Comparison of Healthcare Costs for Women with Treated Versus Untreated Vasomotor Symptoms Due to Menopause

Abstract

Introduction

The study objective was to estimate all-cause healthcare resource utilization (HCRU) and medical and pharmacy costs for women with treated versus untreated vasomotor symptoms (VMS) due to menopause.

Methods

A retrospective study was conducted using US claims data from Optum Research Database (study period: January 1, 2012–February 29, 2020). Women aged 40–63 years with a VMS diagnosis claim and ≥ 12 and ≥ 18 months of continuous enrollment during baseline and follow-up periods, respectively, were included. Women treated for VMS were propensity score matched 1:1 to untreated controls with VMS. Standardized differences (SDIFF) ≥ 10% were considered meaningful. A generalized linear model (gamma distribution, log link, robust standard errors) estimated the total cost of care ratio. Subgroup analyses of on- and off-label treatment costs were conducted.

Results

Of 117,582 women diagnosed with VMS, 20.5% initiated VMS treatment and 79.5% had no treatment. Treated women (n = 24,057) were matched to untreated VMS controls. There were no differences in HCRU at follow-up (SDIFF < 10%). Pharmacy ($487 vs $320, SDIFF 28.4%) and total ($1803 vs $1536, SDIFF 12.6%) costs were higher in the treated cohort. Total costs were 7% higher in the treated cohort (total cost ratio 1.07, 95% CI 1.05–1.10, P < 0.001). The on-label treatment pharmacy costs ($546 versus $315, SDIFF 38.6%) were higher in the treated cohort. Off-label treatment had higher medical costs ($1393 versus $1201, SDIFF 10.4%).

Conclusions

Most women with VMS due to menopause were not treated within 6 months following diagnosis. While both on- and off-label treatment increased the total cost of care compared with untreated controls, those increases were modest in magnitude and should not impede treatment for women who report symptom improvement as a result of treatment.

Key Summary Points

Why carry out this study?

Vasomotor symptoms (VMS) due to menopause are estimated to impact between 40 and 80% of women; approximately 34% of women characterize their symptoms as moderate to severe

VMS often goes untreated, and this can be costly

The cost of VMS medications may be offset by reductions in medical costs related to improved VMS symptoms

What was learned from this study?

Most women did not receive treatment for their VMS due to menopause

Among women who were treated for VMS, their total cost of care was higher than for their untreated counterparts; however, the increase was modest

Introduction

Vasomotor symptoms (VMS), characterized by hot flashes (also known as hot flushes) and/or night sweats, are the predominant symptoms experienced by most women during the menopausal transition [1]. The prevalence of VMS due to menopause varies widely given a range of individual and demographic factors and stage of menopause but is estimated to be between 40 and 80% in the US [2–4]. A recent cross-sectional survey estimated the prevalence of moderate to severe VMS to be 34% in the US [5].

VMS is frequently left untreated [6]. Hormone therapy (HT) is the most commonly prescribed treatment for VMS [7], but many women are averse to HT because of concerns of increased risk of breast cancer or vascular events [5, 8]. A recent study estimated that as many as 54% of women who are eligible for HT are HT-averse [5]. Another study reported that when left untreated, moderate to severe VMS can be costly [9]. Furthermore, healthcare resource utilization (HCRU), healthcare costs, and work productivity loss were all higher in women with untreated VMS compared with women without VMS [9]. To our knowledge, no studies have quantified the healthcare resource and cost burden in women who are treated or untreated for menopause-related VMS. The primary objective of our study, therefore, was to estimate and compare all-cause HCRU and healthcare costs for treated versus untreated women with VMS. It was hypothesized that total 12-month follow-up healthcare costs, net of pharmacy expenditures, would be lower in treated patients than in untreated comparators. A secondary objective was to estimate and compare all-cause HCRU and healthcare costs for on- and off-label VMS treatment subgroups of women with VMS versus untreated women with VMS and was exploratory in nature.

Methods

Study Design and Data Source

A retrospective cohort study of US women enrolled in commercial health insurance plans was conducted using administrative claims data from the Optum Research Database from January, 1, 2012, through February 29, 2020. The Optum Research Database is a large, US-based repository of administrative claims data for > 111 million enrollees representing all 50 states, Washington, DC, and Puerto Rico. IRB approval was not sought because this study is based on previously collected data and does not contain any studies with human participants or animals performed by any of the authors.

Women were included in the analysis population using the first medical claim with a diagnosis code for VMS (index claim) during the identification period (January 1, 2013, through September 1, 2018). The date of the index claim was the index date. The baseline period was 12 months prior to the index date during which demographic and clinical characteristics were examined. To define treated and untreated cohorts, a 6-month treatment assessment period starting on the index date was specified. The follow-up period was 12 months after the end of the treatment assessment period during which differences in all-cause HCRU and costs were assessed (Fig. 1).

Fig. 1.

Study design. HCRU, healthcare resource utilization

VMS diagnosis claims were medical claims with International Classification of Diseases, 10th Revision, Clinical Modification (ICD-CM) codes in any position satisfying either of the following conditions during the identification period: (1) natural or surgical menopausal states (ICD-9-CM 627.2, 627.4/ICD-10-CM N95.1, E89.41) or (2) flushing (ICD-9-CM 782.62/ICD-10-CM R23.2) or hyperhidrosis (ICD-9-CM 780.08/ICD-10-CM R61) and ≥ 1 claim with diagnosis code for natural or surgical menopause or procedure code for surgical menopause on the same date or in the past 12 months. Females aged 40–63 years in the year of the index date who had ≥ 12 months of continuous enrollment in the health plan with medical and pharmacy benefits prior to index date and ≥ 18 months continuous enrollment starting on the index date were included. Women were excluded if they had missing or invalid demographic data; had any of the following claims in the baseline period: ≥ 1 VMS diagnosis claim, ≥ 1 VMS treatment claim, or a malignant solid tumor diagnosis claim; or had ≥ 1 pharmacy claim for an oral contraceptive or NuvaRing^® (Organon; Jersey City, NJ, USA) during the treatment assessment period.

Cohort Assignment

Cohorts (treated and untreated) were defined based on evidence of VMS treatment during the 6-month treatment assessment period after their VMS diagnosis date. Patients treated with a VMS treatment in the treatment assessment period were assigned to the treated cohort. VMS treatments included systemic HT, paroxetine 7.5 mg, selective serotonin reuptake inhibitor (SSRI; excluding paroxetine 7.5 mg), serotonin-norepinephrine reuptake inhibitor, clonidine, gabapentin, pregabalin, oxybutynin, and compounded estradiol pellet. Patients with no evidence of VMS treatment were assigned to the untreated cohort. The treated population was further divided into on- and off-label treatment subgroups for the secondary objective, which compared all-cause HCRU and costs for on-label-treated women with VMS and off-label-treated women with VMS versus untreated women with VMS (controls). We refer to the two products (HT and paroxetine mesylate 7.5 mg) approved for VMS by the Food and Drug Administration as on-label treatments. These were the only products approved for VMS during the study period. The following off-label treatments were included: clonidine, SSRIs (excluding paroxetine mesylate 7.5 mg), serotonin and norepinephrine reuptake inhibitors, gabapentin, pregabalin, oxybutynin, and compounded estradiol pellets. Each subgroup (i.e., on- or off-label treated) was compared with a propensity score (PS)-matched untreated control. All-cause HCRU and costs in each subgroup were analyzed using the same methods as described for the primary objective. Each analysis compared women who were treated for VMS to a matched control with VMS who did not receive treatment within the first 6 months of VMS diagnosis (i.e., the treatment assessment period) as depicted in Fig. 1.

Outcomes

HCRU and costs (means and standard deviations) were summarized for the 12-month baseline and follow-up periods, and standardized differences (SDIFF) were estimated between cohorts (treated versus untreated, primary objective; on-label and off-label versus untreated, secondary objective). HCRU included office visits (i.e., traditional outpatient visits), hospital outpatient visits, emergency department (ED) visits, inpatient stays, and inpatient days. Costs included office, hospital outpatient, ED visit, inpatient stay, and pharmacy costs. Costs were adjusted using the annual medical care component of the Consumer Price Index to reflect inflation to year 2020 [10]. Total costs were estimated using all health plan and patient-paid medical costs (i.e., office, hospital outpatient, ED, inpatient, other medical costs) and pharmacy costs. Other medical costs included costs from services delivered at other venues, such as independent laboratory or radiology facilities, at the patient’s home, or by ambulance personnel.

Statistical Analysis

All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA). PS matching, a widely used method in observational studies for causal inference, was conducted to control for possible confounding of the association between outcomes and treatment status. Covariates included in the PS were baseline demographics (i.e., age, age category, geographic region, and index year), baseline conditions and comorbidity burden (i.e., hysterectomy, postmenopausal bleeding, Quan-Charlson Comorbidity Score, sleep disturbances, anxiety, depression, neuropathy, fibromyalgia, migraine, hypertension, and Agency for Healthcare Research and Quality comorbid conditions), baseline healthcare utilization (i.e., all-cause ambulatory visits, all-cause office visits, all-cause outpatient visits, flag for ER, and flag for inpatient stay), and baseline costs (i.e., all-cause ambulatory costs, all-cause pharmacy costs). After PS models were finalized, treated women were PS matched to untreated controls using a 1:1 ratio within strata of baseline hypertension status (yes or no) by the logit of their PS. A caliper width of 0.20 times the standard deviation (SD) of their logit was used [11]. Treated women for whom a match was not found were excluded from the post–PS-match analysis. The success of the matching procedure was evaluated by comparing the demographic and baseline characteristics between cohorts using the SDIFF. The SDIFF expresses the difference in means in units of the pooled standard deviation [12]. An SDIFF of < |10%| for a measure was considered as an acceptable threshold for balance between the treatment cohorts.

Following the matching procedure, descriptive analyses of HCRU and cost outcomes were performed. The conventional significance level of α = 0.05 was used for all comparisons; however, due to the large sample, the SDIFF was estimated with SDIFF ≥ |10%| representing a potentially meaningful difference and SDIFF < |10%| a small difference [12]. A generalized linear model (GLM) estimated the total cost of care ratio, representing the difference in 12-month follow-up costs between cohorts being compared. A GLM with gamma distribution and log link was specified, and Wald z-tests using robust standard errors were used to calculate P values. Adequacy of the model’s mean–variance relationship was evaluated using the modified Park test [13].

Inclusive Language

In our report, we use gender-specific language, as reflected in the referenced publications and protocols. However, we recognize that some individuals who experience vasomotor symptoms related to menopause may identify differently from gender and pronouns used in this article.

Results

A total of 619,116 women were identified with ≥ 1 VMS diagnosis claim(s) during the identification period from January 1, 2013, to September 1, 2018. After applying inclusion/exclusion criteria, a total of 117,582 women were included in the study. During the treatment assessment period (i.e., 6 months following VMS diagnosis), 20.5% (n = 24,069) of women experiencing VMS initiated VMS treatments and 79.5% (n = 93,513) had no evidence of treatment in their claims. After PS matching, 24,057 of the 24,069 treated women were successfully matched to untreated controls (n = 24,057) and included in the analysis. Of treated women, 62.6% (n = 15,069) received on-label treatments and 37.4% (n = 8988) received off-label treatments (Fig. 2). The PS-matched sample was balanced on all measured baseline characteristics (SDIFF < |10%|). Of 48,114 PS-matched treated and untreated women, the mean (SD) age was 51.6 (4.8) years. Over 40% of women in the post-PS-matched sample were aged 50–54 years (n = 19,557), followed by 26% (n = 12,368) aged 45–49 years, and 20% (n = 9605) aged 55–59 years (Table 1).

Fig. 2.

Cohort identification. ^aOf the 24,069 treated women, 24,057 were successfully matched to 24,057 untreated controls. PS propensity score, VMS vasomotor symptoms

Table 1.

Demographic characteristics and index year post-PS matching

Demographics		Post-PS matching, treated (n = 24,057)	Post-PS matching, untreated (n = 24,057)	Post-PS matching treated vs untreated SDIFF, %
Age group, years
40–44	n	1915	1725	3.0
	%	8.0	7.2
45–49	n	5961	6407	− 4.2
	%	24.8	26.6
50–54	n	9923	9634	2.5
	%	41.3	40.1
55–59	n	4772	4833	− 0.6
	%	19.8	20.1
60–63	n	1486	1458	0.5
	%	6.2	6.1
Region
Northeast	n	1298	1229	1.3
	%	5.4	5.1
Midwest	n	6131	6092	0.4
	%	25.5	25.3
South	n	11,674	11,856	− 1.5
	%	48.5	49.3
West	n	4940	4868	0.7
	%	20.5	20.2
Other	n	14	12	0.4
	%	0.1	0.1
Index year
2013	n	5671	5673	− 0.02
	%	23.6	23.6
2014	n	4784	4800	− 0.2
	%	19.9	20.0
2015	n	3966	3961	0.1
	%	16.5	16.5
2016	n	3626	3648	− 0.3
	%	15.1	15.2
2017	n	3445	3421	0.3
	%	14.3	14.2
2018	n	2565	2554	0.2
	%	10.7	10.6

PS propensity score, SDIFF standardized difference

HCRU and Costs

After PS matching, the treated versus untreated cohorts were balanced on baseline HCRU, with all SDIFF < |10%|. HCRU remained similar between cohorts at follow-up, with no meaningful SDIFF observed.

The post-PS-matched treated versus untreated cohorts were balanced on baseline costs, with all SDIFF < |10%|. At follow-up, pharmacy ($487 versus $320, SDIFF 28.4%, P < 0.001) and total ($1803 versus $1536, SDIFF 12.6%, P < 0.001) costs were higher in the treated cohort (Fig. 3). The difference in total costs was driven by pharmacy costs; difference in medical costs was small ($1316 versus $1217, SDIFF 5.1%, P < 0.001). The GLM results were consistent with the descriptive results (i.e., costs were higher in the treated cohort), with total costs 7% higher in the treated cohort (total cost ratio 1.07, 95% CI 1.05–1.10, P < 0.001).

Fig. 3.

Follow-up VMS all-cause healthcare costs,^a treated vs untreated^b post-PS-matched, USD. ^aNot shown are emergency department, inpatient, and other medical costs due to minimal counts. ^bOf the 24,069 treated women, 24,057 were successfully matched to 24,057 untreated controls. ^cStandardized differences noted are meaningful, i.e., SDIFF ≥ 10%. PS propensity score, SDIFF standardized difference, USD US dollars, VMS vasomotor symptoms

Subgroup Analysis

On-Label Versus Untreated Controls

The post-PS–matched on-label subgroup and untreated controls were balanced on post-PS–matched baseline HCRU and costs. HCRU at follow-up remained similar, with no SDIFF ≥ |10%|. Follow-up all-cause total costs ($1816 versus $1541, SDIFF 12.6%, P < 0.001) and pharmacy costs ($546 versus $315, SDIFF 38.6%, P < 0.001) were higher in the on-label-treated than untreated cohort, as was similarly observed in the primary analysis of treated versus untreated controls (Fig. 4). The GLM estimated total costs were 3.5% higher in the VMS on-label cohort versus matched controls (cost ratio 1.04, 95% CI 1.0–1.1, P = 0.03).

Fig. 4.

Follow-up VMS all-cause healthcare costs^a by on-label^b and off-label treatment^c post-PS-matched, USD. ^aNot shown are emergency department, inpatient, and other medical costs due to minimal counts. ^bOf the 24,057 successfully matched treated cohort, 15,069 received on-label treatment and were successfully matched to 15,069 untreated controls. ^cOf the 24,057 successfully matched treated cohort, 8988 received off-label treatment and were successfully matched to 8988 untreated controls. ^dStandardized differences noted are meaningful, i.e., SDIFF ≥ 10%. PS propensity score, SDIFF standardized difference, USD US dollars, VMS vasomotor symptoms

Off-Label Versus Untreated Controls

Similar to the on-label subgroup, the off-label subgroup and untreated controls were balanced on post-PS–matched baseline HCRU and costs. HCRU at follow-up was higher in the off-label cohort for outpatient visits (4.2 versus 3.3, SDIFF 13.9%, P < 0.001) but was similar between cohorts for all other follow-up HCRU. Follow-up all-cause total costs ($1781 versus $1528, SDIFF 12.7%, P < 0.001), medical costs ($1393 versus $1201, SDIFF 10.4%, P < 0.001), and pharmacy costs ($388 versus $327, SDIFF 10.8%, P < 0.001) were all higher in the off-label cohort (Fig. 4). The GLM estimated total costs were 13.4% higher in the VMS off-label cohort versus matched controls (cost ratio 1.13, 95% CI 1.0–1.2, P < 0.001).

Discussion

In this retrospective study of administrative claims, 117,582 women with VMS diagnosis claims were identified during the 8-year period examined. Of those, only 20.5% initiated VMS treatment within 6 months of diagnosis. The primary objective of the study was to examine differences in HCRU and costs between treated and untreated women with VMS due to menopause. Treated women were PS-matched to untreated controls, and baseline differences in HCRU and costs were small between groups. At follow-up, HCRU differences remained small between groups. Total costs were higher in the treated group, mainly driven by pharmacy costs. Our secondary objective examined the on- and off-label subgroups of the treated cohort versus the untreated cohort. The only meaningful difference observed in follow-up HCRU was more hospital outpatient visits in the off-label subgroup. Similar patterns of higher total costs and pharmacy costs in the on-label subgroup as in the treated cohort overall were observed, suggesting higher pharmacy costs of the on-label treatments drive the higher total costs in the treated cohort. It is also important to note that on-label treatments are specifically approved by the United States Food and Drug Administration for women with moderate to severe VMS [8, 14, 15]. The off-label subgroup had approximately 13% higher total costs than untreated controls, driven mainly by outpatient and pharmacy costs. Since VMS severity was not captured in this analysis, it is possible the higher costs observed in the off-label subgroup were due to more severe VMS.

There is a paucity of available literature that examines the HCRU and costs of VMS due to menopause. Only one other database study was identified that examined incremental direct and indirect costs of untreated VMS [9]. That was a retrospective matched-cohort design using the Optum Health Reporting and Insights Database from 1999 to 2011, but in contrast to the current study, it compared untreated VMS to a control group without VMS. Sarrel et al. [9] found that untreated VMS was associated with higher HCRU, work loss, and cost burden compared with the non-VMS controls. However, when two cohorts with VMS are compared, as was done in the current study, the treated VMS cohort was found to have higher HCRU and costs when compared to the untreated VMS control cohort. Whiteley and colleagues conducted a survey of women contacted from the 2010 National Health and Wellness Survey that evaluated the impact of the presence and severity of VMS on health status, productivity, HCRU, and costs; 3267 women were included in the analysis [16]. They found that VMS was associated with greater costs, physician visits, presenteeism, and work impairment [16]. Those who had more severe VMS had worse outcomes, such as poorer health status, greater work impairment, and greater HCRU, than those with mild symptoms [16]. The current study adds to the limited knowledge regarding the costs of VMS by directly comparing HCRU and costs of treated versus untreated VMS.

Strengths and Limitations

A key strength of this study is the novelty of the research design, as few studies have examined economic impact of VMS due to menopause in a real-world setting. To our knowledge, this is the first to examine the difference in HCRU and costs of treated versus untreated VMS due to menopause using claims from a large database. Furthermore, we are unaware of any studies that directly measured VMS treatment costs for on- and off-label VMS therapies. There are a few limitations that are common within database research studies. The analytic sample included women in the US with commercial insurance coverage; therefore, the results reported may not be generalizable to women with other types of coverage (e.g., Medicaid) or those without health insurance. Because there are no ICD-CM codes for VMS severity, it is not captured in administrative claims databases and therefore cannot be controlled for in the study design or analysis. For this reason, severity of VMS could not be assessed by cohort and it is possible that the treated cohort may have had more-severe VMS than the untreated cohort. This may have masked treatment benefits, as more severe VMS might result in more HCRU and costs. Administrative claims data cannot capture the costs incurred by those who do not receive treatment for VMS, including clothing, appliances (e.g., fans, air conditioners), and products to create a cooler environment [17]. As a result, costs associated with untreated VMS may be underestimated. Finally, claims data only capture condition information if a patient seeks diagnosis or care; therefore, this study does not represent women with VMS who did not seek care or whose providers did not bill to the ICD codes specified for inclusion in the analysis.

Conclusion

Only 20% of women initiated treatment for VMS within 6 months following their diagnosis. This finding indicates that most women did not receive timely treatment for their VMS or chose not to seek treatment after their diagnosis. This may suggest that greater clinical intervention or education relating to the current treatment landscape of VMS due to menopause is needed. While both on- and off-label treatment increased the total cost of care compared with untreated controls, those increases were modest in magnitude and should not impede treatment for women who report symptom improvement as a result of treatment.

Author Contributions

Aki Shiozawa and Shayna Mancuso conceived the study design. Aki Shiozawa, Jennifer Friderici, Summer Tran, and Helen M. Trenz collected and assembled the data. Aki Shiozawa, Shayna Mancuso, Christopher Young, Jennifer Friderici, Summer Tran, and Helen M. Trenz analyzed and interpreted the data. Aki Shiozawa prepared the first draft of the manuscript. All authors contributed substantively to the review and revision of the manuscript, approved the final version of the manuscript for publication, and agreed to be accountable for all aspects of the work.

Funding

Sponsorship for this study and Rapid Service Fee were funded by Astellas Pharma, Inc. (Northbrook, IL).

Data Availability

Researchers may request access to anonymized participant-level data, survey-level data, and protocols from Astellas-sponsored clinical trials at www.clinicalstudydatarequest.com. For the Astellas criteria on data sharing see: https://clinicalstudydatarequest.com/Study-Sponsors/Study-Sponsors-Astellas.aspx.

Declarations

Conflict of Interest

Aki Shiozawa, Shayna Mancuso, and Christopher Young are employees of Astellas Pharma, Inc. Jennifer Friderici, Summer Tran, and Helen M. Trenz are employees of Optum, which received funding for the current study from Astellas Pharma, Inc.

Ethical Approval

This study is based on previously collected data and does not contain any studies with human participants or animals performed by any of the authors.