Diagnosis, Therapy and Evidence

When Dr. Novella recently wrote about plausibility in science-based medicine, one of our most assiduous commenters, Daedalus2u, added a very important point. The data are always right, but the explanations may be wrong. The idea of treating ulcers with antibiotics was not incompatible with any of the data about ulcers; it was only incompatible with the idea that ulcers were caused by too much acid. Even scientists tend to think on the level of the explanations rather than on the level of the data that led to those explanations.

A valuable new book elaborates on this concept: Diagnosis, Therapy and Evidence: Conundrums in Modern American Medicine, by medical historian Gerald N. Grob and sociologist Allan V. Horwitz. They point out that 

many claims about the causes of disease, therapeutic practices, and even diagnoses are shaped by beliefs that are unscientific, unproven, or completely wrong.

While we try to be science-based, we are not always as scientific or as logical as we would like to think. We form hypotheses that are compatible with existing data, and then our assumptions guide our thinking and future research and sometimes interfere with our reception of new data. We must recognize those assumptions and constantly re-evaluate them. It’s important that we look the imperfections of science-based medicine squarely in the face if we are going to have any hope of overcoming them.

Of the therapies recommended in a 1927 textbook only 23 were later validated as effective or preventive. The other 211 were subsequently found to be either harmful, useless, of questionable value, or simply symptomatic.

Medical treatment has had a big impact on human health, but there’s more to the story. We developed effective treatments for ulcers, but the incidence of ulcers was already declining before those treatments had any impact. The decline of rheumatic heart disease is probably not due to antibiotics but may be due to decreased virulence of the causal bacteria. We have no idea why the incidence of stomach cancer has decreased in the US, or why it is so high in Japan.

A popular concept today is that cancer is largely a preventable illness linked to diet, environmental carcinogens and behavior. This is rooted largely in belief and hope rather than fact. Smoking is the one notable exception. Genetic factors and the many physiologic changes of aging may contribute more than we would like to think. To some extent, disease is an unavoidable consequence of life: the idea that science can eventually provide perfect health may be a chimera.

In our efforts to prevent heart attacks we are essentially treating risk factors, without a clear understanding of how they relate to pathophysiology. We are treating hypertension, hyperlipidemia and other risk factors rather than directly treating the cause(s) of cardiovascular disease. We offer behavioral prescriptions based on assumptions derived from inadequate epidemiologic evidence, and this kind of thinking can lead us astray. Recommending a low fat diet helped fuel an epidemic of obesity as people replaced the fat in their diet with extra carbohydrates.

Once we have formed a belief we are slow to respond to new evidence that refutes it. The book covers the history of tonsillectomy. Tonsillectomies remained fashionable long after the evidence showed most of them were useless.

The most interesting question they ask is

How do diagnoses come into existence and why do many disappear with the passage of time?

What ever happened to chlorosis and neurasthenia? The same patient presenting with the same symptoms in 1890 and 2010 would get entirely different diagnoses. The ailments that afflict humans don’t change much; our diagnostic categories do.

Autism, CFS and fibromyalgia are all relatively new diagnoses for conditions that undoubtedly existed long before the diagnostic name was coined. “Their pathobiology remains unknown, and there is little agreement on their diagnostic boundaries. Once given a name, however, the numbers given to each diagnosis have expanded exponentially.”

Psychiatric diagnoses are particularly slippery. Where exactly do you draw the line between normal sadness and depression? Disease occurs on a continuum and we try to fit it into discrete boxes. We organize the data differently at different times as influenced by historical circumstances. The Diagnostic and Statistical Manual of Mental Disorders (in its many iterations, now up to DSM-5) changes as it reflects not only new data but cultural, social, and political forces. There is no evidence that the new DSM categories of anxiety have improved the diagnosis, treatment, or understanding of anxiety disorders. The popularity of the diagnosis of post traumatic stress disorder (PTSD) raises issues about the connection between external causes, individual responses, and resulting symptoms. Broadened criteria for PTSD have made it possible for almost everyone to be diagnosed or considered at risk.

We differentiate between science-based medicine and belief-based medicine, but we mustn’t forget that scientists form beliefs too. Our interpretation of the evidence is influenced by our working hypotheses. We must remember to constantly guard against overinterpretation and to concentrate only on what the evidence actually shows. When we use a diagnosis, we must remember that it is not definitive, but only an artificial category we have imposed on nature to help us understand our patients’ symptoms and provide a framework for treatment decisions. When we have an explanation, we must keep re-evaluating the data to make sure another explanation doesn’t fit the data just as well.

Ionannidis showed that most published studies are wrong. Grob and Horwitz show that many of our current diagnoses, treatments, and ideas about disease may be wrong too.

I suggest that we all repeat the mantra: “I could be wrong” and keep asking “Could any other explanation fit the data?”


[Slashdot]
[Digg]
[Reddit]
[del.icio.us]
[Facebook]
[Technorati]
[Google]
[StumbleUpon]

Related Posts

Comments are closed.