I was asked to assist on a project using ECG and biomarkers to screen racing thoroughbreds and classify them regarding their risk of sudden cardiac death. In human athletes the risk is 1-3 per 100,000 athletes and in horses its approximately 10 per 100,000. One of the dilemmas is the ethics of blinding people to the results of the test when it is felt that action such as restricting athletic activity might save an individuals life if a signal is noted. A study in Italy compared the incidence rate before and after the implementation of a screening program in athletes and compared the incidence rate to the non athletic population. There was a decrease in rate after the program was initiated…but since there was no randomization could there be other factors that precipitated the decline in incidence?
I was wondering in horses if you could randomly assign individual horses to a screening program and compare them to a control group that does not receive the screening program. Then compare the differences in sudden cardiac death between groups. Since the incidence rate is low it would likely require a large sample size and some time but would it more accurately reflect the impact of the screening program? https://jamanetwork.com/journals/jama/fullarticle/203513
I think in this case often one would compare two screening models rather than comparing a screening model with no model. Then one can assess the impact of the screening. It is like a RCT where one would compare against standard of care rather than against no treatment.
If the screening program is involved, one could at least compare to just the numerical risk estimates you have given vs this more involved screening program.
This is an interesting question - in a sense it’s asking whether screening can be considered an intervention, and likewise whether we should be testing our screening models rigorously - very relevant to questions about clinical predictive models.
One of the problems with developing a prediction model is the outcome is sudden death…something we are trying to avoid. If researchers find a suspicious EKG finding what are the ethics with regard to informing the caretakers of the subject? In other words findings of the predictive model could be used to institute actions that prevent the outcome of interest therefore biasing the result. If randomized to standard of care or new screening test then we can determine if the screening test has any utility. The other problem is that due to the low incidence rate a large sample size will be required.
Need an ordinal Y with death (rare) at the top and various life-threatening conditions below that.
Hi James
I can’t copy/paste big sections of UptoDate here, but if you know an MD who has access to it, you might want to have a look at the following section: Screening to prevent sudden cardiac death in competitive athletes.
The upshot is that recommendations for screening in humans don’t seem to be based on RCTs. This is understandable, as the trial would have to be very long and enrol a very large number of athletes in order to accrue a sufficient number of outcomes of interest (though, as you say, sudden death rates are higher in horses than in humans). Clinical equipoise might be another barrier (?), especially if a high proportion of sudden deaths involve underlying pathology that plausibly could be picked up through simple screening measures (?)
Presumably, given breeding practices among racehorses and some domesticated animals, the prevalence of conditions like cardiomyopathies (e.g., hypertrophic cardiomyopathy and arrhythmogenic right ventricular cardiomyopathy) and inherited channelopathies will be higher than in human populations (?) And since many organizations do recommend some type of screening for human athletes (recognizing that these practices are imperfect and deaths can still occur among those who screen “negative”), what is the downside to screening racehorses [other than cost to the (likely very wealthy?) owners] ? Wouldn’t a breeder want to know about conditions like this in order to avoid putting the index case and future generations of horses at risk?
I just found this example of an RCT for evaluating a screening program.
Hi James
Many RCTs have been conducted to evaluate screening programs in human medicine. However, this doesn’t seem to be the case for pre-participation cardiac screening (based on a very cursory check, but I might be wrong…).
In humans, the goal of many screening programs (e.g., for various cancers) is to detect prognostically poor, yet treatable diseases at an earlier stage, thereby improving prognosis. But, in contrast to cardiac screening for athletes, cancer screening is not a “one-off” intervention - it’s a serial/periodic intervention.
Imperatives to consider (for humans or horses) in the case of pre-participation cardiac screening:
- The imperative not to subject a “vulnerable” human/horse to a practice (regular vigorous exertion) that might increase the risk for sudden, premature death;
- The imperative not to label a human/horse as “at risk” and limit athletic participation when the absolute future risk of sudden premature death is actually vanishingly small;
- The potential impact of screening results on insurability;
- The psychological impact of adverse screening results on humans and owners.
Let’s pretend that we could recruit a huge number of racehorses and randomize them, at the start of their careers, to cardiac screening or no screening. If we were to then follow them over several years, we would see, tragically, a very small number, in both arms (see below), die suddenly and prematurely during exercise. But since we already know that absolute rates of sudden death are low in the absence of screening, we already know that the impact of our screening program would also, by definition, be very small in absolute terms.
In situations where the absolute benefit of an intervention is expected to be very small, it’s important to be very clear about about what we’re trying to achieve. In this case, we’re trying to prevent the very rare occurrence of loss of many years of life. Since sudden deaths during sports are still reported very occasionally, even among athletes who have undergone cardiac screening, it’s reasonable to assume that the same would be true for racehorses, even after implementation of a universal cardiac screening program. Accepting our inability to eliminate risk completely, the next question to ask would be what degree of risk reduction would justify implementation of universal cardiac screening for horses, IF such a reduction could be demonstrated through an “ideal” RCT? For example, would a reduction in the rate of sudden death from 10 to 6 per 100 000 (presumably these rates apply over a horse’s racing career) justify such a program? What about a reduction from 10 to 8 per 100 000? Arguably, unless there is a priori consensus around the answer to this question, there’s no point in trying to design an RCT to quantify the “impact” of a screening program.
An alternate way to look at the issue would be to start with the premise that, even though we will never be able to eliminate sudden cardiac death, every such case is tragic enough, in both humans and horses, that we want to do everything possible to reduce the risk. Although screening tests (e.g. ECG, echocardiogram) would be normal in the vast majority of animals, we would expect, occasionally, to identify an occult high-risk condition (e.g., hypertrophic obstructive cardiomyopathy) that would preclude a racing career. In the case of horses, identifying such conditions would presumably also be important to inform breeding decisions.
All things considered, what would you consider to be the downside to universal cardiac screening of racehorses? Would any particular RCT result change this assessment?