Epi in the News

I cannot see the news article unless I get the paid subscription?

Hello everybody... here are my answers...

1) The outcome measure is the occurrence of coronary heart disease in the form of cardiac arrest. Its an incidence measure- they followed healthy ppl for over 16 years and looked if they developed CHD

2) Exposure was the cumulative LDL levels. The exposure here is a combination of high LDL levels and the duration of exposure to high LDL levels- they wanted to look at the duration that a person was exposed to high LDL level and explore a cumulative effect. I did look a bit into the article as well- they calculated cumulative LDL as mg/dL X years- and this they split into 4 quartiles and compared the survival in each of these quartiles.

3) Prospective cohort study, with survival analysis

4) The exposure was cumulative LDL- mg/dL X years- this was divided into 4 quartiles- quartile 4 having the largest cumulative exposure to LDL.

5) So here they calculated the hazard ratios- the lowest quartile would be the baseline group.
The highest group vs lowest quartile is 1.57- which is 57% increased risk of CHD at any one point in time.

6) Since the association did not reach statistical significance, the confidence intervals would include the point of no difference here- which would be one..

7) It does not mean that there is no association- this could just be a false negative finding because of lack of power to evaluate this question.

8) Case control design would be much better for rare outcomes- power wise atleast.

9) Smoking would be quite an important confounder here ... smoking is know to make the effect of LDL worse and reduces HDL. And smoking is independantly associated with CHD due to the chemicals themselves- rather than via cholesterol (so not in the pathway as well). Since it satisfies the criteria for being a confounder, it was best that they adjusted for it...

10) Statins or lipid lowering medication would keep the LDL under control and reduce the incidence of CHD.... so probably its needed to look at the cumulative effect of LDL in those who take medication and those who don't separately- I am guessing they were maybe expecting an effect modification...

11) Well for causal relation- here there is temporality, there is biological reasoning, there is consistency, there is a dose-response relation even- they are showing a strong statistically significant relation that is specific as well (they controlled for other factors), and it goes with previously published facts- sort of like an extension of what is already known.... I think thats more than enough Hills criteria to say causal association. RCT would not be required here...


P.S: If anybody wishes to read another article related to the same study try this-
https://www.tctmd.com/news/cumulative-burden-ldl-cholesterol-increases-chd-risk

The link to the article is here- only the abstract is available though-
https://jamanetwork.com/journals/jamacardiology/article-abstract/2784038

The above 2 links should be enough to answer the questions- in case you are not able to access the New York Times article.

Fathima

Hi Fathima and Sujit,

1. What were the outcomes of interest? Were these prevalence or incidence figures?

Incidence of coronary heart disease, stroke and heart failure

2. What was the exposure, and how was it measured?

At least two LDL cholesterol levels taken between the ages of 18 and 60, one of which was taken between the ages of 40 and 60. Then new variables were created from these readings - cumulative LDL burden, time weighted average LDL, and LDL slope. These new variables were the exposure variables.

3. What is the study design? (Ok, actually, designs)

pooled data from 4 prospective cohort studies

Compared with those in the lowest quartile for cumulative exposure, those in the highest had a 57 percent increased risk.

4. How was the exposure classified?

A measure called LDL-C burden was created - the weighted average amount that LDL-C was greater than 1.8mmol/mmol. The time weighted LDL-C average was calculated. The slope of change of LDL-C over time was calculated. All 3 exposures were tested separately with respect to all 3 outcomes.

5. How did the researchers set up a calculation to get to "57 percent increased risk"?

This is a relative risk. The risk in participants with the highest quartile of LDL-C burden was 1.57 greater than those with the lowest quartile of LDL-C burden.

...their study had too few cases of stroke to achieve statistical significance.

6. What can you assume about the 95% CI for the association between lifelong LDL and stroke?

It is likely that the 95% CI for the relative risk crossed 1

7. Does this mean there is no association between lifelong LDL and stroke?

No. It may be that the study was insufficiently powered to find an association.

8. What epidemiologic research design is (usually) more suitable for investigating risk factors for rare outcomes such as stroke?

Case control study

The study controlled for race and ethnicity, sex, year of birth, body mass index, smoking, high-density lipoprotein (HDL, or “good” cholesterol), blood pressure, Type 2 diabetes and the use of lipid-lowering and blood pressure medicines.

9. Why did the researchers control for smoking?

Smoking is a potential confounder of the association between LDL-C cholesterol ad vascular events. If it is associated both with LDL-C levels and with vascular outcomes, particularly as we know it is not on the causal pathway - it is not caused by smoking. Not taking into account smoking might mean the finding an association when none exists or not finding an association when one exists.

10. Why did the researchers control for lipid-lowering medicine? (I actually don't know - this seems strange to me!)

I think that some people believe that statins have an extra mechanism by which they reduce cardiovascular risk, in addition to their known mode of action of lowering cholesterol. It may be that the researchers did not want this extra pathway to confound the association between lipid levels and CHD, and so controlled for lipid lowering medication.

11. Can we conclude that the researchers found some causal relationships here? Or do we need an RCT?

I could not find the full text of the paper, but cohort studies can never prove a casual relationship, because they are subject to residual confounding. We do need an RCT. I agree with you Fathima - it does seem very likely.

Judith