This post introduces a column I wrote over at TheHeart.org | Medscape Cardiology
The good news is that most people infected with coronavirus don’t need a hospital or doctor. But some do. Some get very ill.
The maddening thing is that doctors don’t have an effective treatment for the virus. There are no cures. The Worldmeter today shows nearly 5 million infections and more than 300,000 deaths. And no effective therapy.
Excluding a possibly modest effect of Remdesivir, our care is supportive, which is medical jargon for giving simple things like oxygen, acetaminophen, IV fluids and letting the body do the rest.
That sentence makes supportive care seem simple, but it is not so with COVID19. The virus can cause havoc in the body. Damage to the lungs (pneumonia) gets most of the attention, but other organs can be harmed.
COVID19 and Clotting:
One system in particular that can go haywire is the clotting system (medical term is coagulation). Many studies have shown that patients ill with COVID19 can have excess clotting.
(It’s not well known, but our bodies do this elegant dance to keep the blood clotting factors and thinning factors in balance. Medical people call this equilibrium or homeostasis of the coagulation system.)
We now have blood tests that give us a window, albeit a somewhat opaque one, onto the clotting system. Also imperfect is the observation that patients with COVID19 can clot off intravenous lines, or on autopsy, clots can be seen in organs.
The worry about excess clotting (medical term is thrombosis) causes doctors to consider using drugs called anticoagulants, which come in many different forms–heparin (IV), low-molecular weight heparin (sub-q shots) and direct-acting oral anticoagulants (DOACs).
“If there are too many clots, you give clot-blocking drugs”…goes the Cartesian body-as-a-machine thinking.
Alas, the problem is a) infection control issues often makes it impossible to scan COVID19 patients, so we don’t know if they actually have clots in their organs, and b) anticoagulants can accentuate bleeding, which can be horrific.
Anticoagulants Work in Some Diseases:
In non-COVID19 patients, we know anticoagulant drugs work because we have done proper trials.
Take anticoagulants for patients with atrial fibrillation. We know these drugs work because in trials with thousands of AF patients who take the drug compared to thousands who take placebo, fewer people have strokes while taking anticoagulants. These trials also tally bleeds and thus doctors can know the net benefits, which favor the anticoagulants in selected patients.
Proper Trials = Knowledge:
The essential part of a proper trial is that the choice to use the drug or not (placebo) is random. That means in a randomized controlled trial (RCT) you end up with two balanced groups of patients. (Well, mostly balanced). This is crucial because if one group does better, you know it was drug, and not because one group was healthier, or sicker.
Trials, of course, take effort. You have to write a protocol, specify what you will measure and the types of patients to include or exclude, get ethical approval, recruit sites, consent patients, to name just a few tasks. You also have to blind the patients and doctors so they don’t know the treatment arms. This takes money and time. RCTs don’t make themselves.
The Limits of Observation:
In the COVID era, people don’t want to wait. So instead, investigators are recording outcomes on what happens to patients who have been treated.
These are called observational studies; people simply observe what happens after the fact.
Observational studies are deeply problematic for deciding what caused what.
A simple example: a heart surgeon decides to do an extra procedure during routine valve surgery, say, a closure of the left atrial appendage in the heart. It takes an extra few minutes and the idea is that closure of the appendage will reduce clots and that will reduce future stroke.
Then researchers look back, and voila, patients who had the extra procedure during valve surgery had fewer strokes than patients who did not have it. The procedure must work. Wrong.
The problem is that the surgeon’s decision to close the appendage was not random: thus, she may have chosen healthier patients to do the procedure on. And that is why the intervention looked good. We call this selection bias.
The Pressure of COVID:
My post over at TheHeart.org | Medscape Cardiology critiqued a study published in an influential cardiology journal. A group of researchers from Mt. Sinai hospital system in New York City looked back on about 3000 patients with COVID19.
They studied the association between the use of anticoagulants (or non-use) and death in the hospital. They specifically considered patients sick enough to require ventilators.
They found that patients who received the anticoagulants had better survival. One of the authors was actually the editor-in-chief of the journal that published the study. His name is Dr. Valentin Fuster and he is one of the most-cited most influential voices in cardiology today.
Dr. Fuster was impressed with the results and went to the media to say:
I can tell you any family of mine who will have this disease absolutely will be on antithrombotic therapy and, actually, so are all of the patients at Mount Sinai now.
The problem was that these observations were biased. The choice to use anticoagulation was not random. So you can’t know if the drugs made survival better or whether it was something else.
Multiple experts, writing on Twitter, identified a well-known bias called immortal time bias. The last link explains the problem. But this was only one of the many biases in this observational study.
The Harm of Observational Studies:
Remember in the beginning I wrote that there was no known therapy for COVID19? One of the main reasons is observational studies.
Take the case of Mt. Sinai hospital. Because of this study, and the influence of the researchers, and publication in a big journal, there is a protocol in which sick patients get treated with anticoagulants. A treatment is codified as standard of care. And if it is codified at a big hospital in NY, it will likely be codified in other hospitals.
Here is the issue: now you want to do a study to find out whether anticoagulants work. A patient is asked to participate, which means they could receive placebo. The patient does a Google search and up pops Dr. Fuster’s comments to the Washington Post alongside a headline saying anticoagulants have promise in COVID19.
In this setting, the patient as well as many doctors will no longer have that feeling of uncertainty about the use of anticoagulants. We call this equipoise. And without it, you can’t ethically do trials.
The lack of equipoise causes us to persist with unproven therapies. Take Hydroxychloroquine–we still don’t have a proper trial.
Smart people, people whom I respect, say that the desperation of COVID19 is such that we should use the data we have.
I do not agree. In my column, I argue that observational data may actually be worse than no data.
The title of it is: Some Data May Be Worse Than No Data in the COVID Era
2 replies on “COVID19 and Finding Effective Medical Therapies”
Yes. Controlled trials are important. So does it not follow that there needs to be a controlled trial to see whether “supportive care in a hospital” is effective?
It is beyond obvious that there are myriads problems with hospitalization.
The beds are awful and do not promote sleep.
The food is awful and not what the patient knows he/she needs,
The chances that a drug error ( omission or commission) will be made is huge.
Some nurses are nasty and mean
The rooms are not temperature adjusted. for your comfort
Unnecessary tests are procedures are done
Infectious disease thrive in hospitals
Hospitalization is expensive
I could go on. “Supportive care” needs to be relegated to the dustbin of medicine ( like blood letting) until it is proven to have value. and effectiveness.
“Observational data may be worse. than no data ”
I agree. But “experts” are bombarding the public with the idea that doctors should “experiment” on patients. and the CDC should promote. such experimentation.