The article “Seven Deadly Medical Hypotheses” by Reynold Spector, MD, in our March/April 2011 issue prompted considerable comment and controversy within the skeptical community. Several physician-skeptics wrote critical blogs, and others let us know they were uncomfortable with much of what Spector had written and the certainty with which he expressed his conclusions. We invited four of the critics to submit short letters for this section. Reynold Spector then responds. —EDITOR
DAVID H. GORSKI
There’s an old recurring Saturday Night Live sketch in which a Scotsman bellows, “If it’s not Scottish, it’s crap!” As I read Reynold Spector’s “Seven Deadly Medical Hypotheses” (Spector 2010), I couldn’t shake the image of him in a kilt roaring, “If it’s not a randomized, double-blind clinical trial based on the hypothetical/inductive method, it’s deadly crap!” Unfortunately, Spector fails to demonstrate convincingly why most of his seven hypotheses even merit the label “deadly.” Sadly, Spector’s overwrought application of the word deadly to hypotheses that are not even, by Dr. Spector’s definition, “deadly” torpedoes what might have been a provocative exercise in skepticism. What is left are weak or erroneous conclusions based on dubious arguments.
Because there are so many questionable arguments in Spector’s article, I will be forced to “cherry pick” a couple. More complete discussions of other examples can be found elsewhere (Gorski 2011a). His first Deadly Medical Hypothesis (DMH) is as good an example as any. Unfortunately, DMH #1 is not even a hypothesis at all but rather two opinions, the second of which is a straw man tacked on to allow Spector to gleefully attack both opinions as one: “Either the investigator does not need a specific hypothesis and/or can use an inadequate method to test the hypothesis.”
Spector cites an editorial as support for the existence of this DMH (Glass and Hall 2008). However, although the authors of the cited editorial did make the provocative argument that in the age of genomic medicine specific hypotheses might not always be necessary anymore and that instead science can focus on answerable questions, nowhere do they advocate using “inadequate methods” to test hypotheses or answer questions. That is Spector’s opinion, and the straw man that gives him the opening to attack “hypothesis-less” studies. Yet hypothesis-generating studies are very important to the process of discovery, because they are how we develop new lines of investigation that might one day come to fruition as useful treatments that can pass FDA muster, standards designed for a very late point in the discovery process.
It’s a long and winding road from hypothesis generation to hypothesis testing to treatment to treatment validation. Most hypotheses fail and are forgotten, and even for those that succeed, the median time between first publication of a discovery or hypothesis and its validation in large randomized clinical trials ranges from fourteen to forty-four years (Contopoulos-Ionnidis et al. 2008). Compared to this, genome-wide association studies (GWAS) have only been in existence for a relatively short time. Ironically, Spector argues that GWAS should be replaced by “direct sequencing of the portions of the genome of interest, comparing patients with proper control subjects.” That is exactly what is happening now (Berger et al. 2011) as next generation sequencing techniques drive the cost of genome sequencing low enough to make such studies feasible. I also note that many of the assumptions behind GWAS that Spector dismisses are also assumptions behind—you guessed it!—next generation sequencing (NGS) experiments of the sort that Spector apparently approves of. Right now, the National Institutes of Health is funding a huge initiative, The Cancer Genome Atlas (TCGA, see http://cancergenome.nih.gov), in which dozens of cancer genomes are being sequenced—without hypotheses!—generating numerous promising hypotheses to test, the results of which might one day be true personalized medicine (which, by the way, was another of Spector’s misguided not-so-deadly hypotheses).
Speaking of cancer, the most offensive DMH is the last one, namely that “from a public health perspective, cancer chemotherapy (chemo) has been a major medical advance.” Note how Spector is careful to insert the phrase “from a public health perspective.” This, too, is a straw man, as no one argues that chemotherapy is a major public health advance; rather it is an advance in the treatment of individual patients with certain kinds of cancer that respond to chemotherapy. I’ve discussed at length the slanted perspective inherent in this DMH, including its misleading characterization of chemotherapy (Gorski 2011a) and why we “haven’t won the war on cancer yet” (Gorski 2011b). If Spector had simply written that chemotherapy, while capable of curing many hematologic malignancies and a handful of solid malignancies (such as testicular and anal cancer), does a poor job on its own of prolonging life in patients with inoperable solid malignancies, I would have had little argument with him. Apparently such a universally accepted conclusion wasn’t provocative enough, however; so he declared chemotherapy an utter failure—from a public health perspective, of course!—ignoring cancers for which chemotherapy improves disease-free and overall survival and quality of life even when they are advanced (Venook 2005) and the usefulness of chemotherapy in adjuvant and neoadjuvant settings.
No doubt Spector thought he was being provocatively “skeptical.” Unfortunately, most of his DMHs are neither deadly nor, truth be told, hypotheses. At worst, they are the sorts of dead ends that science eventually corrects, which is how science works. Apparently, Spector doesn’t have the patience to deal with the messiness of science-based medicine.
Berger, M.F., M.S. Lawrence, F. Demichelis, et al. 2011. The genomic complexity of primary human prostate cancer. Nature 470: 214–20.
Contopoulos-Ioannidis, D.G., G.A. Alexiou, T.C. Gouvias, et al. 2008. Life cycle of translational research for medical interventions. Science 321: 1298–99.
Crislip, M. 2011. Deadly indeed (blog entry). Science-Based Medicine (February 25). Available online at www.sciencebasedmedicine.org/?p=11155.
Glass, D.J., and N. Hall. 2008. A brief history of the hypothesis. Cell 134: 378–81.
Gorski, D.H. 2011a. Skepticism versus nihilism about cancer and science-based medicine (blog entry). Science-Based Medicine (February 28). Available online at www.sciencebasedmedicine.org/?p=11185.
———. 2011b. Why haven’t we cured cancer yet? (blog entry). Science-Based Medicine (February 14). Available online at www.sciencebasedmedicine.org/?p=10761.
Spector, R. 2011. Seven deadly medical hypotheses. SKEPTICAL INQUIRER 35(2): 40–48.
Venook, A. 2005. Critical evaluation of current treatments in metastatic colorectal cancer. The Oncologist 10(4): 250–261.
Wadman, M. 2011. Fifty genome sequences reveal breast cancer’s complexity. Nature (online April 2). doi:10.1038/news.2011.203.
David H. Gorski, MD, PhD, FACS
Managing Editor, Science-Based Medicine
Leader, Breast Cancer Multidisciplinary Team
Co-Leader, Breast Cancer Biology Program, Barbara Ann Karmanos Cancer Institute
Associate Professor of Surgery, Wayne State University School of Medicine
The article “Seven Deadly Medical Hypotheses” has fundamental flaws. The argument, as I understand it, is that medical researchers fail to use the hypothetical/deductive approach to decide upon research agendas and therefore waste a “vast quantity of resource to disprove them.”
The argument as presented is circular. The author uses information from completed studies to prove the studies should not have been done in the first place. He fails to consider the state of knowledge at the time the studies were done and the cumulative nature of medical knowledge.
The author’s criticism of Hormone Replacement Therapy (HRT) is an example of his reasoning and his ignoring the state of medical knowledge. The benefits and risks of HRT were discussed at length with patients and on rounds in the 1980s when I was a medicine resident. Are the benefits of HRT greater than the potential risk of cancer from estrogen? As noted at the time in assessing HRT risks:
Hip fractures, Colles’ fractures, and coronary heart disease, and breast and endometrial cancers are important conditions in postmenopausal women that might be influenced by the use of hormone replacement therapy. . . . A 50-year-old white woman has a 16% risk of suffering a hip fracture, a 15% risk of suffering a Colles’ fracture, and a 32% risk of suffering a vertebral fracture during her remaining lifetime. These risks exceed her risk of developing breast or endometrial cancer. She has a 31% risk of dying of coronary heart disease, which is about 10 times greater than her risk of dying of hip fractures or breast cancer. These lifetime risks provide a useful description of the comparative risks of conditions that might be influenced by postmenopausal hormone therapy. (Cummings et al. 1989)
And the risk for cancer was uncertain, as discussed in a 1989 New England Journal of Medicine editorial: “Evidence that estrogen increases the risk of breast cancer has been surprisingly difficult to obtain” (Barrett-Connor 1989). In the 1980s the risks of HRT were uncertain and the benefits appeared considerable. I spent significant time convincing my mother, based on the information of the time, to take her Premarin.
At no time does Reynold Spector describe, based upon the information we had at the time, why evaluation of HRT, or the other deadly hypotheses, should not have been pursued. It is the data gathered from the present he uses to discredit the past. If he had used the hypothetical/deductive methods based on information known at the time, or better, information from today, to predict what was and what will not be worth pursuing, it would be more impressive. Hindsight is always 20:20.
Spector makes similar arguments for his other six deadly hypotheses. Well, one deadly hypothesis, five rejected hypotheses, and one opinion. The last hypothesis, “cancer chemotherapy (chemo) has been a major medical advance,” which Spector declared false, is not a hypothesis generated by researchers but by the author’s after-the-fact opinion, and he fails to recognize the benefits of failure.
The author states “when one dispassionately weighs the minimal prolongation of ‘good’ [the quotes around “good” demonstrating there is nothing dispassionate in the author’s analysis] life in patients with metastatic cancer . . . versus the very distressing side effects of chemotherapy with ‘targeted’ chemotherapy drugs, the case is close.”
If it had been 1990, Spector could have been writing about AIDS and HIV therapy. AIDS is infinitely simpler than cancer, yet a decade of false starts and dead ends, of failed hypotheses, led to progressive understanding of the pathophysiology and treatment of HIV and resulted in the era of HAART, where the expected life expectancy of those with HIV is often close to normal.
Medicine advances slowly and erratically, and unfortunately more insights are often gained from failure than success. Only after enormous effort can we retroactively identify which medical interventions warranted the gold standard, large, randomized, placebo-controlled trials demanded by Spector. Doing so prospectively is not as easy as he suggests.
Medicine is pushed forward not by pseudoscientists and worse but by caring, hardworking, and committed health care professionals who are doing the best they can under often tight funding limitations and great uncertainty. Spector’s dismissive attitude toward the blood and sweat expended by the researchers—and, often forgotten, the patients who volunteer their life and health to advance medicine by participating in clinical trials—is disappointing.
Cummings, Steven R., Dennis M. Black, and Susan M. Rubin. 1989. Lifetime risks of hip, Colles’, or vertebral fracture and coronary heart disease among white postmenopausal women. Archives of Internal Medicine 149 (11): 2445–48.
Barrett-Connor, Elizabeth. 1989. Postmenopausal estrogen replacement and breast cancer. New England Journal of Medicine 321(August 3): 319–20.
Mark Crislip, MD
Editor, Science Based Medicine (www.sciencebasedmedicine.org)
AVRUM Z. BLUMING and CAROL TAVRIS
In the May/June 2010 issue of the SKEPTICAL INQUIRER, we published “The Alarms of Hormone Replacement Therapy: Are They Supported by the Data?,” which criticized the headlines proclaiming that HRT significantly increases the risk of breast cancer. Readers might understandably be puzzled by the discrepancy between our perspective and that of Reynold Spector’s “Deadly Hypothesis Two,” in which he says that it is a “well-documented fact that estrogen is a carcinogen and causes breast cancer.” To support this claim, Spector relies on the findings of the Women’s Health Initiative (WHI). But the WHI is one of the studies we criticize in great detail, showing that its findings regarding breast cancer have been inconsistent. Virtually all of its alarmist conclusions—that HRT increases the risk of dementia, stroke, cardiovascular disease, breast cancer, and even “deaths from all causes”—were a result of selection bias or post hoc statistical manipulation.
Is estrogen the carcinogen that Spector so unequivocally thinks it is? Our paper reviews the mounting evidence that questions this common wisdom. Even the WHI results have not supported that association, finding a modest (and usually statistically nonsignificant) rise in relative risk only for the combination of estrogen-progestin therapy. In fact, a recent reanalysis of updated WHI data reported that post-menopausal administration of estrogen alone decreases the risk of subsequent breast cancer.
Of course, scientists and other skeptics can and do disagree with one another when it comes to interpreting data. Our primary objection to Spector’s discussion of HRT is not his conclusion but his tone of certainty, his oversimplification of an enormous body of research, and his unwillingness to question received wisdom from the WHI. Even some investigators who share his belief that the relative risks of HRT warrant concern acknowledge that the absolute risks from this treatment are small. In one worst-case analysis, researchers calculated that a fifty-year-old woman taking estrogen and progestin for ten years has a 96 percent chance of remaining free of breast cancer versus a 98 percent risk if she does not take HRT.
We refer interested readers to our SI article and its longer, more detailed version, “Hormone Replacement Therapy: Real Dangers and False Alarms,” The Cancer Journal, March/April 2009, pp. 93–104. This paper contains 210 references and a timeline of studies on HRT from 1942 to the present. The free, full-text article is available online at: http://journals.lww.com/journalppo/Fulltext/2009/04000/Hormone_Replacement_Therapy__Real_Concerns_and.1.aspx.
Avrum Bluming, MD
Oncologist, hematologist, internist
Former clinical professor of medicine, University of Southern California
Carol Tavris, PhD
Social psychologist, writer, and lecturer
CSI fellow, Los Angles, California
When I first skimmed Reynold Spector’s article “Seven Deadly Medical Hypotheses,” I found myself agreeing with most of his points yet feeling somehow disturbed. On a closer reading, I realized my discomfort was because he sounds more like a denialist or a contrarian than like a judicious skeptic. His seven hypotheses are stated in words that subtly misrepresent the truth, and he attacks them with dogmatism where nuance is called for.
His main thesis is that “many medical and nutritional hypotheses are ill-conceived.” That may be true, but scientists are doing the best they can. Does he imagine that they could somehow discipline themselves to only study hypotheses that will turn out to be true?
After falsely characterizing hormone replacement as an avoidable error, he perpetuates the myth that HRT kills 5,000 women yearly from breast cancer. The Women’s Health Initiative (WHI) study showed no difference in overall mortality; so if HRT killed 5,000 women from breast cancer it would have to have saved another 5,000 from other causes of death.
Deadly Hypothesis Four is that “screening tests beyond the standard medical examination are necessary for identifying disease or the risk of disease in apparently healthy, asymptomatic adults.” The utility of screening tests is not an ill-conceived hypothesis, much less a deadly one. A routine history and physical on an asymptomatic patient are next to useless for detecting pre-symptomatic disease or risk of disease.That’s why we have tried to devise useful screening tests to diagnose certain diseases and risk factors before symptoms develop, so early treatment might have a chance of altering the course of disease. No screening test was ever adopted without credible data and plausible reasoning. It is a gross oversimplification to call “screening tests are necessary” a false hypothesis. Instead, Spector should have attacked the myth that screening tests are categorically good and always save lives. Sometimes they fail to improve patient outcomes or even do more harm than good, and we then abandon them. All the tests Spector questions (mammography, PSA tests, genetic screening) were first questioned in the very medical literature he disparages.
What Spector characterizes as “a chronic scandal” is nothing of the sort. It is a reflection of the complexity of the scientific process. Yes, we need to be more skeptical. We need to carefully evaluate the evidence for any medical claim. We need to do better-designed studies. But it’s not always possible to do the ideal study; sometimes we have to go by the best available evidence from epidemiologic and other less desirable studies. We had to figure out that smoking caused lung cancer without forcing subjects to smoke or not smoke in a randomized controlled prospective study.
Spector’s article was doubly disappointing because he had so many excellent points that were sabotaged by the way he presented them. Medical science is far from perfect, but it is a cooperative, self-correcting endeavor that constantly criticizes itself and is constantly improving. We need continued rational skepticism and constructive dialog, not oversimplified contrarian polemics.
Harriet Hall, MD, “The SkepDoc”
Editor, Science-Based Medicine
Contributing Editor, SKEPTICAL INQUIRER
Contributing Editor, Skeptic magazine
REYNOLD SPECTOR, MD, Responds
It is clear from reading these letters that the correspondents do not understand how to evaluate the medical literature—how to discriminate the wheat from the chaff. They do not understand the hierarchical nature of proffered evidence. For example, I doubt any of the correspondents understand Hill’s criteria for assessing causal relationships in epidemiology/observation studies (see below). I recommend they read Spector and Vesell (2006) or the online methodological supplement to my paper on nutrition in SKEPTICAL INQUIRER (Spector 2009). They should also consult the FDA regulations for what constitutes proof in drug development.
Let me be concrete. Three correspondents claim I don’t understand that the evaluation of clinical science is not easy. Gorski talks about the “messiness of science-based medicine.” Crislip states “medicine advances slowly and erratically.” Hall states that “scientists are doing the best they can.” What they do not understand is that one of the reasons for this sorry state of much of the clinical literature is that many published studies do not have a clear hypothesis, are underpowered, or use inadequate methods (e.g., epidemiology/observation studies attempting to “prove” causal connections). Examples include the hundreds of epidemiology/observation studies on hormone replacement therapy (HRT) (see below) or megavitamin therapy (Spector 2009), most of which yielded incorrect or inconclusive results and essentially none of which satisfied the Hill criteria for validity (Spector and Vesell 2000; 2006). Genome-wide association studies, or GWAS, are another example. We have known for decades how to do good clinical science, but the methodologies are often not followed for reasons previously enumerated (Spector and Vesell 2002); flawed studies are published and confusion reigns. Much of the confusion could be eliminated by practicing valid science. The notion that all these unscientific studies are hypothesis-generating is ludicrous.
Bluming and Tavris challenge the notion that HRT increases breast cancer in post-menopausal women. Let me give a broad overview of the multifaceted and overwhelming data that supports the view that HRT causes human breast cancer:
- Estrogen causes cancer of the breast in female animals.
- Unopposed estrogen causes cancer of the uterus in humans with a risk of up to twenty times more than average. That is why unopposed estrogen is not given to any woman with a uterus.
- The Women’s Health Initiative (WHI) and then the Million Woman Study (Beral et al. 2011) both showed that HRT in women with a uterus increased breast cancer by 40–90 percent. The longer the use and the earlier the start of HRT, the higher the risk of developing breast cancer.
- The Million Woman Study also showed about a 40 percent increase in incidence of breast cancer for hysterectomized women who took estrogen for eight years (Beral et al. 2011). The much smaller WHI study of unopposed estrogen was stopped early because of increased strokes for women on estrogen. The women in that study were on estrogen for only three and a half years, and the results are inconclusive (Jungheim and Colditz 2011).
- After the publication of the WHI study on HRT, the use of HRT fell precipitously and soon after breast cancer incidence declined, consistent with HRT causing breast cancer (Beral et al. 2011).
- Anti-estrogens in post-menopausal women decrease breast cancer incidence by a whopping 60–80 percent (Goss et al. 2011).
- The International Agency for Research on Cancer considers HRT (either unopposed estrogen or estrogen/progesterone combinations) human carcinogens (cancer causing agents) (Jungheim and Colditz 2011). So do the FDA and U.S. Supreme Court.
- The labeling of HRT, approved by the FDA, recommends HRT for the shortest possible time at the lowest possible dose.
- Bluming and Tavris concede that in a worst-case scenario, there might be a 2 percent increased risk of breast cancer in women on HRT for ten years. This amounts to two hundred thousand more cases of breast cancer per ten million women on HRT for ten years. At the peak of HRT use, many times more women were taking HRT. What are Bluming and Tavris thinking?
Crislip writes an apologia for why, before the definitive studies of HRT began to be published in 1998, physicians used HRT broadly. I find his argument unconvincing—in fact worrisome. First, we have known for decades that estrogen causes breast cancer in animals and cancer of the uterus in women. There was also substantial data suggesting but not proving that HRT causes breast cancer in women. Second, the FDA wisely rejected (in the 1990s) the manufacturer’s claims that HRT decreased cardiovascular and central nervous system disease. These claims were based on epidemiology/observation studies, but they didn’t meet the Hill criteria and turned out to be wrong. The FDA approved claims only for HRT decreasing post-menopausal symptoms and fractures. The FDA wisely ignored the unbridled industry-driven hype, unlike many gullible physicians. Third, there were many HRT skeptics, including this correspondent (Spector and Vesell 2000) and National Institutes of Health Director Bernadine Healy, who began the WHI studies in 1991 to answer the HRT questions definitively. Now, twenty years later, we know that HRT causes breast cancer, strokes, clotting, uterine cancer (with use of unopposed estrogen), cardiovascular disease, and so forth. HRT should not be used chronically; excellent nonhormonal ways to prevent fracturing have been available since 1996.
In fact, I believe a good physician should use only proven therapies for which the risk/benefit ratio favors the patients. That was never the case with HRT, a known carcinogen. Fortunately, when my sister asked me if she should take HRT for her menopausal symptoms in the mid-nineties, I said no; it’s too risky and its benefits are uncertain. I also wrote about this over ten years ago (Spector and Vesell 2000).
To answer the questions on cancer chemotherapy, I recommend my SKEPTICAL INQUIRER article “The War on Cancer” (Spector 2010). Moreover, the problems with chemotherapy keep surfacing. For example, it was shown that for lung cancer patients, hospice care with minimalist chemotherapy was better than standard chemotherapy in terms of both quality of care and longevity (Temel et al. 2010). The scandals associated with the so-called targeted chemotherapy—such as the ineffective, incredibly expensive, and risky Avastin in breast cancer treatment—escalate (Tucker 2011). Finally, the authors of a recent long, thoughtful article (Smith and Hillner 2011) argue persuasively, as I did, that there is tremendous overuse of chemotherapy. The authors point out that “some oncologists choose chemotherapy in order to maximize their practice income. A system in which one half the profits in oncology [practice] are from drug sales is unsustainable.” Where is the well-being of the patient in this? A thoughtful, empathetic reader will weep.
Beral, V., G. Reeves, D. Bull, et al. 2011. Breast cancer risk in relation to the interval between menopause and starting hormone therapy. Journal of the National Cancer Institute 103: 296–305.
Goss, P.E., J.N. Ingle, J.E. Alès-Martinez, et al. 2011. Exemestane for breast-cancer prevention in postmenopausal women. New England Journal of Medicine, in press.
Jungheim, E.S., and G.A. Colditz. 2011. Short-term use of unopposed estrogen. A balance of inferred risks and benefits. Journal of the American Medical Association 305: 1354–55.
Smith, T.J., and B.E. Hillner. 2011. Bending the cost curve in cancer care. New England Journal of Medicine 364: 2060–65.
Spector, R. 2009. Science and pseudoscience in adult nutrition research and practice. SKEPTICAL INQUIRER 33(3) (May/June): 35–41.
———. 2010. War on cancer: A progress report for skeptics. SKEPTICAL INQUIRER 34(1) ( January/February): 25–31.
Spector, R., and E.S. Vesell. 2000. The pursuit of clinical truth: Role of epidemiology studies. Journal of Clinical Pharmacology 40: 1205–10.
———. 2002. Which studies of therapy merit credence?: Vitamin E and estrogen therapy as cautionary examples. Journal of Clinical Pharmacology 42: 1–8.
———. 2006. Pharmacology and statistics: Recommendations to strengthen a productive partnership. Pharmacology 78: 113–22.
Temel, J.S., J.A. Greer, A. Muzikansky, et al. 2010. Early palliative care for patients with metastatic non-small-cell lung cancer. New England Journal of Medicine 363: 733–42.
Tucker, F.C. 2011. Drugs and profits. New York Times (May 25).
Reynold Spector, MD
Clinical Professor of Medicine, Robert Wood Johnson Medical School (New Jersey)