(Updated on October 25 with a response from James Stein)
We have plenty of sophisticated gear to understand heart problems. For COVID-19, we’ve seen Troponin blood tests, cardiac MRIs and echo. For example, cardiac injury was associated with a higher risk of death in hospitalized patients. And yet, we have no answers to the questions that matter for clinical practice.
Echo and cardiac MRI may well be valuable in the assessment of people with COVID-19 (and indeed many conditions) but we don’t yet have evidence for this. We just assume more detailed information has to be better. This isn’t necessarily true. Test results can trigger further tests or treatments that might lead to more harm than good overall. It might have been better not to know. This concept is a hard sell but surely we should focus our finite resources on things with evidence-based benefits?
So, where should go from here?
If I was a patient (and I suppose in a way, we all are potential patients), I might wonder:
- Has COVID-19 caused me to have heart damage?
- How bad is the heart damage?
- Will my heart recover?
Clinicians think about how to diagnose and treat disease so they might wonder:
- What are the effects of COVID-19 on the heart?
- Who is at risk of heart problems because of COVID-19?
- What cardiac tests should people with COVID-19 have?
- What is the significance of abnormal cardiac tests in people with COVID-19?
- What treatment is there?
To address the above questions, researchers might ask:
- What is the prevalence of cardiac injury in patients with COVID-19?
- What is the mechanism of cardiac injury?
- What is the clinical course of people with cardiac injury and COVID-19?
- Are there features that make some hearts more vulnerable to injury?
- Do people with cardiac injury tend to be people with pre-existing heart problems?
- How do the effects of COVID-19 on the heart compare to the effects of other viruses and to the effects of other critical illnesses?
- Do traditional treatments for cardiovascular disease improve prognosis in people with COVID-19?
There are studies that look at these questions but a systematic assessment of an unbiased sample of people has not been done. In other words, I have not come across a study in which all the people in an unbiased sample have the same set of tests. If you retrospectively collect data you find that some people were so well (or too sick or declined) they didn’t get a blood test or echo and we lose the chance to reliably estimate the prevalence of cardiac injury.
Even everyone having the same set of tests is still not enough! The cardiac injury may occur later in the disease than the time of testing. People have to be tracked over time. This requires a prospective cohort study. And not just at one unit because then the findings might only be applicable to patients from the local population that attend that unit, a population which will be different to the rest of the world.
And when I say all people, I mean people both with and without COVID-19 (the control group!), those with and without hospital admission and with and without symptoms. The selection of the sample makes all the difference to how you interpret the results. This isn’t easy to achieve. It’s much easier to just analyse data you have routinely collected or to do a small study in one unit with “matched controls”. Unfortunately, these study designs are vulnerable to bias.W
We need a prospective cohort study run across the a sample of the whole population
This does not come without substantial planning, coordination and resources, but it (or the closest practical alternative) will come.
The question of treatment can only be answered with randomised controlled trials
You may say I’m idealistic. Of course, for research studies to practically take place, compromises often have to be made. But some compromises bias the results to the extent that they are no longer useful and can seriously mislead us. We shouldn’t expend resources on these studies.
James Stein sent the following response to this post: