Insulins NPH, glargine, and detemir, and risk of severe hypoglycemia among working-age adults*

Figures & data

Table 1. Baseline characteristics of the study population stratified by type of basal insulin (n = 16,985).

Baseline	Patients who initiated detemir n = 3900 (23.0%)		Patients who initiated glargine n = 7499 (44.2%)		Patients who initiated NPH n = 5586 (32.9%)
Variables^a	n	%	n	%	n	%	p value^b
Age, years
18–29	324	8.3%	471	6.3%	140	2.5%	<0.001
30–49	1016	26.1%	1865	24.9%	1275	22.8%
50-	2560	65.6%	5163	68.9%	4171	74.7%
mean (SD)	51.1 (11.9)	52.0 (11.0)	53.9 (9.3)
Gender
Female	1373	35.2%	2369	31.6%	1863	33.4%	<0.001
Type of diabetes
Type 1	755	19.4%	1333	17.9%	516	9.2%	<0.001
Type 2	2940	75.4%	5672	75.6%	4642	83.1%
Undefined	205	5.3%	494	6.6%	428	7.7%
Time since diabetes diagnosis, years
≤1	1548	39.7%	2959	39.5%	2221	39.8%	0.25
>1–≤5	1034	26.5%	1862	24.8%	1436	25.7%
>5–≤10	895	23.0%	1766	23.6%	1271	22.8%
>10	423	10.9%	912	12.2%	658	11.8%
Mean (SD)	4.3 (5.7)	4.4 (5.8)	4.2 (5.3)
Calendar year at start of follow-up
2006	234	6.0%	595	7.9%	2492	44.6%	<0.001
2007	766	9.6%	2091	27.9%	1745	31.2%
2008	1295	33.2%	2437	32.5%	846	15.2%
2009	1605	41.2%	2376	31.7%	503	9.0%
Prior use of nonbasal insulin^c
Yes	396	10.2%	671	9.0%	329	5.9%	<0.001
Prior use of sulfonylurea
Yes	1823	46.7%	3639	48.5%	2863	51.3%	<0.001
Prior hospitalizations due to severe hypoglycemia
Yes	257	6.6%	645	8.6%	468	8.4%	<0.001

^a Baseline is the date of purchase of first basal insulin.

^b p value based on Pearson’s Chi-square test.

^c “Prior use of nonbasal insulin” denotes a prescription for a nonbasal insulin. This consisted mostly of brief use of a short acting insulin during periods of ill health or pregnancy.

Table 2. Factors associated with risk of a severe hypoglycemic event.

Variables (Reference category)	Categories	HR	95% CI	p value
Basal insulin (NPH)	Detemir	0.70	0.52, 0.94	0.018
	Glargine	0.92	0.74, 1.15	0.47
	Unknown	0.96	0.69, 1.33	0.82
Basal insulin (glargine)^a	Detemir	0.76	0.58, 0.99	0.040
Age, years (18–29)	30–49	1.25	0.79, 1.97	0.35
	50-	1.51	0.95, 2.39	0.082
Gender (Male)	Female	0.87	0.72, 1.05	0.13
Type of diabetes (Type 1)	Type 2	1.39	1.02,1.89	0.035
	Undefined	1.21	0.73, 2.03	0.46
Time since diagnosis of diabetes, years (≤1)	>1–≤5	0.86	0.66, 1.13	0.28
	>5–≤10	0.83	0.61, 1.11	0.21
	>10-	0.85	0.56, 1.28	0.43
Calendar year at start of follow-up (2006)	2007	0.83	0.67, 1.04	0.10
	2008	0.71	0.54, 0.92	0.011
	2009	0.56	0.38, 0.81	0.002
Prior use of non-basal insulin at baseline^b (No)	Yes	0.98	0.71, 1.35	0.88
Prior use of sulfonylurea at baseline (No)	Yes	0.93	0.72, 1.20	0.58
Prior hospitalizations due to severe hypoglycemia (No)	Yes	5.71	4.75, 6.87	<0.001
Switch of basal insulin (No)	Yes	1.39	1.00, 1.93	0.052
Concurrent use of sulfonylureas (No)	Yes	1.00	0.78, 1.28
Concurrent use of other insulin (No)	Yes	1.30	1.05, 1.62	0.017

^a Alternative parameterization.

^b Denotes a prescription for a nonbasal insulin. This consisted mostly of brief use of a short acting insulin during periods of ill health or pregnancy.

Table 3. Sensitivity analyses for risk of severe hypoglycemia for the users of insulins detemir and glargine with NPH as reference.

	Detemir		Glargine
Model	Hazard ratio	95% CI, p value	Hazard ratio	95% CI, p value
Model 1: Base model. Follow-up periods with unknown exposure of basal insulin form an unknown exposure category.	0.70	0.52, 0.94, p = 0.018	0.92	0.74, 1.15, p = 0.47
Model 2: Periods with unknown insulin exposure are censored.	0.73	0.55, 0.99, p = 0.041	0.96	0.77, 1.20, p = 0.70
Model 3: Assume use of last purchased basal insulin during periods of unknown insulin exposure.	0.76	0.57, 1.01, p = 0.054	0.97	0.78, 1.20, p = 0.77
Model 4: Time periods with overlapping use of basal insulins are censored.	0.70	0.52, 0.95, p = 0.020	0.94	0.76, 1.18, p = 0.61
Model 5: The follow-up period is censored at the purchase of a basal insulin different to the initial basal insulin (switch).	0.69	0.50, 0.95, p = 0.024	0.94	0.74, 1.18, p = 0.57
Model 6: Analysis using the inverse of the propensity scores as weights (1% of the lowest and highest weights are truncated). Baseline summary table for the PS weighted population is presented in Table 4.	0.74	0.63, 0.86, p < 0.001	0.94	0.82, 1.07, p = 0.34
Model 7a: Stratified analysis: Patients with no prior hypoglycemic hospitalization at baseline.	0.77	0.53,1.11, p = 0.16	0.99	0.74, 1.31, p = 0.92
Model 7b: Stratified analysis: Patients with prior hypoglycemic hospitalization at baseline.	0.52	0.31,0.89, p = 0.017	0.77	0.53, 1.12, p = 0.17

Propensity score weighted analysis:

Propensity scores (PS), the probabilities of initiating each of the basal insulins at the start of follow-up conditional on measured potential confounders (i.e. age, gender, type of diabetes, calendar year on initiation, prior use of sulfonylurea, prior use of other insulin, prior severe hypoglycemic events, and the time since the diagnosis of diabetes), were calculated using multinomial logistic regression. Weighting each subject with the inverse of the PS results in a “pseudo”population where the distributions of the potential confounders (covariates included in PS model) are balanced between exposure groups (17). Utilizing the PS weights corresponds to performing an unweighted analysis on the pseudo population. The distribution of the baseline variables used as covariates in the PS model is presented for the pseudopopulation in Table 4. The risk of severe hypoglycemia for the PS-weighted population is given in sensitivity model 6 in Table 3.

Table 4. The distribution of baseline variables among the inverse propensity score (PS) weighted population stratified by type of insulin.

	Patients who initiated detemir n = 14,527 (31.4%)		Patients who initiated glargine n = 16,603 (35.9%)		Patients who initiated NPH n = 15,099 (32.7%)
Baseline variables^a	n	%	N	%	n	%
Age, years
18–29	1017	7.0%	949	5.7%	630	4.2%
30–49	3770	26.0%	4237	25.5%	3794	25.1%
50-	9740	67.0%	11417	68.8%	10675	70.7%
mean (SD)	51.54 (11.41)	52.11 (10.85)	52.91 (10.15)
Gender
Female	4939	34.0%	5402	32.5%	5040	33.4%
Type of diabetes
Type 1	2946	20.3%	3230	19.5%	1506	10.0%
Type 2	10,722	73.8%	12,235	73.7%	12,590	83.4%
Undefined	859	5.9%	1138	6.9%	1004	6.7%
Time since diabetes diagnosis, years
≤1	5926	40.8%	6671	40.2%	6348	42.0%
>1–≤5	3660	25.2%	4187	25.2%	3745	24.8%
>5–≤10	3272	22.5%	3801	22.9%	3289	21.8%
>10-	1669	11.5%	1944	11.7%	1717	11.4%
mean (SD)	4.35 (6.00)	4.37 (5.90)	4.08 (5.40)
Calendar year at start of follow-up
2006	1327	9.1%	2920	17.6%	3303	21.9%
2007	4109	28.3%	4615	27.8%	4585	30.4%
2008	4566	31.4%	4581	27.6%	4231	28.0%
2009	4525	31.2%	4486	27.0%	2980	19.7%
Prior use of other insulins than basal insulin^a
Yes	1378	9.5%	1432	8.6%	1153	7.6%
Prior use of sulfonylurea
Yes	6795	46.8%	7966	48.0%	7105	47.0%
Prior hospitalizations due to severe hypoglycemia
Yes	1098	7.6%	1329	8.0%	1248	8.3%

A reasonable balance in the distributions between groups is obtained for all baseline variables except for the variable “calendar year at initiation”.

Note all sample sizes result from weighting the original population with the inverse of the PS.

^a Denotes a prescription for a nonbasal insulin. This consisted mostly of brief use of a short acting insulin during periods of ill health or pregnancy.

Rys P, Wojciechowski P, Rogoz-Sitek A, Niesyczyński G, Lis J, Syta A, et al. Systematic review and meta-analysis of randomized clinical trials comparing efficacy and safety outcomes of insulin glargine with NPH insulin, premixed insulin preparations or with insulin detemir in type 2 diabetes mellitus. Acta Diabetologica. 2015;54:649–62.

 Web of Science ®Google Scholar