What was POSCH?

POSCH (The Program on the Surgical Control of the Hyperlipidemia) was a randomized trial published in 1990 (1). It involved 838 subjects with a prior heart attack (secondary prevention), most of whom were men. POSCH was designed to "ascertain whether lowering of total plasma cholesterol is directly associated with a reduction in overall mortality" (2). Thus, the primary aim of the trial was to test the effect of lowering blood cholesterol levels on total mortality:

The primary response variable, or endpoint, of POSCH is death from any cause.

POSCH was unique in the sense that it used a surgical procedure on the small intestine, called "partial ileal bypass," to achieve the lower cholesterol levels. By employing this procedure, participants would experience sustained cholesterol reductions throughout the trial. Also, the effect of cholesterol reductions would not be confounded by other factors, because the surgery does not result in other effects aside from cholesterol lowering. Allegedly.

The Basic Setup

To test the effect of cholesterol lowering, the researchers randomized the 838 participants into two groups. One group received the cholesterol-lowering surgery (421 subjects) whereas the other group did not (417 subjects). Both groups were also given advice to eat a diet lower in fat and cholesterol (American Heart Association Phase II diet). The mean follow-up period of the trial was 9.7 years.

Thus, the trial seemed to be solid test of cholesterol lowering. It was randomized, supposedly involved an intervention that isolated the effects of sustained cholesterol reductions, and lasted almost a decade. Furthermore, if you listen to cholesterol-lowering supporters, you would get the impression that the trial was a resounding success (3):

When reported in November 1991, the study's chief investigator, University of Minnesota surgeon Henry Buchwald, launched a national publicity blitz proclaiming this trial was the most convincing proof yet of the benefits of cholesterol lowering.

But in reality, the results of POSCH were unimpressive, and the trial was methodologically flawed.

First, let us look at what the results actually showed.

The Results

When the formal trial ended after almost 10 years, there were 62 deaths (14.87%) in the control group and 49 deaths (11.64%) in the surgery group. This represents an absolute risk reduction of only 3%. Moreover, the result was not statistically significant (P = 0.17).

Therefore, the trial only showed weak evidence for the surgery, at best. And even if we assume that the mortality reduction was the result of the surgery and not due to random error, the small risk reduction means that the surgery failed to prevent the majority of expected deaths over the given period.

Also, it is difficult to ignore that those assigned to the surgery experienced significant side effects. As such, there was little evidence of improved quality-of-life (4):

The surgery-assigned group experienced diarrhea and an increased incidence of kidney stones and gallstones compared to the control-assigned group.

Keep in mind that these poor results were seen despite the long duration of the trial, the highly selected group of participants (including many with genetic lipid disorders), and the sustained lower total cholesterol levels and LDL cholesterol levels in the surgery group.

To put into context: One year before the publication of POSCH, the DART trial (fish-arm) showed a mortality reduction in just two years (5), and four years after the publication of POSCH, the Lyon Diet-Heart Study also found a mortality reduction in two years (6). Notably, the absolute (and relative) risk reductions in DART and Lyon were larger, statistically significant (P = 0.01 and 0.02), and were seen without lowered cholesterol levels.

In fact, there is a curious finding from POSCH that is rarely mentioned. When an observational analysis of the POSCH population was conducted, the authors found that total cholesterol and LDL cholesterol were not associated with mortality (7):

Only the all visits actual values for triglycerides and VLDL cholesterol are statistically significantly related to overall mortality, with an increase in these values associated with a lower risk.

Thus, POSCH could not even show a harmful association between cholesterol and mortality, let alone causation.

And it gets worse. The POSCH trial (and its data-dredging post-trial papers) should have never been published as evidence for lowering cholesterol in the first place. Why? Because the effect of modulating cholesterol was confounded . . .

Confounding Issues

If you recall, it was stated that the surgery (partial ileal bypass) apparently does not cause other effects aside from cholesterol lowering. Well, this is false. The surgery group in POSCH had a lower BMI, a mean weight loss of 5.3 kg, and a decrease in blood pressure (8):

The intervention group at 5 years had significantly lower body weights and significantly lower systolic and diastolic BP.

It is no wonder that weight loss surgeon, Dr. John Morton, described POSCH as a "randomized trial comparing traditional medical management versus surgical weight loss" (9). Furthermore, weight loss may have led to another associated benefit reported later on: a reduction in diabetes (10).

Therefore, the confounders in POSCH are obvious, as long noted in 1993 (11):

Most cholesterol lowering interventions can influence coronary heart disease and overall mortality risk through mechanisms other than the cholesterol lowering itself. Examples include the fact that partial ileal bypass results in weight loss, blood pressure reduction, and improved glucose tolerance, in addition to cholesterol lowering.

In other words, the reductions in BMI, weight, blood pressure, and diabetes invalidate POSCH as a "cholesterol-lowering" trial.

There are also other potential confounders that may be overlooked, such as plant sterols. Plant sterols are linked to adverse health effects (12,13), and partial ileal bypass is one method of reducing their levels (14):

It is noteworthy that partial ileal bypass surgery has been known to, and infrequently used to, decrease plant sterol levels.

Of course, these confounding factors are all fatal flaws by themselves, but the trial also suffered from major methodological issues.

Methodological Problems

One major limitation of POSCH is the lack of blinding. The patients, clinical personnel, and physicians all knew which participants belonged to surgery group and the control group. Although surgical trials are often difficult or impossible to blind, lack of blinding is an inherent limitation that cannot be dismissed.

It is often stated that blinding does not matter with objective endpoints such as mortality. But this claim is based on observational (meta-epidemiological) studies — i.e., weak observational data (15). And as others already pointed out, blinding does matter, even for objective outcomes like mortality (16,17):

It is not hard to see that differential attention and ancillary care across treatment groups can influence even hard endpoints such as mortality.

For example, there were more control patients taking at least one cholesterol-lowering medication, which cannot be assumed as an advantage. Cholesterol-lowering drugs available at the time did not reduce mortality and were actually linked to adverse effects in some cases. Not to mention the potential adverse effects of mixing medications.

The main point is: If cholesterol-lowering medications could be so imbalanced between the groups — due to lack of blinding — then other factors affecting mortality could have been too (i.e., the risk of "performance bias" is high).

Also, POSCH did not have adequate allocation concealment. POSCH used a restricted form of randomization (permuted blocks), which is no longer an acceptable method in non-blinded trials (18). As Berger notes (19):

Improper randomization, such as permuted block randomization, invites the type of selection bias that can masquerade as a treatment effect even when the treatment in fact is no more effective than the control (or placebo).

Lastly, there is no information on dietary and lifestyle habits. Adherence to dietary instructions were only documented in patient interviews by a study nurse practitioner, but there is no published information whatsoever. Moreover, no information exists on other lifestyle habits (Buchwald, personal communication). Therefore, confounding by dietary and lifestyle habits cannot be excluded.

Conclusions

The POSCH trial is often hailed as strong evidence for lowering blood cholesterol levels. But the truth is, the trial was of such low quality, that it was closer to an uncontrolled epidemiological study than a controlled scientific experiment.

On top of this, there has never been any attempted replication of the trial, which alone is good reason to view the results with skepticism. In fact, looking outside the bubble of POSCH, the vast majority of cholesterol-lowering trials have failed to produce a mortality benefit (a topic for future posts!)

Thus, POSCH cannot be taken as good evidence for partial ileal bypass. Nor could it be taken as evidence for cholesterol lowering. Under the conditions, POSCH did not and could not assess the effect of lowering cholesterol levels.


References

1) Buchwald, H., Varco, R. L., Matts, J. P., Long, J. M., Fitch, L. L., Campbell, G. S., ... & POSCH Group*. (1990). Effect of partial ileal bypass surgery on mortality and morbidity from coronary heart disease in patients with hypercholesterolemia: report of the Program on the Surgical Control of the Hyperlipidemias (POSCH). New England Journal of Medicine, 323(14), 946-955.

2) Buchwald, H., Matts, J. P., Fitch, L. L., Varco, R. L., Campbell, G. S., Pearce, M., ... & POSCH Group. (1989). Program on the Surgical Control of the Hyperlipidemias (POSCH): design and methodology. Journal of clinical epidemiology, 42(12), 1111-1127.

3) Moore, T. J. (1994). Lifespan: New Perspectives on Extending Human Longevity. Touchstone Books.

4) Buchwald, H., Fitch, L. L., Matts, J. P., Johnson, J. W., Hansen, B. J., Stuenkel, M. R., & Brooks, H. B. (1993). Perception of quality of life before and after disclosure of trial results: a report from the program on the surgical control of the hyperlipidemias (POSCH). Controlled clinical trials, 14(6), 500-510.

5) Burr, M. L., Gilbert, J. F., Holliday, R. A., Elwood, P. C., Fehily, A. M., Rogers, S., ... & Deadman, N. M. (1989). Effects of changes in fat, fish, and fibre intakes on death and myocardial reinfarction: diet and reinfarction trial (DART). The Lancet, 334(8666), 757-761.

6) De Lorgeril, M., Renaud, S., Salen, P., Monjaud, I., Mamelle, N., Martin, J. L., ... & Delaye, J. (1994). Mediterranean alpha-linolenic acid-rich diet in secondary prevention of coronary heart disease. The Lancet, 343(8911), 1454-1459.

7) Buchwald, H., Boen, J. R., Nguyen, P. A., Williams, S. E., & Matts, J. P. (2001). Plasma lipids and cardiovascular risk: a POSCH report. Atherosclerosis, 154(1), 221-227.

8) http://www.jarcet.com/articles/Vol3Iss2/Buchwald.htm

9) http://www.cardiosource.org/News-Media/Publications/CardioSource-World-News/2013/June/Feature-Interview-John-Morton-MD-MPH.aspx

10) Buchwald, H., Oien, D. M., Schieber, D. J., Bantle, J. P., & Connett, J. E. (2017). Partial ileal bypass affords protection from onset of type 2 diabetes. Surgery for Obesity and Related Diseases, 13(1), 45-51.

11) Smith, G. D., Song, F., & Sheldon, T. A. (1993). Cholesterol lowering and mortality: the importance of considering initial level of risk. British Medical Journal306(6889), 1367-1373.

12) Köhler, J., Teupser, D., Elsässer, A., & Weingärtner, O. (2017). Plant sterol enriched functional food and atherosclerosis. British journal of pharmacology, 174(11), 1281-1289.

13) Makhmudova, U., Schulze, P. C., Lütjohann, D., & Weingärtner, O. (2021). Phytosterols and Cardiovascular Disease. Current Atherosclerosis Reports, 23(11), 1-8.

14) Goldstein, M. R., Mascitelli, L., & Pezzetta, F. (2008). Point: Statins, plant sterol absorption, and increased coronary risk. Journal of clinical lipidology2(4), 304-305.

15) Herbert, R. D. (2020). Controversy and Debate on Meta-epidemiology. Paper 2: Meta-epidemiological studies of bias may themselves be biased. Journal of clinical epidemiology, 123, 127-130.

16) Berger, V. W. (2011). Can objective endpoints be manipulated in unmasked trials?. Statistics in medicine30(30), 3573-4.

17) Berger, V. W. (2016). Risk of bias: Worse than it appears. International journal of nursing practice, 22(1), 111-111.

18) Berger, V. W., & He, X. (2015). Concealing block sizes is not sufficient. Clinics in orthopedic surgery7(3), 422-423.

19) Berger, V. W. (2016). Risk of selection bias in randomized trials: further insight. Trials17(1), 1-5.