A recent publication, Mortality and Other Important Diabetes-Related Outcomes With Insulin vs Other Antihyperglycemic Therapies in Type 2 Diabetes, is an excellent example of the truism that "correlation does not imply causation."
The authors of this study conclude that “In people with T2DM, exogenous insulin therapy was associated with an increased risk of diabetes-related complications, cancer, and all-cause mortality.” True, insulin therapy was associated with increased risk of bad things. But as I indicate later in this discussion, insulin therapy was also associated with higher A1C levels.
The lead author is quoted as saying that "By reviewing data… between 1999 and 2011 we've confirmed there are increased health risks for patients with type 2 diabetes who take insulin to manage their condition." Sure, but more importantly, it’s well known that there are increased health risks for people with higher A1C levels, which is a point that the study barely mentions. The patients on insulin therapy had baseline A1C levels averaging 9.5 (for insulin monotherapy) and 9.6 (for insulin plus metformin combination therapy), whereas for the patients not on insulin therapy, the baseline A1C levels were 8.5 (metformin monotherapy), 8.7 (sulfonylurea monotherapy) and 9.0 (metformin plus sulfonylurea combination therapy).
The study didn’t prospectively randomize patients with T2DM to the various therapies and observe them for major adverse cardiovascular events and death over a period of years, which would be the best way to study their hypothesis. Instead, the authors did a retrospective study using data from a UK database, the Clinical Practice Research Datalink. Their objective “was to compare the clinical outcomes of people with T2DM treated with insulin with those in patients treated with other glucose-lowering regimens.” You’ll notice that their assumption from the get-go was that insulin therapy is bad – both in the objective I just quoted, and the title of the article.
My biggest concern is that the authors didn’t make any effort to see if the patients who had lousy control at the beginning had excellent or lousy control at the end of the period of observation. If, for instance, the patients who at study entry were on insulin with lousy A1C levels of 9.5 and later had excellent A1C control with levels of (let’s say) 6.5, and still had complications at a higher rate than other patients who were on pills with lousy A1C levels of 9.5 and later had excellent A1C control with levels of (let’s say) 6.5, then we’d have learned something. But I can’t see such data in the study’s publication.
When one recalls that high A1C levels correlate with high risk of complications, which was proven decades ago in the DCCT (which studied the effects of glucose control in T1DM) and the UKPDS (which studied glucose control in T2DM) studies, the design of the study falls apart. Clearly, if you choose to look at diabetic patients who had higher A1C levels at the beginning of the study, you’d automatically be identifying patients who would be at higher risk of complications and death. And if, at the end of the study, the patients who had the highest A1C levels at entry into the study turn out to have the highest rate of complications, and there’s no data about whether their diabetes control improved or deteriorated during the study, whether they were on insulin or other therapies seems to me to be irrelevant.