Author + information
- Received December 14, 2014
- Revision received February 9, 2015
- Accepted February 11, 2015
- Published online April 28, 2015.
- Ann Marie Navar-Boggan, MD, PhD∗∗ (, )
- Eric D. Peterson, MD, MPH∗,
- Ralph B. D’Agostino Sr., PhD†,‡,
- Michael J. Pencina, PhD∗ and
- Allan D. Sniderman, MD§
- ∗Duke Clinical Research Institute, Duke University Medical Center, Durham, North Carolina
- †Department of Mathematics and Statistics, Boston University, Boston, Massachusetts
- ‡Harvard Clinical Research Institute, Boston, Massachusetts
- §Mike Rosenbloom Laboratory for Cardiovascular Research, McGill University Health Centre, Royal Victoria Hospital, Montreal, Canada
- ↵∗Reprint requests and correspondence:
Dr. Navar-Boggan, Duke Clinical Research Institute, Duke University Medical Center, 2400 Pratt Street, Durham, North Carolina 27705.
Background New cholesterol guidelines emphasize 10-year risk of cardiovascular disease (CVD) to identify adults eligible for statin therapy as primary prevention. Whether these CVD risk thresholds should be individualized by age and sex has not been explored.
Objectives This study evaluated the potential impact of incorporating age- and sex-specific CVD risk thresholds into current cholesterol guidelines.
Methods Using data from the Framingham Offspring Study, this study assessed current treatment recommendations among age- and sex-specific groups in 3,685 participants free of CVD. Then, it evaluated how varying age- and sex-specific 10-year CVD risk thresholds for statin treatment affect the sensitivity and specificity for incident 10-year CVD events.
Results Basing statin therapy recommendations on a 10-year fixed risk threshold of 7.5% results in lower statin consideration among women than men (63% vs. 33%; p < 0.0001), yet most of the study participants who were 66 to 75 years of age were recommended for statin treatment (90.3%). The fixed 7.5% threshold had relatively low sensitivity for capturing 10-year events in younger women and men (40 to 55 years of age). Sensitivity of the recommendations was substantially improved when the treatment threshold was reduced to 5% in participants who were 40 to 55 years of age. Among older adults (66 to 75 years of age), specificity was poor, but when the treatment threshold was raised to 10% in women and 15% in men, specificity significantly improved, with minimal loss in sensitivity.
Conclusions Cholesterol treatment recommendations could be improved by using individualized age- and sex-specific CVD risk thresholds.
The new American College of Cardiology/American Heart Association (ACC/AHA) blood cholesterol guidelines strongly recommend statin therapy consideration in adults with elevated 10-year risk of atherosclerotic cardiovascular disease (CVD) (1), which is defined as a 7.5% or higher risk of experiencing a cardiovascular event over the next decade. The guidelines also provide an option to treat those adults with a risk between 5% and 7.5%.
These latest guidelines increase the number of adults in the United States recommended for statin therapy by nearly 13 million (1,2). The majority of the increase is among adults 60 years of age or older, in whom statin therapy is now recommended for 80% or more of patients (3). These newer guidelines base their primary prevention treatment recommendations on the 10-year predicted CVD risk model using the Pooled Cohort Equations. The thresholds for therapeutic consideration were selected on the basis of 3 statin primary prevention trials: 1) MEGA (Primary Prevention of Cardiovascular Disease with Pravastatin in Japan); 2) AFCAPS/TexCAPS (Air Force/Texas Coronary Atherosclerosis Prevention Study); and 3) JUPITER (Justification for the Use of Statins in Prevention: An Intervention Trial Evaluating Rosuvastatin). This selection of treatment thresholds has been a point of controversy (4). Ideally, one would want to treat a majority of those patients who will subsequently develop a CVD event (high sensitivity) while avoiding unnecessary treatment among those patients not who will not develop a CVD event (high specificity). Yet to date, it is unclear whether the guideline-recommended thresholds achieve this goal among certain subgroups, such as younger and older adults, and women and men. Furthermore, it is possible that the performance of the guidelines could be improved if treatment thresholds were age and sex specific.
Using data on adults from the Framingham Offspring Study, we estimated the sensitivity and specificity of current ACC/AHA cholesterol treatment guidelines for identifying adults at risk for CVD. We also estimated the impact of varying the 10-year risk thresholds that are used to identify adults for statin therapy across age and sex groupings, to determine how sensitivity and specificity may be affected by using different risk thresholds.
Our study population included adults from the Framingham Offspring Study, a longitudinal population-based study that started in 1975, enrolling the children of the original Framingham Heart Study cohort and their spouses. We evaluated adults 40 to 75 years of age who were free of CVD at examination cycle 3 or 6. Baseline characteristics (at examination 3 or 6), including blood pressure, blood pressure treatment, diabetes status, smoking status, sex, and lipid levels, were used to calculate a person’s 10-year estimated risk on the basis of the Pooled Cohort Equations. We excluded those adults with data missing for variables in the Pooled Cohort Equations (n = 43). The basis of the statin treatment recommendations was the 2013 ACC/AHA cholesterol guideline (1). The number and percentage of adults recommended for statin therapy were first calculated on the basis of the 2013 ACC/AHA cholesterol guideline for all patients and then calculated by age-specific (40 to 55 years, 56 to 65 years, and >65 years) and sex-specific (women and men) subgroups.
Adults were followed prospectively for new-onset CVD events over the subsequent 10 years. These events were defined as a nonfatal myocardial infarction, death from coronary heart disease, fatal or nonfatal stroke, peripheral arterial disease, or heart failure. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the guideline statin treatment recommendations were calculated on the basis of a patient’s predicted likelihood for subsequent CVD events, while taking into account the time-to-event nature of the data (5). Analyses were stratified by age (40 to 55 years, 56 to 65 years, and 66 to 75 years) and sex. Next, we varied the treatment threshold for statin therapy from 3% to 20% and determined the effect these changes had on guideline performance characteristics (e.g., sensitivity, specificity, PPV, NPV).
Framingham Study participants provided written informed consent for participation. This analysis was approved by the Duke University Institutional Review Board. All analyses were performed using SAS software version 9.3 (SAS Institute, Inc., Cary, North Carolina).
Our study population included 3,685 adults; their average age was 57.2 years. Table 1 shows the characteristics of the study population at the “baseline” visit (examination 3 or examination 6). Overall, 46.8% of adults met criteria for statin therapy on the basis of the 10-year risk threshold of 7.5% used in the current guidelines. Treatment recommendations varied substantially by age and sex. Among adults 40 to 55 years of age, 21.0% met criteria for statin therapy, compared with 55.2% of adults 56 to 65 years of age and 90.3% of adults 66 to 75 years of age. The 10-year CVD event rate also increased across age groups (12.5%, 17.5%, and 23.0%, respectively), but not as sharply as the rate at which statin therapy was recommended. In women, the 10-year CVD rate was 11.0% versus 22.6% in men.
Tables 2 and 3 display the proportion of women and men, respectively, recommended for statin therapy, along with sensitivity, specificity, PPV, and NPV of the guidelines on the basis of varying treatment thresholds ranging from 3% to 20%, stratified by age group. Differences in sensitivity and specificity by age group in women and by age group in men are shown in the Central Illustration.
In younger women and younger men (40 to 55 years of age), the guideline sensitivity for identifying adults who would develop CVD over the next 10 years by using a cutoff of 7.5% was relatively low, with only 36% of future cases identified as treatment candidates in women and 49% of future cases identified in men. Reducing the 10-year risk threshold to 5% in women and men who were 40 to 55 years of age markedly improved sensitivity (from 36% to 48% in women and from 48% to 71% in men), with only a modest reduction in specificity (90% to 87% in women and 72% to 56% in men).
More marked differences were seen in guideline performance between men and women who were 56 to 65 years of age. Because 84% of men were recommended for statins in this age group at a cutoff of 7.5% versus only 29% of women, sensitivity was markedly higher in men compared with women (90% vs. 49%, respectively). Using a risk threshold of 7.5%, the guidelines were able to identify 9 in 10 men who would develop CVD over the next 10 years, but fewer than 1 in 2 women. When the optional cutoff of 5% for statin use was extended to women 56 to 65 years of age, sensitivity improved to 65%.
In men 66 to 75 years of age, guideline sensitivity and specificity varied little across treatment thresholds because nearly all men in this age group were recommended for statin therapy—even at a risk threshold of 10% (97% recommended). Consequently, specificity for a threshold of 10% or less was extremely low (3%). Raising the 10-year risk cutoff in older men to 15% led to a reduction in the number of men recommended for statin therapy (97% to 89%), with no change in sensitivity (96% to 96%) but improved specificity (3% to 14%). Even when the threshold was increased to 20% (thereby reducing the number treated with a statin to 68%), sensitivity was reduced only to 87%, whereas specificity improved to 40%. Similarly, among older women, increasing the threshold from 7.5% to 10% reduced the sensitivity only slightly from 95% to 87%, but it nearly doubled the specificity from 17% to 34%.
The newest ACC/AHA cholesterol guidelines substantially increase the number of individuals recommended for statin therapy, particularly among older women and men and among adult men of all ages. Our study evaluated the treatment thresholds used by the new guidelines to determine the number treated who are likely to develop actual CVD events over the next decade. We found that guideline treatment performance varied in clinically significant ways across age- and sex-specific groups. Existing thresholds had decreased sensitivity for predicting future CVD events in women and younger adults, and they had poor specificity for predicting such events in men and older adults. Most importantly, we found that guideline treatment performance could be improved by selecting age- and sex-specific thresholds for statin initiation.
Previous work has questioned the accuracy of the Pooled Cohort Equations in predicting 10-year atherosclerotic CVD risk, with criticism focusing on the calibration and discrimination of the risk estimates (6,7). However, as applied in the new AHA/ACC cholesterol guidelines, the Pooled Cohort Equations are used not to predict precise risk, but rather to stratify adults into “high-risk” (≥7.5% 10-year risk) or “low-risk” categories using binary risk categories. Therefore, the continuous precision of the Pooled Cohort Equations may actually be less important than its ability to divide patients appropriately into those above and below a particular risk threshold. Our objective was not to examine the calibration of the Pooled Cohort Equations, but rather to focus on the practical issue of altering statin treatment thresholds by age- and sex-specific groups to determine the performance characteristics of these treatment recommendations.
Our study highlights 2 key concerns regarding the use of a fixed risk threshold in any type of treatment guideline. First, the 7.5% threshold adopted by the current cholesterol guidelines may not be optimal for identifying younger adults in need of statin therapy. Using the “optional” 5% threshold, we observed an improvement in guideline performance sensitivity in identifying younger adults (40 to 55 years of age) who are at risk for premature CVD (Central Illustration). This sensitivity improvement supports the use of the optional 5% threshold proposed in the current ACC/AHA guidelines. Furthermore, our results underscore the importance of efforts aimed at increasing the ability to detect younger adults who are at increased risk for CVD, beyond what is offered by current 10-year risk estimation methods. Similarly, the sensitivity of the guidelines was poor among women 56 to 65 years of age, a finding suggesting that this is another group deserving of improved risk prediction models. Measures of subclinical disease, comprehensive lipid profile, biomarkers, and algorithms focused on 30-year or lifetime risk should be examined for their impact on the sensitivity of future CVD detection, particularly in these risk groups.
Our second major concern is that although most older adults (66 to 75 years of age) were recommended for statin therapy, that recommendation was made significantly more for men than for women in this age group. Because most (9 in 10) of the older men had a 10-year risk of 7.5%, the specificity of this recommendation was poor among older men, thus leading to high rates of potentially unnecessary statin treatment. We found that guideline specificity in older men can be substantially improved with no impact on sensitivity by raising the treatment threshold to 15%. Nonetheless, the high proportion of those patients recommended for treatment and the low specificity achieved in this age group suggest that efforts should be dedicated to improving discrimination between those who require statin treatment and those who do not.
We found that a slight adjustment of classification thresholds can improve the overall operating characteristics of the current fixed 7.5% threshold guideline. For example, when we used a lower threshold of 5% in adults 40 to 55 years of age, a 7.5% threshold in those 56 to 65 years of age, a 10% threshold in women 66 to 75 years of age, and a 15% threshold in men 66 to 75 years of age, the overall number of adults recommended for treatment increased by only 2% (from 47% to 49%), whereas the overall guideline sensitivity increased from 71% to 77%, with only a small decrease in specificity (from 58% to 56%).
We acknowledge that there is no such thing as a perfect threshold; improved sensitivity for predicting future events comes with the tradeoff of reduced specificity and vice versa. As a result, providers and patients must weigh potential risks of statin therapy (including cost considerations) against the perceived benefit of these drugs when deciding whether or not to use statin therapy for primary prevention. As the guidelines suggest, the thresholds proposed should be used as starting points in a conversation between patients and providers, with a focus on shared decision making; however, our study demonstrates that these treatment thresholds should not be fixed at an arbitrary number, but rather determined with consideration of a patient’s age and sex.
First, our study population, adults in the Framingham Offspring Study, is not nationally representative. Given that the Pooled Cohort Equations are race specific, similar analyses using alternate, racially diverse populations should be performed to evaluate the potential need for race-specific thresholds as well. Second, because adults in the Framingham Offspring Study contributed to the derivation cohort used for the Pooled Cohort Equations, the Pooled Cohort Equations may be more accurate in this population than in a nationally representative cohort. Nevertheless, older age and male sex have been shown to be risk factors for CVD across multiple cohorts, so the overall magnitude and direction of our estimates (higher sensitivity in older adults and men) are not likely to change when applied to a broader sample. Finally, adults included in our study were predominantly enrolled in the Framingham Offspring Study in the 1990s. Contemporary changes in risk factor distribution and improvements in CVD prevention efforts may contribute to lower overall rates of CVD than in those studied in our cohort. This would lead to an overestimation of PPV and an underestimation of NPV but should not affect guideline sensitivity and specificity estimates.
Considering the Pooled Cohort Equations in the context of current cholesterol guidelines highlights how differences in the age- and sex-specific distribution of adults recommended for statin therapy affect guideline performance. Establishing age-and sex-specific 10-year risk thresholds to identify adults for statin therapy may improve the balance between avoided CVD events and unnecessary therapy, as well as enhance overall guideline performance.
COMPETENCY IN PATIENT CARE: The decision to begin statin therapy should be shared between patient and provider and should consider differences in risk and guideline performance related to age and sex.
TRANSLATIONAL OUTLOOK: More information about age- and sex-specific CVD risk thresholds warranting therapy could improve individualized implementation of future cholesterol treatment guideline recommendations.
The authors thank Erin Hanley, MS, for her editorial contributions to this manuscript. Ms. Hanley did not receive compensation for her contributions, apart from her employment at the institution where this study was conducted. The Framingham Heart Study is conducted and supported by the National Heart, Lung, and Blood Institute (NHLBI) in collaboration with Boston University (Contract No. N01-HC-25195).
This study was supported by unrestricted grants from the Doggone Foundation and with research funds from the Duke Clinical Research Institute. The Framingham Heart Study is conducted and supported by the National Heart, Lung, and Blood Institute (NHLBI) in collaboration with Boston University (contract N01-HC-25195). This article has been reviewed by Framingham Heart Study investigators for scientific content and consistency of data interpretation with previous Framingham Heart Study publications. Dr. Peterson has received funding for research grants from Eli Lilly and Janssen Pharmaceuticals; and funding for serving as a consultant/participant on advisory boards for Merck, Sanofi, Janssen Pharmaceuticals, and Boehringer Ingelheim. Dr. Pencina has received funding for serving as a consultant for AbbVie. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
Drs. Pencina and Sniderman contributed equally to this work.
- Abbreviations and Acronyms
- cardiovascular disease
- negative predictive value
- positive predictive value
- Received December 14, 2014.
- Revision received February 9, 2015.
- Accepted February 11, 2015.
- 2015 American College of Cardiology Foundation
- Stone N.J.,
- Robinson J.,
- Lichtenstein A.H.,
- et al.
- Andrus B.,
- Lacaille D.
- Goff D.C. Jr..,
- Lloyd-Jones D.M.,
- Bennett G.,
- et al.
- Cook N.R.,
- Ridker P.M.