Contra Science-Based Medicine

Epistemic status: hand-wavy, but making a serious point

TW: diets.

I recently did some reading about ketogenic diets for cancer, and I’d like to compare and contrast my approach with the explanation on the blog Science-Based Medicine, which consistently presents the “skeptical” perspective on alt-med questions.

David Gorski is a cancer biologist himself, as I am not; his posts are always informative, and I have no quarrel with his facts. I read the studies mentioned in the post, so we’re using pretty much the same set of data points. And I agree with the broad outlines of his claims: ketogenic diets have some promising but by no means conclusive preclinical evidence for brain cancers; they’re definitely not a substitute for chemotherapy in general; and Dr. Seyfried has been overselling his research as a cure for cancer in disreputable alt-med venues.

But I want to pick apart some points of perspective and interpretation.

The first part of the post is all about painting Seyfried as disreputable because of his associations with alt-med institutions. Gorski says of the American College for Advancement in Medicine, “this is not an organization with which a scientist who wishes to be taken seriously by oncologists associates himself.”

Now, I’m not defending the cancer quacks mentioned; these are people who pitch chelation and coffee enemas, things that are pretty clearly scientifically disproven.  However, I’m suspicious of the rhetorical trick of guilt by association and argument from consensus.  Surely we care about whether Seyfried is correct, not whether he is “taken seriously”, “reputable”, or “legitimate.”  These are all social words, not scientific ones, and constitute an emotional appeal to social conformity and authority.

To his credit,  Gorski doesn’t stop there; he does make substantive criticisms of Seyfried’s work.  But I think it’s worth pointing out when, as happens so often in the biomedical world, a social argument is conflated with a scientific one.

Gorski goes on to criticize Seyfried for “exaggerating how hostile the cancer research community is towards metabolism as an important, possibly critical, driver of cancer” when cancer metabolism is, in fact, an active area of current cancer research.  He goes on to say, “Dr. Seyfried, in my readings, appears all too often to speak of “cancer” as if it were a monolithic single disease. As I’ve pointed out many times before, it’s not. Indeed, only approximately 60-90% of cancers demonstrate the Warburg effect.”

None of these facts are wrong, but the interpretation is misleading. Cancer metabolism and metabolic mechanisms for cancer treatments are, in fact, common topics of cancer research; but this ought to be evidence in favor of Seyfried’s hypothesis, that it’s within the range of mainstream science and is supported by many cancer biologists, rather than being pure invention like most alt-med “cancer cures.”

I’d also argue with the statement that “cancer isn’t one disease.”  It’s true that not all cancers demonstrate the Warburg effect, but 60-90% is a lot of cancers; a drug that was effective in 60-90% of cancers would be as revolutionary an advance as chemotherapy.  An antibiotic that killed 60-90% of bacteria could fairly be said to “kill bacteria.” When most (if not all) cancers have structural features in common, that indicates that talking generally about “cancer” is meaningful, and that it doesn’t make sense to treat every sub-sub-type as though it is a completely different disease.  Cancer has both unity and diversity.  Saying “cancer isn’t one disease” is a rhetorically loaded move that means “don’t generalize from one type of cancer to another.”  But it’s not correct to never generalize; that would utterly paralyze research.  How much it’s safe to generalize depends on how common the relevant feature is across cancers; in the case of the Warburg effect, that’s a matter of current debate, but it’s fair to call it pervasive.

I don’t have much criticism of the way Gorski handles the ketogenic diet studies. He’s on the skeptical side, but skepticism is warranted. Mouse studies very frequently don’t generalize to humans; they’re suggestive, but only weak evidence. And while there were two case studies of patients who did notably better than typical glioblastoma patients on ketogenic diets, we don’t have enough patients to be confident that the improvements were a result of the diet.

But then Gorski says, “Clearly, ketogenic diets are not ready for prime time as a treatment for cancer.”

Now, wait a minute. What does that even mean?

As a cancer patient, does it make sense for you to try a ketogenic diet?  Well, there’s a plausible mechanism for it to work (particularly in brain cancer), there’s some suggestive evidence in mice and a few humans with brain cancers, and — crucially — it’s just a diet.  People go on ketogenic diets all the time, for no other reason than wanting to lose weight. It’s even been shown medically safe (though apparently hard to comply with) in cancer patients. Trying a special diet is pretty low risk, and a reasonable person aware of the evidence might very well choose to try it.

It doesn’t make sense to use a ketogenic diet as a replacement for chemotherapy or radiotherapy in cancers where those treatments work. That would be very unsafe.  But for certain advanced brain cancers, chemo barely extends life if at all, and is very unpleasant. If there’s anyone who has a good reason to refuse chemotherapy, it’s someone who’s almost certain to die soon and doesn’t want their last few months to be agonizing.

Is a ketogenic diet for cancer something that every oncologist in the world should be prescribing for his patients? No way. Should it be the “standard of care”? No; there isn’t enough evidence that it helps.  But is it worth trying for an individual who wants to? Quite possibly.

The distinction here is about where you put the reader’s locus of identity. Is a reader supposed to imagine herself as a potential cancer patient, considering whether or not to try the diet? Or as a potential administrator, considering whether or not to make the diet a policy for everyone?  The rhetorical trick Gorski’s using here is in identifying the reader with a nebulous “we”, as in “should we put cancer patients on ketogenic diets?”  You are meant to imagine a consensus, or an authoritative body.  The medical profession, the government, something like that.  This imagined “we” is the mirror image of the nebulous “they” that conspiracy theorists believe in, the “they” who doesn’t want you to know about cancer cures.

The overall effect of believing in an imagined “we” or an imagined “they” is to make social reality the primary reality.  “We” or “they” represents a vague model of “society” — the “respectable” people, the “legitimate” and “reputable” people, the “consensus”. In other words, the tribal elders. If you have a positive association with “the consensus”, as Gorski clearly does, then you want to expel the “disreputable” from the consensus.  If you have a negative association with “the consensus”, then you mistrust anything that sounds official and look for fellow mavericks and outsiders.  In neither case are you primarily evaluating claims of fact; you are evaluating people.

For instance, the existence of Phase I/II trials of the ketogenic diet on glioblastoma ought to be good news for ketogenic diets.  More evidence will soon come in; and the fact that the studies exist at all is further evidence that ketogenic diets are taken seriously by mainstream cancer researchers.  However, Gorski treats this as an indictment of Seyfried, because he wanted to do an (uncontrolled) case series of ketogenic diets rather than the more thorough controlled studies.  The overall intent of the blog post is to communicate Seyfried is disreputable, cancer is complicated, people who believe in cancer cures are beyond the pale, when one could have used exactly the same facts to make the point ketogenic diets are an exciting possibility for glioblastoma and the preliminary evidence is encouraging.

My own perspective can perhaps be summarized as “a contrarian worldview from mainstream sources.”  Looking at ordinary sources like journal articles and historical primary sources, looking at uncontroversial claims of fact, often gives me a view of the world that is quite different from the “we”-based view where “society” is more or less getting things right.  My object-level beliefs are rarely that unusual; the connotation of those beliefs is where I differ from most people.  I don’t feel myself to be safely nestled in the lamplit circle of “we”; I feel like I’m outside, tumbling in the abyss, with only the frail spark of my mind to illuminate a small patch around me.  And I think that, ultimately, the abyss is real, and the lamplit circle is imaginary.

14 thoughts on “Contra Science-Based Medicine

  1. When you start checking references you realize that high level summaries are just stories. Stories are always false, but they are false to different degrees. We have some good mechanisms for causing more predictive stories to win out over time, and we’ve also gotten lucky a few times. But overall we do not yet have a reliable machine that translates data points into stories in a way that doesn’t have lots of leaks between abstraction levels. It’s a tough problem, but I am heartened to see that *awareness that it is a problem* is at least slowly spreading.

  2. Please continue writing this blog! I really liked reading your posts, especially Aesthetics are Moral Judgments.

  3. Thank you for this! It’s accurate, important, and beautiful.

    I’ve been studying the topic as an unfunded side-project and would be happy to swap references with you.

    Interestingly, it appears that mainstream oncology is actually much *more* amenable to investigating potential benefits of keto diet than some other fields for which there is already stronger evidence, such as diabetes.

    The reasons for this appear to be to be social power dynamics, plus potentially a large dose of profit-motivated corruption.

    But in any case, the numerous ongoing registered trials of keto against cancer are a very promising sign of change.

    Richard Feinman (the living biochemist, not the dead Nobel laureate) wrote in his recent book something like “if it turns out that we need to address cancer in order to accept the obvious treatment for diabetes, that won’t be the strangest thing that’s ever happened in science”.

    • Yeah, I have a doctor friend who helped run low-carb clinics for diabetics; it seems pretty intuitive and it worked well for them, and I’m surprised that it hasn’t gotten more traction. (Haven’t read up on it, so it’s still possible it doesn’t work in trials?)

      • I think the problem lies in the assumption that diabetes is a progressive disease. This is how the American Diabetes Association views it. So the main focus is management rather than a cure. Since standard diabetic treatment necessitate the consumption of carbohydrates, the end results is the further deterioration of the patient. It’s only when you view diabetes (type II) as a totally curable, temporary condition, do low carb and ketogenic diets make sense. But in the eyes of the American Diabetes Association type II diabetes can’t be cured – only manged into remission. Which is absurd. If I break my arm, put in a cast, and 12 weeks later I have full use of my arm, my break didn’t go into remission. It healed. It’s no longer broken. Even cancer is considered “cured” after 5 years of no symptoms. Not so with diabetes. Once a diabetic, always a diabetic the experts say.

      • No, it works in trials. (Even though, in my opinion, the trials that have been done so far all have design flaws, some of which I think bias them in favor of the control group.)

        Here’s a good recently trial: it’s an RCT, it was large (n=115), and long-running (52-wk):

        Tay-2015-“Comparison of low- and high-carbohydrate diets for type 2 diabetes management: a randomized trial”

        Patients in the low-carb arm had improved glycemic control, even while simultaneously reducing their medications, and had improved markers of cardiovascular health (which is consistent with other experiments in non-diabetics — see (self-citation)).

        My caveat about this Tay-2015 study is that the diets (of both arms) were low-calories diets. This confounds the results about carbs per se. Also it is unrealistic because patients, with very few exceptions, cannot go on a life-long semi-starvation diet, as has been demonstrated many times in both clinical observation and in RCTs. Unfortunately anecdote, folklore, and mainstream nutritional pseudo-science continue to tell us that prolonged semi-starvation is not only possible, and not only healthy, but even necessary! Which is why it often shows up in experiments like this.

        A final objection to the low-calorie confounder is that it biases the experiment to make the control group look better, since the high-carb control arm also reduced their carb intake.

        There are indicators of this confounding in the results — the low-carbers didn’t lose body fat any better than the high-carbers did. (In fact, the _high-carb_ arm lost more body fat and less muscle than the low-carb arm, non-statistically-significantly.) This is inconsistent with numerous other RCTs (in non-diabetics) where low-carb almost always does better at shedding excess body fat and at preserving or building lean mass.

        I interpret this as meaning that both groups were rapidly losing both body fat and muscle due to the semi-starvation and due to being obese to start with. Such rapid weight loss under semi-starvation has a lot of consequences that could confound some of these results (and isn’t healthy and shouldn’t be recommended to patients).

        But, let’s not lose track of the main point here: evidence from well-controlled studies such as this one supports the hypothesis that low-carb is superior to the alternative for glycemic control and other clinically important effects in diabetics. I’m just quibbling about how _much_ better I think the effects would be in my ideal experiment.

        Here are more studies that are worse in various ways — smaller, shorter-running, and in some cases not full RCTs — but that are in my opinion worth reading if you’re interested in this topic:

        * Saslow-2014-“A Randomized Pilot Trial of a Moderate Carbohydrate Diet Compared to a Very Low Carbohydrate Diet in Overweight or Obese Individuals with Type 2 Diabetes Mellitus or Prediabetes”

        * Westman-2008-“The effect of a low-carbohydrate, ketogenic diet versus a low-glycemic index diet on glycemic control in type 2 diabetes mellitus”

        * Gannon-2004-“Effect of a High-Protein, Low-Carbohydrate Diet on Blood Glucose Control in People With Type 2 Diabetes”

        * Krebs-2016-“A randomised trial of the feasibility of a low carbohydrate diet vs standard carbohydrate counting in adults with type 1 diabetes taking body weight into account”

        * Nielsen-2012-“Low carbohydrate diet in type 1 diabetes, long-term improvement and adherence: A clinical audit”

        Click to access 1758-5996-4-23.pdf

        Read the first three paragraphs of the Discussion section!

        * Maekawa-2014-“Retrospective study on the efficacy of a low-carbohydrate diet for impaired glucose tolerance”


        A “position paper” by some of the leading lights of the heterodox position on this (these ones happen to be from New Zealand). The arguments are very strong!

        * Schofield-2016-“Very low-carbohydrate diets in the management of diabetes revisited”

        A longer, more detailed position paper by the (mostly-American) leaders, also packed with compelling arguments:

        * Feinman-2015-“Dietary carbohydrate restriction as the first approach in diabetes management: Critical review and evidence base”

        (That’s the Feinman that I mentioned in my earlier comment, as having contrasted the sociology of mainstream cancer researchers with that of mainstream diabetes researchers.)

        P.S. See what you did in your comment, when you said “perhaps it doesn’t work in trials?”? You were guessing at an explanation that would be consistent with the hypothesis that mainstream diabetes medical theory and practice is based on good science. I’m arguing that there isn’t such an explanation. 😉

  4. Update: (not that I know anybody is reading this) is indeed a good study, in my view. It compares ad libitum medium-carb (incorrectly termed “low-carb” in the paper) against calorie-restricted high-carb. Despite the carb restriction being only mild (200g/d to 120g/d), the HbA1c of the medium-carb arm fell from 7.6% to 7.0% in 6 months. Not bad.

    It may be confusing to the reader that the *ad libitum* medium-carb dieters consumed (in some subgroups, on average) fewer calories than the calorie-restricted high-carb dieters. That isn’t too surprising if you’re familiar with the numerous other studies showing a similar effect. If you have to consciously restrict your intake despite hunger then this means your body needs more (and/or different) nutrition than it is getting (perhaps because you’re feeding it a high-carb diet that is distorting your metabolism and hunger signals).

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s