Are patients with hyperlipidemia undertreated?

METHOD

We conducted the study at an acute care community hospital in Winnipeg, Man. Before we started the study, we obtained approvals from the University of Manitoba Health Research Ethics Board and from the hospital. The Research Ethics Board waived the need for patient consent to access medical charts. We reviewed consecutive charts with admission diagnoses of acute MI, unstable angina, or acute coronary syndrome from the 2002–2003 and 2003–2004 fiscal years. Charts were excluded if lipid levels had not been measured within 24 hours of admission, if there was inadequate history to establish risk status, if patients were not expected to survive for 1 year due to noncardiac illness, or if patients were treated with only palliative measures.

We used a “treat to target” assessment tool for calculating the 10-year risk of coronary artery disease in nondiabetic people, as recommended by the Canadian Working Group on Hypercholesterolemia and Other Dyslipidemias.8 Depending on the person’s risk category, the tool directs users to the recommended target levels for LDL-C. We considered patients with diabetes as very high risk in keeping with the Canadian guidelines and based on the observation that diabetic patients without overt vascular disease appear to share the same risk as nondiabetic patients with documented vascular disease.12

Using a moderate effect size of 0.6 standard deviations1,5,13 and an α set at .05 (1-tailed), power analysis revealed that we needed a sample size of 26 subjects per risk group (low, moderate, high, and very high) to achieve a power of 90%14 in detecting a significant difference between actual and target LDL-C levels.

Pertinent demographic information was recorded along with the information necessary to complete a “treat to target” tool based on Framingham risk calculations. Each subject’s risk points were added together to determine their 10-year risk of having coronary artery disease, allowing for risk categorization. Target LDL-C levels were recorded for each subject based on the 2000 Canadian guidelines.8

Our initial intent was to review charts until we had adequate data for analysis in each risk category. It became apparent, however, that very few patients fell into the high-risk category, and further culling of charts to locate sufficient patients in that category would have exceeded our resources. Hence, the 3 records in the high-risk group were not subjected to further analysis. Data were entered into an Excel 5.0 spreadsheet and analyzed using the Statistical Analysis software (SAS Institute).

Descriptive statistics were used to report numerical data on age, total cholesterol, high-density lipoprotein cholesterol, LDL-C, blood pressure (BP), calculated risk, and length of stay in hospital. Categorical data, such as primary diagnosis, smoking status, diabetes, and outcome, were described by relative frequencies expressed in percentages. Relative frequencies of risk categories and use of lipid-lowering agents before admission were calculated. One-sample t tests were used to determine the difference between actual and target LDL-C levels in each risk category. A 1-tailed α of .05 was used to determine statistical significance.

Limitations

This study was a retrospective chart review in a hospital. Office records, including patients’ lipid levels and the duration of lipid-lowering therapy were generally unavailable. Framingham-based risk assessment, although recommended by guidelines, is neither the only nor necessarily the best predictor of risk. Some risk factors, such as family history, were not taken into account.

Blood pressure at time of presentation to hospital might not represent “usual” BP. We attempted to correct this by using the first documented daytime BP on the second day of admission when patients were free of pain. Given patients’ recent physical stress and emotional reactions, their BP might still have been higher than usual. On the other hand, many patients would have received new or increased doses of BP-lowering medications and would have been on bed rest, which might have led to underestimation of true BP. Most patients still fell into the very high-risk category, suggesting that the effect of any underestimation of BP on our results was negligible.

With the exception of the high-risk group, the sample size was adequate to determine statistical significance. We cannot comment further on the high-risk group, but this has no effect on the validity of our findings concerning the other groups. There were only 24 patients in the moderate-risk group when we stopped collecting data, but the difference between their mean LDL-C level and the target level was highly statistically significant; finding 2 more such patients would have taxed our resources with no chance of having a material effect on this finding.

Patients in previously published surveys7,15 have been identified from pools of stable outpatients. Our study adds to the current literature by identifying patients at a different point in the spectrum of hyperlipidemia and coronary artery disease: on admission to hospital with coronary events, at which time we were able to estimate their risk of such events.