Higher insulin doses do not raise the risks for mortality or major adverse cardiovascular events (MACE) in patients with type 2 diabetes, new research indicates, after careful adjustment for potential confounders.
The results, from a cohort study of primary-care records from the UK-based Clinical Practice Research Datalink, were published online November 16 in Lancet Diabetes & Endocrinology by John-Michael Gamble, PhD, of the School of Pharmacy at Memorial University of Newfoundland, St John's, Newfoundland and Labrador, and colleagues.
The study offers at least a partial explanation for opposing findings between previous observational studies — suggesting that insulin may dose-dependently raise cardiovascular risk in type 2 diabetes — and randomized clinical trials such as the UK Prospective Diabetes Study (UKPDS) and Outcome Reduction With Initial Glargine Intervention (ORIGIN) suggesting that insulin is not cardiotoxic.
"Our findings provide additional evidence regarding the overall and cardiovascular safety of insulin treatment in type 2 diabetes. Our study does not provide complete reassurance, as no single study can, but provides convincing evidence that confounding may be partially responsible for the observed dose-dependent association between insulin and mortality," Dr Gamble told Medscape Medical News.
He added, "The findings should not change our approach to prescribing insulin but provide further evidence supporting the current clinical practice guidelines for using insulin in the management of type 2 diabetes."
Nevertheless, he cautioned, "As always, it is still important to not overtreat with insulin, given the known risks such as hypoglycemia."
T2D Patients Taking Insulin Die Sooner, but Is Insulin at Fault?
In the current study, both raw and adjusted data did suggest an increase in death and MACE with increasing insulin doses in type 2 diabetes patients. However, that association disappeared after Dr Gamble and colleagues applied a statistical approach called "marginal structural models," which adjusts for factors that change over time and that are affected by insulin dosage — such as number of hypoglycemic events, body weight, and glycemic control.
Those same factors also in turn influence insulin dosage and are related to the outcomes of interest — in this case, death — and therefore may be responsible for a spurious association between insulin dosage and death, Dr Gamble explained.
In an accompanying editorial, Amanda Adler, MD, PhD, of the Wolfson Diabetes and Endocrine Clinic at Addenbrooke's Hospital, Cambridge, United Kingdom, points out: "People with type 2 diabetes who take insulin die sooner than people with type 2 diabetes who do not take insulin. This is accepted. What is not accepted is whether insulin is at fault."
Dr Adler comments that not only is the new study unique in the application of marginal structural models but that the use of patients for whom insulin is added to metformin makes it "more generalizable" than prior studies of patients on insulin monotherapy, although it may not apply to patients on other treatment regimens.
But, she cautions, "marginal structural models depend on assuming no unmeasured confounding, and, alas, this is not empirically verifiable."
At the same time, she adds, randomized clinical trials to answer this question are probably not feasible, given the ethical issues and the difficulty in blinding when administering increasing insulin doses.
Successive Adjustments Reduced Insulin Dose Risk
Study subjects were 6072 new users of insulin added to existing metformin therapy during 2001–2012, divided into four groups based on the mean daily insulin dose during follow-up.
Relative differences in mortality and MACE (nonfatal myocardial infarction, nonfatal stroke, cardiovascular-related mortality) were assessed. A subgroup of 3599 patients had linked hospital and death-certificate data. Median follow-up time was 3.1 years.
Crude mortality rates by daily insulin dose per 1000 person-days were 46 for less than 25 units per day, 39 for 25 to 49 units, 27 for 50 to 74 units, 34 for 75 to 99 units, and 32 for 100 or more units. The overall crude mortality rate during follow-up was 31 deaths per 1000 person-years.
After adjustment for baseline covariates, including age at insulin initiation, sex, smoking, HbA1c, body-mass index, blood pressure, number of physician visits in the year before insulin initiation, medication use, and comorbidities, higher insulin doses were still significantly associated with increased mortality (P for trend = .006). Sensitivity analyses produced consistent results.
However, the magnitude of the association with increased mortality diminished following further time-dependent adjustments for changes in glycemic control, body weight, frequency of hypoglycemic events, and occurrence of cardiovascular events, to the point that only the two highest insulin-dose categories remained significantly associated with increased mortality.
Compared with less than 25 units a day, hazard ratios were 1.25 and 1.23, respectively, for 25 to 49 and 50 to 74 units per day, not achieving significance.
For the upper two doses, the relationship was still significant, with hazard ratios 1.71 and 1.88 for 75 to 99 and >100 units per day, respectively.
But after the application of marginal structural models, the association between insulin dose and all-cause mortality disappeared completely, Dr Gamble and colleagues report.
Can This Be Proved Definitively: Why Doesn't Insulin Prolong Life?
In her editorial, Dr Adler points out that because the hazard ratios all exceeded 1.0, "it seems possible that a bigger study would find significant associations suggesting harm but also that residual confounding would remain.
"Only when confounding does not interfere and the study is big enough can one . . . conclude that insulin is not harmful to people with type 2 diabetes."
Dr Gamble told Medscape Medical News that he thought the editorial was "thoughtfully written," and he agrees that "the question at hand is difficult to answer using a [randomized controlled trial] design, given blinding would not be feasible between groups with varying intensities of insulin treatment. This reinforces the importance of assessing the safety of insulin using observational methods."
He also acknowledged that the study does not distinguish among the various types of insulin, noting, "Future research is needed to assess potential outcome differences between basal and prandial insulins, as well as the newer insulin products on or slated to arrive on the market."
Dr Adler concludes her editorial with a different thought — should there ever be data to definitively prove the safety of insulin in type 2 diabetes patients, "one could also wonder why more insulin does not make people with type 2 diabetes live any longer."
Dr Gamble has no relevant financial relationships. Disclosures for the coauthors are listed in the article. Dr Adler has no relevant financial relationships.
Lancet Diabetes Endocrinol. Published online November 16, 2016.
New Insight Into Insulin Dose and Mortality in Type 2 Diabetes. Medscape. Dec. 2, 2016.