Hormone exposure and venous thromboembolism in commercially insured women aged 50 to 64 years

Abstract

Background

Menopausal hormone therapy (MHT) can elevate venous thromboembolism (VTE) risk, but less is known about formulations and routes of exposures.

Objective

To estimate hormone-associated VTE risk by route and formulation in exposed and unexposed women aged 50 to 64 years in the US.

Methods

In a nested case-control study of US commercially insured women aged 50 to 64 years (2007-2019), cases were defined as incident VTE diagnoses and matched to 10 controls by date of VTE and age, excluding prior VTE, inferior vena cava filter placement, or anticoagulants. Filled prescriptions in the prior year defined hormone exposures. International Classification of Diseases and Current Procedural Terminology codes identified risk factors and comorbidities.

Results

Odds ratios (ORs) were estimated with conditional logistic regression controlling for differences between cases (n = 20,359) and controls (n = 203,590) in comorbidities and VTE risk factors. For exposures within 60 days, oral MHT risk was almost twice as high as transdermal MHT (OR = 1.92; 95% CI, 1.43-2.60); transdermal MHT did not elevate risk compared with no exposure (unopposed OR = 0.70; 95% CI, 0.59-0.83; combined OR = 0.73; 95% CI, 0.56-0.96). Risk was highest for MHT combinations with ethinyl estradiol, followed by conjugated equine estrogen (CEE) (ethinyl estradiol-CEE: OR = 1.55; 95% CI, 1.07-2.25), and lowest for estradiol (CEE-estradiol: OR = 1.33; 95% CI, 1.02-1.72). Combined hormonal contraceptives elevated risk 5 times higher than no exposure (OR = 5.22; 95% CI, 4.67-5.84) and 3 times higher than oral MHT (OR = 3.65; 95% CI, 3.09-4.31).

Conclusion

The risk of VTE is much lower with MHT than combined hormone contraceptives and varies by hormone formulation and route of exposure. Transdermal MHT did not elevate risk. Oral MHT combinations with estradiol were lower risk than other forms of estrogen. Oral combined hormone contraceptives had much higher risk than oral combined hormonal MHT.

Essentials

• •Venous thromboembolism from menopausal hormone therapy (MHT) varies by type and route.
• •US women aged 50 to 64 years (cases = 20,359; controls = 203,590) were identified from administrative claims data.
• •Transdermal MHT did not elevate risk; among oral combined MHT, estradiol had the lowest risk.
• •Oral combined hormone contraceptives had a much higher risk than oral combined hormonal MHT.

1. Introduction

Two decades ago, the Women’s Health Initiative (WHI) [1,2] and the Heart and Estrogen-Progestin Replacement Study (HERS) [3,4] found that menopausal hormone therapy (MHT) lacked cardiovascular benefit and increased incidence of thrombotic events, particularly venous thromboembolism (VTE). Oral conjugated equine estrogen (CEE) approximately doubled the relative risk of VTE, with slightly lower risk when initiated closer to menopause [5,6] New evidence on VTE risk from UK and European observational studies suggests that hormone type (estrogen alone or combined), route, and formulation may affect risk of adverse outcomes [[7], [8], [9], [10], [11], [12], [13]]. Specifically, results suggest that VTE risk is lower for estradiol than for CEE and that transdermal MHT does not raise risk.

The question is whether these results replicate in other national contexts. US formularies, prescribing patterns, and VTE incidence differ from those in Europe. For example, dydrogesterone, a lower risk progestin, is unavailable and incidence of VTE is much higher in the US than in Europe [14]. The WHI and HERS results vastly reduced use of MHT [[15], [16], [17]], but approximately 8% of US women aged ≥50 years continue to use noncontraceptive estrogens [17], and little data exist as to how many additional women in this age range use hormonal contraceptives. Common indications for MHT include hot flashes, vaginal dryness, disruptions in sleep patterns, and inability to concentrate. However, many women who experience menopausal symptoms are not prescribed MHT because US clinical guidelines discourage its use [18].

This study examined VTE risk from hormone exposures within the US context. Using a large, US population-based administrative health database, risk was assessed in commercially insured US women aged 50 to 64 years without a prior VTE. A nested case-control design with matched cases and controls estimated VTE risk of hormone exposures by administration route and formulation, while controlling for other risk factors. The study focused on women entering the menopausal period, because they may be the most likely to desire and use MHT.

2. Methods

2.1. Data

Optum’s deidentified Clinformatics Data Mart database contains patient-level, longitudinal administrative health claims for one of the largest US commercially insured populations. This deidentified longitudinal database of administrative, medical, pharmacy, and laboratory claims represents approximately 62 million unique members (2007–2018) across all US regions.

A base cohort of women aged 50 to 64 years was selected based on the median age of menopause at 49.6 years [19] and the large transition of enrollees from commercial insurance to Medicare at the age of 65 years. Women with at least 1 year of enrollment between January 1, 2007 and December 31, 2019 were eligible. Anyone with missing information for sex (0.01%) or age (0.00007%) was excluded prior to selection of the cohort. This was a records-based study without direct patient or public involvement. Because analyses were conducted with de-identified patient data and results were to be reported only in aggregate form, the University of Texas Medical Institutional Review Board approved the study without consent.

2.2. Cases

Using the base cohort, acute VTE diagnoses were identified with International Classification of Diseases (ICD) codes (see Supplementary File for details on all variable definitions, including ICD codes), regardless of the setting of the diagnosis (inpatient or outpatient). Cases required a VTE diagnosis plus filling of an anticoagulant prescription (excluding heparin flushes), placement of an inferior vena cava (IVC) filter, or death within 30 days following diagnosis to minimize the inclusion of suspected or unconfirmed cases [14]. VTE case identification in administrative claims data using this method has shown good accuracy [20]. Cases were the first occurrence of a VTE after 12 months of continuous enrollment without a VTE or IVC filter. VTE date defined the index date. Those with anticoagulant long-term use (in past 12 months) or exposure within 14 days of the index date (heparin flushes excepted) were excluded.

2.3. Controls

Controls were matched to cases (10:1) on index date and age (±2 years) and selected at random. A minimal matching strategy was used to increase generalization to the population. Controls had to be continuously enrolled in the 12 months prior to the index date with the same exclusion criteria as cases.

2.4. Hormone Exposures

For cases and controls, all prescriptions filled within the year prior to the index date containing estrogen and/or progestogen were recorded, as well as hormone type (unopposed estrogen, estrogen-progestogen combinations, progesterone only), administration route (oral, transdermal, vaginal), estrogen formulation (estradiol, CEE, ethinyl estradiol, etc.), and progestogen formulation (progesterone, norethindrone, medroxyprogesterone acetate [MPA], etc.). Low-dose vaginal estrogens used exclusively for treatment of local genitourinary symptoms were excluded due to their lack of effect on systemic estradiol levels [21]. Timing of the most recent exposure was coded as within 30 days, 31 to 60 days, 61 to 90 days, or 91 to 365 days prior to the index date. All 30- and 90-day prescriptions were considered for subsequent 30- or 90-day exposure periods. Estrogen-progestogen contraceptives were categorized separately. Those without a hormonal prescription in the previous year were considered unexposed and formed the reference category.

Women with overlapping prescriptions for an estrogen and a progestogen (eg, transdermal estrogen with an oral progestogen) were considered exposed to both. For women exposed to estrogen through 2 different routes of administration (oral and either transdermal or vaginal) in the same month, the exposure route was considered oral (n = 5 in 0-60 day period).

2.5. Covariates

Known VTE risk factors were recorded: hospitalization/surgery or trauma in the previous 30 days; malignancy, varicose veins, other thromboses/thrombophilia, and smoking documented in the previous 12 months (see Supplementary File). In addition, coronary artery disease (CAD), stroke/transient ischemic attack (TIA), and Elixhauser comorbidities (31 conditions, such as hypertension, diabetes, and liver disease that are predictive of mortality) [22,23] in the past 12 months were included to control for general health status between cases and controls (possible sources of indication bias).

2.6. Analysis

Relative risks for hormone exposures were estimated with odds ratios (ORs) and 95% Wald CIs from conditional logistic regression models compared with no exposure in the past year. Analyses were adjusted for known VTE risk factors, CAD, stroke/TIA, Elixhauser comorbidities, age, and US region of residence. The Elixhauser comorbidities were reduced into a single index: none, 1 or 2, and ≥3. Cancers and coagulopathy were removed from the Elixhauser index and analyzed as separate variables as known VTE risk factors. Because previous studies have varied in their definition of the timing of exposure (within 1 month [11], 3 months [13,24], and 6 months [25] of the index date), models explored timing of hormone exposures to identify the highest risk exposure period (0-30, 0-60, or 0-90 days before the index date). Separate models assessed the risk associated with the type of therapy (estrogen with/without other hormones), route of administration, and formulation, while controlling for other risk factors. Stratified analyses explored possible subsample differences, first by excluding anyone with cancer or other thromboses/thrombophilia and second by stratifying by age (<58 and ≥58 years), because most women reach menopause by 58. All statistical analyses were performed with SAS version 9.4 [26]. Number needed to harm (NNH) [27] was estimated assuming a population VTE incidence of 180 per 100,000 for women in this age range [14] and the adjusted ORs from conditional logistic regression models.

3. Results

3.1. Samples

From the base cohort of women, 20,359 cases of acute VTE met inclusion criteria and were matched with 203,590 controls (Figure). Among cases, 76% were primary diagnoses, 89.37% (18,194) received anticoagulant therapy, 12.00% (2444) had an IVC filter placed, and 5.88% (1197) died within 30 days of diagnosis. There were 10,995 (54.01%) cases of pulmonary embolism (with or without deep vein thrombosis) and 9364 (45.99%) cases of deep vein thrombosis (without pulmonary embolism). Cases had more comorbidities and risk factors than controls: 55.94% of cases and 22.45% of controls had ≥3 Elixhauser comorbidities; 26.00% of cases and 5.59% of controls had cancer in the previous year; and 30.22% of cases and 2.29% of controls were recently hospitalized or had surgery (Table 1).

Figure.

Selection of incident cases by Dx, diagnosis; IVC, inferior vena cava; VTE, venous thromboembolism and matched controls.

Table 1.

Covariates.

Variable	VTE cases (n = 20,359)	Controls (n = 203,590)
Region
1 Northeast	8.3% (1686)	8.9% (18,010)
2 Midwest	27.8% (5661)	25.6% (52,021)
3 South	43.8% (8912)	44.8% (91,138)
4 West	20.0% (4067)	20.1% (40,947)
5 Unknown	0.2% (33)	0.7% (1474)
Agea
50-55	33.4% (6789)	34.8% (70,927)
56-60	34.8% (7084)	33.5% (68,237)
61-65	31.9% (6486)	31.6% (64,426)
EH comorbidities
None	13.8% (2812)	37.3% (75,981)
1-2	30.3% (6159)	40.2% (81,909)
3+	55.9% (11,389)	22.5% (45,700)
Cancer
None	74.0% (15,065)	94.4% (192,220)
Nonmetastatic	11.1% (2262)	4.8% (9764)
Metastatic	14.9% (3032)	0.8% (1606)
Hospital/surgery	30.2% (6152)	2.3% (4672)
Trauma	15.0% (3060)	3.0% (6030)
Varicose veins	2.7% (556)	0.9% (1790)
Other thromboses or thrombophilia	11.0% (2237)	1.3% (2551)
CAD	13.5% (2743)	4.8% (9775)
Stroke/TIA	7.3% (1495)	2.4% (4886)
Smoking	23.1% (4710)	9.5% (19263)

CAD, coronary artery disease; EH, Elixhauser; TIA, transient ischemic attack; VTE, venous thromboembolism.

^aCases and controls were matched on age ±2 years.

3.2. Timing of hormone therapy exposure

Because VTE risk diminished 60 days after discontinuation, we used hormone exposures within 60 days of the index date as the focus of the analysis. Within the year prior to the index date, hormone exposures occurred in 11.19% (25,053/223,949) of women; most were continuous users (96.54%), and the remainder were discontinuers, with very few initiating therapy within 60 days. Within the most recent 60 days, 8.73% (10.46% cases, 8.56% controls) were exposed. Those with a filled prescription resulting in exposure within 30 days of the index date had an OR of 1.53 (95% CI, 1.45-1.62) compared with those without hormone exposure in the past year (Table 2). The OR for those without a filled prescription in the previous 30 days but who had filled a prescription resulting in exposure within 31 to 60 days was elevated (OR = 1.22; 95% CI, 0.99-1.50) but not significantly different from no exposure. The OR for a 0 to 30 days exposure was 26% higher than that for the 31 to 60 days exposure (OR = 1.26, 1.02-1.56), and the OR for the 31 to 60 days exposure was 35% higher than that for the 61 to 90 days exposure (OR = 1.35, 0.95-1.92). ORs for exposures 61 to 90 days before the index date were not elevated compared with the past 91 to 365 days (OR = 1.04, 0.77-1.41) or compared with no exposure in the past year (OR = 0.90, 0.68-1.19). Thus, current exposure was defined as any exposure within 60 days of the index date, and “past” use was defined as any exposure in the previous 61 to 365 days. The OR for current exposure to any estrogen or progestogen was 1.51 (95% CI, 1.42-1.59) compared with no exposure. Current users tended to be slightly younger (44% of exposed and 34% unexposed were aged 50-55 years).

Table 2.

Odds ratios for VTE by recency of exposure.

Last exposure	Adjusted ORa (95% CI)
0-30 d	1.53 (1.45-1.62)
31-60 d (no 0–31 d)	1.22 (0.98-1.48)
61-90 d (no 0-60 d)	0.90 (0.68-1.19)
91-365 d (no 0-90 d)	0.94 (0.83-1.06)
0-60 d	1.51 (1.43-1.59)
61-365 d (no 0-60 d)	1.93 (0.83-1.04)

Reference category is no exposure in the past year.

OR, odds ratio; VTE, venous thromboembolism.

^aAdjusted for all variables in Table 1.

3.3. Type of hormone therapy exposure

Although VTE risk was elevated with any hormone exposure occurring within 60 days of the index date compared with no exposure in the past year, risk varied by type of exposure. After adjustment for all covariates, the OR for MHT exposure with unopposed estrogen was 1.13 (95% CI, 1.04-1.23; exposed n = 9650) and 1.22 (95% CI, 1.10-1.38; n = 5160) with estrogen-progestogen combinations compared with no exposure. The highest OR was from estrogen-progestogen contraceptive use (OR = 5.22; 95% CI, 4.67-5.84; n = 2476). The OR for progestogens alone was 1.49 (95% CI, 1.24-1.78; n = 1486) and 0.49 (95% CI, 0.34-0.71; n = 786) for estrogen-testosterone (with or without a progestogen). Table 3 shows crude (unadjusted) ORs and ORs adjusted for covariates in comparisons between treatments and the unexposed reference group (no exposure in the past year). Controlling for covariates had the largest effect on treatments with oral ethinyl estradiol and on small subgroups.

Table 3.

Odds ratios for VTE by specific hormones.

Hormonal exposures	N	Controls (203,590)	Cases (20,359)	Crude OR (95% CI)	Adjusted ORa (95% CI)
ANY 0–60 d estrogen or progestogen	19,558	17,428	2130	1.25 (1.19-1.31)	1.51 (1.43-1.59)
Estrogen only (MHT)	9650	8735	915	1.07 (0.99-1.15)	1.13 (1.04-1.23)
Oral	6638	5907	731	1.26 (1.17-1.36)	1.33 (1.21-1.46)
Estradiol	3680	3294	386	1.20 (1.08-1.33)	1.24 (1.09-1.40)
CEEb	2746	2426	320	1.34 (1.19-1.51)	1.46 (1.28-1.68)
Other estrogens	212	187	75	1.36 (0.90-2.07)	1.39 (0.87-2.24)
Transdermal/estradiol	2885	2713	172	0.65 (0.56-0.76)	0.70 (0.59-0.84)
Vaginal/estradiol esterc	106	97	9	0.95 (0.48-1.87)	1.04 (0.44-2.46)
IM/Estrogend	21	18	3	1.68 (0.50-5.70)	0.38 (0.08-1.77)
Estrogen + progestogens (MHT)	5160	4727	433	0.94 (0.85-1.03)	1.22 (1.10-1.38)
Oral	3947	3582	365	1.04 (0.93-1.16)	1.40 (1.24-1.59)
Estradiol + any progestogen	1995	1844	151	0.84 (0.71-0.99)	1.14 (0.95-1.37)
Estradiol + norethindrone	1014	945	69	0.75 (0.58-0.95)	1.03 (0.78-1.35)
Estradiol + other	981	899	82	0.93 (0.74-1.17)	1.25 (0.97-1.61)
CEE/MPAe	1519	1363	156	1.17 (0.99-1.38)	1.52 (1.25-1.84)
Ethinyl estradiol + norethindrone	413	358	55	1.57 (1.18-2.09)	2.35 (1.71-3.25)
Other estrogens/other progest	20	17	3	1.81 (0.53-6.18)	1.33 (0.25-7.14)
Transdermal/estradiol + progestf	1213	1145	68	0.61 (0.48-0.78)	0.73 (0.55-0.96)
+Progesterone	487	465	22	0.48 (0.32-0.74)	0.55 (0.34-0.89)
+Micronized progesterone	262	249	13	0.54 (0.31-0.94)	0.63 (0.34-1.19)
+ Norethindrone	304	288	16	0.57 (0.35-0.95)	0.65 (0.36-1.14)
+MPA	55	49	6	1.26 (0.54-2.94)	1.76 (0.69-4.50)
+other	105	94	11	1.18 (0.63-2.20)	1.74 (0.87-3.46)
Estrogen+progest (contraception)	2476	1915	561	3.14 (2.84-3.46)	5.22 (4.67-5.84)
Oral/ethinyl estradiolg	2399	1862	537	3.08 (2.79-3.41)	5.11 (4.57-5.73)
Transdermal/ethinyl estradiol	13	11	2	1.92 (0.42-8.65)	5.34 (1.06-26.79)
Vaginal/ethinyl estradiol	64	42	22	5.60 (3.34-9.39)	9.26 (5.31-16.15)
Estrogen/CEE+testosterone (±P)h	786	745	41	0.56 (0.41-0.77)	0.49 (0.34-0.71)
Progestogen-only	1486	1306	180	1.43 (1.22-1.67)	1.48 (1.24-1.78)
Progesterone	588	550	38	0.71 (0.51-0.98)	0.83 (0.57-1.20)
Micronized progesterone	223	201	22	1.13 (0.73-1.75)	1.20 (0.73-1.97)
Norethindronei	232	198	34	1.83 (1.27-2.64)	1.99 (1.30-3.03)
MPA	443	357	86	2.51 (1.98-3.17)	2.20 (1.67-2.92)
ANY past estrogen or progest (61-365 d)	5494	4995	499	1.02 (0.93-1.12)	0.94 (0.84-1.05)
NO exposure (ref)	198,897	181,167	17,730	1.00	1.00

CEE, conjugated equine estrogen; IM, intramuscular; MHT, menopausal hormone therapy; MPA, medroxyprogesterone acetate; OR, odds ratio.

^aModels adjusted for all risk factors in Table 1 and reference is no hormone use in the past year (n = 198,897).

^bIncludes esterified estrogen (n = 86).

^cVaginal estrogen only (FemRing) also includes those with oral progestogen (n = 17).

^dIncludes IM estrogen + progestogen (n = 1).

^eIncludes other progestogens (n = 98); includes esterified estrogen (n = 15).

^fIncludes transdermal estrogen-progestogen and transdermal estrogen with oral progestogen.

^gIncludes oral mestranol (n = 6), estradiol + norgestimate with levonorgestrel (plan B) (n = 1), and estradiol ester (n = 1).

^hIncludes some with progestogens (n = 146).

ⁱIncludes other progestogens (n = 1).

For MHT, oral exposures had a higher risk than transdermal or vaginal exposures compared with unexposed (Table 3). The OR for oral, unopposed estrogen was 1.33 (95% CI, 1.21-1.46; n = 6638) and 1.40 (95% CI, 1.24-1.59; n = 3947) for oral estrogen-progestogen combinations. Transdermal estradiol, whether unopposed (OR = 0.70; 95% CI, 0.59-0.84; n = 2885) or combined with a progestogen (OR = 0.73; 95% CI, 0.55-0.96; n = 1213) had the lowest risk. Transdermal estradiol-progestogen (n = 380) and transdermal estradiol with an oral progestogen (n = 833) were combined, as neither was significantly different from no exposure or from one another. Estrogen-only vaginal exposure (FemRing) also may be low risk (OR = 1.04; 95% CI, 0.44-2.46; n = 106), but the subsample was small. Categories for menopausal vaginal exposures with unopposed estrogen (n = 89) and those supplemented with an oral progestogen (n = 17) were combined, as neither was significantly different from no exposure or from one another. In a comparison between oral and transdermal exposures (Table 4 shows direct comparisons between treatments), the ORs for oral exposures were almost twice as high as transdermal exposures (estrogen-only OR = 1.90; 95% CI, 1.56-2.32; combined therapy OR = 1.92; 95% CI, 1.43-2.60).

Table 4.

Direct comparisons of treatments.

Treatment exposure	Treatment exposure	Adjusted OR
Any oral exposure (estrogen only)	Any transdermal exposure (estrogen only)	1.90 (1.56-2.32)
Any oral exposure (estrogen + progestogen)	Any transdermal exposure (estrogen + progestogen)	1.92 (1.43-2.60)
Oral CEE exposure (estrogen only)	Oral estradiol exposure (estrogen only)	1.19 (0.99-1.43)
Oral CEE exposure (estrogen + progestogen)	Oral estradiol exposure (estrogen + progestogen)	1.33 (1.02-1.72)
Oral ethinyl estradiol exposure (estrogen + progestogen)	Oral CEE exposure (estrogen + progestogen)	1.55 (1.07-2.25)
Oral ethinyl estradiol + norethindrone exposure	Oral estradiol + norethindrone exposure	2.28 (1.50-3.48)
Any contraceptive (estrogen + progestogen)	Any MHT (estrogen + progestogen)	4.24 (3.64-4.98)
Oral contraceptives (estrogen + progestogen)	Oral MHT (estrogen + progestogen)	3.65 (3.09-4.31)
Transdermal (estrogen + progestogen)	Transdermal (estrogen only) + oral (progestogen)	0.82 (0.46-1.45)
Vaginal (estrogen only)	Vaginal (estrogen only) + oral (progestogen)	0.57 (0.05-6.32)

CEE, conjugated equine estrogen; MHT, menopausal hormone therapy; OR, odds ratio.

The estrogen MHT formulation also affected risk. With oral unopposed estrogen, the OR was 1.46 with CEE (95% CI, 1.28-1.68; n = 2746) and 1.24 with estradiol (OR = 1.24; 95% CI, 1.09-1.40; n = 3680) compared with no exposure (Table 3), although the 2 treatments were not significantly different from one another (OR = 1.19; 95% CI, 0.99-1.43; Table 4). Among women exposed to unopposed oral estrogen, for every 2320 women taking estradiol or 1211 taking CEE, 1 additional woman will experience a VTE. ORs for oral estrogen-progestogen combinations were 1.14 with estradiol (95% CI, 0.95-1.37; n = 1995), 1.52 with CEE (95% CI, 1.25-1.84; n = 1519), and 2.35 with ethinyl estradiol (95% CI, 1.71-3.25; n = 413) compared with no exposure. Comparisons between oral combined hormonal MHT treatment groups showed that the lowest risk was with estradiol: The OR was significantly higher for combinations with CEE than those with estradiol (OR = 1.33; 95% CI, 1.02-1.72) and for combinations with ethinyl estradiol than those with CEE (OR = 1.55; 95% CI, 1.07-2.25); the OR for oral ethinyl estradiol-norethindrone was almost 3 times as high as that for estradiol-norethindrone (OR = 2.88; 95% CI, 1.07-2.25). Among women exposed to oral combined MHT, for every 3976 women taking estradiol + progestogen (NNH = 3976), 1071 women taking CEE + progestin (NNH = 1071), or 413 women taking ethinyl estradiol + progestin (NNH = 413), 1 additional woman will experience a VTE.

Risk was very high for contraceptive exposure (almost exclusively ethinyl estradio l + progestin combinations) compared with no hormone exposure. Contraceptive users tended to be younger (94% were aged 50-54 years) and most had oral exposures. The OR for contraceptive exposures was 5.11 (95% CI, 4.57-5.73; n = 2399). ORs were 5.34 for transdermal exposures (OR = 5.34, 95% CI, 1.06-26.78; n = 13) and 9.26 for vaginal contraceptives (NuvaRing, 95% CI, 5.31-16.15; n = 64). For every 132 women exposed to estrogen-progestogen contraceptives, 1 additional woman will experience a VTE (NNH = 132). A comparison between any combined hormone contraceptive exposure and any estrogen-progestogen MHT exposure (all routes) showed a 4-times-higher OR with contraceptives (OR = 4.24; 95% CI, 3.64-4.98); oral combined hormone contraceptives had a 3.64-times-higher OR than oral combined MHT (OR = 3.65; 95% CI, 3.09-4.31). Increased risk with contraceptives may be due to formulation and higher dosage; a direct comparison between oral combined hormone contraceptives and oral combined MHT, both containing ethinyl estradiol + norethindrone showed the contraceptive form more than doubled the OR (OR = 2.16; 95% CI, 1.50-3.10).

Because there were many progestogens with small subgroup sizes, we present—for descriptive purposes only—the 3 categories most sensitive to physician choice: progestogens combined with oral estradiol, or with transdermal estradiol, and progestogens alone. Oral estradiol was most often combined with norethindrone (n = 1014), but also with MPA (n = 465), progesterone (n = 280), micronized progesterone (n = 143), levonorgestrel (n = 1), norgestimate (n = 38), and drospirenone (n = 56). Oral CEE was most often combined with MPA (n = 1421). When progestogens were given alone, ORs were highest for MPA (OR = 2.20; 95% CI, 1.67-2.92; n = 443), followed by norethindrone (OR = 1.99; 95% CI, 1.30-3.03; n = 232), micronized progesterone (OR = 1.20; 95% CI, 0.73-1.97; n = 223), and progesterone (OR = 0.83; 95% CI, 0.57-1.20; n = 588) compared with no exposure in the past year. A similar ordering was observed when progestogens were combined with transdermal estradiol: ORs were highest with MPA and lowest with progesterone compared with no exposure.

Subanalyses checked model stability and robustness across subgroups. When those with cancer or other thrombosis/thrombophilia (cancer only [n = 15,156], thrombosis/thrombophilia only [n = 3280], or both [n = 1508]) were excluded, results from the full sample were similar for MHT exposures but showed some increases for CEE and contraceptive exposures (Table 5). When stratified by age (<58 and ≥58 years), transdermal exposures remained without elevated risk across subsamples, but the VTE risk from MHT was slightly higher in younger women. In contrast, risk from contraceptive exposure was markedly elevated in older women (OR = 8.95; 95% CI, 4.59-17.44) than in younger women (OR = 4.83; 95% CI, 4.31-5.41). The difference in ORs for contraceptives across the 2 age groups may be due to the very small sample size: most contraceptive exposures (principally ethinyl estradiol + norethindrone) were in women aged <58 years (only 48 women in the older group were exposed to contraceptives), although risk may indeed be higher. However, the OR for oral ethinyl estradiol + norethindrone MHT was stable across the 2 age subgroups and had approximately half the exposed group in each of the 2 age categories.

Table 5.

Subsample results focusing on route and formulation of estrogen exposures.

Hormonal exposures	Total samplea (N = 223,949)	No cancer, no thrombosesb (n = 204,005)	Age < 58 ya (n = 106,484)	Age ≥ 58 ya (n = 117,465)
Referent: no exposure in the past year	1.00	1.00	1.00	1.00
Estrogen only (MHT)
Oral
Estradiol	1.24 (1.09-1.40)	1.27 (1.11-1.46)	1.54 (1.29-1.84)	1.17 (0.98-1.39)
CEEc	1.46 (1.28-1.68)	1.53 (1.32-1.78)	1.75 (1.42-2.16)	1.41 (1.17-1.69)
Other estrogens	1.39 (0.87-2.24)	1.52 (0.91-2.53)	1.08 (0.42-2.77)	1.87 (1.08-3.25)
Transdermal/estradiol	0.70 (0.59-0.83)	0.69 (0.57-0.84)	0.78 (0.61-0.99)	0.63 (0.48-0.81)
Vaginal/estradiol esterd	0.96 (0.43-2.14)	0.87 (0.36-2.07)	1.46 (0.48-4.44)	0.62 (0.18-2.13)
IM/Estrogene	0.38 (0.08-1.77)	0.34 (0.06-2.11)	0.79 (0.07-8.46)	0.31 (0.04-2.23)
Estrogen + progestogen(MHT)
Oral
Estradiol	1.14 (0.95-1.37)	1.18 (0.97-1.45)	1.14 (0.88-1.47)	0.97 (0.74-1.28)
CEE/MPAf	1.52 (1.25-1.84)	1.69 (1.38-2.07)	1.55 (1.18-2.04)	1.38 (1.06-1.80)
Ethinyl estradiol + norethindrone	2.35 (1.71-3.25)	2.41 (1.70-3.41)	2.13 (1.33-3.43)	2.11 (1.36-3.30)
Other estrogens	1.33 (0.25-7.14)	1.67 (0.27-10.36)	1.40 (0.06-35.21)	1.07 (0.15-7.59)
Transdermal/estradiolg	0.73 (0.56-0.96)	0.70 (0.51-0.96)	0.62 (0.42-0.92)	0.78 (0.53-1.15)
Estrogen + progestin(contraceptives)	5.22 (4.67-5.83)	5.59 (4.96-6.30)	4.83 (4.31-5.41)	8.95 (4.59-17.44)
Estrogen/CEE + testosterone (±Progest)h	0.49 (0.34-0.71)	0.46 (0.30-0.71)	0.49 (0.28-0.85)	0.46 (0.28-0.77)
Progestogens only	1.48 (1.23-1.78)	1.45 (1.18-1.77)	1.70 (1.37-2.12)	1.29 (0.92-1.81)
Past hormone therapy (61-365 d)	0.94 (0.84-1.05)	0.88 (0.77-1.01)	1.01 (0.87-1.17)	1.95 (0.80-1.14)

CEE, conjugated equine estrogen; IM, intramuscular; MHT, menopausal hormone therapy; MPA, medroxyprogesterone acetate.

^aModels adjusted for all risk factors in Table 1 and reference is no use in the past year.

^bModels adjusted for risk factors in Table 1, except cancer and other thromboses/thrombophilia; reference is no use in the past year.

^cMay include esterified estrogen.

^dVaginal estrogen only (FemRing) also includes some with oral progestogen.

^eMay include IM estrogen + progestogen (n = 1).

^fMay include other progestogens (n = 98); may include esterified estrogen.

^gMay include transdermal estrogen-progestogen and transdermal estrogen with oral progestogen.

^hMay include progestogens (n = 146) and may include esterified estrogens.

4. Discussion

4.1. Principal findings

Since the landmark trials revealed the potential for thrombotic risk with MHT, large observational studies have provided convincing evidence that transdermal estradiol, whether unopposed or combined with a progestogen, did not increase VTE risk, and for oral therapy, estradiol was safer than CEE.

This study strengthens that evidence and uniquely adds information regarding risk within the context of US prescribing patterns. Results indicated that more than two-thirds of transdermal estradiol exposures were combined with an oral progestogen. Transdermal exposures remained lower risk, but risk may increase when combined with a higher-risk oral progestin (eg, MPA). In addition, MHT with ethinyl estradiol may be more common in the US than in Europe and increased risk more than CEE. VTE risk, however, appeared to diminish 60 days after hormone discontinuation.

A striking finding in this study was the 5-fold increase in VTE risk with estrogen-progestin contraceptives compared with no hormone exposure (NNH = 132) and a 4-fold increase compared with combined MHT. In other studies and depending on the progestin, oral combined contraceptives have had an approximately 4-fold increase in VTE risk for women aged 15 to 50 years [28,29] and a 6-fold increase in women aged ≥50 years [11]. In this study, contraceptives likely increased risk in women compared with MHT because of the estrogen formulation (ethinyl estradiol) and dosage. Within MHT formulations, ethinyl estradiol had the highest risk, and in a direct comparison between oral combined contraceptives and oral combined MHT with the same formulation (ethinyl estradiol + norethindrone), contraceptives doubled risk. The observed 9-fold increase in women aged >58 years may be due to true increased risk in older women or small sample size (n = 48).

4.2. Strength and weaknesses

A principal limitation of this study is that insurance claims data may not capture important information. Some risk factors (eg, smoking) are challenging to accurately capture in claims data and may be underestimated. We lacked detailed personal information, such as socioeconomic status and age at menopause. Exposure was estimated from filled prescriptions without information on adherence, although most women who filled hormone prescriptions did so continuously. We also lacked data on the indications for prescriptions, particularly limiting interpretation for women on progestogen-only regimens, because potential indications range from contraception to abnormal uterine bleeding to menopausal symptoms. Although these limitations may result in some residual confounding, it is unlikely that they would be systematically different between cases and controls.

The results of this study may be limited by the observational design. Randomization in clinical trials alleviates bias and confounding, but such trials are costly, time consuming, and unlikely to be large enough to test various formulations and routes of exposures. Insurance claims data provide sufficiently large samples for more nuanced analyses. In addition, the nested case-control design used in this study with a time-restricted approach offers a strong and efficient method that results in unbiased estimation of exposure risk for the population in the base cohort. The base cohort, however, may not be representative of the US population, so the results are most relevant to insured women. We statistically controlled for factors associated with the outcome variable (VTE risk factors) and factors likely to be associated with exposure (CAD, stroke/TIA, and the Elixhauser index) to minimize indication bias. Nevertheless, in an observational design, it is difficult to disentangle prior differences between subgroups of women from therapeutic effects so that results, especially possible protective effects from transdermal estradiol or estrogen–testosterone combinations, are only suggestive until tested in a clinical trial.

Primary data collection efforts assessing thrombotic risk due to MHT have had limited sample sizes to explore exposure routes and formulations. VTE incidence is infrequent (129/100,000) [14] and large medical record studies and meta-analyses have facilitated assessment of risk from various types of exposures. VTE risks in this study are consistent with those from a recent meta-analysis [30] and large UK clinical databases [13] for oral unopposed estrogen and transdermal estrogen. For oral combined MHT, the meta-analysis found heterogeneity across studies. Our results were similar to UK results (1998–2017) [13], but we used a more restrictive VTE definition and age range and found lower risk for oral CEE + MPA. Our CEE MHT estimates were somewhat lower than a similar study (1998–2001) [31] in a US health maintenance organization that used clinical records to verify cases. At that time, however, more than a third of women used MHT; now far fewer women use MHT, and women who are at higher risk may be discouraged from using MHT. Our results also suggest that transdermal VTE risk may increase when combined with higher risk oral progestins, such as MPA. Progestogens and specifically MPA can increase the thrombotic risk of estrogens [[10], [11], [12], [13]]. In this study, MPA had the highest risk, whereas the effect of oral norethindrone was unclear. In the UK study, progestogen-only exposures had higher risk than in this study, possibly because of differences in the prescribed progestogens.

4.3. Implications and future research

To our knowledge, this is the largest, detailed study on hormone risk in US women aged 50 to 64 years. The study confirms prior findings that estradiol, and in particular transdermal estradiol, offers MHT with lower risk of VTE. However, more research is needed on the effects of various oral progestogens when combined with transdermal estradiol because two-thirds of the transdermal estradiol exposures were combined with oral progestogens. In addition, the results suggest that ethinyl estradiol carries an even higher VTE risk than CEE.

Evidence has accumulated on differences in risk by type, route, and formulation of hormones, and our study highlights that some women are exposed to riskier forms of hormones than alternatives. A striking finding was a 4-fold higher VTE risk from contraceptive exposures compared with MHT for women aged ≥50 years. The markedly increased VTE risk from combined hormonal contraceptives—whether oral, transdermal, or vaginal—suggests that strong caution should be taken in prescribing contraceptive-grade hormone therapy for perimenopausal symptoms, when conception risk is low. Women should be counseled on the risks of continuing hormonal contraceptives into older ages and that benefits may be outweighed by risk at these dosages. Where possible, clinicians should consider withdrawing contraceptive therapy to assess for clinical/laboratory markers of menopause so that women can transition to MHT in place of contraceptives. Clinicians should be thoughtful about baseline risk—including older age—as well as risk from specific products. Further research needs to explore hormonal risks across a wider age range from the same cohort, including women aged >50 years for contraceptive exposures and women aged <50 years for MHT exposures.