The Case For Making Dosing Decisions Based On Risk … Not LDL Measurements
Author:
Stuart T Haines, PharmD, BCPS, BC-ADM
I’ve long believed that statin therapy should be guided by a clinical risk assessment – not LDL cholesterol goals. None of the landmark clinical trials demonstrating the benefits of statin therapy have titrated the dose to an LDL goal. None. Participants were eligible for the studies based on clinical risk factors and, in most cases, a baseline LDL measurement. In the JUPITER study1, they used hs-CPR to determine a patient’s risk of CVD. But, once the participants met the eligibility criteria, everyone received a fixed dose of a statin. They didn’t titrate the dose or add agents or switch agents when participants didn’t meet the lipid goal. The LDL goals established by the National Cholesterol Education Program Adult Treatment Panel III (NCEP III) are based on the assumption that achieving the MEAN LDL cholesterol that was observed in the randomized clinical trials will produce similar results in practice. But what if that assumption is incorrect? Perhaps it’s the STATIN DOSE that’s most important (not the LDL target).
A recent analysis published in the Annals of Internal Medicine challenges the current standard of practice by examining the potential benefits of treating patients with fixed doses of statins based on risk rather than titrating the dose based on LDL cholesterol monitoring.2 Using data from NHANES III (National Health and Nutritional Survey III) and the Framingham study, the authors estimated the baseline cardiovascular risk of a simulated population of patients between the ages of 30 to 75 years old who had no history of myocardial infarction. Next, they constructed two models of care. In the Treat-to-Target strategy, the dose of the statin was actively adjusted based on LDL monitoring to achieve traditionally recommended NCEP III LDL goals (standard model) or optional LDL goals (intensive model). The statin drug choice and dose under each of these models would be titrated, starting with simvastatin 20mg and increasing in dose and potency to atorvastatin 80mg. The Treat-to-Target model is the most common method of managing patients with elevated LDL cholesterol levels today. Of course, there were certain assumptions built into the model – such as the starting dose most patients would receive, the number of dose titrations that would be required to achieve the goal LDL, and the number of LDL cholesterol measurements that would be obtained over a 5-year observation period. These assumptions were subsequently tested using a sensitivity analysis. Under the second patient management strategy, the Tailored Treatment model, all patients with an estimated 5-15% coronary artery disease (CAD) risk over 5 years received a 40mg fixed dose of simvastatin. Those with a >15% risk of CAD were treated with a 40mg fixed dose of atorvastatin. The Tailored Treatment approach did not involve titrating the statin therapy or measuring lipid profiles.
After constructing the two models of care, the investigators estimated the percentage of patients in the population who would receive treatment and the potential benefits (e.g., reduction in CVD events) under each model. The estimated benefits were based on the Heart Protection Study (HPS), Incremental Decrease in End Points Through Aggressive Lipid Lowering (IDEAL) trial, and Treat to New Targets (TNT) study – all landmark trials.3,4,5Further, they estimated treatment discontinuation rates and the disutility (e.g. the adverse effects of treatment and CVD events) for each of the treatment strategies. Again, the authors had to make assumptions to arrive at these estimates and these assumptions were subsequently tested in their sensitivity analysis.
Under the Treat-to-Target approaches, the number of Americans between 30 and 75 years old who would be treated was 37.9 million (standard approach) to 53.4 million (intensive approach). Under these two Treat-to-Target approaches, millions of CVD events (1.67 to 2.39 million, respectively) would be prevented and there would be substantial increase in quality adjusted life years (QALYs). Under the Tailored Treatment strategy, the number of Americans who would be treated was 53 million, but, as dictated by the treatment approach, none would receive atorvastatin 80mg. Surprisingly, the Tailored Treatment approach resulted in the greatest number of CVD events averted (2.82 million) and the most QALYs saved. Most of us don’t treat millions of patients in our practices, so let’s examine the data:
Table 1: Outcomes prevented per 5 years of treatment; Adults aged 30-75 years old with no history of MI
Rate per 1000 person treated
Events |
Deaths |
QALYs Saved |
|
Treat-to-Target (Standard) |
44.0 |
1.9 |
48 |
Treat-to-Target (Intensive) |
44.9 |
1.9 |
45 |
Tailored Treatment |
53.2 |
2.4 |
55 |
Nearly as many patients would be treated under the Tailored Treatment approach when compared to the Intensive Treat-to-Target approach, but more events would be prevented. And the quality of life years would be greater if the Tailored Treatment approach was adopted. Moreover, the sensitivity analysis was rather robust – no single assumption had a large effect on the results. Even a 3-way sensitivity analysis, adjusting for the three most powerful factors in the model, failed to favor the Treat-to-Target approach. Thus, under no reasonable circumstance would the Treat-to-Target approach result in more events prevented or QALYs saved. If you factor in the extra health care visits and lab tests required under the Treat-to-Target approach, the Tailored Treatment strategy wins hands down.
So what does this all mean? I know this is a heresy, particularly to those of you who manage lipid clinics, but I believe it’s time to stop making dosing decisions based on LDL cholesterol levels. There has never been a clinical trial that used a treat-to-target strategy (despite its name, the TNT trial did not actually titrate the dose of atorvastatin). So why are we doing this in practice? Now, I’ll admit that lipid clinics can play a role in educating patients, assessing cardiovascular risk, and assuring adherence to pharmacologic strategies and lifestyle modification. But, I believe these functions can be managed by pharmacists working in primary care practices or, better yet, community pharmacies. We don’t need fasting lipid panels to make treatment decisions for most patients. It adds unnecessary costs and diverts our attention from more important things. Let’s stop titrating statins and prescribe target doses based on risk, similar to what we do for ACE inhibitors for heart failure and antiplatelet therapy for acute coronary syndromes.
I welcome your comments!
Stuart: I certainly agree
Stuart: I certainly agree with your analysis. In May 2006 Andreas Laupacis and colleagues published an effectiveness and efficiency study of various lipid guidelines in BMJ. The US guidelines were the worst. The New Zealand guidelines performed the best in number of deaths avoided. They use a strategy almost identical to the recent Annals paper; lower risk = simva 40 and highest risk = atorva 40 with no dose titrating. I think that the influence of PHARMA on the guideline members in the US delayed this more evidence-based approach.
There are currently discussions ongoing about developing an accreditation council for guidelines in the US. Organizations would send their methodology to this group for review. If the guidelines met certain evidence-based standards they would be an “accredited” guideline. The current NCEP-ATP, JNC VII, AACE Lipid guidelines and the ADA diabetes guidelines would not meet these criteria. They are considered more “opinion-based” guidelines.
Jupiter trial under attack
I’m curious if anyone is changing their thought process on this in light of the articles just published in Arch Intern Mede Lorgeril M et al “Cholesterol lowering, cardiovascular diseases, and the rosuvastatin-JUPITER controversy” Arch Intern Med 2010; 170:1032-1036.
The usual arguments are being made about conflict of interest and bias. Questions about the decision to stop the Jupiter trial early are being asked, and what the reasons were as they were not well delineated. In the article by de Lorgeril, they point to the fact that the curves for all cause mortality appear to have been converging when the trial was prematurely ended, and if follow up had continued, the significant (borderline) difference may well have vanished. Additionally, there is question about the statistical data presented in Jupiter. de Lorgeril and colleagues found no difference in CV mortality with rosuvastatin vs. placebo. The differences in fatal vs. non-fatal MI also seem off.
Thoughts?