Last week I got myself into a little trouble. I came across this article in NEJM journal watch and it immediately struck me. I’ve always objected to the quality measure of treatment with high dose statins in diabetic patients. It just doesn’t make sense to me. Particularly if they have normal cholesterol levels and no documented coronary artery disease. The alerts from the electronic medical record that get sent my way for these patient’s have always annoyed me.
When I came across this article I thought, “now’s my chance.” I got excited and started sending emails to administrators stating we should re-evaluate the use of this quality measure in our patient population.
Whenever I do something like this it ultimately leads to more work, which in this case it did. The committee asked me to present the article and my recommendations at their next hearing. I figured my blog was perfect practice to work out my thoughts prior to the presentation.
This was a retrospective matched cohort study. The data used comes from the US Veteran’s Affairs agency. The database is the most extensive in the US. It is the beginning for many publications. A matched retrospective cohort means they created two separate groups within the database, matched them for specific characteristics, then observed the frequency of the endpoint in the two groups. This is a common retrospective study design.
The observation period was 2003-2015. Participants had to be over 30 years old and had to meet criteria for a regular VA health system user. They had to be diabetic and started on a statin during the study period. Participants that were prescribed a statin 12 months prior to the study period were excluded. The time was divided between a baseline period and a follow-up period based on the index date of statin prescription.
Diabetes progression was defined as therapy intensification, or new persistent hyperglycemia. There were several secondary outcomes. Two important secondary outcomes were, the number of drug classes used to control diabetes during the study and new insulin starts.
The control group was compared to high intensity, moderate intensive, and low intensity statin to reduce confounding and establish treatment effect.
When reviewing a retrospective cohort design there are a few questions that need to be answered to establish the validity of the methods. First, did participants in each group have the same risk of experiencing the outcome? If not, was statistical adjustment used to match each group.
A matched cohort is an attempt to recreate randomization retrospectively. You take data from a large database and randomly select a group of individuals to be the control group and compare the occurrence of common risk factors to the experimental group. In this matched cohort the only difference was slightly higher population of men in the control group and higher cholesterol in the statin group. Otherwise they were comparable.
Next, was the outcome measured equally in both groups? Since all of the participants were registered in the VA database, and the outcome was fairly objective, it is safe to say the outcome was measured equally.
Finally, was the follow-up period substantial enough to detect the outcome. In this study the mean duration of statin use was 5.5 years. Which is likely adequate to detect a difference across the cohort.
Statin users were more likely to experience diabetes progression in all outcome categories. The Odd’s ratio for diabetes progression was 1.37 (1.35 to 1.40 CI). New insulin starts were 1.16 (1.12 – 1.19 CI).
The raw numbers were: Statin users – (46,434)/(83,202) = 0.558
Non-Statin -- (39,868)/(83,202) = 0.479
In this case the number needed to harm (NNH):
NNH = 1 / (0.558-0.479) = 12.7
For every 13 patients treated with high dose statin, one experiences progression of their diabetes.
The Odd’s ratio worsened from low intensity statin to high intensity statin.
This is a retrospective observational study, so concluding with certainty that stains are harmful in diabetics is not possible. Also, the study investigated diabetes progression and did not consider the benefit statins may have in reducing cardiovascular events. It is possible that an unmeasured variable is responsible for the effect not seen in the data. However, the suspicion that this treatment could be harmful should eliminate it from mandated use in diabetics.
Some patients with diabetes should be on statins. But not all. Statins have revolutionized the treatment of vascular disease. Particularly, patients with high cholesterol, smokers, and those with coronary artery disease have seen their life expectancies extended by them. We know with certainty through randomized trials that cardiovascular events are prevented by statins. Not for all, but for some patients that cross a certain risk threshold.
Like most VA studies this was well executed. The methods were solid. The duration of follow-up was adequate for each participant to experience the outcome. The treatment effect was significant with a potential NNH of 12.7. This result can be trusted with some degree of certainty. This data should be used by policy makers, insurance companies, and individual physicians to help guide appropriate statin use.