Dear JULES,
Risk factors for women & heart disease
Executive Summary
* A study was published in JAMA Cardiology on January 20th, 2021.
* The study included 28,000 women from the US Women’s Health Study, who were followed-up for 21 years. The goal was to examine the first diagnosis of Coronary Heart Disease (CHD) and to see which risk factors this was most associated with.
* The study reviewed women in four age groups – under the age of 55; 55 to under 65; 65 to under 75 and 75 or older.
* Reassuringly, 94.5% of women did not develop heart disease during 600,000 person years of follow-up.
* Having diabetes at the start of the study was the single biggest risk factor for developing CHD in any age group and especially when younger. Women with diabetes were nearly 11 times more likely to develop heart disease before the age of 55 than women without diabetes.
* Metabolic syndrome was the second highest risk factor – dwarfing even smoking. Blood lipid measures (cholesterol) were inconsequential in comparison with blood glucose measures.
* A 1997 Scottish paper undertook a similar exercise by ranking risk factors for all-cause mortality in men and women separately. This also found that there were far greater risk factors than cholesterol and yet cholesterol (different measures thereof) has been the major focus for heart disease in the almost 25 years between these two papers.
* This metabolic vs lipid finding should be a game changer, but it probably won’t be.
Introduction
Many thanks to Dr Eric Westman for this week’s topic. Eric sent me a paper which was published in JAMA Cardiology on January 20th, 2021. It was called “Association of lipid, inflammatory, and metabolic biomarkers with age at onset for incident Coronary Heart Disease in women” (Ref 1). The goal was to examine first incidence of heart disease in women and to see which baseline characteristics of those women were associated with heart disease. Women were examined in four groups, providing an age-dimension to the study. The age groups were under the age of 55; 55 to under 65; 65 to under 75 and 75 or older.
With the caveat that this was only a study of women, and only a study of American women (and thus not generalisable to men and/or other populations) the study produced some remarkably interesting findings, which might well be more widely applicable.
The study
The researchers reviewed 28,024 women from the US Women’s Health Study. The average (median) follow-up was 21.4 years. The women were aged 45 or older. They did not have cardiovascular disease (CVD) at the start of the study. The researchers were interested in more than 50 lipid, inflammatory, and metabolic risk factors, and biomarkers (these included factors from BMI to cholesterol and from activity to diabetes).
The outcome of interest was the first confirmation from medical records of Coronary Heart Disease (CHD). CHD was defined as any of the following: a first heart attack, percutaneous coronary intervention (a stent being fitted), coronary artery bypass grafting (heart bypass surgery), or death from CHD.
The results
The first interesting and reassuring finding was how few women experienced CHD even during a large study with over 21 years of follow-up. Table 1, the characteristics table, reported the number of women who had confirmed CHD during the study by age group. I’ve extracted the relevant numbers below and then calculated these as a percentage of total women:
The table above shows that just 63 women had CHD before the age of 55 – that represented 0.2% of all women. There were 384 women diagnosed with CHD between the ages of 55 and 65. There were 654 women diagnosed with CHD between the ages of 65 and 75 and then 447 aged 75 or older. The age group 65 to under 75 had the highest incidence of CHD at 2.3% – interestingly not the 75 or over age group. Overall, 94.5% of women did not develop CHD. Remember that this was over an average 21 years of follow-up and hence there were 1,548 cases of CHD in approximately 600,000 person years of follow-up.
The characteristics table confirmed that, as ever, there was a healthy person confounder. The women most likely to develop CHD were less educated, with a higher BMI, more than three times as likely to be a current smoker, far more likely to be inactive, eight times more likely to have diabetes, two and a half times more likely to have metabolic syndrome, and twice as likely to have high blood pressure. (These were all raw data relativities before adjustment).
These factors are usually adjusted for, but this paper was trying to see which risk factors were associated with CHD, so it measured these differences rather than adjusting for them. It did adjust for factors that weren’t the health measures under examination. Hence it adjusted for race/ethnicity, educational level, menopause, postmenopausal hormone use, and interactions between the risk factor of interest and age groups. For example, on the latter, blood pressure tends to increase with age, so it would adjust for that to try to establish the relationship between CHD incidence and blood pressure aside from age.
The risk factors
The study performed an interesting statistical analysis using the standard deviation. The standard deviation is a measure of the amount of variation in a set of values. If everyone in a school class is aged 11-12 – the standard deviation for age (in years) will be small as the average age is similar. If everyone in the whole school is aged 11-18, the standard deviation (in years) will be greater because the variation from the average age is greater.
In a normal distribution, approximately 68% of the data fall within one standard deviation of the average (mean). We don’t know if the many markers examined in this study followed a normal distribution, but we can assume with some certainty that most results fell within one standard deviation. Those with a result one standard deviation or more away from the average, therefore, were measurably different to the average person.
This study used the standard deviation of each marker to assess the association between CHD and that marker. For example, the average systolic blood pressure (BP) (the higher of the two readings) was 125 mm Hg in the women who did not develop CHD. The standard deviation for BP was 13.7 mm Hg (supplementary file). The researchers found that women with systolic blood pressure 13.7 mm Hg higher were 2.2 times more likely to develop CHD by the age of 55 (Table 2 main paper).
Now that we know the statistical technique used, let’s just drop the standard deviation terminology and report the associations between risk factors and CHD. In the table below, I have extracted key risk ratios from Table 2 in the main paper. I have omitted the confidence intervals to keep it simple. All the risk ratios extracted are statistically significant. I have sorted the table in order of the worst risk factor for having CHD before the age of 55.
Here are the most striking results:
– Diabetes was the single highest risk factor bar nothing. Women with diabetes were 10.7 times more likely to develop CHD before the age of 55. They were 10.9 times more likely to develop CHD between the ages of 55 and 65. They were still between 3.5 and 4.5 times more likely to develop CHD over the age of 75 or between the ages of 65 and 75, respectively.
– Diabetes dwarfed even smoking. Being a current smoker gave risk factors of approximately 2-4 times higher risk of developing CHD across age groups.
– Metabolic syndrome was the second highest risk factor. Metabolic syndrome was defined according to the consensus criteria of the American Heart Association and the National Heart, Lung, and Blood Institute (Ref 2).
– Physical inactivity did not make a significant difference (hence why it is not in the above table, which only includes significant risk factors).
– Total cholesterol and LDL-cholesterol were inconsequential compared to diabetes and metabolic syndrome. The cholesterol marker most associated with CHD was triglycerides, which is a useful marker for carbohydrate intake (Ref 3).
– There is a confounder with the cholesterol numbers, which will make the risk ratios higher than they should be. One of the criteria for incidence of CHD was having a stent fitted. This is something recommended by a physician if they think the patient is at higher risk. If the woman has higher cholesterol (total and LDL especially), the physician is more likely to think that she is at risk. There is thus a confounding association between having higher cholesterol and having an intervention that qualifies as a CHD incident. This makes the diabetes/metabolic syndrome even more of a greater risk factor relative to lipid profiles.
– In almost all cases, the risk factor numbers reduce (or stay the same) as you read across each row. This means that, for example, metabolic syndrome is a greater risk for the under 55 age group than it is for the 55-65 age group, than it is for the 65-75 age group, than it is for the over 75 age group. This is probably because you can only have the first incident once and if you have a significant risk factor (like metabolic syndrome) it is likely that you will have a first incident younger than the woman who doesn’t have this risk factor.
– The risk ratios for the top risk factors are so large that they are likely to be causal rather than mere association. The Bradford Hill criteria start with the strength of association. If this is double or greater, it is worth examining the other criteria. The diabetes risk factor is 10 times stronger – that’s five times double! (Ref 4)
Eric shared this as “This may be the first study in CHD to suggest that: Treating metabolic syndrome is more important than TChol, LDL; and treating obesity is more important than TChol, LDL.” Eric might well be right in terms of how clearly this paper compares the risk of metabolic syndrome vs the ‘risk’ of cholesterol measures. However, long before we used terms such as metabolic syndrome and while cholesterol was just total cholesterol and not LDL-cholesterol (let alone small dense/large fluffy), there was evidence that cholesterol was not the be-all-and-end-all of heart disease…
The Scottish Heart Health Study
This paper reminded me of a study I first saw presented by Dr Malcolm Kendrick at a conference we hosted in Birmingham in 2013 (Ref 5). It was called “Comparison of the prediction by 27 different factors of coronary heart disease and death in men and women of the Scottish Heart Health Study.” The (1997) paper contained the following diagrams illustrating which risk factors were associated with the most important measure of all – all-cause mortality.
The first chart is for women. This shows that many factors were not significant, including total cholesterol, energy intake, alcohol and BMI (BMI being non linear as opposed to non signficiant). Interestingly blood glucose and diabetes were also not significant, but this was almost 25 years ago and these were far less prevalent conditions then. The important risk factors for women in the Scotland study back in 1997 were having had a previous heart attack, having a sedentary job, serum cotinine levels (which reflect exposure to cigarette smoke) and smoking. Being inactive in leisure time, blood pressure and triglycercides were also signficiant risk factors.
The second chart is for men. This shows that many factors were not significant, including total cholesterol, energy intake, alcohol, and BMI. For men, diabetes was significant – and number six on the list of risk factors. The top risk factors for men in the Scotland study back in 1997 were having had a previous heart attack, plasma fibrinogen (clotting factors), serum cotinine levels (exposure to cigarette smoke) and smoking. Being inactive at work and leisure were also significant risk factors for men.
Eric’s interesting paper has shown that blood glucose measures (diabetes and metabolic syndrome) were far higher risk factors than blood lipid measures (cholesterol) in a study of American women. The risk factors were so large that causation would be reasonably inferred, and the results may well, therefore, be generalisable beyond American women. This should be a game changer in terms of heart disease treatment, but arguably we should not have been so focused on cholesterol from the outset. We have had at least one study, almost 25 years before Eric’s one, showing that total cholesterol was not a risk factor in all-cause mortality. Yet, in between these studies, cholesterol (total cholesterol and then more recently LDL-cholesterol) has been heralded as THE major risk factors for heart disease and deaths.
This may well have been to the detriment of focus on a more important risk factor – glucose measures. There is a double whammy to this detriment. The cholesterol theory demonises fat (the diet heart hypothesis) while the glucose theory demonises carbohydrate. If glucose is more of an issue than cholesterol, then carbohydrate is more of an issue than fat. The triglycerides measure showing up as a risk factor confirms that we should be more worried about carbohydrates than fat. Oh, what damage have we done?!
Until the next time
All the best – Zoë
References
Ref 1: Dugani et al. Association of Lipid, Inflammatory, and Metabolic Biomarkers With Age at Onset for Incident Coronary Heart Disease in Women. JAMA Cardiology. January 2021. https://pubmed.ncbi.nlm.nih.gov/33471027/
Ref 2: “The metabolic syndrome is a constellation of interrelated risk factors of metabolic origin—metabolic risk factors—that appear to directly promote the development of atherosclerotic cardiovascular disease (ASCVD). Patients with the metabolic syndrome also are at increased risk for developing type 2 diabetes mellitus. Another set of conditions, the underlying risk factors, give rise to the metabolic risk factors. In the past few years, several expert groups have attempted to set forth simple diagnostic criteria to be used in clinical practice to identify patients who manifest the multiple components of the metabolic syndrome. These criteria have varied somewhat in specific elements, but in general they include a combination of both underlying and metabolic risk factors.
“The most widely recognized of the metabolic risk factors are atherogenic dyslipidemia, elevated blood pressure, and elevated plasma glucose. Individuals with these characteristics commonly manifest a prothrombotic state and a pro-inflammatory state as well. Atherogenic dyslipidemia consists of an aggregation of lipoprotein abnormalities including elevated serum triglyceride and apolipoprotein B (apoB), increased small LDL particles, and a reduced level of HDL cholesterol (HDL-C).”
https://www.ahajournals.org/doi/10.1161/circulationaha.105.169404
Ref 3: Parks EJ. Effect of Dietary Carbohydrate on Triglyceride Metabolism in Humans. The Journal of Nutrition 2001
Ref 4: https://www.zoeharcombe.com/2016/09/the-bradford-hill-criteria/
Ref 5: Tunstall-Pedoe et al. Comparison of the prediction by 27 different factors of coronary heart disease and death in men and women of the Scottish Heart Health Study: cohort study. BMJ 1997.