Thank you to my friend and colleague Paul Ricard, DPT, CCS, clinical specialist at John’s Hopkins Medical Center and Director of the Critical Care Fellowship in PT for sending me a recent article about the relationship between cardiac stress test performance and mortality. The study (abstract here) is based on a very large sample of subjects (~58,000) receiving cardiac stress testing (for a variety of indications - but all related to testing for cardiovascular disease). It is a retrospective cohort study, and there is a temporal sequence between the events. The events were when subjects took their stress test and when they may have had an event - in this analysis the event of interest was death as they were coming up with a prediction equation for 10 year survival. The authors admit in one sentence of the discussion that the results are most relevant to people that have been referred to receive a cardiac stress test. In other words, we do not know, based on this analysis whether we can apply this model to someone referred to receive a stress test for the purpose of the prediction equation; or whether it applies to a sample of subjects without cardiac symptoms or risks that would typically result in referral for a stress test. That is a welcome and honest caveat by the authors of the study.
There is one other thing to point out in this study - at least in terms of the practical implications (practice suggestions) being made by the study authors. The study hopes that the prediction model could be used by health care providers to encourage patients to exercise more after a stress test based on their 10 year survival score. The study is not making (and has not tested) an explicit causal pathway of fitness to survival. They are, however, implicitly assuming such a causal association exists. The fact of an association between fitness and cause can be explained by two causal directions and each direction has at least two causal pathways - and they are not mutually exclusive.
Causal direction 1: From fitness to survival
Causal direction 2: From survival to fitness
For causal direction 1 we have possible pathways:
Being fit leads to increased probability of living
Being unfit leads to decreased probability of living (that is increased probability of dying)
It is important to recognize these as separate, and both capable of driving an association. For example, just because being obese leads to higher rates of CVD does not mean that not being obese lowers your risk in an absolute way. It simply means it lowers your risk in a relative way - relative to someone obese. So it should not be surprising when someone that is “not obese” has a cardiac event.
For causal direction 2 we have possible pathways:
Having a higher probability of living leads to being fit
Having a higher probability of dying leads to being less fit
Since these 4 contingencies are not mutually exclusive, all of them may be driving the association that is allowing the prediction equation to meet a level of discrimination worthy of consideration for use (statistically speaking). Remember the statistics are based on the association, not the underlying causal structure. But practice is about the underlying causal structure. How we can apply the association depends greatly on the causal structure. In this particular situation there is basically no risk in assuming causal direction 1 and pathway 1; which lead to the unquestionable recommendation to exercise to become more fit to increase the probability of survival, or to exercise to avoid becoming less fit to reduce the risk of death (non survival). This was implicit in the paper’s intended practical use of the predication model - I simply wanted to make it explicit.
I suppose now my blog will have to come with a warning label since it is “EXPLICIT”
Please also keep in mind, my pointing this out does not mean that I disagree with the prediction model, or advocating that people exercise. It is simply to point out our use of causal reasoning when we go from studies to practice as a matter of pedagogy.