It is essential that clinical researchers turn their attention and ingenuity to forming an emerging paradigm for medicine that emphasizes new ideas and methods. One reason for this is that standard late modern methods cannot address chronic ambient poisoning for reasons such as the following:
Exposures Can’t Be Assessed
Investigators could study any exposure with emerging methods if they could find a way to measure exposure accurately and reliably, and so consistently compare people with more exposure to those with less. More and more, however, the exposures that our species creates are dynamic, complex, and mysterious, and their assessments are speculative and uncertain. That is to say: because many important hazards are ubiquitous, cumulative, and modified by the body, we cannot assess exposure after the fact. Worse, we are guessing as to what dose of which hazard may correlate with which ailments. Does peak dose matter? Frequency? Cumulative dose? Combined doses? We don’t know; we guess. Then we gather data on our guess, and if we find nothing, conclude that there is no effect. In other words, we over-interpret the negative study and glean false reassurance.
For example, when Vietnam veterans noted health problems in their children, they lobbied the government to study the health effects of wartime exposures to Agent Orange. The researchers doing these studies had no way of knowing which veterans had been exposed or how much exposure they had sustained. Cynics saw these studies as impossible and therefore an end run around activists, studies that ultimately showed no effect. Vietnam veterans eventually negotiated a political solution in spite of the results of large, socio-politically motivated studies—not because of them.
Unexposed Groups Are Disappearing
Studies that guide late modern policies compare risks of disease in people who are exposed to a source of potential harm with the risks of disease in those who are not. That is, if you collect data on a hazard and an ailment in a group of people, you can then divide them into four groups: those who have the ailment or don’t, and, within each group, those who have been exposed or not. The result is a fourfold table:
| AILMENT | EXPOSED | |
| Yes | No | |
| Yes | a | b |
| No | c | d |
The odds ratio estimates the degree to which a and d dominate: ad/bc. As sources become ubiquitous, the unexposed group disappears and renders the method moot—if everyone is exposed to the hazard, you get everyone divided by no one. You get nonsense. If you break exposure into categories, you may have no one who is normal enough to be deemed unexposed, and little import spread across the other categories.
If the source is doing real harm, the causal web of the disease it causes will shift, baseline rates will increase, and we will view them as randomly distributed aspects of the new normal. In other words, we accept the new baseline and miss clues that would enable cure.
Obversely, large prospective studies of health benefits ensuing from behaviors like increased physical activity randomly assign participants to intervention and comparison groups. Those assigned to the comparison group tend to learn of the intervention and adopt it themselves. This equalizes the groups, spoiling the experiment and its results.
Relying on Hypothesis Rejection and Statistics
Case counts are an early modern method of measuring rates of illness. They are useful tools, as during the COVID-19 pandemic. Case counts from the mid-nineteenth century documented the first interruption of cholera transmission by Dr. John Snow, and the first use of denominators to reveal patterns of disease in urban Boston as a means of guiding public health interventions. In the mid-twentieth century, doctors like Sir Richard Doll and Arthur Herbst used statistics to relate lung cancer to smoking and vaginal adenocarcinoma to the use of diethylstilboestrol by their mothers. Risk factor modification became a cornerstone of the new specialty of Preventive Medicine.
As time passed and illness evolved, however, methods lagged. Statistics were used as the means for justifying policy choices and became more entrenched, elaborate, and divorced from reality. Now, such methods are being used for problems that they do not fit. Consequently the scientific process in which a problem is presented and suitable methods are devised to solve it has gone by the wayside.
Statistical significance is determined by sample size; if you have enough study subjects, your estimates—however modest—will turn out to be statistically “significant.” This is like a popularity contest: A hot proposal that gets big funding will be able in the end to claim validation by fact of size alone. That is—tongue firmly in cheek: Big Data = Little Importance + Opportunities for Confirmation Bias. You can “prove” any number of contradictory findings.
On the bright side, the social consensus surrounding modern statistics is fast losing traction, particularly with the growing recollection that not all conditions are infectious or the result of single insults such as drugs, alcohol, or tobacco. In these instances, single factors have such dominant effects as to be evident in crude studies.
Prevailing Consenses on Proof Unravel
In the 1950s, Sir Richard Doll linked cigarette smoking to lung cancer with the aid of simple statistical methods that rapidly disseminated and dominated standards of proof. These methods continue to be of use in studies of lone factors with lone effects, especially the effects of medical drugs, devices, and procedures on prevailing complications. An example is the detection of higher rates of vaginal adenocarcinoma in daughters of mothers who took prenatal of diethylstilbestrol, called DES.
These standards represent a poor fit to the exploration of what Dr. Alvan Feinstein and colleagues called the menace of daily life. More elaborate statistical methods function poorly, even if they fit the causal web model proposed by Dr. Mervyn Susser in Causal Thinking in the Health Sciences. The absence of a ready replacement for the early, simple approach has prolonged our reliance on standard study methods and criteria for proof that conceal as much as they reveal.
Anyone who has seen a diagram of a biochemical process like the Krebs cycle can infer that the relation of an individual human body or the body of a species to the body of life will be too complex and protean to describe in terms of one putative hazard and one adverse consequence. The results of large prospective studies of the effects of alcohol and cigarette abuse show that single hazards lead to many consequences, and that hazards synergize. Dr. Kenneth Rothman’s more recent causal pie model reveals that causes that act together appear to result from a lone cause when that cause is least prevalent and thus most limiting.
While the question of how to assess the consequences of environmental exposures remains open, researchers continue to impose statistical models both simple and elaborate on complex deterministic processes. For example, the Human Genome Project enabled diabetes researchers to look for the one or two genes they hoped might account for most cases of adult onset diabetes. They didn’t find one or two; they found hundreds.
In studies of rare conditions like amyotrophic lateral sclerosis—that is, ALS or Lou Gehrig’s disease—men who use agricultural chemicals at work and who have a gene that makes them susceptible to those chemicals may develop ALS. The problem is that as researchers work their way back to the exposure, they may forget that the use of one or more agricultural chemicals may lead to any number of diseases, including Parkinsonism, as reported in Sierra Magazine by Joy Horowitz.
Clinical Science is Missing
Clinicians are no longer learning from life. If you’re afraid to reveal the truth in a medical record (e.g. to record legally embarrassing phrases like “I don’t know,” “haven’t got a clue,” “No idea what’s going on here,” “all I can do here is try to stop the patient’s complaining,” or “this patient is not capable of comprehending simple instructions; meth addict? illiterate? brain damage?”), your record will document a virtual reality that obscures and erodes clinical acumen. Lawyers need cases; doctors need truth and integrity.
If you’re afraid to say “I don’t know” in your own mind, you shouldn’t do research. You won’t engage the unknown openly, try and find out what’s going on, or develop a new gold standard. Forget about defining a diagnosis or discerning the etiopathogenesis. Your best effort will be no more than full-price, useless folly à la the Emperor’s new clothes.
If your experience is insufficient to delineate a causal web that features twenty interacting factors, you’re not ready for a ‘big’ study. If you do use statistics be aware that the thin and fragile foundation of Popperian logic, which supports testing the null hypothesis, makes no sense in real complex systems where context is key. Also, quantitative proof will be robust only when you study risk factors that account for an overwhelming proportion of cases of rare conditions—i.e. to problems of little of no present import.
The Fallacy of Molecularism
For decades there has been a hot molecule or set of molecules that everyone wants to research; right now, it’s genes. When poisons are ubiquitous, the genes that determine the kind of harm those poisons do—the common variations in enzymes that break down one or more poisons, or that make one or more end organs susceptible—come to the foreground of our attention. This would only make sense if poisons were essential and genes a matter of choice, or if the many genes that make a difference could be grouped in some useful way. Researching this would be fascinating but would not obviate the need for a better option.
These and other points, when carefully considered, add up to the good news that we humans can stop using our present ideas and methods; they are outdated and have baffled us. They were good first efforts that deserve to evolve, as I have evolved through my experience. The fields of medicine and public health are likewise ready to evolve into forms that can solve current problems.
Some text excerpted from Medical Detective: Chronic Ambient Poisoning, pp. 89-91.