There’s something appealing about a threshold-based approach to treatment: if your cholesterol is over 200, it’s high. If your blood pressure is over 140, it’s high. If you A1C is over 7, it’s high. It’s clean, easy to understand and easy to communicate. But is it helping to improve meaningful outcomes, and is it doing so equitably? Does 140 for me create the same problem as 140 for you? Is 141 worlds away from 139? These aren’t easy questions to answer. It complicates the already complicated conversations around taking medications for prevention.
We’re starting to see evidence across different conditions that risk-based approaches to treatment, i.e. those that treat based on aggregation of multiple risk factors into a predicted risk of some adverse outcome, might be better. The first major test-case for this was lipid guidelines, but recommendations for other conditions are starting to change too:
- The 2013 ACC/AHA lipid guidelines changed practice from treating to an LDL target to treating based on calculated 10-year risk of atherosclerotic cardiovascular disease. This is appealing on the surface: the medication is intended to lower risk, so let’s use that risk to assess the need for it. Considerably more patients are statin-eligible under the risk-based guidelines, though, so it’s not a minor change. Certainly this change was controversial, and realized benefits to clinical outcomes have yet to be thoroughly analyzed (time since the change has been widely implemented is still short). The risk factors in the formula include age, sex, smoking, diabetes, blood pressure, and lipid levels. But they don’t include a lot of other things that also shape risk.
- Hypertension guidelines have always varied some degree with age and comorbidies, like laxer targets for older people and stricter targets for people with diabetes. But even still, we’ve typically set “treat to target” goals. However, a recent analysis of a huge pooled sample suggests that a risk-based strategy rather than a threshold strategy could prevent more CV events (or treat ferwer people and prevent the same number of events). This idea is still in infancy, but the writing is on the wall, if you ask me.
- The American College of Physicians has issued new guidance for diabetes management, suggesting personalizing goals and perhaps aiming for 7-8% rather than the more typical strict target of >7. If patients are expected to live less than 10 years, the document suggests, forget A1C targets altogether and treat to avoid symptoms. Personalizing targets isn’t a new idea, but this is a bolder statement about it.
These new guidelines and studies across conditions emphasize shared decision-making, which brings us back to the idea risk comunication. It’s easy to say “your numbers are higher than the cutoff, so this is a good idea.” It’s harder to have a genuine conversation about lowering risk, especially when statistical numeracy among providers is bad and among patients is worse. If patients don’t really understand risk, it’s no wonder they don’t adhere to risk-reducing treatments. Then, there’s a question of how much explaining we do and should do: do we say, “this risk calculation is incomplete and based largely on people who aren’t like you in important ways?” Do we say, “these calculations look at surrogate values and pharmacologic treatments, but don’t have a good way of estimating the effects of powerful lifestyle factors like diet and activity level on outcomes like how you feel”? In other words, do these risk calculations steer us towards recommending interventions that we can clearly see relfected in our equations, whether or not they’re the best strategy for the patient in front of us?
To understand the impact of this kind of strategy, we have to ask ourselves: Are the trials we’re using for evidence looking at patient-oriented, clinically relevent endpoints? Are the trials representing the populations we care for adequately?
For example, traditional CV risk factors may not account for female-specific factors (preganancy-related risk, autoimmune disease, psychosocial issues, treatment difference). What about race/ethnicity? The BPLTTC reports have not included race data in the all-important table 1, and it’s not included in many risk calculations (or if it is, it’s black/white/other, like in the ACC/AHA pooled cohort equation risk calculator).
Finally, most patients have more than one issue. While a risk calculator might ask about diabetes or high blood pressure, what about the myriad of other health problems people live with? In a PLOS medicine article, Rahimi, Lam, and Steinhubl note: “understanding multimordibity requires dealing with complexity and understanding patients’ experiences, emphasizing the importance of interdisciplinary research.” Indeed.
Providers have (hopefully) always considered individual risk and benefit when making treatment decisions. But now more than ever, we have evidence that understanding individual risk is paramount to safe and effective care. So let’s make sure that we take the current evidence for what it is, and we create our future evidence to reflect our patients and their needs better.