The incidence of ischemic heart disease (IHD) among women has increased in several Western countries [1, 2]. Compared with men, the diagnosis of IHD occurs later in life and the symptoms are more ambiguous, which results in underdiagnosis and undertreatment [3, 4]. More knowledge of predictive and modifiable risk factors in middle-aged women is therefore needed in order to identify high-risk individuals and introduce preventive actions at an earlier stage [4, 5].

Blood cholesterol is regarded as one of the most important risk factors for IHD, but the recommendations for clinical use of different lipid measures to predict cardiovascular risk have changed over the years and still diverge [6, 7]. Risk assessment models are useful for clinicians, especially in primary care, to identify high-risk individuals, but the models differ in the choice of lipid measure [8]. Total cholesterol is widely used in Sweden and many other countries [9], but European and American guidelines also recommend non-high-density-lipoprotein-cholesterol (non-HDL-C) for cardiovascular risk assessment, especially in people with diabetes mellitus, obesity or low levels of low-density-lipoprotein-cholesterol (LDL-C) [10‐12]. As non-HDL-C includes both LDL-C, very low-density lipoprotein cholesterol (VLDL-C) and intermediate lipoproteins, non-HDL-C is likely to be more atherogenic than either lipoprotein alone and may therefore have a higher predictive ability of IHD [10, 13]. A recent multinational study found that increasing concentrations of non-HDL-C can predict long-term cardiovascular risk, particularly when the increase is modest at a young age [14]. Other studies have suggested that the ratio of total cholesterol (TC) and HDL-C (TC/HDL-C) is the most efficient IHD predictor [15‐17]. In a previous study of the same cohort, we showed that TC/HDL-C was a strong predictor for acute myocardial infarction in middle-aged women [18]. However, the clinical use of TC/HDL-C ratio is not widely spread and long-term studies of non-HDL-C and TC/HDL-C in relation to IHD in women are relatively scarce [6]. A more comprehensive picture of cholesterol measures is therefore needed for prediction of IHD risk in middle-aged women.

The first aim was to examine the association between both TC/HDL-C and non-HDL-C and IHD during a 17-year long follow-up of a large sample of middle-aged women, adjusting for a comprehensive set of potential confounders, i.e. age, exercise, smoking, waist-hip ratio, diabetes mellitus type 2 and blood pressure. The second aim was to compare the predictive ability for IHD between TC/HDL-C and non-HDL-C. We also estimated the 5-, 10- and 15-year risks of IHD after baseline adjustments.

Data for this cohort study is based on the Women’s Health in Lund Area (WHILA) study, which is a prospective cohort study of middle-aged women in southern Sweden. The details of the study have been described previously [3, 19]. Briefly, between 1995 and 2000, a total of 6916 women aged 50–59 years underwent a health survey after prior invitation to the study. Due to missing values in several variables and exclusion of those with IHD before baseline, a sample of 6537 women was included in the present study. After written consent, a physical examination with measurement of minimal waist and maximal hip circumference was performed. Blood pressure (mm Hg) was measured twice in the right arm, after 15 min and 20 min rest in sitting position, and the average was used. Non-fasting serum levels of TC and HDL-C were measured with a Cholestech LDX-instrument (Cholestech Corporation, Hayward, CA, USA) on capillary whole blood [19, 20]. Furthermore, the women completed a self-administered questionnaire about medical history, medication, lifestyle and sociodemographic data. The study was approved by the Regional Ethical Review Board in Lund (approval no. 2011/494).

The baseline characteristics of the women are shown by the categorised variables of TC/HDL-C (Table 1) and non-HDL-C (Table 2). The women were, on average, 56 years of age at the screening and 80% were menopausal. Smoking, high WHR, high blood pressure, diabetes, and low level of exercise were more prevalent in the higher categories of TC/HDL-C and non-HDL-C, respectively. During the mean follow-up period of 17 years, a total of 551 women (8.4%) suffered from IHD. The means of TC and TC/HDL-C were higher in the higher quintiles, and HDL-C was lower (Table 3).

In Table 4, hazard ratios (HR) of TC/HDL-C and non-HDL-C for IHD are presented, with adjustments for age and for all the included variables. We found an increasing association by quintile between TC/HDL-C ratio and IHD and between non-HDL-C and IHD. The highest HR in quintile 5 was 2.30 (95% CI: 1.70–3.11) for TC/HDL-C and 1.67 (95% CI: 1.25–2.24) for non-HDL-C, after adjustments for all the included variables.

The continuous relationships of TC/HDL-C and non-HDL-C with IHD are shown in Fig. 1 and Fig. 2. We found that non-HDL-C was linearly related to IHD (p = 0.58); however, TC/HDL-C was not to the same extent linearly related to IHD (p = 0.07).

In Table 5 the models are evaluated by Harrell’s C, to assess the IHD predictive ability of the lipid measures. There was a significant difference between TC/HDL-C and non-HDL-C in the age-adjusted models (p = 0.003), which indicates that TC/HDL-C has a higher predictive ability for IHD than non-HDL-C, but this difference disappeared in the full model (p = 0.061). Moreover, the full models had a significantly higher predictive ability than the age-adjusted models. The AIC test also showed a better fit (higher predictive ability) for TC/HDL-C than for non-HDL-C in the age-adjusted models (AIC was smaller for TC/HDL-C than for non-HDL-C).

In Table 6, the risk of IHD (%) at 5, 10 and 15 years is presented by quintile of TC/HDL-C and non-HDL-C. The risk calculation is based on the full models and adjusted for baseline values, i.e. it shows the independent effect of the cholesterol measures. The risks at 5, 10 and 15 years increased by increasing level of TC/HDL-C and non-HDL. The risk of IHD at 15 years was 7.0%; 9.0% in quintile 4 (TC/HDL-C; non-HDL-C) and 12.9%; 10.6% in quintile 5.

In this cohort study of middle-aged women with a mean follow-up of 17 years, we found a strong association with IHD for both TC/HDL-C ratio and non-HDL-C, after adjustment for age, exercise, smoking, waist-hip ratio, blood pressure and diabetes. The results showed that the higher TC/HDL-C or non-HDL-C, the higher the risk of IHD. Moreover, in comparisons between the two lipoprotein measures, we found that TC/HDL-C had a slightly higher predictive ability for IHD than non-HDL-C. However, this difference disappeared when other confounding factors were taken into account. The absolute risk of IHD over the 15 years of follow-up was 12.9% and 10.6%, respectively, in quintile 5. The strong association indicates that these lipoprotein measures are an important part of cardiovascular risk assessment and a cornerstone of preventive efforts to middle-aged women in clinical practice.

Previous studies have mostly had a shorter follow-up or studied older men and women individuals than in the present study [6, 14]. A cohort study of American middle-aged women, who were followed for 10 years, found that non-HDL-C and TC/HDL-C had higher predictive ability of cardiovascular events than total cholesterol [24], which is recommended by SCORE guidelines [25]. In contrast, a study of older adults (mean age 69 years) found that non-HDL-C was not associated with IHD in women, while TC/HDL-C was [26]. Our results are in line with a multinational study that was published in Lancet in 2019, which analysed the long-term relationship between non-HDL-C and cardiovascular disease in middle-aged individuals, and found that the cardiovascular risk was strongly differentiated by non-HDL-C levels, particularly beyond 10 years [14]. By use of a derivation and validation design, the study simulated a primary preventive risk-modelling, and found that a 50% reduction of non-HDL-C was associated with reduced risk of cardiovascular disease by the age of 75 years, and the earlier the cholesterol was reduced, the greater was the risk reduction. Another meta-analysis also found an age-dependent association between IHD and non-HDL-C and IHD and TC/HDL-C, where the risk increase was higher in younger individuals [27].

Some previous studies have stated that TC/HDL-C is the best predictor for IHD [27, 28] and others have preferred non-HDL-C [14]. A comparison between the two measures has only been addressed in a few previous studies. One UK study made a similar comparison as we did with AIC, and found that TC/HDL-C is a stronger predictor of CHD risk than non-HDL-C, but this study was restricted to men and women with type 2 diabetes [29]. Another study of men and women from the multicenter Atherosclerosis Risk in Communities (ARIC) study found that 21% of individuals with low non-HLD-C had high levels of TC/HDL-C, which was associated with a 29% greater risk of incident atherosclerotic cardiovascular disease [30]. The results of the present study indicate that the predictive ability of TC/HDL-C is higher than that of non-HDL-C for IHD in women, but the difference was not large, which means that either of the lipid measures may be used in clinical practice.

We believe that the results of the present study have relevant clinical implications. Some women have high total cholesterol and LDL-C, but also high HDL-C. For these women, it is important to calculate TC/HDL-C or non-HDL-C to predict their cardiovascular risk more accurately, as the risk may not be increased at all. These women probably do not benefit from, e.g. medication with statins. The recommendations for which lipid measure to use for cardiovascular risk prediction has changed over the years, and in many countries total cholesterol is widely used, e.g. in the SCORE risk prediction model [25]. Early identification of high-risk individuals is important to be able to offer preventive efforts, and long-term risk prediction models are of importance [31]. A recent systematic review of cardiovascular disease risk prediction models identified 363 different models, and total cholesterol was the most common measure for blood lipids, followed by HDL-C [8]. The more recently developed risk prediction model from the PREDICT study in New Zealand, uses TC/HDL-C ratio instead of total cholesterol, and the study concluded that older risk prediction models need to be updated to better fit the modern population [17].

The role of lipoproteins in the development of IHD involves lipid deposition in the arterial wall and the growth and progression of atherosclerotic plaques, with subsequent plaque rupture and formation of an arterial thrombosis [11]. The cumulative exposure to lipoproteins is probably related to the total atherosclerotic plaque burden (concentration and duration); thus, a healthy lifestyle and, sometimes, lipid lowering medication may reduce the risk for IHD. In primary care, it is essential with risk factor assessment to identify high-risk individuals, and subsequent primary prevention with encouragement of physical activity and a healthy diet, low in saturated fat and with a focus on wholegrain products, vegetables, fruit and fish [11, 32]. A recent study of national trends of different cholesterol measures showed that non-HDL-C and TC/HDL-C have decreased in most Western countries, Japan and South Korea between 1980 and 2015, in contrast to China, where TC/HDL-C and non-HDL-C have increased [33]. Dietary changes in Western countries, and to a certain degree the use of statin medication, are important explaining factors behind this trend. Low levels of the healthy HDL-C are associated with obesity and intake of trans-fats and carbohydrates, whereas high levels are associated with physical activity, alcohol consumption and intake of total and unsaturated fat [33‐37]. Several randomised controlled trials have concluded that changing dietary saturated fat to polyunsaturated vegetable oil is associated with an improved lipoprotein profile and may reduce the risk of cardiovascular disease by 30% [36].