Saturday, October 15, 2011

Is Elevated Cholesterol Beneficial?

The study:

Is The Use Of Cholesterol In Mortality Risk Algorithms In Clinical Guidelines Valid? Ten Years Prospective Data From The Norwegian HUNT 2 Study Journal of Evaluation in Clinical Practice, September 2011

It found higher mortality in men with low cholesterol (less than ~193 mg/dl) and high cholesterol (greater than ~270 mg/dl) compared to men with levels in between (193 to 270). If you visualize that, it looks like a U-shaped curve.

Here's one graph from the study. The U-shape in the men's curve is evident. (You can convert mmol/L to mg/dl by dividing by 0.0259, so 5 mmol/L divided by 0.0259 = 193 mg/dl.)

You can see from this graph that women with high cholesterol fared even better than men. At about the 250 mg/dl mark, men's death risk was rising, but women's risk was still falling. (I thought this data had wide confidence intervals, those vertical bars. I think of that as a wide margin of error. The data was rather scattered and a good portion of people ended up with an increased risk, not a decreased risk, as their cholesterol rose.)

The authors say that since, in their study, risk of death did not go up as cholesterol went up, using this number is not a good gauge of early death, it overestimates risk. I imagine pharmaceutical companies would be loathe to hear this since it might reduce the number of candidates for cholesterol-lowering statin drugs.

I compared this cholesterol study with the vitamin study in my previous post:

There are some features which I thought reduced the internal validity of this cholesterol study, at least in comparison to the vitamin study:
  • It based findings on one cholesterol reading taken when the participant was healthy and projected it across 10 years. A lot can happen to that number in 10 years. A lot can happen to it in a few months. What was their cholesterol in the weeks and months leading up to their death?

  • It didn't collect data about cholesterol-lowering drugs. Some statins act as anti-inflammatories, which could improve the takers' health profile apart from effect on cholesterol. Given the current guidelines, many statin-takers would fall in the lower-risk (beneficial) valley of this U-shaped curve.

  • It didn't adjust for several potential confounders, as you can see from the table. An important one was diet. How do you know that those with the supposedly beneficial cholesterol (193-270 mg/dl) weren't eating more fruits and vegetables? Weren't eating less fast food? Weren't drinking less alcohol? No data on food intake was collected.
Poor generalizability is an aspect that reduces the external validity of both studies (I mentioned this in the vitamin study). The population in this cholesterol study was fairly cloistered so it's difficult to generalize the findings to other groups:
"The HUNT 2 population is ethnically homogeneous (dominated by individuals of Nordic origin) and has been considered fairly representative of the total Norwegian population with respect to demography, socio-economic factors, morbidity and mortality."
So, these results do not apply to anyone not of Nordic origin who lived in Norway in the early 2000s. You cannot say that a total cholesterol of 170 mg/dl in an Amazonian native is indicative of early death. You cannot say that a cholesterol of 250 mg/dl is beneficial for a middle-aged Hispanic woman living in San Diego.

Along these lines, those of generalizability, you must also consider that Norwegians are relatively wealthy, healthy, and have good health care:
"Norway is an affluent country, and Norwegians are currently one of the longest lived people in the world. The rate of smoking among men is relatively low, by international comparison. The stable social structure could also play a part, including a well-functioning health care system with good access and coverage for all."
The US has relatively poor life-expectancy among developed countries, ranking 38th in the world. And its poverty rate is over 15%, with some regions even higher. The poverty rate here in Philadelphia is over 25%. Millions of Americans are uninsured or underinsured. So, in Norway, good healthcare or good genes could be buffering negative effects of elevated cholesterol. You don't know.

I don't think either of these studies are junk science, as I've seen the vitamin study described. Just my opinion. I think both of them are valid epidemiological studies, and raise the specter that both taking vitamins, and gauging death risk using total cholesterol*, are not good ideas. But I suppose if your criteria for "junk" is weak associations, limited statistical adjustment, and minimal input, then this cholesterol study, compared to the vitamin study, is junkier than junk.

* Certainly, there are other subparticles that may prove more indicative, perhaps HDL or Lp(a). This study assessed total serum cholesterol only.


Dr. Mel said...

Well, that's different! Look forward to reading more carefully after I finish grading student papers (sigh).

Bix said...

From what I've read, total cholesterol alone doesn't seem to tell a good enough story.

Dr. Mel said...


Laurie Endicott Thomas said...

Cholesterol can be low because of liver disease. People who die of cirrhosis of the liver often have arteries that are clean as a whistle. Ask any pathologist.

Be very careful about interpreting the results of any population study unless you know what's going on in that population.