Why Do We Really Need Clinical Trials?

A point I make over and over again when talking about new or alternative therapies that are not supported by good clinical trial evidence is that lower-level evidence, such as theoretical justifications, anecdotes, and pre-clinical research like in vitro studies and animal model testing, can only be suggestive, never reliable proof of safety or efficacy. It is necessary to begin evaluating a new therapy that does not yet have clinical evidence to support it by showing a plausible theory for why it might work and then moving on to demonstrate that it actually could work through pre-clinical research, which includes biochemistry, cell culture, and animal models. These sorts of supporting preclinical evidence are what we refer to when we refer to the “prior plausibility” of a clinical study. But this kind of evidence alone is not sufficient to support using the therapy in real patients except under experimental conditions, or when the urgency to intervene is great enough to balance the significant uncertainty about the effects of the intervention.

In support of this conclusion, we can consider the inherent unreliability of individual human judgments and all the many ways in which inadequately controlled research can mislead us. And we can reflect on how promising results in early trials often melt away when better, larger, more rigorous studies are done that better control for bias (the so-called Decline Effect). And it is not at all difficult to compile a large list of examples of the harm inadequately studied medical interventions can cause.

But what I’d like to do here is focus on a particularly good specific example of why thorough clinical trial evaluation of promising ideas is not just a nice extra to confirm what we already believe is true, it is the only way to genuinely know whether our treatments to more good than harm.

SELECT: What Is It About?

The Selenium and Vitamin E Cancer Prevention Trial (SELECT) was initiated because, according to the National Cancer Institute (NCI) at the National Institutes of Health (NIH), “Evidence from epidemiologic studies, observational studies, and clinical trials looking at preventing cancers other than prostate cancer had suggested that selenium and/or Vitamin E might prevent prostate cancer.” The NCI describes some of the specific studies that suggested selenium and Vitamin E supplementation might protect against prostate cancer.


Selenium is a nonmetallic trace element found in food, especially plant foods such as rice and wheat, seafood, meat, and Brazil nuts. Selenium is an antioxidant and may help control cell damage that could lead to cancer.

The Nutritional Prevention of Cancer (NPC) study, first reported in 1996, included 1,312 men and women who had a history of non-melanoma skin cancer. Results of the trial showed that men who took selenium to prevent new non-melanoma skin cancers received no benefit from selenium in preventing that disease. However, approximately 60 percent fewer new cases of prostate cancer were observed among men who had taken selenium for six and one-half years than among men who took placebo (1). In a 2002 follow-up report, the data showed that men who took selenium for more than seven and one-half years had about 52 percent fewer new cases of prostate cancer than men who took placebo (2). This trial was one of the reasons for studying selenium in SELECT.

Vitamin E

Vitamin E is in a wide range of foods, especially vegetables, vegetable oils, nuts, and egg yolks. Vitamin E, like selenium, is an antioxidant, which may help control cell damage that can lead to cancer.

In the 1998 study of 29,133 male cigarette smokers in Finland (known as the Alpha-Tocopherol, Beta-Carotene Trial, or ATBC), 32 percent fewer new cases of prostate cancer and 40 percent fewer deaths from prostate cancer were observed among men who took Vitamin E in the form of alpha-tocopherol to prevent lung cancer than among men who took a placebo.

So in 2001, the trial began enrolling subjects in a randomized, blinded, placebo-controlled prospective clinical trial to see if supplementation of selenium and Vitamin E, or a combination of the two, would reduce the incidence of prostate cancer. By 2004, over 35,000 men in the United States, Canada, and Puerto Rico had been enrolled. The subjects were randomly assigned to receive:

  • Selenium and Vitamin E
  • Selenium and a placebo
  • Vitamin E and a placebo, or
  • Two placebos. Two placebos were used in the trial: one looked like a selenium capsule; the other looked like a Vitamin E capsule. Each placebo contained only inactive ingredients. Neither the participants nor the researchers knew who received the selenium and Vitamin E, or the placebos, a process known as blinding or masking.

The trial was originally planned for between 7 and 12 years of supplementation and then subsequent followup to monitor for the development of prostate cancer, and it was designed to detect a 25% reduction in the incidence of prostate cancer, which was felt to be an achievable and  clinically meaningful difference.

So How Did It Work Out?

In 2008 the investigators reviewed the data up to that point and found no sign of a protective effect from selenium or Vitamin E, separately or taken together. It was judged unlikely that the 25% reduction in risk that was the criteria for effect in the study would be seen even if the study continued, so the supplementation portion of the study as halted. The subjects were told which supplement or placebo they had been given and were instructed to stop taking their supplements.

At that time, a greater number of cases of prostate cancer had developed in the group taking only Vitamin E than in the other groups. However, the difference did not reach statistical significance, so the investigators could not say whether it was a meaningful difference or only the result of chance.

Subsequently, the men have continued to be monitored, and the difference in the rate of prostate cancer between men who had been taking only Vitamin E supplements and the other men in the study has continued to grow even after they stopped taking the supplement. The difference is now statistically significant, so it likely represents a real biological phenomenon, not simply random chance. Men who were in the Vitamin E alone group have developed prostate cancer at a rate 17% higher than the men in the other groups.

Those men who took selenium alone or selenium and Vitamin E also have a higher rate of prostate cancer than those men who received the placebos, but this difference is not yet statistically significant. No other differences suggesting a beneficial or harmful effect of either supplement have been detected.

So Why Didn’t It Work Out As Expected?

From a biological point of view, the reason why Vitamin E supplementation turned out to be a risk factor for development or prostate cancer instead of a protective factor aren’t yet known. Nor is it clear why a statistically significant increase in risk hasn’t yet been found for those men who took Vitamin E and selenium together. There are, of course, theories as to how precisely Vitamin E influences prostate cancer risk, and the NCI has solicited research proposals from other investigators to use the data and samples collected in SELECT to investigate this question.

But from the point of view of epistemology, how we figure out what works and what doesn’t, there are a number of reasons why the finding shouldn’t be entirely surprising. For one thing, new ideas often turn out to be wrong. The degree of plausibility to the underlying theory and the amount of supportive pre-clinical research evidence are useful in determining roughly how promising an idea is, but ultimately the proof is in the pudding, with the pudding being multiple replicated independent high-quality randomized clinical trials. Given the incredible number of complex, interacting factors involved in the development of each particular disease, it shouldn’t surprise us when even a reasonable hypothesis with limited support from observational trials and other lower level evidence turns out to be false.

And in the case of this particular hypothesis, the preliminary evidence which first suggested the need for the SELECT in the first place didn’t work out so well either. As already mentions, “results of the [Nutritional Prevention of Cancer] trial showed that men who took selenium to prevent new non-melanoma skin cancers received no benefit from selenium in preventing that disease.” And there has since been some indication of potential harm, though the data is conflicting; “Since the start of SELECT, four studies have been published on the effect of selenium on blood glucose and risk of diabetes. Two studies suggested that higher levels of selenium taken from supplements or received naturally were associated with an increased risk of diabetes. One study showed no such association, and one showed that people with higher levels of selenium in their blood had a reduced risk of diabetes (3-6).”

As for Vitamin E, the NCI states, “There are no clinical trials that show a benefit from taking vitamin E to reduce the risk of prostate cancer or any other cancer or heart disease (7, 8, 9-13).” And the subjects in the  HYPERLINK “” ATBC trial who took Vitamin E had a 50% increase in the occurrence of hemorrhagic strokes, an 18% increase in the occurrence of lung cancer, and an 8% overall greater mortality than control subjects. So not only were the initial promising results suggesting a benefit for prostate cancer not correct, there were fewer benefits and greater risks than expected in terms of other diseases as well.

The Bottom Line

The moral of this story is simple and shouldn’t be controversial: While a plausible theory and suggestive preclinical evidence may suggest a new therapy could be beneficial, only multiple, rigorous, high-quality clinical trials can determine if this is actually true. Good ideas turn out to be wrong often enough that our initial enthusiasm with the ideas themselves and with supportive preliminary research data should always be tempered by an understanding that these lower levels of evidence are inherently less reliable than the slow, expensive, complicated process of rigorous clinical studies. And bad ideas, those without a plausible theoretical foundation or supportive pre-clinical evidence, are even less likely to bear fruit when properly examined.

Of course, such high-quality studies may not always be possible. As a veterinarian, I can never expect to have a 7-year prospective study of 35,000 subjects to rely on.  So the temptation to act on lesser evidence is understandable. But therapies used without higher-quality evidence to support them are inevitably going to be more likely to fail or to cause harm than therapies that have been more thoroughly studied. We must accept this and factor that knowledge into our clinical decisions. If the need to intervene is great enough, it can be justifiable to employ therapies without high-quality clinical trial evidence to support them. However, we must convey the real degree of uncertainty to our patients or clients, and we must be especially vigilant and honest about potential failures or unintended consequences.

It is common to see sweeping and confident claims of benefit and safety for alternative therapies that, when examined, turn out to be based on suggestive evidence often much weaker than that which led the NCI to study the effect of selenium and Vitamin E on prostate cancer risk. CAM advocates have gotten quite good at throwing up publication references to support their claims, and the hassle factor involved in examining those papers in detail to see what level of evidence they actually provide can discourage anyone from thoroughly investigating these claims. This, of course, leaves the impression that there is good evidence to support what CAM advocates are saying, not only in the minds of the general public but in the minds of physicians, veterinarians, nurses, and others who should know better.

I think much of the success of the “integrative medicine” meme has been based on the lack of an adequate understanding among health professionals about the serious limitations of low-level evidence. The SELECT illustrates nicely how even a plausible intervention with enough low-level evidence to justify a major clinical trial can prove not only less helpful than originally hoped but even actively harmful. The same principle applies to an even greater degree to less plausible hypotheses. High-quality clinical trials are not simply icing on the cake confirming what we already know, they are the cake without which we know a lot less than we usually think.


  1. Clark LC, Combs GF Jr., Turnbull BW, et al.  HYPERLINK “″ Effects of selenium supplementation for cancer prevention in patients with carcinoma of the skin. A randomized controlled trial: Nutritional Prevention of Cancer Study Group. JAMA 1996; 276(24):1957-1963
  2. Duffield-Lillico AJ, Reid ME, Turnbull BW, et al.  HYPERLINK “″ Baseline characteristics and the effect of selenium supplementation on cancer incidence in a randomized clinical trial: A summary report of the Nutritional Prevention of Cancer Trial. Cancer Epidemiology, Biomarkers & Prevention 2002; 11(7):630-639.
  3. Stranges et al.  HYPERLINK “″ Effects of Long-Term Use of Selenium Supplements on the Incidence of Type 2 Diabetes. Annals of Internal Medicine; 147:217-233, 2007.
  4. Bleys J et al.  HYPERLINK “″ Serum Selenium and Diabetes in U.S. Adults. Diabetes Care; 30:829-834, 2007.
  5. Rajpathak et al.  HYPERLINK “″ Toenail Selenium and Cardiovascular Disease in Men with Diabetes. Journal of the American College of Nutrition; 24: 250-256, 2005.
  6. Czernichowet et al.  HYPERLINK “” Antioxidant supplementation does not affect fasting plasma glucose in the Supplementation with Antioxidant Vitamins and Minerals (SU.VI.MAX) study in France: association with dietary intake and plasma concentrations. American Journal of Clinical Nutrition; 84:395-9, 2006.
  7. Lippman SM, Klein EA, Goodman PJ, et al. Effect of selenium and vitamin E on risk of prostate cancer and other cancers. JAMA 2009; 301(1). Published online December 9, 2008. Print edition January 2009.
  8. EA Klein, IM Thompson, CM Tangen, et al. Vitamin E and the Risk of Prostate Cancer: Results of The Selenium and Vitamin E Cancer Prevention Trial (SELECT). JAMA 2011; 306(14) 1549-1556.
  9. Yusuf S, Dagenais G, Pogue J, et al. Vitamin E supplementation and cardiovascular events in high risk patients. The Heart Outcomes Prevention Evaluation Study Investigators. New England Journal of Medicine. 2000;342:154-60.
  10. Sesso HD, Buring JE, Christen WG, et al. Vitamins E and C in the prevention of cardiovascular disease in men: the Physicians’ Health Study II randomized controlled trial. JAMA. 2008; 300(18):2123-33.
  11. Lee IM, Cook NR, Gaziano JM, et al. Vitamin E in the primary prevention of cardiovascular disease and cancer: the Women’s Health Study: a randomized controlled trial. JAMA. 2005; 294(1):56-65.
  12. Lonn E, Bosch J, Yusuf S, et al. Effects of long-term vitamin E supplementation on cardiovascular events and cancer: A randomized controlled trial. JAMA 2005; 293(11):1338-1347.
  13. Miller ER III, Pastor-Barriuso R, Dalal D, et al. Meta-analysis: High-dosage vitamin E supplementation may increase all-cause mortality. Annals of Internal Medicine 2005; 142(1):37-46.

Posted in: Cancer, Clinical Trials, Herbs & Supplements, Science and Medicine

Leave a Comment (11) ↓

11 thoughts on “Why Do We Really Need Clinical Trials?

  1. rork says:

    I ofcourse agree with most of this, but:

    1) For vitamin E, might the smoker’s benefit have been real? That’s a little question. I read the link, but it didn’t ask that question, but did mention the doses were higher than in the ATBC trial (maybe too high being the possibility for failure).

    2) more serious and often asked: “If the need to intervene is great enough, it can be justifiable to employ therapies without high-quality clinical trial evidence to support them.” Even if you aren’t doing it as part of a trial? Have you not just written Burzynski’s (and every other quack who wants to follow) free pass to do whatever seems plausible next? Oh, I think the wording “need to intervene” is overstatement of “situation is dire”, and curiously manages to add the idea that intervening helps, when that is the very question being asked.

  2. Harriet Hall says:

    Just one minor quibble: I don’t think you necessarily have to have a plausible theory for why a new treatment might work. If there were enough credible preliminary evidence that it did work, that would justify further study. If the evidence from further studies were strong enough, we would be justified in using the treatment before we understood the mechanism.

  3. @rork

    1) One additional piece of information about the ATBC Trial:

    “The lower prostate cancer incidence rates in participants taking alpha-tocopherol supplements during the trial returned toward normal soon after the trial ended, but remained below the placebo group rates throughout the six-year post-intervention period.”

    This raises the possibility that, while the lower rate of prostate cancer in the treatment group might have been due to the Vitamin E, it also could be a selection bias artifact. The rate remained lower in the treatment group compared to the placebo group even after supplementation was stopped. So either the Vitamin E had some lasting effect even when the subjects were no longer taking it, or the subjects in the treatment group had an intrinsically lower prostate cancer rate for other reasons. Randomization doesn’t, of course, completely guarantee the elimination of unknown confounding factors.

    In any case, a single trial is rarely sufficient evidence to conclusively demonstrate an effect unless the effect is unusally large, and it can never definitively confirm a hypothesis that it was not initially designed to test. That is, after all, why the SELECT was undertaken. The balance of the evidence, especially the lack of benefit and an indication of possible increased risk in the trial specifically designed to test the effect of Vitamin E supplementation of prostate cancer risk, seems pretty solidly against the hypothesis despite the preliminary results of the ATBC Trial.

    2) As for the question of when does the urgency of the clinical situation justify acting on the basis of imperfect evidence, I think it is a complex one without a definitive objective answer. Of course, people can use the desperation of life-threatening illness to justify doing ridiculous things. On the other hand, most of conventional chemotherapy and cardiac surgery and other clearly beneficial interventions were initially used on patients with no other good options because these therapies killed a lot of patients. The long-term result has been to identify true benfits and to improve our understanding of the risks so we can make rational decisions about balancing the two. And granted, some of the means by which we got to this point might not be considered ethical today. But I don’t think science-based medicine necessarily requires never intervening unless the evidence for our therapy is ironclad. That is, afterall, a charicature of evidence-based medicine that CAM proponents use to argue that it is worthless because it makes the perfect the enemy of the good.

    Certainly, in my field, I would be unable to treat almost every clinical problem I see if I had to eschew interventions for which there was only low-level supporting evidence. Many such interventions may welll turn out not to be as useful as I think they are, but there is a wide spectrum between gold-stanard clinical trial eviddence on one hand and pure irrational quackery on the other, and though some like Burzynski exploit the gray areas in that, it doesn’t mean that any therapy with imperfect or incomplete supporting evidence is quackery.

  4. Harriet,

    I mostly agree with you. Lots of plausible theories turn out to be wrong, and as you say sometimes things turn out to work even though we don’t understand all the details of why.

    I would say that preclinical evidence, such as in vitro studies of the effect of a drug on tissue, lab animal model studies, etc, usually exists in the first place because someone had a hypothesis about why the intervention might work, so for most things I imagine there is at least a theory behind the idea before it reaches the stage of clinical trials. If that theory is highly implausible, that weakens the value of the preclinical evidence, though it doesn’t invalidate it entirely. That does then raise the isse of whether investing resources in clinical trials, and exposing patients to potential risks, is justified when there is no theoretical reason to think the intervention might be effective.

    As I was trying to indicate to rork, the situation in which I think it might make sense to study something that seems implausible is when the need is great and the options currently available are few. Such a situation tips the balance towards pursing a therapy even without plausibility and with only pre-clinical evidence.

    Still, most wacky ideas really do turn out to be wrong. So while plausibility, as Sir Austin Bradford Hill said, “depends upon the biological knowledge of the day,” and we should bear in mind such knowledge is itself imperfect, I do think plausibility is a important factor in helping us decide how to allocate limited resources for investigating novel therapies. However, I probably did overstate the case when I implied it was absolutely necesary.

  5. Harriet Hall says:

    I certainly agree about the importance of prior plausibility. I was thinking about things like minoxidil (Rogaine). It was a blood pressure medication, and users noticed that increased hair growth was a side effect. Now it is being sold to promote hair growth, although the mechanism is still not understood.

  6. annappaa says:

    “Plant foods such as rice and wheat … seafood?! … meat?! … and Brazil nuts”?

    But to respond seriously, this is really interesting. I am currently researching the evidence behind the idea that eating yogurt will prevent yeast infections, and while it does seem that there is some interesting pre-clinical evidence, there doesn’t seem to be much of any in the way of quality clinical trials in humans. The ones that do exist are poor quality and the results are conflicting. But there are studies that suggest that some of the “probiotics” in yogurt can interfere with the growth or attachment of Candida albicans. And there are also studies that show that these probiotic bacteria can’t adhere to vaginal cells either! But these were all performed in a lab, not in human subjects.

    These issues are similar to those I came across while researching the piece I wrote on cranberry juice for urinary tract infections last year. Really intriguing pre-clinical studies seem to show that chemicals in cranberries can physically alter the shape of E. coli‘s appendages, rendering them unable to attach to cells in the urinary tract. So we seem to have a plausible underlying mechanism by which cranberries can be helpful … But so far the clinical trials performed in humans have yielded incredibly underwhelming results.

    I’m all for home remedies if they actually work, so it’s disappointing that the evidence just isn’t there!

  7. rork says:

    Perhaps Dr. Hall was pointing at me. I did say plausible in perhaps not quite the most common manner (for here) it’s true, and shoulda done better. By plausible, today, I meant how likely I think it is to work, which encompasses knowledge of mechanism, and empirical data. For example antibody against ERBB2 might work on patient’s tumor with high ERBB2, even if it isn’t breast cancer. Mechanism there is almost as nothing (to me, fairly ignorant of all but the most obvious details) compared to the data from using in on breast cancer patients. Hell, you could even call it something bewildering like herceptin or trastuzumab and I wouldn’t care.

    It’s the von Hoff paper that is what is still nagging on me (pubmed 20921468). I worry it will inspire heroic doctors to “experiment” outside the context of trials. I worry it already has. I don’t see where the breaks on this vehicle are.

  8. pmoran says:

    .There will always be tension between our profession’s dual obligations.

    1. To maintain and enlarge a body of reliable medical knowledge. .

    2. To wholeheartedly serve the interests of that person sitting in our office.

  9. David Weinberg says:

    Nice post and an important subject.

    There is a similar story for carotenoids and lung cancer. There was theoretical and epidemiologic evidence that these supplements might reduce the rate of lung cancer in high risk patients. 2 randomized trials showed that, paradoxically there was an increased risk of lung cancer. One trial was performed in Finland:

    And one in the USA

  10. anoopbal says:

    Nice article! And this is the crux of the problem: People don’t understand why science evolved in the firs place and the pt falls of anecdotal evidence.

    Another example is the anti-arrhythmia drugs in the 80’s. It killed more people than the Vietnam war and everyone was convinced about their efficacy based on biological plausibility and observations – until we did an RCT.

Comments are closed.