Prevention guidelines: bad process, bad outcome.

Abstract

Thomas Jefferson, the third American President and author of the Declaration of Independence, famously opined, “Governments derive their just powers from the consent of the governed.” The principle underlying this simple statement is an essential component of any collection of rules that govern how individuals conform to behavioral standards. Clinical practice guidelines should explicitly or implicitly follow the same principles. Guidelines are effective only when they involve the participation and consent of the stakeholders whose behavior they intend to govern. The 2011 Institute of Medicine (IOM) report1 in 2011 (Clinical Practice GuidelinesWe Can Trust) emphasizes transparency and public involvement as integral to thedevelopmentprocess. In this context, therecently issuedguidelinesforcholesterol reduction2 represent an important failure of guideline governance and oversight process. Rather than forging a consensus on cholesterol management, the guidelines have further polarized the debate on appropriate use of statinmedications. How did the process go awry? Many of the current concerns stem from a process lacking openness and transparency, as recommended by the IOM. The Prevention Guidelines were originally commissioned in 2008 by the National InstitutesofHealth (NIH), the supervisingauthority for all previous lipid guidelines.Writing efforts lastedmore than5years from initiation to completion. During the development process, several committee members resigned, although the exact reasons for theirwithdrawal havenever been revealed.An adequate explanation for the longperiodof development also has not been provided. Although 4 separate guidelines were issued, despite years in development, the guidelines are narrow, not comprehensive. The main drug therapy recommendations address only 3 “critical questions,” primarily describing which patients should receive statins. After commissioning the guidelines, in June 2013, the NIH withdrew as the supervising authority.3 As a consequence of the NIH decision, the guideline committee was abandoned as a stateless entity, a group of writers with no oversight or commissioning organization. Rushing to fill the void, the American College of Cardiology (ACC) and American Heart Association (AHA), assumed responsibility for oversight, but instead of reconstituting the committee and recommissioning the document (in the usual thoughtful and deliberate fashion), they apparently just accepted the result “as is.” Worse yet, the AHA pressed to release the guidelines in advance of the AHA Scientific Sessions to take advantage of the publicity provided by such an important document. These decisions, the abandonment by the NIH, the unquestioning acceptance by the ACC and AHA, and the rush by AHA to release the document had unfortunate and preventable untoward effects. The “governed” (practicing physicians) were left in a bewildered state, eroding the confidence required for successful implementation. Oneof themost seriousproblemsof theguidelines is carefully analyzed in the article by Cook and Ridker4 in this issue of JAMA Internal Medicine. The guidelines included a newly developedandheretoforeunpublishedcardiovascular risk calculator,designedto identifypatientsmost likely tobenefit from statins. In thecurrentmanuscript,CookandRidker4 apply this new risk calculator to apopulationwith knownoutcomes and showquite convincingly that the risk calculator supplied is inaccurate. Using the Women’s Health Study cohort, the observed risk is substantially lower than the riskpredictedby the guidelines. The extent of miscalibration is substantial, overestimating risk by at least 50%. This is not a trivial problem. While statins arevaluabledrugs,particularly in secondaryprevention, they do have downsides, and prudence requires not administeringdrugs topatientswhowill likelynotbenefit. The implicationsof theoverestimationof risk areprofound.A50% overestimationbytheguideline riskequationswould likelyadd millions of Americans to the roles of patients for whom statins are recommended. The poor performance of the risk calculator is selfevident to any clinician who attempts to make a calculation for sample patients. A 65-year-old white woman with a total cholesterol levelof 175mg/dL,anhigh-density lipoproteincholesterol (HDL-C) level of 54mg/dL, and a low-density lipoprotein cholesterol (LDL-C) level of 96 mg/dL, a blood pressure (BP)of 134mmHg(while taking lisinopril, 5g/d) reachesa“predicted risk” of 7.7%, above the 7.5% level at which treatment with a statin is recommended. Similarly, a 47-year-old African American man with total cholesterol level of 170 mg/dL, an HDL-C level of 45 mg/dL, an LDL-C level of 95 mg/dL, and aBPof 129mmHg(while takinghydrochlorothiazide, 25mg/d) exceeds the 7.5% threshold for lifelong treatment. Most practitioners would not treat such patients. (To convert total, HDL-C, and LDL-C cholesterol to millimoles per liter, multiply by 0.0259.) In theirpublic response to this criticism, theguidelinewriters and professional societies organized a scrambled defense basedmostlyonhubris rather than thoughtful analysis.Rather than investigating the concerns, several authors5 stated that the guidelines were never intended to be taken at face value. Rather, theywerea“startingpoint” for a conversationwith the patient about treatment. As suggested by Montori et al,6 it is critically important that any decision to begin statins be discussed with patients, consistent with shared decisionmakingprinciples. Such advice is sensible, but issuing recommendations using a flawed risk calculator undermines the Related article page 1964 Research Original Investigation Cardiovascular Risk Calculator Controversy