January 8, 2016
Personalization of Treatment Recommended More
I admit I am happy to see more articles about this and more writers emphasizing personalization of diabetes treatment repeatedly. Doctors may not be happy about this, but it is of their own doing, as they only operate on a one-size-fits-all theory.
I am surprised that the Veterans Health Administration supplied the data for the retrospective cohort study. This was done to examine the rate at which physicians discontinued or deintensified therapy when their diabetes patients' glycemic or blood pressure (BP) levels were low enough to do this. Participants included one cohort of 211,667 patients with diabetes who were older than 70 years and were receiving active treatment for hypertension, and a second cohort of 179,991 patients receiving active treatment for hyperglycemia.
Patients were considered eligible for deintensification if they had low BP or a low glycated hemoglobin (A1c) level in their last measurement in 2012 (index measurement). Very low BP was defined as less than 120/65 mm Hg and moderately low as systolic BP of 120 to 129 mm Hg or diastolic BP less than 65 mm Hg. Very low A1c was defined as less than 6.0%, and moderately low as 6.0% to 6.4%. The main outcome measure was discontinuation or treatment dose reduction within 6 months after the index measurement. The investigators also examined whether life expectancy played a role in the rate of deintensification.
Survey responses indicated that "risk of hypoglycemia" and "life expectancy" were considered the most important parameters and thus received the highest weights. "Important comorbidities," "complications," and "cognitive function" were moderately important, while "disease duration" and "resources and support system" received the lowest weights.
To use the algorithm, a clinician simply scores each of these eight items as low, moderate, or high risk, and the algorithm generates an A1c target. In the six original cases as well as the three new cases, the algorithm suggested a target that was nearly identical to the A1c values recommended by 57 experts.
Clinical decision-making is a tricky process, especially when the patient is complex. Because diabetes affects essentially every major organ and system of the human body, few diseases are more complex. To simplify treatment, professional organizations such as the American Diabetes Association (ADA) and the European Association for the Study of Diabetes (EASD) develop guidelines and recommendations that are based on the best available evidence. When new evidence emerges that conflicts with existing guidelines, the complexity of treating complicated patients is amplified.
There is no better example than with diabetes, where lower glycemic and blood pressure levels proved not to be ideal for everyone, although there were subsets for which lower targets were better. Another possible explanation for lack of deintensification is that presenting patients seemed okay so clinicians chose the conservative path of leaving things alone. As Sussman and colleagues conclude, however, clinicians "should assess the potential harms of intensive therapy just as they do the benefits," a process that may demand new tools and a personalized approach.
An important concern with Cahn and colleagues' algorithm (and others that might be developed in similar ways) is that it is entirely based on expert opinion. There is no empirical data to verify that the identified A1c target is indeed the "right" one for the patient at hand. Furthermore, the very nature of personalization makes such evidence impossible to obtain. Because clinicians are scientists who are used to practicing evidence-based medicine, making such an important decision without clear-cut evidence may be uncomfortable. Still, the result could make the patient more comfortable. Isn't that the goal?