Use of fertility treatment modalities in a large United States cohort of professional women.

Infertility influences millions of couples each year( 1 ), with recent national estimates of infertility ranging from 6.0%( 2 ) to 15.5%( 3 ). In 2010, over 147,000 in-vitro fertilization (IVF) procedures to treat infertility were administered in the United States( 4 ). While the number of women utilizing IVF treatment has steadily risen over the past three decades( 4 ), there is not a clear understanding of the prevalence and patterns of different fertility treatment modalities among women utilizing fertility treatments since there are no established registries for tracking such information. Using the Nurses’ Health Study II cohort, we evaluated patterns of fertility treatment modalities for the purposes of informing public health planning, estimating some of the costs associated with infertility, and evaluating the long-term health outcomes. The Nurses’ Health Study II is a unique data source with information on a variety of prescription fertility treatments reported by participants on regular questionnaires from 1993 through 2009. The population utilizing infertility treatments in the U.S. tends to be of high socio-economic and educational status ( 5 ), and thus, our population of health professionals may fairly well-represent use of fertility treatments by the general population of fertility patients. We present data from 10,036 women who reported use of fertility treatment from 1993-2009. We hypothesized that financial factors, such as household income and state mandated insurance coverage, as well as, biologic or temporal patterns such as parity, age, and decade of treatment will influence treatment utilization patterns and thus we conducted stratified analyses by these variables.

The Nurses’ Health Study II is a prospective cohort study which began in 1989, when 116,430 registered nurses aged 25-42 years returned a mailed questionnaire regarding their health and lifestyle. At recruitment, women lived in one of fourteen states: California, Connecticut, Indiana, Iowa, Kentucky, Massachusetts, Michigan, Missouri, New York, North Carolina, Ohio, Pennsylvania, South Carolina, and Texas; however, the participants have since moved to all 50 states. Follow-up questionnaires are sent biennially. The follow-up rate from the original cohort is 92%. The study is approved by the Institutional Review Board of Brigham and Women's Hospital. The current analysis is an investigation of patterns in reported utilization of fertility treatment in the Nurses' Health Study II. Follow-up for the current analysis began in June 1993, when participants were first asked about treatments for infertility; data are included from the 1993 through 2009 questionnaires, which covers the time period from June 1993 to May 2011 (when the 2009 questionnaire cycle ended). For the purpose of this manuscript, we defined fertility treatment as clomiphene, and gonadotropin injections (either with or without IUI, and IVF/ART). Women were asked if they had taken “clomiphene or gonadotropins (yes/no) to induce ovulation” over the previous two years, on six questionnaires in 1993, 1995, 1997, 2001, 2005, and 2009. Women who marked “yes” were then asked to separately provide the number of months of clomiphene and of gonadotropin usage since the previous mailed questionnaire. Additionally, in 2009, women were asked a more detailed question specifically regarding their lifetime history of gonadotropin injections to treat infertility, including specific information on the type of treatment in which gonadotropins had been used (ie, gonadotropin injection alone, gonadotropins for intrauterine insemination [IUI]; gonadotropins for in vitro fertilization [IVF]). With these data, we categorized women's use of each of four treatments, defined as report at any time from 1993 through 2011 of clomiphene, gonadotropin injections alone, gonadotropin injections with IUI, or gonadotropin injections with IVF. Although we thought it was unlikely that women would not accurately report their use of fertility treatments, since it is usually a major event in a woman's life, we evaluated the reliability and validity of self reported fertility treatment. We compared gonadotropin use reported on each of the regularly mailed questionnaires from 1993-2009 with the single item in 2009 regarding lifetime history of gonadotropin use; we found very high reliability of reporting (concordance > 84%) for the prospective reports versus the lifetime history question. In a validation study, we obtained medical records regarding fertility from 44 participants (with their signed permission); all of the records which mentioned fertility treatment (74% of the records) confirmed women's reported treatment, while the remaining records generally contained no information on specific treatments. We evaluated the distribution of characteristics of women in our cohort at the time of first report of fertility treatment using data from the biennial questionnaires. These characteristics included body mass index, cigarette smoking status (categorized as current, former, and never smoking), and race/ethnicity (categorized as white, black, American Indian, Asian, Hawaiian, multi-racial, other). We considered several biological, economic, and other factors that might modify treatment patterns, including parity in 2009 (as a proxy for severity of underlying infertility), infertility insurance coverage, household income, decade of initial treatment, and age at first reported fertility treatment. Specifically, information on pregnancies was collected on each biennial questionnaire; thus, updated information was available on parity, which was categorized in 2009 as nulliparous or parous. In 2001, women were asked about pre-tax annual household income. We created five categories: less than $50,000, $50,000-75,000, $75,000- $99,999, $100,000-150,000, and greater than $150,000. For assessing insurance coverage of fertility treatment, we utilized our data on women's state of residence at the time that they reported initial fertility treatment; women were considered from an un-insured state if their state did not have mandated coverage of fertility treatment of any kind during the year that they first reported fertility treatment during the 1993-2009 follow-up period. For assessing age at fertility treatment, we utilized data on women's age when they reported first fertility treatment. Women were then categorized according to SART (Society of Assisted Reproductive Technologies) guidelines of less than 35 years, 35-37 years, 38-40 years, and greater than 40 years old. For assessing time period of fertility treatment, we utilized the date when they first reported fertility treatment. Women were categorized as having had treatment during the 1990's or after 2000; 2000 represented the approximate mid-point of our study follow-up. In analyses comparing fertility treatment utilization across categories, we used a chi-square test to assess statistical differences between groups.

In our population of women reporting fertility treatment at anytime during follow-up (n=10,036) ( Table 1 ), mean age at first time of reported fertility treatment was 37.7 years (SD=4.3). The mean BMI at first time of reported fertility treatment was 25.4 kg/m2 (SD= 5.9), with 36.9% overweight or obese (>=25.0 kg/m2). At initial time of reported treatment, 7.8% of women reported current cigarette smoking. The cohort was primarily Caucasian, with 4.3% of women identifying as minority race or ethnicity. The majority of women lived in states without mandated insurance coverage of fertility treatments, with 16.2% of women having mandated coverage at the time of the initial report of fertility treatment. Overall, 13.1% of women had a household income <$50,000, 25.4% of women had household income between $50,000-75,000 per year, 21.9% of women had household incomes between $75,000 and 100,000, 23.8% had household income between $100,000 and 150,000 and 15.8% of women had household incomes greater than $150,000 per year. Approximately 27.9% of the women in our cohort who reported fertility treatment started treatment below age 35 years, 23.9% of women started treatment between the ages of 35 and 37 years old, 15.8% of women started treatment between 38-40 years old, and 32.4% of women started treatment at age greater than 40. 94.5% of our population reported starting treatment during the 1990's, although over 500 women (5.5%) started treatment after 2000. Approximately 79.5% of those using fertility treatments reported a term pregnancy as of the 2009 questionnaire. Of women who used fertility treatment, 89.6% reported having failed to become pregnant after more than 1 year of trying. 46.9% cited ovulatory disorder as a reason for the infertility. Approximately 20.7% reported spouse/partner as a reason for infertility, and 19.4% reported endometriosis. Approximately 9.9% of women reported menstrual cycle lengths of greater than 40 days, while 16.3% of women reported irregular menstrual cycles. When we evaluated use of each fertility treatment reported during our follow-up period, from 1993-2009 ( Table 2 ), most women (93.9%) had used clomiphene. In total, 12.7% of women had used gonadotropin injections alone (i.e., not as part of IUI or IVF), which was similar to the prevalence of gonadotropin injection for IUI treatment (10.9%) or IVF (11.0%). In addition to overall prevalence of each treatment, we also examined patterns of consecutive treatments ( Figure 1 ). Treatment with only clomiphene and no other subsequent medications/procedures was the most common, with 72.9% reporting clomiphene as their only form of fertility treatment. However, 22.3% of women who had used clomiphene subsequently utilized gonadotropin injections (i.e., gonadotropins alone, or as part of IUI or IVF). Slightly more patients followed clomiphene with IUI (6.5%) than with IVF (4.6%), and the most common treatment directly subsequent to clomiphene was gonadotropin injections alone, without IUI or IVF (11.2%). Moreover, of women treated with clomiphene, few subsequently used two or more additional treatment modalities: for example, 2.7% of clomiphene users followed with gonadotropin injections alone and then IUI; 0.5% with gonadotropin injections alone and then IVF; and 1.4% with gonadotropin injections alone and then IUI and IVF. As expected, few women (6.1 % of all women receiving fertility treatment) initiated treatment with gonadotropin injections (i.e., gonadotropin alone, or as part of IUI or IVF) ( Figure 1 ). Among women starting treatment with gonadotropins as their first modality, a similar number utilized injections alone initially (36.0%) as utilized IUI initially (31.0%) or IVF initially (33.0%). We examined several variables which might affect these treatment patterns ( Table 3 ). First, since financial ability may affect treatment choices, we separately considered women whose states mandated insurance coverage of fertility treatment and women of varying household incomes ( Table 3 ). The 16.2% of women in states with mandated insurance for fertility treatment were more likely to utilize treatments with gonadotropins than women living in states without mandated coverage. For example, 16.6% of women in states with treatment insurance reported IVF compared to 9.9% of women living in other states (P <0.0001) ( Table 3 ). A similar pattern of increased IUI utilization was seen among those in states with mandated insurance coverage compared to those in other states (P< 0.0001). Additionally, women with higher household income were more likely to utilize gonadotropin treatments. For example, report of IVF varied from 17.7% of women with a household income greater than $150,000 to 8.4% of those with a household income less than $50,000 (P <0.0001) ( Table 3 ). A similar pattern of increased IUI utilization was seen with increasing household income (P= 0.03). In additional analyses ( Table 3 ), we used parity in 2009 as a surrogate for underlying severity of infertility and examined whether treatments may differ in women who remained nulliparous (ie, never achieved a full-term pregnancy). As expected, nulliparous women utilized gonadotropin treatments more commonly than other women; 20.2% of nulliparous women reported IVF treatment, 18.0% reported IUI treatments and 16.4% reported gonadotropin alone compared to 8.7%, 9.1%, and 11.7%, respectively, among parous women (all P <0.0001). Age at time of initial reported treatment also influenced treatment utilization patterns ( Table 3 ). Older women were less likely than younger women to report treatment with gonadotropins; 7.6% of women above the age of 40 used IVF compared to 11.9% of women who started treatment before the age of 35 (P <0.0001). A similar pattern of increased IUI utilization was seen with decreasing age at reported treatment (P <0.0001). Lastly, time period of treatment appeared to modify treatment modalities ( Table 3 ). Among women initiating treatment during the 1990's, 97.2% of women undergoing fertility treatment used clomiphene, compared to 83.0% of women undergoing treatment after 2000 (P<0.0001). Women initiating treatment after 2000 were more likely to use gonadotropin injections alone and in combination with IUI and IVF. For example, 26.9% of women who were treated after 2000 used IVF compared to 9.5% of women treated in the 1990's (P <0.0001); IUI was also used less in the 1990's as compared to after 2000 (P <0.0001). Among this large cohort of health professionals reporting fertility treatment from 1993-2009, the vast majority used clomiphene for treatment. Of women reporting use of gonadotropin injections, similar utilization was reported of injections alone (i.e., without IUI or IVF) compared to injections as part of IUI or IVF procedures. To our knowledge, these are the first data to assess use of fertility treatments in this detail in a large population of women. We found that financial status, such as household income or insurance for fertility procedures, was associated with treatment patterns to a similar extent as reproductive history and age of patient (i.e., variables that likely represent more biologic factors). For example, IVF procedures were approximately twice as prevalent in women with higher household income, in women living in states with mandated insurance for fertility treatment, women under 36, women treated after 2000, and nulliparous women, than in their counterparts. Our findings across all these variables suggest that many factors may influence treatment choices besides medical necessity– which should be considered further by physicians, patients, and the broader healthcare community. Our findings on financial determinants of treatment utilization are consistent with results from studies that have found that household income( 6 ) and mandated insurance coverage were associated with increased utilization of IVF and IUI. ( 6 - 8 ) Our strongest finding, surprisingly, was that women who first reported treatment after 2000 were less likely to use clomiphene than women treated in the 1990's; however, this finding should be interpreted with caution, since these women represent less than 10% of our study population. We are not sure what might lead to large differences in treatment patterns across time since the types of treatments and their success have not changed drastically over this time period.( 9 ) It is possible that the decreased use of clomiphene may be due to a shift from generalists to specialists as providers, or changes in treatment seeking behaviors of fertility patients. In addition, we found a somewhat higher prevalence of tubal factor infertility in our cohort in the later time period (before 2000: 8.3% reported tubal factor; after 2000: 12%), which might partially explain the lower use of clomiphene; however, this could not completely explain the approximately 2- to 3-fold higher prevalence of gonadotropin use in the later time period. Certainly, the average age at initiation was greater in the women who initiated treatment after 2000 (mean age=37.5 years for women treated in 1990's, and mean age=40.6 for women treated after 2000); however, since we found overall that older women were less likely to use gonadotropins than younger women (see table 3 ; for example, in women >40 years, 7.6% utilized IVF and in women <35 years, 11.9% utilized IVF), this age difference could not explain the observed findings. Our study has several strengths. The large sample of women allows detailed estimation of the prevalence of different treatment modalities and patterns, including in several subgroups. There are also limitations. Treatment data are self-reported, which may result in misclassification of treatment status. However, this is unlikely to meaningfully influence our results since we found high reliability and validity of the self-reported treatments. Generalizability of results may also be an issue. Our population receiving care is predominately Caucasian and may not represent utilization patterns of different racial/ethnic groups.( 10 ) Participants are all nurses, and may have different utilization of fertility treatments than the larger population of infertile women. It is impossible to determine to what extent this may have influenced our findings; however, the general population of women treated for infertility tends to be of high socioeconomic status and educational attainment, thus study participants may not dramatically differ from other women treated for infertility in the United States. Nonetheless, our findings should be generalized with caution, specifically to under-reprepresented racial and ethnic groups. In addition, since the cohort is comprised of women originally from 14 states in the US, rather than all states, the findings may not apply to the overall United States population. However the cohort does represent women from both urban and rural environments and geographically diverse states across the East, West, South and North regions of the United States. Finally, our subgroup analyses of variables which may alter treatment patterns (i.e., income, parity in 2009, etc.) may not be representative of the true underlying construct of the ability to pay for treatment and severity of infertility. Specifically, use of parity in 2009 as marker for infertility severity may accurately quantify our cohort's lifetime history of live births, but it does not take into account many factors that may influence one's choice to become pregnant or to maintain pregnancy; nonetheless, these are difficult variables to measure perfectly, and those findings should also be interpreted with caution. In conclusion, these data represent an important step in beginning to evaluate utilization of different fertility treatments in the United States. Future research should be conducted to examine racial disparities in fertility treatment. To our knowledge, this represents the first study of fertility treatment patterns in the US, and could be helpful in public health planning and in better estimating potential costs associated with infertility.