Risk Markers and Intermediate Outcomes in Research and Clinical Practice

In adult cardiovascular disease, researchers have introduced a plethora of new imaging and biomarkers in hopes that these novel risk markers will either improve risk prediction for individuals, lend biological plausibility to modifying underlying behaviors or more aggressively treating traditional risk factors like blood pressure and lipids, or offer insights into pathophysiology that is amenable to drug therapy.¹ One of these novel factors is retinal vessel diameters, which predict future cardiovascular outcomes, at least in women.^2,–4 In this issue of Pediatrics, Köchli et al⁵ examine retinal vessel diameters in children. They posit that physical activity, BMI, and blood pressure affect retinal vessel diameters, which reflect early signs of microvascular disease, which, in turn, leads to adult cardiovascular outcomes. With this article, we provide an opportunity to consider when novel risk factors are useful either in making clinical decisions or in research.

Pediatric clinicians often use risk markers, by themselves or in combination with others, to stratify patients into risk groups and then intervene on higher-risk children to reduce adverse longer-term health outcomes. For example, clinicians measure BMI to categorize children as having obesity. Obesity predicts many adverse health outcomes; along with recent studies revealing that intensive treatment is effective, clinicians now have evidence that routinely measuring BMI in childhood is reasonable.⁶ For a new marker to be clinically useful, it must be a better or less expensive predictor or easier to obtain than what already exists. It may be a single predictor, as with BMI, or part of a multivariable score as in adult cardiovascular risk calculators.⁷ Ultimately, before including a new measure in practice, clinicians should have evidence that intervening on individuals they identify as high-risk results in better outcomes. Sometimes, markers used to categorize individual risk reasonably fail the intervention test and thus are not useful in practice. A (controversial) example in adult cardiovascular disease prediction is plasma C-reactive protein.⁸ Among moderate risk adults, C-reactive protein predicts cardiovascular outcomes. However, in trials in which C-reactive protein is lowered, it is not clear that lowering C-reactive protein is the explanation for better outcomes. Hence, guideline committees differ as to whether they recommend its use.⁸ 

As one might expect, given the long follow-up needed, no authors of longitudinal studies have directly addressed associations of childhood retinal vessel diameters with adult cardiovascular outcomes or even with the same measure in adulthood. In addition, although adding adult retinal vessel diameters to existing cardiovascular risk prediction scores may have marginal benefit for women, it appears not to be helpful for men.^4,9 Therefore, even if childhood retinal vessel diameters strongly predicted its counterpart in adulthood, it is unclear whether it would be a useful measure for pediatric clinicians trying to group their patients into high- and low-risk of eventual strokes and heart attacks.

In addition to amassing an evidence base for prediction, researchers investigate etiology and pathogenesis. Some novel factors lie on a causal pathway between modifiable exposures and actual health outcomes and thus are intermediate outcomes, often called mediators.^10,11 They are common in pediatrics in part because, increasingly, pediatric researchers are concerned with prevention of future adult disease, which is distant from what we can easily measure among youth.¹² An example is screening and intervention for smoking during adolescence. Researchers assume that interventions to prevent smoking in adolescence will lead to lower rates of smoking in young adulthood¹³ and, hence, reduce the risk of cancers in adulthood. Thus, pediatric researchers use smoking rates among adolescents, an intermediate outcome, as a study end point to obviate the need to follow large populations for decades to determine actual health outcomes. Because there is no evidence of the relationship between childhood retinal vessel diameters and either its adult counterpart or health outcomes, using it as an outcome in studies of modifying childhood physical activity, BMI, or blood pressure is not currently justified.

Although retinal vessel diameters are not currently a strong candidate for clinical risk prediction or study outcomes, it has intuitive appeal because the eye provides the opportunity to view systemic arteries and veins directly. As technological advances make this novel assessment more available to clinicians,¹⁴ we hope that such availability is accompanied by additional evidence for or against its usefulness among clinicians and researchers.