In the wonderful world of modern medicine, there are pesky little exceptions to every rule.  For example, cardiologists tell us that people with high LDL (bad) cholesterol numbers are more at risk for heart attack. Yet we know that not everybody with dangerously high cholesterol levels suffers a heart attack, and not all heart attack patients have high cholesterol. In France, cardiologists puzzle over the “French Paradox” in which French citizens who eat more high-fat dairy foods, smoke more, and exercise less than all of their European neighbours have one of the lowest rates of heart disease in Europe.

This is what statisticians call intermediate or surrogate endpoints. It means you can’t really predict the final consequences based on how things are going along the way.

Consider the 2010 World Cup in South Africa for a similar illustration of intermediate endpoints. Sports pundits, for example, insist that, generally speaking, the team that can maintain possession of the soccer ball for most of the match will be the team that ultimately wins.  Makes sense, doesn’t it?  However, ball possession alone may not actually mean much at all to the end result.  Here’s why: 

First, let’s look at some early World Cup matches that did indeed seem to prove the accepted theory.

• Chile owned the ball for 56% of their match against Honduras, and won 1-0.
• The Netherlands had 58% ball possesion against Denmark, and won 2-0.
• Germany controlled the ball for 55% of the match against  Australia, and won 4-0.

But on the other hand, consider these clear exceptions to this theory:

• England possessed the ball for 54% of their match against the U.S.A. but only managed a 1-1 tie.
• France had 53% ball possession during their game against Mexico, but lost 2-0.
• The #2 ranked team from Spain dominated ball possession for 63% of their match with Switzerland, yet the Swiss won 1-0 in a stunning upset.

Similar puzzling results have come in medical research, explains senior reporter Andrew Holtz at the medical information network MDiTV.

For example, he cites recent studies on patients with diabetes that included aggressive control of blood sugar, high blood pressure and cholesterol in people considered to be at very high risk for heart attacks.  Oddly enough, this research showed that:

• strict management of blood sugar did not reduce heart attack deaths
• reduction in high blood pressure levels did not reduce heart attack deaths
• controlling high LDL cholesterol numbers with the use of statin drugs did not reduce heart attack deaths

Holtz explains that lab results may not actually be accurate predictors of mortality – they are merely intermediate or surrogate endpoints along the way.  And just because a drug improves lab test results doesn’t mean it saves lives – despite the efforts of Big Pharma to convince drug prescribers otherwise.

A perfect example of this is the patient information that the drugmaker Pfizer includes on every single bottle of its cholesterol-lowering blockbuster statin drug Lipitor:

“Lipitor has not been shown to prevent heart disease or heart attacks.”

I find this astonishing – given that Lipitor has been the biggest-selling drug on the planet, prescribed by millions of physicians for one reason only: so that their patients won’t have a heart attack!  Big Pharma has yet again shown remarkable success in convincing doctors to prescribe a drug that the drug’s own manufacturer tells us does not achieve the insinuated end results.

This is truly a triumph of marketing-based medicine.

Similarly, clinical trials showed that the statin drug simvastatin (Zocor) was effective in reducing cholesterol (the surrogate endpoint) without showing directly that simvastatin prevents cardiac events. In fact, proof of Zocor’s efficacy in reducing cardiovascular disease was finally only presented by Norwegian researchers five years after its original introduction, and then only for secondary prevention in those already diagnosed with heart disease.(1)

In another statin drug case, AstraZeneca has been accused of marketing rosuvastatin (Crestor) without providing hard endpoint data, relying instead on surrogate endpoints. The company simply replied that it had been tested on larger groups of patients than any other drug in the same statin class, and that its effects should be comparable to the other statins.

Dr. Laurence S. Freedman of the National Cancer Institute in the U.S. wrote about intermediate endpoints in the American Journal of Epidemiology back in 1992.  He explained that, when studying a potential intermediate endpoint in a clinical drug trial for example, these five questions must be asked:

• Does the treatment affect the intermediate endpoint?
• Is the intermediate endpoint associated with risk factors?
• Is the intermediate endpoint associated with the main outcome?
• Is the treatment effect on the main outcome mediated by the intermediate endpoint?
• Are the prognostic or risk factor effects mediated by the intermediate endpoint?

But ultimately, just as in World Cup football, all that really counts is the final score.

◊    ◊   ◊

For more World Cup wisdom off the soccer pitch, see also: Adidas vs Nike: the World Cup Battle of the Brands

For a shocking example of how an ineffective drug for acute heart failure was approved based on intermediate endpoints, see: How Did This Heart Drug Get Approved In the First Place?

.

(1) Pedersen TR, Olsson AG, Faergeman O, et al. (1998). “Lipoprotein changes and reduction in the incidence of major coronary heart disease events in the Scandinavian Simvastatin Survival Study (4S)”. Circulation 97 (15): 1453–1460.

.