidence of diabetic complications (e.g. blindness or amputation). Some, in fact, have been shown to be actively harmful, e.g. rosiglitazone increased deaths caused by heart disease.

Furthermore, smaller, shorter studies are of little use in revealing rare but serious adverse drug effects, or the cumulative effects of long-term exposure – data which are instead accrued through repeated observations for years following the drug’s approval. By sampling bias alone, smaller studies are also more likely to report findings of significant effect size, which may in reality arise by random chance.

The use of surrogate end-points to determine clinical trial success is widespread, and usually for reasons of cost- and time-efficiency, as well as the apparent advantage of exposing fewer patients to the potential hazards of an experimental treatment. This is particularly salient for companies attempting to gain accelerated approval (which is granted in some cases) for new drugs to treat chronic diseases, where there is a lengthy period before a significant clinical benefit is seen. For example, early cheers of success over the novel oral treatment for cystic fibrosis, ivacaftor (trade name Kalydeco), have been based on gross extrapolations of benefit demonstrated on a few surrogate markers measured in small numbers of patients, from trials of relatively short duration.

In some cases, it may even be seen as unethical to wait for the results of large definitive trials using hard outcomes, instead of the easier-to-obtain surrogates, because of the perception that assumedly beneficial treatments are withheld from those in control/placebo groups for long periods of time.

Indeed, the trial literature is heavy with reports that use surrogate end-points alone. By publishing so extensively using such soft markers, the short-term goals of scientists (trying to prove the worth of their pre-clinical work), clinical investigators (trying to be the first to bring new and better treatments into practice), and pharmaceutical companies (trying to convince everyone, themselves included, of the value of their product), are furthered. It should be clear that this is not always in the interests of improving clinical outcomes that are of material interest to patients.

Instead, by changing the tone of the health conversation in such a pervasive way, toward a focus on surrogates, such as maintaining blood tests within normal ranges, the health beliefs and demands of healthcare users are greatly influenced, and not for their own good. Rather, this can equate to the “medicalisation” of asymptomatic states, such as “pre-diabetes” or “pre-hypertension”, or even variants of normal physiology, on the grounds of very poorly validated surrogate markers of disease. The world of illness encroaches into the lives of more and more people, and the range and scale of interventions patients come to expect from their healthcare providers only increase. The patient who presents saying “Doctor, I want you to check my blood pressure, and give me this drug to get it down”, is not an easy patient to convince otherwise, when in fact the evidence may support not treating hypertension in the individual of his/her risk level.

It is easy to convince the public of the breakthrough nature of new medications, the injustice of withholding “miracle cures” from gravely ill and needy patients, and the stifling power of statutory regulation on industrial innovation and economic growth. It is less easy, on the other hand, to say “No, we just don’t know enough about these drugs – without more information, we can’t say they won’t do you more harm than good in the long-run”. To ethically safeguard the public against unforeseen potential harms, hard outcomes must be mandatory for full market authorisation, and we could incentivise this by extending patent times for drug companies who carry out such important and meaningful trials.