Maybe Old is Gold? Newer Insulins Might Not Be Better – Just More Expensive

Download the podcast patient case:  NPH vs Insulin Analogs 

Fredrick Banting, the Canadian scientist who discovered insulin in 1921, sold the patent for just $1 to the University of Toronto and made it available to pharmaceutical companies royalty-free.  Today he would be disappointed to know that the high cost of insulin is now a major barrier to treatment. The American Diabetes Association recently reported that ~$35 billion was spent on high-price tag medications like the insulin analogs, including long-acting (glargine U-100 or detemir) and ultra-long acting insulins (glargine U-300 or degludec).¹ The average price of insulin has nearly tripled, from $4.34/ml in 2002 to $12.92/ml in 2013.² One 10 ml vial of NPH costs approximately $25 compared to $180 for insulin glargine and $250 for insulin detemir.³ Insulin’s high cost affects everyone: (1) uninsured patients, (2) insured patients with high co-payments and deductibles, (3) Medicare beneficiaries with coverage gaps and fixed income, and (4) everyone else paying higher premiums to offset the insurers’ expenditures. Are the newer insulins really worth the extra cost?

To understand the differences in health outcomes associated with use of lower and higher cost insulin products, investigators at Kaiser Permanente Northern California compared the rates of hypoglycemia-related emergency department (ED) visits or hospital admissions and changes in glycemic control after initiation of long-acting insulin analogs (glargine or detemir) compared with human Neutral Protamine Hagedorn (NPH) insulin among patients with type 2 diabetes mellitus (T2DM)⁴. The study was a retrospective observational study using electronic medical records over a nearly 10-year space of time from January 1, 2006 and September 30, 2015.  The median follow-up after treatment initiation was 1.71 years. Patients included in the analysis were initiated on either NPH insulin, insulin glargine, or insulin detemir (See Table 1). The predicted probability of initiating a long-acting insulin analog was calculated for each patient. Quintiles of the propensity score were created based on the distribution of the scores among those patients who initiated insulin analogs. Using frequency matching, 500 patients who initiated NPH insulin were selected from each of the quintiles, creating a population in which the distribution of covariates in the NPH insulin cohort was similar to those in the insulin analog cohort.  This was done to minimize the impact of observed confounders. The study excluded patients with type 1 diabetes mellitus (T1DM), who had any insulin prescription fill within the prior 12 months, and those using prandial or premixed insulin at baseline. Patients were censored at death, loss of health plan coverage, initiation of prandial insulin, or at the end of the study.  

Table 1. Baseline Characteristics  

^{a Patients who initiated NPH insulin were frequency matched with patients initiating insulin analogs based on propensity score quintile.}  

^{b After frequency matching; an absolute value < 0.1 indicates a negligible difference in the mean or prevalence of a covariate between groups.}  

^{c Based on the modified version of the Deyo Charlson Score. Possible scores ranged from 0-17 and represent the number of selected comorbid conditions (other than diabetes) during the 2 years prior to baseline.} 

^{d including α-glucosidase inhibitors, amylin and meglitinides} 

The primary outcome was the time to hypoglycemia-related ED visit or hospital admission after initiation of insulin therapy (See Table 2 and 3). The difference in occurrence of the primary outcome between the two groups was not statistically significant.  The between-group difference was 3.1 events (95% CI, −1.5 to 7.7) per 1000 person-years (P = .07). The secondary outcome was the change in hemoglobin A1c and a >0.5% change between the groups is typically considered clinically important. Participants with missing data for A1c at baseline (n = 402) and those censored within 90 days of baseline (n = 3665) were excluded (n = 4067) from this part of the analysis.  The difference-in-difference for glycemic control was statistically significant favoring NPH insulin (−0.22% greater decrease with NPH, 95% CI, −0.09% to −0.37%) but this difference is arguably not clinically important.  

Table 2. Primary and Secondary Outcomes^*

^{*results reported from full cohort prior to frequency-matching}

Table 3. Sensitivity Analysis

^{*The hazard ratios in the full cohort used traditional regression adjustment.  The hazard ratios in the frequency-matched sample used 1000 bootstrap regressions. Risk conferred by initiating long-acting insulin analog vs neutral protamine Hagedorn insulin. Hazard ratio >1 favors NP insulin.}

^{**Adjusted for prior severe hypoglycemia-related ED visits or hospital admissions, time-dependent indicators for oral diabetes therapy use, and additional unbalanced baseline covariates.  The covariates were baseline diabetes treatment regimen, statin use, visual impairment, hospital use, outpatient medical visits, duration of diabetes, body mass index, year of index prescription, patient insulin co-pay, Kaiser Permanente of Northern California (KPNC) service area, prescribing clinician specialty, and medication nonadherence.}

 ^{***Adjusted for prior severe hypoglycemia-related ED visits or hospital admissions, time-dependent indicators for oral diabetes therapy use, and additional unbalanced baseline covariates.  The covariates were outpatient medical visits, KPNC service area, and year of index prescription.}

Strengths of this observational study include real world data from a large cohort and propensity score-matching to control for potential confounding factors. Important limitations to this study include the potential for residual confounding due to imbalances in unobserved covariates; however given the null results it is unlikely they would have resulted in a meaningful difference. It is also important to note that patients included in this study had poorly controlled diabetes with an average A1c of 9.4% at baseline and may have relatively lower risk of hypoglycemia compared to a patient population with better-controlled T2DM. Findings of this study can’t be applied to T1DM or those requiring basal-bolus insulin regimens. Future studies would need to evaluate benefits that insulin analogs provide over NPH in these patient populations.   

The study did not assess the rate of nocturnal hypoglycemia which could have occurred without triggering an ED visit or hospitalization. Prospective clinical trials have shown that long-acting insulin analogs modestly reduce the rate of nocturnal hypoglycemia, but not severe hypoglycemic episodes, compared to NPH insulin.^5,6 A recent systematic review showed that insulin glargine use was associated with higher probability of reaching target A1c level without hypoglycemia compared to NPH insulin.⁷ Also, clinical trials comparing long-acting vs ultra-long acting insulin analogs show that the latter is associated with fewer nocturnal as well as severe hypoglycemic episodes in T1DM and T2DM patients.^8,9,10 Currently there are no comparative safety and efficacy studies evaluating differences between ultra-long acting insulin analogs and NPH insulin.  Moreover, “glycemic variability”, defined as fluctuations in blood glucose levels between hyperglycemia and hypoglycemia, may potentially worsen cardiovascular complications of diabetes.¹¹ This may serve as another variable consider when evaluating the safety and efficacy of insulin products. 

The use of NPH insulin in the study population was high (92%) and the results may not applicable to other healthcare systems where clinicians may not be familiar with NPH insulin. A high degree of familiarity with NPH — how to dose and titrate it — may have contributed to lower rates of hypoglycemia and improved glycemic control with NPH in this study.  However, the outcomes observed in this >20,000 patient cohort highlight how familiarity with NPH prescribing, monitoring, and patient education can result in lower costs without sacrificing safety or efficacy. We all should be trained on how to use NPH and this is a potential role of clinical pharmacists!  Eight percent of the study population were prescribed an insulin analog in this study.  It would be helpful to better understand what patient-specific features lead clinicians to select an insulin analog over NPH in these cases.

The authors recommend that NPH insulin be used in the majority and insulin analogs be reserved for selected patients. This recommendation is contradictory to current consensus guidelines which favor long-acting and ultra-long acting insulin analogs.  To be fair, this study lacks data comparing NPH to ultra-long acting insulins and did not differentiate between insulin glargine and insulin detemir, combining the data for both insulin analogs. Clinicians should be mindful of this prior to changing their practice.

If additional studies fail to demonstrate a clear clinical benefit from insulin analogs, clinicians will need to be able to identify patients who can safely use NPH and those who really should be prescribed an insulin analog. Key patient characteristics likely include cognitive ability and physical capacity, consistency with food intake and timing of meals, and motivation to monitor blood glucose readings and adjust doses.  Do the insulin analog provide sufficient value to justify their high cost? Are you willing to use NPH first?