A cost comparison of long-acting insulin analogs vs NPH insulin-based treatment in patients with type 2 diabetes using routinely collected primary care data from the UK

Figures & data

Table 1. Baseline characteristics.^a

	Glargine	Detemir	NPH	Total	p value*
People, n (%)	794 (53.79)	252 (17.07)	430 (29.13)	1476 (100)
Age (years)	57.38 (11.57)	56.36 (12.75)	56.01 (12.18)	56.81 (11.97)	0.129
Sex (% male)	56.42	53.57	50.93	54.34	0.177
Townsend score (quintiles, 0–5)^b	2.88 (1.42)	2.77 (1.43)	2.83 (1.33)	2.85 (1.40)	0.020
Clinical
HbA1c (%)^c	9.85 (1.64)	10.23 (1.80)	9.95 (1.70)	9.94 (1.70)	0.022
BMI (kg/m^2)^d	29.06 (5.23)	29.16 (5.07)	28.82 (4.59)	29.02 (5.05)	0.815
Weight (kg)^d	82.72 (16.62)	84.42 (18.51)	82.24 (16.13)	82.91 (16.85)	0.497
Hypoglycemic events—all^e
Number of episodes	6	1	2	9	0.736
Mean no. episodes per patient	0.01 (0.09)	0.00 (0.06)	0.005 (0.07)	0.01 (0.08)
Hypoglycemic events—severe^e
Number of episodes	0	0	0	0	–
Mean no. episodes per patient	–	–	–	–	–
Comorbidities
Number	58	14	37	109	0.849
Mean no. per patient	0.07	0.06	0.09	0.07

^aBaseline characteristics for the quarter (3 month period) prior to insulin initiation. Data are mean ± SD unless specified otherwise.

^bData missing for 11 people. Townsend score 1 = lowest and 5 = highest level of deprivation for geographical areas.

^cData missing for 466 people.

^dData missing for 729 people.

^eData missing for 1467 people. Events recorded over the 12 months prior to insulin initiation.

*One-way analysis of variance, F-statistic (Prob. > F) for between groups variation and Chi² statistic.

Table 2. Baseline characteristics after propensity score measure (mean).

	Treated	Control	p value*
Age (years)	58.39	57.26	0.001
Sex (% male)	56	51	0.001
Townsend score^a	2.86	2.83	0.531
HbA1c (%)	8.90	9.00	0.001
BMI^a (kg/m^2)	29.65	29.73	0.566
Comorbidities	0.04	0.06	0.001
FPG^a (mmol/l)	9.59	9.72	0.258

*<0.05 suggests groups are well balanced.

^aThe absolute imbalance is small.

BMI, body mass index; FPG, Fasting Plasma Glucose.

Table 3. Multivariate analyses of total cost, £ (mean): propensity score matched.

	Analogs (a) (n = 430)	NPH (b) (n = 430)	Difference (a − b), mean (SE)*
yr0 (year prior to insulin initiation)	594	489	105 (33)
yr1 post-initiation	1129	909	220 (46)
yr2 post-initiation	1052	885	168 (55)
yr3 post-initiation	1045	898	146 (43)
Change from baseline (mean)
yr1 vs yr0	534	420
yr2 vs yr0	458	396
yr3 vs yr0	451	409
Change from baseline (mean) (PS included in multivariate analysis)
yr1 vs yr0	437	313
yr2 vs yr0	364	294
yr3 vs yr0	39	315

SE, standard error; yr, year; yr0, baseline.

*Differences at each year are statistically significant, p < 0.05.

Table 4. Analyses of total cost, £ (mean, SE).

Univariate	Glargine (n = 794)	Detemir (n = 252)	NPH (n = 430)	Prob. > F-statistic (pairwise significance)^a
yr0 (year prior to insulin initiation)	614 (11)	624 (20)	489 (9)	<0.001 (glargine vs NPH and detemir vs NPH SS)
yr1 post-initiation	1101 (10)	1259 (27)	909 (18)	<0.001 (all comparisons SS)
yr2 post-initiation	1046 (11)	1140 (28)	885 (24)	<0.001 (glargine vs NPH and detemir vs NPH SS)
yr3 post-initiation	1040 (10)	1081 (22)	898 (17)	<0.001 (glargine vs NPH and detemir vs NPH SS)
Univariate change from baseline (mean)^b
yr1 vs yr0	486	635	420	<0.05 (glargine vs detemir and detemir vs NPH SS)
yr2 vs yr0	432	516	396	0.175 (no comparisons SS)
yr3 vs yr0	425	457	409	0.714 (no comparisons SS)
Statistical significance of changes vs baseline^c	p < 0.001	p < 0.001	p < 0.001
Multivariate change from baseline (mean)
yr1 vs yr0	414	462	312
yr2 vs yr0	347	343	290
yr3 vs yr0	332	316	309

Yr, year; y0, baseline.

^aStatistical significance (SS) is based on the 5% (0.05) level of testing. Probability >F-statistic is for the null hypothesis that there is equality in the means of the three groups. Statistics based on one-way analysis of variance models of combined imputation data-sets; pairwise comparisons are based on Bonferroni, Scheffe, and Sidak tests. SS of between-group comparisons confirmed using regression analysis on imputation data-sets with treatment group as a factor variable.

^bChange calculated from point estimates of mean cost in each period. SS of between-group differences in changes from baseline are based on one-way analysis of variance models with change in total cost as the dependent variable.

^cBased on regression analysis with period as a factor variable.

Figure 1. Selection of patient cohort.

Figure 2. Unadjusted change in HbA1c level (%) from baseline (before) and after initiation of NPH insulin, glargine or detemir.

Table 5. Health outcomes.^a,b,d,e

	Glargine (n = 794)	Detemir (n = 252)	NPH (n = 430)	Prob. > F-statistic (pairwise significance)^c
HbA1c (%)
Baseline^b	9.90 (9.78, 10.03)	10.25 (9.98, 10.53)	10.04 (9.82, 10.25)	0.010 (glargine vs detemir SS)
after 1 yr	8.62 (8.54, 8.70)	8.79 (8.61, 8.96)	8.76 (8.63, 8.89)	0.013 (glargine vs nph SS)
after 2 yr	8.70 (8.61, 8.82)	8.84 (8.63, 9.05)	8.84 (8.72, 8.97)	0.027 (glargine vs nph SS**)
after 3 yr	8.72 (8.61, 8.82)	8.92 (8.63, 9.21)	8.82 (8.67, 8.98)	0.009 (glargine vs detemir SS)
baseline vs yr 1	−1.29	−1.47	−1.27	0.300 (none SS)
baseline vs yr 2	−1.20	−1.42	−1.20	0.211 (none SS)
baseline vs yr 3	−1.19	−1.33	−1.22	0.488 (none SS)
Statistical significance of change vs baseline	p < 0.05 all years	p < 0.05 all years	p < 0.05 all years
BMI (kg/m^2)
Baseline^b	28.98 (28.63, 29.34)	28.61 (27.98, 29.25)	28.79 (28.28, 29.30)	0.526 (none SS)
after 1 yr	29.86 (29.50, 30.22)	29.64 (29.03, 30.24)	29.90 (29.41, 30.39)	0.754 (none SS)
after 2 yr	30.10 (29.75, 30.44)	29.97 (29.23, 30.70)	30.08 (29.60, 30.57)	0.925 (none SS)
after 3 yr	30.14 (29.78, 30.49)	29.80 (29.08, 30.52)	30.36 (29.82, 30.90)	0.322 (none SS)
baseline vs yr 1	0.87	1.03	1.11	0.041 (glargine vs NPH SS)
baseline vs yr 2	1.11	1.35	1.29	0.179 (none SS)
baseline vs yr 3	1.15	1.19	1.57	0.019 (glargine vs NPH SS)
Statistical significance of change vs baseline	p < 0.05 all years	p < 0.05 all years	p < 0.05 all years

yr, year; y0, baseline. Baseline refers to the quarter prior to insulin initiation.

^aData reported as Mean (95% CI).

^bBaseline values reported here are in reference to the analysis of the multiple imputation data-sets; baseline values reported in Table 1 reflect estimates of the un-adjusted (non-imputed) data.

^cStatistical significance (SS) is based on the 5% (0.05) level of testing. Probability >F-statistic is for the null hypothesis that there is equality in the means of the three groups. Statistics based on one-way analysis of variance models of combined imputation data-sets; pairwise comparisons are based on Bonferroni, Scheffe, and Sidak tests. SS of between-group comparisons confirmed using regression analysis on imputation data-sets with treatment group as a factor variable.

^dChange calculated from point estimates of mean cost in each period. SS of between-group differences in changes from baseline are based on one-way analysis of variance models, with change in total cost as the dependent variable.

^eBased on regression analysis with period as a factor variable.

**SS at the 10% level.

Figure 3. Unadjusted change in body mass index (kg/m²) from baseline (before) and after initiation of NPH insulin, glargine or detemir.

Table 6. Univariate analyses of total cost by category^a, £ (mean per patient).

	OADs	Basal insulin	AHAs	SMBG	Bolus insulin	Lab tests	Hospital attendance	Sharps	GP surgery visits	GP home visits	SNV	PNV	LLDs	AHDs	Basal + Bolus
Glargine
Yr 0	182	0	0	100	0	29	51	7	109	1	0	0	83	12	0
Yr 1	71	376	4	275	42	28	23	39	119	4	0	0	93	13	418
Yr 2	60	370	1	203	50	30	38	27	82	0	0	0	97	13	420
Yr 3	55	375	1	193	64	33	33	28	62	1	0	0	101	15	438
Detemir
Yr 0	168	0	0	89	0	44	66	29	108	0	0	0	72	14	0
Yr 1	82	379	3	359	77	35	12	89	111	0	0	0	85	15	455
Yr 2	56	328	1	250	82	27	26	75	69	5	0	0	74	13	410
Yr 3	46	330	1	240	96	32	41	28	55	0	0	1	72	12	426
NPH
Yr 0	108	0	1	95	0	20	32	6	113	1	0	1	42	8	0
Yr 1	32	171	1	271	187	22	11	40	98	0	0	0	57	11	358
Yr 2	26	161	1	223	191	19	53	26	71	0	0	0	63	11	352
Yr 3	25	199	1	233	199	21	11	25	58	0	0	0	67	14	398

AHAs, anti-hypoglycemic agents; AHDs, anti-hypertensive drugs; LLDs, lipid lowering drugs; OADs, oral anti-diabetic agents; PNV, practice nurse visit; SMBG, self-monitoring of blood glucose; SNV, specialist nurse visit; yr, year; yr0, year prior to insulin initiation.

^aThese values are estimated from the non-imputed data-set.