May 03, 2016
7 min read

Analysis: Pooled-cohort risk equation overestimates CV risk in real-world cohort

You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact

Applying the American College of Cardiology/American Heart Association pooled-cohort risk equation to a real-world cohort resulted in a significant overestimation of actual 5-year CVD risk in adults without diabetes, researchers found.

The researchers estimated risk using the pooled-cohort risk equation for 307,591 adults (mean age, 55 years; 62% women) who were in Kaiser Permanente Northern California databases in 2008.

“The 2013 ACC/AHA pooled-cohort risk equation was a commendable step forward, as it helps to estimate risk for both [CHD] events and ischemic stroke,” Jamal S. Rana, MD, PhD, told Cardiology Today. “However, the equation was developed from several groups of enrolled volunteers primarily conducted in the 1990s with limited racial and ethnic diversity, so there has been a concern that its accuracy may vary in current community-based populations.”

Jamal S. Rana

All participants had complete 5-year follow-up and met the following criteria at baseline: aged 40 to 75 years, no known atherosclerotic CVD, no diabetes, LDL 70 mg/dL to 189 mg/dL and no use of lipid-lowering therapies before or during the study period.

Rana and colleagues documented atherosclerotic CVD events from 2008 to 2013 and compared predicted vs. observed 5-year risk for atherosclerotic CVD. They also performed a separate analysis of 4,242 adults with diabetes who met all the other criteria.

Rana, a cardiologist at Kaiser Permanente Northern California, Oakland, and adjunct investigator with its division of research, and colleagues documented 2,061 atherosclerotic CVD events during 1,515,141 person-years in the cohort without diabetes.

Observed rates lower

The researchers wrote that in each category of predicted 5-year risk for atherosclerotic CVD, observed rates were greatly lower:

  • in the group with predicted risk of less than 2.5%: mean expected risk, 1.04%; observed risk, 0.2% (95% CI, 0.2-0.25);
  • in the group with predicted risk of 2.5% to less than 3.75%: mean expected risk, 3.08%; observed risk, 0.65% (95% CI, 0.55-0.7);
  • in the group with predicted risk of 3.75% to less than 5%: mean expected risk, 4.34%; observed risk, 0.9% (95% CI, 0.75-1); and
  • in the group with predicted risk of at least 5%: mean expected risk, 8.72%; observed risk, 1.85% (95% CI, 1.75-1.95).

Calibration was poor and discrimination was moderate (C statistic = 0.74), Rana and colleagues wrote.

Results were consistent across ethnic groups (C statistic range, 0.7 to 0.74) and socioeconomic status, according to the researchers.

“Substantial overestimation based on the current risk calculator translates into the likelihood of overtreating a lot of people who really are at low [CV] risk,” researcher Alan S. Go, MD, chief of cardiovascular and metabolic conditions research at Kaiser Permanente Northern California Division of Research, told Cardiology Today. “Our study provides critical evidence to support recalibration of the risk equation in real-world populations.”

Alan S. Go, MD

Alan S. Go

In the cohort with diabetes, 2% had acute MI, 1.3% experienced CHD-related death and 0.2% had ischemic stroke during the 5-year period. Rana and colleagues reported that calibration was fair between predicted vs. observed risk for atherosclerotic CVD events, but discrimination was modest (C statistic = 0.64).

Participants treated with statins for primary prevention during follow-up were separately analyzed, and the researchers found that their observed rates of atherosclerotic CVD were approximately six times lower than expected. Calibration was poor and discrimination was modest in these participants with (C statistic = 0.61) and without (C statistic = 0.68) diabetes, they wrote.

“We believe our findings have important clinical and health policy implications for determining who should be recommended for lifelong statin treatment,” Rana told Cardiology Today. “We endorse physician–patient shared decision-making and are not recommending any sudden change in current practice. However, given that statin therapy is a mainstay of treatment for millions of Americans, our study highlights the importance of ongoing research and dialogue in this area to provide more rigorous evidence to guide treatment for the patients in contemporary diverse populations.”

Divergent conclusions

In a related editorial, Michael J. Blaha, MD, MPH, wrote that the pooled-cohort equation has performed well in some cohorts but not in others, and “the main reason studies have reached different conclusions is the choice of study population.”

Michael Blaha, MD, MPH

Michael J. Blaha

Blaha, from the Johns Hopkins Ciccarone Center for the Prevention of Heart Disease, wrote that although “we should remain somewhat skeptical of the magnitude of these findings” because of the potential for missed events, the results are consistent with the continued decline in atherosclerotic CVD rates.

He concluded that, “The best solution is to incorporate discussion of confidence (or lack thereof) in risk estimation and the potential for overestimation into the clinician-patient discussion.” – by Erik Swain

Disclosure: Rana is an employee of Kaiser Permanente Northern California. Please see the full study for the other researchers’ relevant financial disclosures. Blaha reports no relevant financial disclosures.