Contraceptive options for women with insulin-dependent diabetes mellitus are limited because of concerns about potential vascular and metabolic effects associated with hormonal methods.1 However, effective contraception is important for these women. Pregnancy may affect the progression of diabetes. In addition, excellent glucose control around the time of conception and during early pregnancy reduces the risk of congenital anomalies in their offspring.2 Hence, pregnancies should be planned in women with insulin-dependent diabetes mellitus.
Copper intrauterine devices (IUDs) are an appealing contraceptive for women with diabetes; they feature superior efficacy, no metabolic side effects, and excellent cost-effectiveness.3 According to the World Health Organization (WHO), women with insulin-dependent diabetes mellitus who want to use a copper IUD fall into Category 1 (no restrictions on the use of the method).1 Results of 2 recent prospective studies have shown no significant differences in copper IUD continuation rates between women with and women without diabetes.4,5 Results of another case series report indicate that the copper TCu-380A IUD is also safe and well tolerated in women with insulin-dependent diabetes mellitus.6
In contrast, women with insulin-dependent diabetes mellitus who want to use the levonorgestrel intrauterine system fall into Category 2 (benefits generally outweigh risks) of the WHO medical eligibility criteria because of the possible influence of levonorgestrel on carbohydrate and lipid metabolism. However, the WHO notes that “Whether the amount of LNG [levonorgestrel] released by the IUD causes such change [in metabolism] is unclear.”1 Some uncertainty also exists concerning the effects of low systemic levels of levonorgestrel from other hormonal contraceptives.7 Hence, we conducted this randomized controlled trial to examine the effect of the levonorgestrel intrauterine system, when compared with a nonhormonal IUD, on glucose metabolism in women with uncomplicated insulin-dependent diabetes mellitus.
We conducted this randomized controlled trial in the Outpatient Department of the Research Center of Obstetrics, Gynecology, and Perinatology of the Russian Academy of Medical Science (Moscow, Russia) in collaboration with Family Health International (Research Triangle Park, NC, USA). The protocol and informed consent documents were approved by Family Health International's Protection of Human Subjects Committee and by the Ethics Committee of the Research Center in Moscow.
Eligible participants were women aged 18 to 45 years who had well-controlled insulin-dependent diabetes mellitus. Each potential participant was seen by a gynecologist, an ophthalmologist, and a diabetologist. Women with normal glucose and glycosylated hemoglobin (HbA1c) levels and without evidence of retinopathy or nephropathy were invited to join the study. All participants had a physical examination, Pap test, and tests for chlamydia and gonorrhea. We inserted the IUDs during the first 7 days after the start of menses.
Women were assigned to receive either the levonorgestrel intrauterine system (Leiras Oy, Turku, Finland) or the copper TCu-380A IUD (Schering AG, Berlin, Germany). Participants were assigned to treatment using random permuted blocks with block sizes of 4 and 6, randomly varied. Computer-generated random numbers were used to select the blocks. Allocation concealment was achieved by having method indicator cards in sequentially numbered, sealed, opaque envelopes that were opened just before intrauterine contraceptive insertion. Participants were not told which contraceptive had been inserted, although complete treatment blinding was unlikely because of the different bleeding patterns related to each contraceptive. The enrollment period was October 1999 to June 2000.
The primary outcome measure was glycosylated hemoglobin levels.8 For participants in both groups, glycosylated hemoglobin levels, fasting serum-glucose levels, and daily insulin requirements were determined at the screening visit and then repeated at 6 weeks, 6 months, and 12 months. The secondary outcome measure was continuation rates at 12 months.
We followed the CONSORT (Consolidated Standards of Reporting Trials) guidelines9 and performed an intention-to-treat analysis. A method of repeated-measures analysis of variance was used to compare glycosylated hemoglobin levels, fasting serum glucose levels, and daily insulin requirements between groups across the study period. We used 2-sample t tests to compare means of these measures at each visit. Time to discontinuation was assessed by the product-limit method, by the log-rank test and by Wilcoxon test. Box plots were used for the better presentation of the results.
Because the study was designed for a repeated-measures analysis, the sample size calculations had to take into account any correlations between measurements within participants.10 We assumed that the correlation coefficient for analysis of glycosylated hemoglobin levels in the blood between baseline and 12-month follow-up was 0.5 in both study groups. Based on a study by Kimmerle et al,5 we also assumed that the standard error for the analysis of glycosylated hemoglobin levels was 1.2% for both baseline and 12-month follow-up for both study groups. With an alpha of 0.05, a sample size of 23 would provide 80% power to detect a 1.0% difference in the mean level of glycosylated hemoglobin between baseline and 12-month follow-up within the levonorgestrel group. With a correlation coefficient of 0.5 for both study groups, a sample size of 17 for each study group would provide 80% power to detect a 1.0% difference in mean glycosylated hemoglobin level from baseline to 12-month follow-up between the 2 groups. We estimated that 10% of participants would be lost to follow-up and an additional 15% would discontinue contraceptive use because of pain, bleeding, or other complaints. Therefore, we planned to enroll 62 women, which would expect to yield a total of 46 women who continued use for 12 months.
Sixty-seven eligible women were invited to participate. Five declined participation for personal reasons. Sixty-two women were enrolled, and 31 were randomly assigned to each group. One participant assigned to the levonorgestrel group did not have the contraceptive inserted; she was discontinued from the study. One woman in each group was lost to follow-up with no outcome data available. Only partial follow-up data were available for 3 women in the levonorgestrel group and 2 in the copper IUD group. Therefore, complete follow-up data over the 12-month trial were available for 26 women in the levonorgestrel group and 28 in the copper IUD group.
Randomization produced treatment groups similar in all important respects (Table 1). Most women were in their early 30s, near 60 kg in weight, and parous. The duration of their diabetes and their educational attainment were also comparable.
The levonorgestrel intrauterine system had no effect on glycosylated hemoglobin levels when compared with the copper IUD over the 12 months of observation (Fig. 1). In the levonorgestrel group, the mean glycosylated hemoglobin level at baseline was 5.6%, standard deviation (SD) ± 1.3. Values at 6 weeks, 6 months, and 12 months were 5.8%, SD ± 1.4, 6.1%, SD ± 1.2, and 6.3%, SD ± 1.5, respectively. The mean value in the IUD group at baseline was 5.5%, SD ± 1.4. Values at 6 weeks, 6 months, and 12 months were 5.5%, SD ± 1.6, 6.0%, SD ± 1.6, and 6.3%, SD ± 1.3, respectively. The differences between groups in mean values at each observation were not statistically significant.
Similarly, the levonorgestrel intrauterine system had no adverse effect on fasting serum-glucose levels when compared with the copper IUD (Fig. 2). In the levonorgestrel group, the mean serum glucose level at baseline was 5.2 mM, SD ± 0.9. Values at 6 weeks, 6 months, and 12 months were 6.5 mM, SD ± 3.4, 6.6 mM, SD ± 3.1, and 7.4 mM, SD ± 4.2 respectively. Corresponding values for the copper IUD group at the four observations points were 5.0 mM, SD ± 0.6, 6.5 mM, SD ± 3.9, 6.2 mM, SD ± 3.4, and 7.5 mM, SD ± 4.2. No statistically significant differences in these levels between treatment groups were observed at any time.
Use of the levonorgestrel intrauterine system also had no effect on women's daily insulin requirements when compared with the copper IUD (Fig. 3). In the levonorgestrel group, the mean daily dose of insulin at baseline was 35.2 units, SD ± 12.7. Mean doses at 6 weeks, 6 months, and 12 months were 33.4 units, SD ± 12.7, 34.3 units, SD ± 12.2, and 35.1 units, SD ± 12.8, respectively. In the copper IUD group, the mean dose at baseline was 36.4 units, SD ± 9.7. Mean doses at 6 weeks, 6 months, and 12 months were 36.1 units, SD ± 8.6, 35.5 units, SD ± 8.9, and 36.4 units, SD ± 9.0, respectively. No significant differences in mean insulin doses between treatment groups were seen at any time.
When the correlations between measurements on the same study subjects over time are taken into account using repeated measures analysis of variance, in both treatment groups, the mean of glycosylated hemoglobin and fasting serum-glucose levels increased significantly from baseline to 12 months (P = .01 and P = .002 respectively). The increases were similar between the groups (P = .90 and P = .99 respectively). No differences in daily insulin requirements from baseline to follow-up endpoints were seen between groups (P = .98). We also found no evidence of group differences in glycosylated hemoglobin levels (P = .61), fasting serum-glucose levels (P = .76), and daily insulin requirements (P = .31) during the follow-up interval. No significant time-by-treatment interaction was evident. These findings were consistent with the results based on time-specific comparisons provided previously.
Women in both treatment groups had similar likelihoods of contraceptive discontinuation (P = .69). The Kaplan-Meier estimate of cumulative discontinuation probability was 0.13 for those assigned to the levonorgestrel group and 0.10 for those assigned to the copper IUD. The continuation rates per 100 women were 86.7%, (95% confidence interval 68.8–100%) for the levonorgestrel group and 90.3% (95% confidence interval 73.7–100% for the copper IUD group). Comparison of survival curves indicated no substantial difference between the two treatment groups (data not shown). The P values for the rank tests for homogeneity were 0.69 for the log-rank test and 0.72 for the Wilcoxon test. No pregnancies occurred during the trial.
Significant changes in hemoglobin levels were observed in both groups at 12 months. In the levonorgestrel group hemoglobin increased from 12.86 g/dL, SD ± 1.0 at baseline to 13.09 g/dL, SD ± 1.13 at 12 months. In the copper IUD group it decreased from 12.67 g/dL, SD ± 0.94 at baseline to 12.28 g/dL, SD ± 1.29 at 12 months.
No important differences in glycosylated hemoglobin levels, fasting serum-glucose levels, or daily insulin requirements emerged between women using an levonorgestrel intrauterine system and women using a nonhormonal copper IUD over 12 months of observation. The mild alteration in glycemic control in both groups is probably of little clinical importance, because the mean glycosylated hemoglobin levels remained within the normal range (< 8%), and the mean daily insulin requirements did not increase during the 12-month follow-up. Therefore, control of diabetes was not adversely affected by either contraceptive.
Our trial had several strengths. Randomization produced treatment groups similar in all important demographic respects, and baseline values for glycosylated hemoglobin levels, serum-glucose levels, and daily insulin requirements were comparable. Randomization also reduced the possibility of known or unknown confounding, and we avoided ascertainment bias by having objective, measurable outcomes. In addition, only 1 participant did not receive the intended treatment, and losses after randomization were infrequent and similar between the treatment groups.
Low systemic levels of levonorgestrel seem unlikely to impair glucose metabolism in women with diabetes. For example, subdermal levonorgestrel rods for contraception have not been associated with adverse metabolic effects.11,12 Even combined oral contraceptives have not been linked with progression of diabetes in women.13,14 In our trial, the small amounts of levonorgestrel absorbed systemically did not impair glucose metabolism. This lack of effect was evident even at the 6-week visit, when serum levels of levonorgestrel are higher than in later use.15 In our trial, as has been previously reported,16 the use of the levonorgestrel intrauterine system was associated with a significant increase in hemoglobin levels.
Overall, our results indicate that both the copper IUD and the levonorgestrel intrauterine system are safe contraceptive methods for women with diabetes. The WHO Category 2 rating for use of the levonorgestrel intrauterine system by women with insulin-dependent diabetes mellitus therefore is overly cautious. The levonorgestrel intrauterine system has several appealing noncontraceptive benefits, including a decrease in menstrual blood loss and dysmenorrhea. In addition, it can be used to treat heavy uterine bleeding, whether idiopathic in origin or associated with adenomyosis and leiomyomata.16,17 Based on these benefits and our new evidence, criteria for use of the levonorgestrel intrauterine system should be liberalized, and the WHO Category 2 rating should be changed to Category 1 for women with diabetes.1
1. World Health Organization. Improving access to quality care in family planning: medical eligibility criteria for contraceptive use. 2nd ed. Geneva (Switzerland): World Health Organization; 2000.
2. Klinke J, Toth EL. Preconception care for women with type 1 diabetes. Can Fam Physician 2003;49:769–73.
3. Chiou CF, Trussell J, Reyes E, Knight K, Wallace J, Udani J, et al. Economic analysis of contraceptives for women. Contraception 2003;68:3–10.
4. Skouby SO, Molsted-Pedersen L, Petersen KR. Contraception for women with diabetes: an update. Baillieres Clin Obstet Gynaecol 1991;5:493–503.
5. Kimmerle R, Weiss R, Berger M, Kurz KH. Effectiveness, safety, and acceptability of a copper intrauterine device (CU Safe 300) in type I diabetic women. Diabetes Care 1993;16:1227–30.
6. Kimmerle R, Heinemann L, Berger M. Intrauterine devices are safe and effective contraceptives for type 1 diabetic women. Diabetes Care 1995;18:1506–7.
7. Konje JC, Otolorin EO, Ladipo OA. Changes in carbohydrate metabolism during 30 months on Norplant. Contraception 1991;44:163–72.
8. Schroeder B, Hertweck SP, Sanfilippo JS, Foster MB. Correlation between glycemic control and menstruation in diabetic adolescents. J Reprod Med 2000;45:1–5.
9. Altman DG, Schulz KF, Moher D, Egger M, Davidoff F, Elbourne D, et al. The revised CONSORT statement for reporting randomized trials: explanation and elaboration. Ann Intern Med 2001;134:663–94.
10. Overall JE, Doyle SR. Estimating sample sizes for repeated measurement designs. Control Clin Trials 1994;15:100–23.
11. Diab KM, Zaki MM. Contraception in diabetic women: comparative metabolic study of Norplant, depot medroxyprogesterone acetate, low dose oral contraceptive pill and CuT380A. J Obstet Gynaecol Res 2000;26:17–26.
12. Curtis KM. Safety of implantable contraceptives for women: data from observational studies. Contraception 2002;65:85–96.
13. Klein BE, Moss SE, Klein R. Oral contraceptives in women with diabetes. Diabetes Care 1990;13:895–8.
14. Garg SK, Chase HP, Marshall G, Hoops SL, Holmes DL, Jackson WE. Oral contraceptives and renal and retinal complications in young women with insulin-dependent diabetes mellitus. JAMA 1994;271:1099–102.
15. Xiao BL, Zhou LY, Zhang XL, Jia MC, Luukkainen T, Allonen H. Pharmacokinetic and pharmacodynamic studies of levonorgestrel-releasing intrauterine device. Contraception 1990;41:353–62.
16. Hubacher D, Grimes DA. Noncontraceptive health benefits of intrauterine devices: a systematic review. Obstet Gynecol Surv 2002;57:120–8.
17. Grigorieva V, Chen-Mok M, Tarasova M, Mikhailov A. Use of a levonorgestrel releasing intrauterine system to treat bleeding related to uterine leiomyomas. Fertil Steril 2003;79:1194–8.