Factors Associated with Early In Vitro Fertilization Treatment Discontinuation

Abstract

Objective:

To investigate factors associated with early in vitro fertilization (IVF) treatment discontinuation.

Design:

Retrospective cohort study.

Setting:

High-volume Midwestern academic medical center.

Patient(s):

Six hundred sixty-nine first-attempt IVF patients who did not have a live birth.

Intervention(s):

None.

Main Outcome Measure(s):

Treatment discontinuation and time to return for a second IVF cycle.

Result(s):

Women without IVF insurance coverage were more likely to discontinue treatment than women with insurance coverage (adjusted odds ratio [aOR] 3.12, 95% confidence interval [CI] 2.22–4.40). African American women were more likely to discontinue treatment (aOR 2.95, 95% CI 1.54–5.66) and returned for treatment more slowly (adjusted hazard ratio [aHR] 0.44, 95% CI 0.28–0.71) than non-Hispanic white women, regardless of IVF insurance coverage or income. Women with a poor prognosis were more likely to discontinue treatment than women with a good prognosis. Older women with IVF insurance coverage or a good prognosis had a shorter time to return for a second IVF cycle than older women without IVF insurance coverage or with a poor prognosis. Estimated income, distance to clinic, fertility diagnosis, number of oocytes retrieved, and history of previous live birth were not associated with treatment discontinuation or time to return for a second IVF cycle after adjustment for covariates.

Conclusion(s):

IVF insurance coverage, race, age, and future treatment prognosis are associated with IVF treatment discontinuation and time to return.

Introduction

Although in vitro fertilization (IVF) has become increasingly desired, available, and effective in the United States and globally,(1, 2) between 17% and 60% of women discontinue treatment.(3–10) Given that the cumulative probability of having a live birth with IVF increases with repeated IVF cycles(11–14) and that approximately 40% of patients who discontinue treatment are not satisfied with their decision,(15) we must identify factors that contribute to treatment discontinuation to improve patient outcomes.

Although previous studies have examined factors associated with treatment discontinuation, most only included patients with IVF insurance coverage.(3–11, 16–24) Therefore, a significant gap in our understanding of IVF treatment discontinuation persists given that the majority of Americans do not have IVF insurance coverage(25) and have limited access to fertility treatment. (26)

The primary aim of this study was to investigate the association between IVF insurance coverage and treatment discontinuation over one year following any first IVF cycle that did not result in a live birth. As delaying treatment is associated with declining prognosis and success rate,(27) our secondary aim was to determine patient characteristics associated with time to return for a second IVF cycle. Our clinic is located on the border between Illinois, a state that mandates IVF insurance coverage, and Missouri, a state that does not mandate coverage. Therefore, we are ideally situated to investigate the association between IVF insurance coverage and IVF treatment discontinuation.

Material and Methods

Study Design

This is a secondary analysis of a retrospective cohort study of women who underwent IVF at Washington University School of Medicine from 2001 to 2014.(28, 29) In brief, women starting IVF treatment at our center between January 1, 2001 and December 31, 2010 were followed through December 31, 2014 to study the association between IVF insurance coverage and cumulative chance of a live birth over multiple IVF cycles.

Women were included in this analysis if their first IVF cycle at our center was unsuccessful. We defined an IVF cycle as controlled ovarian stimulation initiated for a fresh embryo transfer. Women who had previously undergone IVF at another clinic, who resided more than 100 miles from our clinic, or who used donor oocytes or gestational carriers were excluded from this analysis.

Patient demographics, IVF insurance status, and clinical characteristics were extracted from medical and billing records, as detailed previously.(28, 29) Variables of interest included state of residence, race, age at cycle start, infertility diagnosis, distance to clinic, estimated household income, parity, number of oocytes retrieved, whether or not excess embryos were cryopreserved, and whether the cycle resulted in a chemical or clinical pregnancy loss. Number of oocytes retrieved and cryopreservation of excess embryos were used as proxies for patient prognosis. Up to three infertility diagnoses could be listed for each patient. Distance from clinic was determined by using a patient’s ZIP code of residence. Estimated household income was defined as the annual average for a patient’s ZIP code area and was obtained from the U.S. Census Bureau’s 2010 American Community Study.(30) This study was approved by the Washington University Human Research Protection Office.

Outcomes of Interest

Outcomes of interest were treatment discontinuation and time to return for a second IVF cycle. Women were defined as discontinuing treatment if they did not return to our clinic within 365 days of an unsuccessful first cycle. Time to return was defined as the time between the start of the first IVF cycle and the second IVF cycle, which included fresh and frozen embryo transfers.

Statistical Analysis

Univariate associations with treatment discontinuation were assessed by chi-square or Fisher’s exact tests for categorical variables and t-test or Mann-Whitney U tests for continuous variables, as appropriate.

To evaluate predictors for discontinuing treatment, we performed a stepwise backwards elimination multiple logistic regression including the following characteristics: IVF insurance status, age, race, number of oocytes retrieved, whether or not excess embryos were cryopreserved, history of prior live birth, pregnancy loss after the first IVF cycle, distance to clinic, and estimated income. All two-way interactions between these main effects were assessed to determine how one variable modified the effect of another variable on discontinuation.(31)

A survival analysis using stepwise backwards elimination Cox proportional hazards regression was performed to analyze time to return. Women were censored at 366 days if they did not return within one year. Main effects and two-way interactions listed above were assessed for inclusion in this model.

For all analyses, P <0.05 was considered statistically significant. SAS Version 9.4 (SAS Institute Inc., Cary, NC) was used for all analyses.

Results

Patient Characteristics

A total of 669 women were included in this study, and 249 (37%) discontinued treatment within one year of their first failed cycle (Table 1). Women in our cohort were between 21 and 44 years of age. The majority (87%) identified as non-Hispanic white and 7% as African American. In our cohort, 56% of women had IVF insurance coverage, including 78% of women from Illinois and 32% of women from Missouri. Whereas 31% of women from Illinois discontinued treatment, 45% of women from Missouri discontinued. Women from Missouri had an 18% higher estimated median income than women from Illinois, and they lived closer to the clinic; women in Illinois had a median travel distance of 18 miles, whereas women in Missouri had a median travel distance of 7 miles. Within each state, women who had IVF insurance coverage had a 5% lower estimated median income than women without IVF insurance coverage. IVF insurance coverage also varied by race/ethnicity: 56% of non-Hispanic white women, 71% of African American women, and 37% of women who identified as another race or Hispanic had IVF insurance coverage. Twenty-five percent of women previously had a live birth, and of those, 74% had only one child. The most common infertility diagnoses were male factor (31%), ovulation disorder (24%), tubal factor (23%), and unexplained (22%).

Table 1:

Demographic and clinical characteristics by 1-year treatment continuation status

Variable	Discontinued Treatment (n=249)	Continued Treatment (n=420)	Unadjusted P-value
IVF insurance coverage	100 (40)	275 (65)	<0.001a
Age (years)	34.37 ± 4.95	33.50 ± 4.53	0.022
Race/Ethnicity			0.009
Non-Hispanic white	205(82)	376 (90)	0.008
African American	26 (10)	19(5)	0.003
Other	18 (7)	25 (6)	0.515
History of prior live birth	65 (26)	102 (24)	0.599
Had embryos cryopreserved	47 (19)	131 (31)	<0.001
Oocytes retrieved	10.51 ± 5.80	11.53 ± 6.30	0.036
First IVF cycle outcome			0.224
No pregnancy	204 (82)	359 (85)
Clinical or chemical loss	45 (18)	61 (15)
Fertility diagnosisb
Adhesions	21(8)	27 (6)	0.331
Ovulation Disorder	62 (25)	99 (24)	0.697
Endometriosis	42 (17)	69 (16)	0.883
Fibroids	10 (4)	10 (2)	0.246
Male Factor	69 (28)	139 (33)	0.146
Tubal Factor	60 (24)	92 (22)	0.513
Tubal Ligation	13 (5)	13 (3)	0.214
Unexplained	50 (20)	94 (22)	0.529
Other	12 (5)	14 (3)	0.408
Estimated income (thousand U.S. dollars)	62 (52–75)	62 (51–78)	0.955
Distance to clinic (miles)	34 (17–89)	30 (17–66)	0.268

Footnote:

Categorical data represented as n (%), and continuous variables represented as mean ± standard deviation or median (interquartile range) for skewed distribution.

^aP-value <0.05 between the two groups.

^bOne woman had an unknown fertility diagnosis. Up to three fertility diagnoses could be listed per patient.

Predictors of Treatment Discontinuation

Women without IVF insurance coverage had a three times higher odds of treatment discontinuation as women with IVF insurance coverage (aOR 3.12, 95%CI 2.22–4.40, P<0.001) (Table 2). Compared to non-Hispanic white women, African American women also had a three times higher odds of treatment discontinuation (aOR 2.95, 95% CI 1.54–5.66, P=0.001). Having no embryos cryopreserved was significantly associated with treatment discontinuation, and this association was more pronounced with age (interaction-P=0.002). For example, the aOR of treatment discontinuation for 35-year-old women without embryos cryopreserved was 3.02 (95% CI 1.81–5.04), whereas the aOR was 6.48 (95% CI 2.65–15.83) for 40-year-old women.

Table 2:

Univariate and multivariate logistic models for treatment discontinuation

	Unadjusted analysis				Adjusted analysis
Variable	B	OR	95 %CI for OR	P-Value	B	OR	95 %CI for OR	P-Value
No IVF insurance coveragea	1.04	2.83	2.05–3.91	<0.001b	1.14	3.12	2.22–4.40	<0.001
Age (years)	−0.04	1.04	1.01–1.08	0.02	−0.28	0.76	0.63–0.91	0.004
Race/Ethnicitya
African American	0.92	2.51	1.34–4.64	<0.001	1.08	2.95	1.54–5.66	0.001
Other	0.28	1.32	0.70–2.48	0.387	0.08	1.08	0.55–2.12	0.819
No prior live birtha	0.10	1.1	0.77–1.58	0.599
No embryos cryopreserveda	0.70	1.95	1.33–2.85	<0.001	−4.24	0.01	0.001–0.35	0.009
Oocytes retrieved	−0.03	0.97	0.95–0.998	0.037
Pregnancy loss after first IVF cyclea	0.26	1.3	0.85–1.98	0.225
Estimated income (in $10,000)	−0.020	0.99	0.94–1.04	0.558
Distance to clinic (in 25 miles)	−0.10	0.91	0.75–1.10	0.315
Interaction between no embryos cryopreserved and age					0.15			0.002

Note:

OR, odds; CI, confidence interval.

^aReference groups are as follows: for No IVF insurance coverage: IVF insurance coverage; for Race/Ethnicity: Non-Hispanic white; for No prior live birth: Prior live birth; for No embryos cryopreserved: Excess embryos cryopreserved; for Pregnancy loss during first IVF cycle: Did not become pregnant during first IVF cycle.

^bP-value <0.05 between the group and the reference group.

Other assessed interaction terms, such as between race and insurance, were not statistically significant. Furthermore, there were no statistically significant associations between treatment discontinuation and prior live birth, distance to clinic, estimated income, or pregnancy loss after the first IVF cycle. To address changes in discontinuation rates over time, period effects were assessed and were also not significant.

Time to Return for a Second IVF Cycle

For the women who continued treatment, the median time to return was 3.5 months, and 80% returned within 6 months. As with treatment discontinuation, IVF insurance coverage, race, cryopreserved embryos, and age were associated with time to return (Table 3). Pregnancy loss in the first IVF cycle was also associated with a longer time to return but not with treatment discontinuation.

Table 3:

Cox proportional hazard models for time to return

	Unadjusted analysis				Adjusted analysis
Variable	B	HR	95 %CI for HR	P-Value	B	HR	95 %CI for HR	P-Value
No IVF insurance coveragea	−0.65	0.52	0.43–0.64	<0.001b	0.90	2.45	0.53 – 11.42	0.253
Age (years)	−0.02	0.98	0.96–1.003	0.09	0.23	1.26	1.15 – 1.43	<0.001
Race/Ethnicitya
African American	−0.69	0.50	0.32–0.80	0.003	−0.81	0.44	0.28 – 0.71	<0.001
Other	−0.23	0.79	0.53–1.19	0.260	−0.15	0.95	0.57 – 1.30	0.472
No prior live birtha	0.002	1.002	0.80–1.25	0.987
No embryos cryopreserveda	−0.38	0.68	0.56–0.84	<0.001	2.44	11.49	1.93 – 68.40	0.007
Oocytes retrieved	0.02	1.02	1.001–1.03	0.036
Pregnancy loss after first IVF cyclea	−0.40	0.68	0.52–0.89	0.005	−0.47	0.63	0.48 – 0.83	<0.001
Estimated income (in $10,000)	0.002	1.00	0.97–1.03	0.940
Distance to clinic (in 25 miles)	0.04	1.04	0.93–1.16	0.476
Interaction between no embryos cryopreserved and agec					−0.09			0.002
Interaction between no IVF
insurance coverage and agec	−0.05							0.042

Note:

HR, hazard ratio; CI, confidence interval.

^aReference groups are as follows: for No IVF insurance coverage: IVF insurance coverage; for Race/Ethnicity: Non-Hispanic white; for No prior live birth: Prior live birth; for No embryos cryopreserved: Excess embryos cryopreserved; for Pregnancy loss during first IVF cycle: Did not become pregnant during first IVF cycle.

^bP-value <0.05 between the group and the reference group.

^cSee Table 4 for interaction-specific HR.

Age modified the effects of both IVF insurance coverage and having at least one embryo cryopreserved on time to return. Women without IVF insurance coverage had a longer time to return than women with insurance coverage. This effect became more pronounced with increasing age, as aHR was less than 1.0 and decreased with age (Table 4). Similarly, women over 27 had a longer time to return if they had no embryos cryopreserved than if they did, and age amplified this effect (Table 4).

Table 4:

Hazard ratios for interactions between age, IVF insurance coverage, and whether embryos were cryopreserved

Variable	aHR	95% CI for aHR
No IVF insurance coveragea
Age 25	0.76	0.49–1.19
Age 30	0.60	0.46–0.79
Age 35	0.48	0.39–0.59
Age 40	0.38	0.27–0.53
No embryos cryopreservedb
Age 25	1.27	0.80–2.04
Age 30	0.82	0.63–1.06
Age 35	0.53	0.41–0.68
Age 40	0.34	0.22–0.53

Note:

aHR, adjusted hazard ratio; CI, confidence interval.

^aReference group: IVF insurance coverage.

^bReference group: Excess embryos cryopreserved.

Race was significantly associated with time to return (P=0.002); African American women had a longer time to return than non-Hispanic white women (aHR 0.44, 95% CI 0.28–0.71, P<0.001). Furthermore, women who experienced a pregnancy loss in their first IVF cycle had a longer time to return than those who did not become pregnant during their first cycle (aHR 0.63, 95% CI 0.48–0.83, P<0.001).

Discussion:

This study examined the effects of IVF insurance coverage, socioeconomic factors, and cycle characteristics on both treatment discontinuation and time to return after a failed first IVF cycle. Consistent with previous work,(7, 23) 37% of women in our cohort did not return after their failed first IVF cycle. Women without IVF insurance coverage had a three times higher odds of treatment discontinuation than women with IVF insurance coverage, regardless of income. Given that cumulative live birth rates increase with number of IVF cycles, this finding emphasizes the importance of IVF insurance coverage in achieving a live birth.

Previous studies have found that older women are more likely to discontinue fertility treatment, and this trend has been largely attributed to the worsening of prognosis with age. (3–11, 13, 18, 21, 23, 24) In our study, having a poor prognosis was associated with a greater likelihood of discontinuing treatment, and this association was stronger with increasing age. Furthermore, having a good prognosis had a stronger effect on time to return with increasing age. This suggests that prognosis may affect women’s decisions to continue treatment differently at different ages. Similarly, the effect of IVF insurance coverage on time to return varied with age. These findings suggest a complex association between age and treatment discontinuation, which is supported by prior studies. For example, Troude et al. found a J-shaped curve with the lowest rate of discontinuation between 30 and 34 years,(7) and Rajkhowa et al. found that younger patients were more likely to discontinue treatment than older patients.(22)

Minority women are more likely to experience infertility than non-Hispanic white women, but they also are less likely to use assisted reproductive technology services.(32, 33) Two Department of Defense studies demonstrated that African American women undergo IVF treatment in equal proportion to white women when given equal access to care, suggesting that disparities arise from socioeconomic, rather than cultural, factors.(34, 35) However, these studies only examined women at their first IVF cycle, not whether they returned for subsequent cycles.

African American women in our study were the most likely racial group to have IVF insurance coverage. Nevertheless, they were less likely to return and returned more slowly than non-Hispanic white women irrespective of IVF insurance coverage, income, or distance to clinic. The majority of African American women in our cohort lived in Illinois, where the insurance mandate covers four IVF cycles.(25) Therefore, these women are presumed to have had coverage for additional cycles. Whereas African Americans comprise 18% of the St. Louis metropolitan area,(36) only 7% of the women in our cohort were African American. This percentage is within range of other studies, (37–39) which highlights the disparity in access to fertility treatment. Like the Department of Defense studies, our findings demonstrate the importance of IVF insurance coverage for African American women in accessing IVF services.(34, 35) However, our study suggests that factors other than IVF insurance coverage and income affect whether African American women return for a second IVF cycle. The relatively small sample size of non-white women in our cohort limits our ability to make broad conclusions, and thus future studies should examine the barriers that non-white women encounter in IVF treatment continuation.

The ability to afford IVF treatment is impacted by a variety of factors, including insurance coverage, income, and family financial responsibilities. In this study, estimated income, having previous children, and IVF insurance coverage were used as measures of affordability. Insurance coverage was the only factor that influenced treatment discontinuation or time to return, and neither estimated income nor having children modified these effects. This modeling result suggests that insurance status is the primary economic indicator for both treatment discontinuation and time to return for women who failed their first IVF treatment cycle.

2Several limitations of our study must be considered. First, we do not know patients’ reasons for discontinuing treatment; some patients may have discontinued treatment for reasons unrelated to the variables considered, such as conceiving spontaneously or psychological burden. While stated stress and emotional burden were not explicitly known for patients, we included variables associated with stress, such as insurance coverage, income, and prognosis.(9, 17, 21, 22) However, the relationship between psychological burden and discontinuation in uninsured populations warrants further investigation, and future studies should incorporate questionnaires to capture this information. Additionally, we were unable to determine whether patients who discontinued treatment at our center sought treatment at another clinic, which is a common limitation in single-center studies examining follow-up. Nevertheless, transferring care is a clinically important outcome, as continuity of care improves fertility-patient satisfaction.(40, 41)

Although using ZIP code level data as a proxy for patient income is well established, this method is not as precise as individual level data and thus may not capture the effect of disposable resources a patient has for IVF treatment. While our modeling assessed other factors that influence affordability, some factors, such as other financial constraints, were unknown. Additionally, our cohort was constrained to a Midwestern population between 2001 and 2010, which limits its generalizability.

We note three key strengths of this study. To our knowledge, this is the first study to examine factors associated with IVF treatment discontinuation and time to return for a second IVF cycle in a population with variable IVF insurance coverage. Furthermore, this is the first study to explicitly examine the association between IVF insurance status and either treatment discontinuation or time to return. Given that the majority of Americans do not have IVF insurance coverage, this study contributes significantly to our understanding of early IVF treatment discontinuation and its association with insurance coverage. Lastly, by incorporating interaction terms in addition to main effects, our multiple regression models more realistically represent women’s decision-making.

Conclusions:

In summary, women who did not have IVF insurance coverage were more likely to discontinue IVF treatment than women who had coverage. African American women were more likely to discontinue IVF treatment and returned more slowly than non-Hispanic white women regardless of insurance coverage or income. Women returned for a second IVF cycle more quickly if they had IVF insurance coverage or a good prognosis than if they did not, and these effects were stronger with increasing age.