Don’t clinical trial if you can’t formulate the disease model
The last edition of Critical Care Medicine brings a pristine example of an RCT that shouldn’t have ever existed.
It is easy to blog about misconceptions in current critical care research. I don’t need to look for older papers. In the months before launching my Substack, I was afraid of running out of ideas so I searched for and spared 20 ill-conceived NEJM and JAMA critical care papers to comment on. However, it is not necessary to use them. A TBS article appears to fit or inspire a comment every week or two.
Research producers and consumers in our field are utterly ignorant of how medical advances are possible. The consequence is a monumental waste of money and careers and no advance.
Beneath the superficial layer of testing an intervention, clinical trials are testing the validity of a disease model. This week, I am thinking of disease models, and guess what? Critical Care Medicine's latest edition gave me a perfect illustration - The ABC Sepsis study is an example of science without a theory.
Let’s examine the study, define disease models, and examine how authors without a theory behave ex-post.1
Title: Albumin Versus Balanced Crystalloid for the Early Resuscitation of Sepsis: An Open Parallel-Group Randomized Feasibility Trial— The ABC-Sepsis Trial.
Population: Adult Septic patients with NEWS Score > 5 and needing IV fluids.
Interventions: Albumin solution or balanced crystalloid.
Outcome: 30-day Mortality
I will not ask the authors about the disease model they are testing because I know they never thought of having one. It’s not their fault. It is the final result of decades of normalizing nonsensical research, starting with the first ARDS trials in the 60s.
The ABCs of disease models
The example I like most is the evolution of the peptic ulcer disease model. There were many different disease models in the past, but eventually, the luminal acid model prevailed. We didn't learn it in Med school, but several other models disputed the pathobiology of peptic disease. Some of them were:2
The vascular theory, where ulcers were secondary to venous stasis and interstitial hemorrhages, triggered by inflammation or traumatism
The inflammatory theory. Based on the concomitance between ulcers and gastritis/duodenitis, scientists speculated there was a causal link between inflammation and ulceration.
The mechanical theory. Scientists pontificated ulcers were triggered by small traumatic erosions, lately turned into ulcers by the action of the stomach acid.
The gastric hypersecretion theory, where chloridric acid causes ulcerations.
The luminal acid model prevailed because scientists learned that acid is generally necessary to have peptic ulcers. Vagectomy, the section of the vagus nerve before the stomach, was enough to validate the model. The surgery causes a decrease in acid production and thus eases the clinical manifestations. Hence the aphorism “No acid, no ulcer”.
However, is acid sufficient to cause a peptic ulcer? No. Everyone has acid in the stomach, and most people will never have an ulcer. So was the acid model enough? No, it wasn't. The incredible discovery of H. pylori and its integration into the disease model of peptic ulcer benefited millions of patients. Indeed, it benefited humanity. We won't have a better disease model for peptic ulcers in the foreseeable future.
We could also think of ST-elevation myocardial infarction. Do you think it was easy to causally correlate the clinical and electrocardiographic presentation with the finding of an intracoronary thrombus? In hindsight, it appears obvious that a thrombus is necessary and sufficient to cause a myocardial infarction. However, this model evolved for decades during the 20th century to be validated by thrombolysis only in the late 80s. See for yourself. Go to NEJM, JAMA, or The Lancet websites and search for articles about myocardial infarction in reverse chronological order.
I have presented two brilliantly developed and validated disease models. Now, let's return to the embarrassing reality of Critical Care and this ABC Trial that doesn't grasp the ABCs of anything. It's dismaying.
The absent disease model in the Jacksons’ ABC-Sepsis study
The ABC trialists (should I call them the Jacksons?) assume albumin solution and balanced crystalloid possess inherent differences that are both extreme and related to mortality drivers. However, it is not biologically plausible and the authors may have never considered the plausibility. They present no biological plausibility to justify the study.
Instead, they offer the results of other studies equally blind to the marginal utility of the intervention. If they had a disease model in mind, it would invoke an intrinsic quality of albumin that is absent in balanced crystalloids, and explain how that quality would cause a pivotal change in patients' evolution by acting upon a dominant driver of the disease.
It was interesting to see the word “marginal” in their reasoning. They state there are marginal differences when you use albumin instead of a crystalloid in critically ill patients. In the marginal utility framework, marginal means not dominant. Due to the exponential distribution of treatment effects, a marginal effect is not clinically relevant and shouldn't be tested. A link to the framework is presented below.
The assumption of causality is an assumption of dominance. Now, let's recall the disease models of peptic ulcer and myocardial infarction. Both models offer dominant causes to be treated. The causality is strong. You can see by contrast what is lacking in the ABC-Sepsis Trial. Is the difference in fluid composition necessary or sufficient to cause a mortality split between the groups? Of course, it isn't. I think it's needless to elaborate more on it.
The deceiving Task Force and a bad coping strategy
Like the Robinsons, the Jacksons ignore the Task Force's plan to deceive them and sink their research careers.
The Task Force is the group of scientists pushing the failed sepsis definitions since the early 90s. The Task Force doesn't present a disease model to the researchers. Instead, they gave us vague and useless statements like “dysregulated immune response”, and the most extravagant bullshit ever presented as medical science: they told us to diagnose a disease by its prognosis using triage tools like qSOFA. Tell a cardiologist that heart failure is not heart failure until the patient has a bad prognosis.
Finally, the readers of this Substack know how much I worry about the future of our specialty. Instead of acknowledging the limitations imposed by the lack of a workable disease model, the Jacksons may double down in denying it. The path is already clear. They have two options that are latent in the critical care literature.
First, they may provide a Bayesian Spin on the results, stating that according to their Bayesian analysis, the "real" effect is on the upper side of their unity-crossing relative risk confidence interval. Everything is possible in the Magical Kingdom of Bayesian Analysis.
Second, they may amplify the likelihood of a “positive” result using a distorted Win Ratio statistic like the one used in the DEFENDER Trial, linked below.
Well, from Robinsons to Jacksons, our specialty is paralyzed by the lack of valid disease models. I am so envious of other specialties!!
Thanks for reading The Thoughtful Intensivist.
Excelente. 🙌🏻
I don't think there is a "y" in H pylori!