Infertility treatment and risk of severe maternal morbidity: a propensity score-matched cohort study.

BACKGROUND: The extent to which infertility treatment predicts severe maternal morbidity is not well known. We examined the association between infertility treatment and severe maternal morbidity in pregnancy and the postpar- tum period.

We conducted a cohort study using population-based registries from Ontario between 2006 and 2012. Preg- nancies achieved using infertility treat- ment (ovulation induction, intrauterine insemination or in vitro fertilization with or without intracytoplasmic sperm injec- tion) were compared with unassisted pregnancies using propensity score matching, based on demographic, reproductive and obstetric factors. The primary outcome was a validated com- posite of severe maternal morbidity or maternal death from 20 weeks’ gestation to 42  days postpartum. We also calcu- lated the odds ratio of a woman having 1, 2, or 3  or more severe maternal morbid- ity indicators in relation to invasive (e.g., in vitro fertilization) or noninvasive (e.g., intrauterine insemination) infertility treatment.

We matched 11 546 infertility treatment pregnancies with 47 553 untreated pregnancies. Severe mater- nal morbidity or maternal death occurred in 356 infertility-treated preg- nancies (30.8 per 1000  deliveries) ver- sus 1054 untreated pregnancies (22.2 per 1000  deliveries); relative risk 1.39 (95% confidence interval [CI] 1.23– 1.56). The likelihood of a woman hav- ing 3 or more severe maternal morbid- ity indicators was increased in women who received invasive infertility treat- ment (odds ratio [OR] 2.28, 95%  CI 1.56–3.33) but not in those who received noninvasive infertility treat- ment (OR 0.99, 95% CI 0.57–1.72). INTERPRETATION: Women who undergo infertility treatment, particularly in vitro fertilization, are at somewhat higher risk of severe maternal morbidity or death. Efforts are needed to identify patient- and treatment-specific predic- tors of severe maternal morbidity that may influence the type of treatment a woman is offered. RESEARCH CMAJ | FEBRUARY 4, 2019 | VOLUME 191 | ISSUE 5 E119 Severe maternal morbidity refers to a broad set of conditions that identify women who experience a near-fatal event during, or within 42 days of, a pregnancy. Severe maternal morbidity is more common than maternal mortality: rates in Canada are about 15 per 1000 and 10 per 100  000 hospital deliveries, respectively. 10,11 Thus, severe maternal morbidity is the preferred indicator for assessing risk factors or interventions that can influence maternal health.11 The association between assisted reproductive technolo- gies and severe maternal morbidity has been investigated recently in 3  studies in the United States, each reporting an approximate doubling of the risk of severe maternal morbidity among women with pregnancies conceived through assisted reproductive tech- nologies compared to those with pregnancies conceived with- out.12–14 However, we are not aware of any studies that have assessed whether severe maternal morbidity risk varies by type of infertility treatment while simultaneously addressing confounding by indication. Additionally, it is unclear if women who conceived using infertility treatment have more indicators of severe maternal morbidity (i.e., a marker of greater severity). Our aim was to address these questions within a large Canadian population.

We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline for reporting of observational studies.15 Study population and data sources We considered women aged 18–60 years with a hospital delivery, who were residents of Ontario and registered within Ontario’s universal health insurance program (OHIP). We excluded women whose pregnancy ended as an ectopic pregnancy, induced abor- tion or miscarriage, and randomly sampled 1  live birth or still- birth at 20 or more weeks’ gestation per woman (Figure  1). To address potential confounding by indication, we matched about 5 untreated pregnancies to each infertility treatment pregnancy using a propensity score. The propensity score was generated by including all baseline characteristics in a logistic regression model; pregnancies were matched without replacement using the greedy method, with a caliper width score of ± 0.2 standard deviations.16 We identified all live births and stillbirths in Ontario between Apr. 1, 2006, and Mar. 31, 2012, within the Better Outcomes Regis- try & Network (BORN) Ontario legacy data sets (www.bornontario. ca/en/data/data-dictionary/legacy-datasets/). Linked data beyond fiscal year 2012 were unavailable in the registry at the time of this study. This registry captures more than 99% of hospi- tal births in the province, and has been previously validated for data completeness and accuracy. 17 We used the Registered Per- sons Database, the Permanent Resident Database (Immigration, Refugees and Citizenship Canada) and the Discharge Abstract Database (Canadian Institute for Health Information) to capture maternal demographics, preexisting health conditions, and All hospital births in Ontario from 2006 to 2012 N = 813 719 Excluded n = 42 146 • Invalid record number n = 31 274 • Maternal age 60 yr, non-Ontario resident or gestational age at birth < 20 weeks n = 10 775 • Ectopic pregnancy or therapeutic abortion n = 97 Random selection of 1 birth per woman n = 771 573 A/f_ter propensity score matching n = 11 546 • Invasive treatment n = 5974 • Noninvasive treatment n = 5572 A/f_ter propensity score matching n = 47 553 Excluded n = 311 410 • Repeat births to same woman n = 182 254 • Unknown infertility treatment n = 2134 • Invalid birth admission to hospital n = 127 022 No infertility treatment n = 448 198 n = 460 163 Infertility treatment n = 11 965 Figure 1: Flow diagram for the creation of the propensity score–matched study cohort. RESEARCH E120 CMAJ | FEBRUARY 4, 2019 | VOLUME 191 | ISSUE 5 diagnoses and procedures documented during an admission to hospital (Appendix  1, available at www.cmaj.ca/lookup/suppl/ doi:10.1503/cmaj.181124/-/DC1). The data sets were linked using unique encoded identifiers and analyzed at ICES. Diagnostic codes were based on the Canadian version of the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10-CA), and procedural codes were based on the Canadian Classification of Health Interventions (CCI). Exposure and outcome variables We categorized a pregnancy as “exposed” if conceived using any form of infertility treatment, including in vitro fertilization and related intracytoplasmic sperm injection, intrauterine insemina- tion or ovulation induction alone. In addition, we assessed the association between infertility treatment and the outcome of severe maternal morbidity, stratifying by invasive treatment (in vitro fertilization with or without intracytoplasmic sperm injec- tion) or noninvasive treatment (intrauterine insemination or ovu- lation induction alone). The primary study outcome was the occurrence of a compos- ite severe maternal morbidity or maternal death between 20 weeks’ gestation and 42 days after hospital discharge follow- ing delivery. The list of conditions included in the composite, developed by the Canadian Perinatal Surveillance System 18 and recently updated, comprises 44  unique indicators of severe maternal morbidity (Appendix  1). Previous validation of this composite measure included assessments of case fatality rates and maternal length of hospital stay greater than 7  days for each component indicator. 18 Secondary outcomes included the 5  most common individual indicators of severe maternal mor- bidity and all-cause maternal mortality between 42 and 365 days postpartum. Statistical analysis We used standardized differences to compare assisted and unas- sisted pregnancies, with a value of greater than 0.10 suggesting a meaningful difference. 19 Rates of all outcomes were expressed per 1000 deliveries. We used modified Poisson regression to gen- erate relative risks (RRs) and 95% confidence intervals (CIs) for severe maternal morbidity or death, comparing the treated group with their counterparts who conceived without infertility treatment (the reference group). We identified the 5  most com - mon indicators of severe maternal morbidity in the study popu- lation and calculated the RR for each. In addition to examining overall infertility treatment, we estimated RRs for noninvasive (intrauterine insemination and ovulation induction) and invasive treatment (in vitro fertilization with or without intracytoplasmic sperm injection) separately. The main model was further strati - fied by maternal age (< 40 or ≥  40 yr) and by singleton or multiple gestation, with propensity score rematching in each stratum. We considered women with more than 1 indicator of severe maternal morbidity to have more severe morbidity and a higher likelihood of death than women with a single indicator. Accord- ingly, we used multinomial logistic regression (accounting for matching by whether pregnancy had no infertility treatment) to estimate the odds ratio (OR) of a woman having 1, 2, or 3 or more indicators of severe maternal morbidity (compared with none), in relation to infertility treatment versus no infertility treatment. We repeated the same modelling approach when comparing treatment subtype (invasive v. noninvasive infertility treatment) versus no treatment (the referent). In these models, death was not included in the outcome. Ethics approval The use of data in this project was authorized under section 45 of Ontario’s Personal Health Information Protection Act, which does not require review by a research ethics board.

We identified 813 719 births in Ontario during the study period (Figure  1). After exclusions, and before matching, there were 11 965 women with a pregnancy conceived through infertility treatment and 448  198 women with a pregnancy conceived without treatment (Figure 1, Table 1). The women given infertil- ity treatment were older and had higher incomes, and a greater proportion were nulliparous or pregnant with multiple ges tation (Table 1). After propensity score matching, there were 11 546 pregnancies that resulted from infertility treatment and 47 553 untreated pregnancies, with minimal differences in char- acteristics between the 2  matched groups (Figure  1, Table  1, Appendix  3, available at www.cmaj.ca/lookup/suppl/ doi:10.1503/cmaj.181124/-/DC1). In the study cohort before propensity score matching, there were 387 pregnancies achieved through infertility treatment that were affected by severe maternal morbidity or death (32.3 per 1000), compared with 6689 cases of severe maternal morbidity or death in the untreated group (14.9 per 1000), yielding a crude RR of 2.17 (95%  CI 1.96–2.40). Death at 42  days postpartum to 365 days postpartum occurred in 6 women who were given infer- tility treatment (rate 0.3 per 1000) and in 97 women who did not (0.2 per 1000), a crude RR of 1.54 (95% CI 0.57–4.20). In the propensity-matched cohort, severe maternal morbid- ity or death occurred in 356 pregnancies resulting from infertil- ity treatment (30.8 per 1000) and in 1054 untreated pregnancies (22.2 per 1000); adjusted RR  1.39 (95%  CI 1.23–1.56; Figure  2). Severe maternal morbidity or death occurred in 121  pregnan- cies achieved through noninvasive treatment and 235 pregnan- cies achieved through invasive treatment (21.7 and 39.3 per 1000, respectively), yielding adjusted RRs of 0.98 (95% CI 0.81– 1.18) and 1.77 (95% CI 1.54–2.03), respectively, compared with untreated pregnancies (Figure  3). The absolute risk of severe maternal morbidity or death in relation to use of infertility treatment was more pronounced in those 40  years and older and in those with a multiple pregnancy (Table 2). However, rel- ative rates of severe maternal morbidity or death by treatment status were similar among women less than 40 versus 40 or more years of age, and among women with singleton versus multiple pregnancies. In the matched cohort, death from 20 weeks’ gestation up to 42  days postpartum occurred in fewer than 6  women who used infertility treatment (≤ 0.5 per 1000), and in fewer than 6 women RESEARCH CMAJ | FEBRUARY 4, 2019 | VOLUME 191 | ISSUE 5 E121 Table 1 (part 1 of 2): Baseline characteristics of the cohort before and after propensity score matching Characteristic Unmatched cohort Propensity score–matched cohort No. (%)* of women with infertility treatment n = 11 965 No. (%)* of women with no infertility treatment n = 448 198 Standardized difference No. (%)* of women with infertility treatment n = 11 546 No. (%)* of women with no infertility treatment n = 47 553 Standardized difference Year 2006 1225 (10.2) 62 249 (13.9) 0.11 1197 (10.4) 5055 (10.6) 0.01 2007 1387 (11.6) 70 515 (15.7) 0.12 1351 (11.7) 5712 (12.0) 0.01 2008 1647 (13.8) 73 990 (16.5) 0.08 1609 (13.9) 6882 (14.5) 0.02 2009 2150 (18.0) 78 806 (17.6) 0.01 2066 (17.9) 8148 (17.1) 0.02 2010 2708 (22.6) 80 154 (17.9) 0.12 2588 (22.4) 10 637 (22.4) 0.0 2011 2848 (23.8) 82 484 (18.4) 0.13 2735 (23.7) 11 119 (23.4) 0.01 Maternal age at index pregnancy, yr Mean ± SD 34.2 ± 4.9 30.1 ± 5.5 0.79 34.0 ± 4.8 33.6 ± 5.0 0.09 18–24 213 (1.8) 74 300 (16.6) 0.53 213 (1.8) 1791 (3.8) 0.12 25–34 6226 (52.0) 277 239 (61.9) 0.20 6115 (53.0) 24 787 (52.1) 0.02 35–39 3913 (32.7) 79 466 (17.7) 0.35 3758 (32.5) 15 214 (32.0) 0.01 40–44 1327 (11.1) 16 489 (3.7) 0.29 1224 (10.6) 5422 (11.4) 0.03 ≥ 45 286 (2.4) 704 (0.2) 0.20 236 (2.0) 339 (0.7) 0.11 Neighbourhood income quintile† 1 (lowest) 1370 (11.5) 100 445 (22.4) 0.30 1343 (11.6) 5797 (12.2) 0.02 2 1952 (16.3) 90 596 (20.2) 0.10 1903 (16.5) 7970 (16.8) 0.01 3 2598 (21.7) 92 896 (20.7) 0.02 2509 (21.7) 10 480 (22.0) 0.01 4 3231 (27.0) 92 285 (20.6) 0.15 3104 (26.9) 12 653 (26.6) 0.01 5 (highest) 2789 (23.3) 69 907 (15.6) 0.20 2663 (23.1) 10 541 (22.2) 0.02 Maternal world region of origin†‡ Africa 134 (1.1) 8 281 (1.8) 0.06 129 (1.1) 534 (1.1) 0.0 Canada 8967 (74.9) 323 062 (72.1) 0.06 8650 (74.9) 35 591 (74.8) 0.0 Caribbean 102 (0.9) 7504 (1.7) 0.07 98 (0.8) 414 (0.9) 0.0 East Asia 444 (3.7) 27 705 (6.2) 0.11 437 (3.8) 1882 (4.0) 0.01 Hispanic America 233 (1.9) 9945 (2.2) 0.02 223 (1.9) 943 (2.0) 0.0 Middle East 352 (2.9) 12 168 (2.7) 0.01 337 (2.9) 1363 (2.9) 0.0 South Asia 1080 (9.0) 40 382 (9.0) 0 1039 (9.0) 4324 (9.1) 0.0 Western nations and Europe 653 (5.5) 19 122 (4.3) 0.06 633 (5.5) 2502 (5.3) 0.01 Comorbidities BMI, median (IQR), kg/m2§ 25 (22–30) 25 (22–29) 0.08 25 (22–30) 25 (22–30) 0.09 Obesity at any visit (BMI > 30 kg/m 2) 731 (6.1) 22 383 (5.0) 0.05 716 (6.2) 2842 (6.0) 0.01 Cardiac disease¶ 411 (3.4) 12 177 (2.7) 0.04 394 (3.4) 1511 (3.2) 0.01 Stroke 12 (0.1) 403 (0.1) 0 12 (0.1) 40 (0.1) 0.01 Hypertension 701 (5.9) 16 267 (3.6) 0.11 657 (5.7) 2515 (5.3) 0.02 Diabetes mellitus 1478 (12.4) 33 524 (7.5) 0.16 1412 (12.2) 5342 (11.2) 0.03 Renal disease 17 (0.1) 448 (0.1) 0.01 16 (0.1) 54 (0.1) 0.01 HIV 9 (0.1) 411 (0.1) 0.01 9 (0.1) 32 (0.1) 0.0 Tobacco use** 408 (3.4) 54 507 (12.2) 0.33 406 (3.5) 1820 (3.8) 0.02 RESEARCH E122 CMAJ | FEBRUARY 4, 2019 | VOLUME 191 | ISSUE 5 who did not (≤ 0.1 per 1000), an adjusted RR of 4.12 (95% CI 1.19 to 14.22). Death at 42  days postpartum to 365  days postpartum occurred in fewer than 6  women who received infertility treat- ment (≤ 0.5 per 1000) and in 11 women who did not receive treat- ment (0.2 per 1000), an adjusted RR of 1.12 (95%  CI 0.31–4.03) (data not shown). The most common individual indicators of severe maternal morbidity were (in descending order) severe postpartum hemor- rhage (i.e., requiring transfusion of red blood cells or other inter- ventions), maternal admission to an intensive care unit (ICU), puerperal sepsis, hysterectomy and cardiac conditions (Figures 2 and 3). Each was more common in pregnancies achieved through infertility treatment than in untreated pregnancies (Figure  2). Significant associations between infertility treatment and the 3  most common indicators of severe maternal morbidity were observed among pregnancies achieved through invasive treat- ment, whereas no statistically significant associations were observed for noninvasive treatment (Figure 3). The odds of having a greater number of indicators of severe maternal morbidity were higher among women who used infer- tility treatment compared with those who did not (Table 3). For example, the adjusted OR for having 3 or more indicators was 1.65 (95% CI 1.18–2.30). Furthermore, women who received inva- sive treatment had the highest OR for having 3 or more indicators Table 1 (part 2 of 2): Baseline characteristics of the cohort before and after propensity score matching Characteristic Unmatched cohort Propensity score–matched cohort No. (%)* of women with infertility treatment n = 11 965 No. (%)* of women with no infertility treatment n = 448 198 Standardized difference No. (%)* of women with infertility treatment n = 11 546 No. (%)* of women with no infertility treatment n = 47 553 Standardized difference Major psychiatric condition 1967 (16.4) 73 967 (16.5) 0 1887 (16.3) 7376 (15.5) 0.02 Endometriosis 139 (1.2) 1435 (0.3) 0.1 133 (1.2) 301 (0.6) 0.06 Polycystic ovarian syndrome 6 (0.1) 15 (0.0) 0.03 ≤ 5 (0.0) 6 (0.0) 0.0 Pregnancy characteristics Parity† 0 8543 (71.4) 209 443 (46.7) 0.52 8164 (70.7) 32 630 (68.6) 0.05 1 2786 (23.3) 154 585 (34.5) 0.25 2748 (23.8) 12 061 (25.4) 0.04 ≥ 2 635 (5.3) 84 095 (18.8) 0.42 633 (5.5) 2861 (6.0) 0.02 No. of previous live births, median (IQR) 0 (0–1) 1 (0–1) 0.57 0 (0–1) 0 (0–1) 0.05 No. of previous miscarriages† 0 8431 (70.5) 348 836 (77.8) 0.17 8162 (70.7) 34 234 (72.0) 0.03 1–2 3022 (25.3) 90 648 (20.2) 0.12 2903 (25.1) 11 596 (24.4) 0.02 ≥ 3 511 (4.3) 8618 (1.9) 0.14 480 (4.2) 1722 (3.6) 0.03 Multiple gestation in the index pregnancy 3025 (25.3) 14 462 (3.2) 0.66 2663 (23.1) 7115 (15.0) 0.21 Type of infertility treatment in the index pregnancy Intrauterine insemination 2892 (24.2) 0 (0.0) 0.8 2778 (24.1) 0 (0.0) – Ovulation induction 2820 (23.6) 0 (0.0) 0.79 2794 (24.2) 0 (0.0) – In vitro fertilization 5630 (47.1) 0 (0.0) 1.33 5366 (46.5) 0 (0.0) – Intracytoplasmic sperm injection 623 (5.2) 0 (0.0) 0.33 608 (5.3) 0 (0.0) – No. of antenatal visits in the index pregnancy, median (IQR) 22 (17–29) 16 (13–20) 0.87 22 (17–28) 19 (16–25) 0.25 Note: BMI = body mass index, IQR = interquartile range, SD = standard deviation. *Unless specified otherwise. †Data missing in < 1%. ‡Appendix 2 (available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.181124/-/DC1) contains the definitions of each of the maternal world regions of origin. §Data were available for BMI for 383 543 pregnancies in the unmatched cohort and 47 939 pregnancies in the matched cohort. ¶Congestive heart failure, congenital heart disease, coronary artery disease, cardiac dysthymia, chronic rheumatic heart disease or previous myocardial infarction. **Defined as any use (yes/no) as of 20 weeks’ gestation. RESEARCH CMAJ | FEBRUARY 4, 2019 | VOLUME 191 | ISSUE 5 E123 Table 1 (part 2 of 2): Baseline characteristics of the cohort before and after propensity score matching Characteristic Unmatched cohort Propensity score–matched cohort No. (%)* of women with infertility treatment n = 11 965 No. (%)* of women with no infertility treatment n = 448 198 Standardized difference No. (%)* of women with infertility treatment n = 11 546 No. (%)* of women with no infertility treatment n = 47 553 Standardized difference Major psychiatric condition 1967 (16.4) 73 967 (16.5) 0 1887 (16.3) 7376 (15.5) 0.02 Endometriosis 139 (1.2) 1435 (0.3) 0.1 133 (1.2) 301 (0.6) 0.06 Polycystic ovarian syndrome 6 (0.1) 15 (0.0) 0.03 ≤ 5 (0.0) 6 (0.0) 0.0 Pregnancy characteristics Parity† 0 8543 (71.4) 209 443 (46.7) 0.52 8164 (70.7) 32 630 (68.6) 0.05 1 2786 (23.3) 154 585 (34.5) 0.25 2748 (23.8) 12 061 (25.4) 0.04 ≥ 2 635 (5.3) 84 095 (18.8) 0.42 633 (5.5) 2861 (6.0) 0.02 No. of previous live births, median (IQR) 0 (0–1) 1 (0–1) 0.57 0 (0–1) 0 (0–1) 0.05 No. of previous miscarriages† 0 8431 (70.5) 348 836 (77.8) 0.17 8162 (70.7) 34 234 (72.0) 0.03 1–2 3022 (25.3) 90 648 (20.2) 0.12 2903 (25.1) 11 596 (24.4) 0.02 ≥ 3 511 (4.3) 8618 (1.9) 0.14 480 (4.2) 1722 (3.6) 0.03 Multiple gestation in the index pregnancy 3025 (25.3) 14 462 (3.2) 0.66 2663 (23.1) 7115 (15.0) 0.21 Type of infertility treatment in the index pregnancy Intrauterine insemination 2892 (24.2) 0 (0.0) 0.8 2778 (24.1) 0 (0.0) – Ovulation induction 2820 (23.6) 0 (0.0) 0.79 2794 (24.2) 0 (0.0) – In vitro fertilization 5630 (47.1) 0 (0.0) 1.33 5366 (46.5) 0 (0.0) – Intracytoplasmic sperm injection 623 (5.2) 0 (0.0) 0.33 608 (5.3) 0 (0.0) – No. of antenatal visits in the index pregnancy, median (IQR) 22 (17–29) 16 (13–20) 0.87 22 (17–28) 19 (16–25) 0.25 Note: BMI = body mass index, IQR = interquartile range, SD = standard deviation. *Unless specified otherwise. †Data missing in < 1%. ‡Appendix 2 (available at www.cmaj.ca/lookup/suppl/doi:10.1503/cmaj.181124/-/DC1) contains the definitions of each of the maternal world regions of origin. §Data were available for BMI for 383 543 pregnancies in the unmatched cohort and 47 939 pregnancies in the matched cohort. ¶Congestive heart failure, congenital heart disease, coronary artery disease, cardiac dysthymia, chronic rheumatic heart disease or previous myocardial infarction. **Defined as any use (yes/no) as of 20 weeks’ gestation. of severe maternal morbidity (adjusted OR  2.28, 95%  CI 1.56– 3.33), whereas there was no association for noninvasive treat- ment (adjusted OR 0.99, 95% CI 0.57–1.72; Table 3). Interpretation In this large representative sample of women who became pregnant through all forms of infertility treatment, the risk of severe maternal morbidity or death was modestly increased relative to a propensity score–matched comparison group without infertility treatment. The higher risk of severe maternal morbidity was notable for severe postpartum hemorrhage, ICU admission and sepsis. Receipt of in vitro fertilization conferred the greatest burden of severe maternal morbidity or maternal death, whereas such increased risks were not evident for noninvasive infertility treatment. After propensity score matching, the RR of severe maternal morbidity or death was attenuated from 2.17 to 1.39, suggesting that some of the association between infertility treatment and severe maternal morbidity is explained by maternal factors rather than by the treatment. We and other researchers 10 have shown that maternal age over 40 years confers a higher absolute risk of severe maternal morbidity or death, as does a multiple pregnancy. Nevertheless, the significant difference in the effect of in vitro fertilization compared with noninvasive treatment sug- gests that factors associated with in vitro fertilization may con- tribute to severe maternal morbidity. This study used comprehensive and validated data sets cap- turing more than 99% of hospital births. 22 The broad array of maternal variables included in the propensity score used for matching should have minimized confounding by indication. 23 0.5 No infertility treatment: n = 1054/47 553 (22.2) Infertility treatment: n = 356/11 546 (30.8) No infertility treatment: n = 320/47 553 (6.7) Infertility treatment: n = 132/11 546 (11.4) No infertility treatment: n = 219/47 553 (4.6) Infertility treatment: n = 80/11 546 (6.9) No infertility treatment: n = 144/47 553 (3.0) Infertility treatment: n = 54/11 546 (4.7) No infertility treatment: n = 172/47 553 (3.6) Infertility treatment: n = 45/11 546 (3.9) No infertility treatment: n = 94/47 553 (2.0) Infertility treatment: n = 30/11 546 (2.6) No infertility treatment: n < 6/47 553 (≤ 0.1) Infertility treatment: n < 6/11 546 (≤ 0.5) 4.12 (1.19 to 14.22) 1.32 (0.87 to 1.98) 1.08 (0.78 to 1.50) 1.54 (1.13 to 2.10) 1.51 (1.17 to 1.94) 1.70 (1.39 to 2.08) 1.39 (1.23 to 1.56) Adjusted relative risk (95% CI) Cardiac conditions Postpartum hemorrhage with RBC transfusion, procedures to the uterus or hysterectomy Maternal ICU admission Puerperal sepsis Hysterectomy Any SMM All-cause maternal death from 20 wk gestation to 42 d postpartum 1.0 2.0 3.0 5.0 10.0 15.0 No. of women with outcome/no. of women at risk (rate per 1000) Adjusted RR (95% CI) Figure 2: Risk of a woman having severe maternal morbidity or death between 20 weeks’ gestation and 42 days after the index birth, in relation to receipt of infertility treatment within the index pregnancy, relative to no infertility treatment (propensity-matched cohort). Data are shown for the 5 most common subtypes of severe maternal morbidity. Note: CI = confidence interval, ICU = intensive care unit, RBC = red blood cells, RR = relative risk, SMM = severe maternal morbidity. RESEARCH E124 CMAJ | FEBRUARY 4, 2019 | VOLUME 191 | ISSUE 5 We evaluated not only severe maternal morbidity but also the number of severe maternal morbidity indicators. The list of severe maternal morbidity indicators used for Canadian perina- tal surveillance is similar to that proposed for use in the US, including ICU admission, 24 which is strongly predictive of mater- nal mortality.25 Others have previously examined the association between infertility treatment and severe maternal morbidity. In 2016, a ret- rospective cohort study reported that, among 1  million births in the US, including 14 761 pregnancies achieved through assisted reproductive technologies, the adjusted RR for severe maternal morbidity was 2.3  (95%  CI 2.1–2.7) when comparing assisted reproductive technologies with no treatment.12 Unlike that study, which used a data source employing International Classification of Diseases, 9th revision (ICD-9) coding, our data source employed ICD-10-CA coding, which better classifies severe preeclampsia, procedures used to manage severe hemorrhagic complications and other indicators of severe maternal morbidity. In addition, we included socioeconomic factors in propensity score matching, such as income quintile and maternal country of origin, which also strongly predict severe maternal morbidity.20,21 Our study included only 1 randomly selected delivery per woman, thereby avoiding any potential overestimation of the risk of infertility treatment in repeat pregnancies. 12 Another small single-centre study also 0.5 5.01.0 10.0 15.0 20.02.03 .0 No infertility treatment: n = 1054/47 553 (22.2) Noninvasive treatment: n = 121/5572 (21.7) 0.98 (0.81 to 1.18) Invasive treatment: n = 235/5974 (39.3) 1.77 (1.54 to 2.03) No infertility treatment: n = 320/47 553 (6.7) Noninvasive treatment: n = 41/5572 (7.4) 1.09 (0.79 to 1.51) Invasive treatment: n = 91/5974 (15.2) 2.26 (1.80 to 2.85) No infertility treatment: n = 219/47 553 (4.6) Noninvasive treatment: n = 26/5572 (4.7) 1.01 (0.68 to 1.52) Invasive treatment: n = 54/5974 (9.0) 1.97 (1.46 to 2.64) No infertility treatment: n = 144/47 553 (3.0) Noninvasive treatment: n = 21/5572 (3.8) 1.24 (0.78 to 1.95) Invasive treatment: n = 33/5974 (5.5) 1.82 (1.25 to 2.66) No infertility treatment: n = 172/47 553 (3.6) Noninvasive treatment: n = 14/5572 (2.5) 0.70 (0.41 to 1.20) Invasive treatment: n = 31/5974 (5.2) 1.44 (0.99 to 2.10) No infertility treatment: n = 94/47 553 (2.0) Noninvasive treatment: n = 13/5572 (2.3) 1.18 (0.66 to 2.11) Invasive treatment: n = 17/5974 (2.8) 1.44 (0.86 to 2.41) No infertility treatment: n < 6/47 553 (≤ 0.1) Noninvasive treatment: n < 6/5572 (≤ 1.1) 3.41 (0.66 to 17.58) Invasive treatment: n < 6/5974 (≤ 1.0) 4.78 (1.14 to 19.97) Adjusted relative risk (95% CI) Any SMM No. of women with outcome/no. of women at risk (rate per 1000) Adjusted RR (95% CI) Cardiac conditions Postpartum hemorrhage with RBC transfusion, procedures to the uterus or hysterectomy Maternal ICU admission Puerperal sepsis Hysterectomy All-cause maternal death from 20 wk gestation to 42 d postpartum Figure 3: Risk of a woman having severe maternal morbidity or death between 20 weeks’ gestation and 42 days after the index birth, in relation to receipt of noninvasive or invasive infertility treatment within the index pregnancy, each relative to no infertility treatment. Data are shown for the 5 most common indicators of severe maternal morbidity. Note: CI = confidence interval, ICU = intensive care unit, RBC = red blood cells, RR  = relative risk, SMM = severe maternal morbidity. RESEARCH CMAJ | FEBRUARY 4, 2019 | VOLUME 191 | ISSUE 5 E125 showed an increased risk of severe maternal morbidity related to infertility treatment (OR  2.40, 95%  CI 1.10–5.23). 13 A third cohort study from Massachusetts that compared women who conceived through assisted reproductive technologies with those who were subfertile, untreated and pregnant also yielded an increased risk of severe maternal morbidity among singletons for both vaginal deliveries (OR  1.97, 95%  CI 1.30–3.00) and cesarean deliveries (OR  1.75, 95%  CI 1.30–2.35). 14 Differences in access to pregnancy care between US and Canadian health care systems, as well as patient demographics, may explain slight differences in estimates between our study and those of US-based researchers, heighten- ing the importance of our findings to Canadian women consider- ing infertility treatment and to their health care providers. In Canada, recent provincial funding has permitted greater access to infertility treatment. 8 Because such treatment is often accessed by perimenopausal women 26 and those with reversible causes of infertility, including obesity, 27 those receiving treat- ment have a higher prevalence of conditions such as diabetes mellitus and chronic hypertension. About 1 in 5 live births in Can- ada are to women over the age of 35  years, 28 and 13% of all in vitro fertilization cycles in 2014 were performed solely for “advanced female age,” furthering the risk of maternal and peri- natal complications.29 A better understanding of the contribution of infertility treatment to severe maternal morbidity may lead to improved preventive and surveillance strategies for women who are at risk. Table 2: Relative risk of a woman experiencing severe maternal morbidity or death between 20 weeks gestation and 42 days after the index birth, in relation to receipt of infertility treatment within the index pregnancy and relative to no infertility treatment Stratification variable Women who received infertility treatment* Women who did not receive infertility treatment* Relative risk (95% CI) of SMM or death No. with SMM or who died No. at risk (rate per 1000) No. with SMM or who died No. at risk (rate per 1000) Age < 40 yr 279 10 070 (27.7) 866 42 448 (20.4) 1.36 (1.19–1.55) Age ≥ 40 yr 73 1443 (50.6) 155 5046 (30.7) 1.64 (1.25–2.16) Singleton pregnancy 210 8933 (23.5) 798 40 693 (19.6) 1.20 (1.03–1.39) Multiple pregnancy 135 2515 (53.7) 235 6696 (35.1) 1.52 (1.23–1.88) Note: CI = confidence interval, SMM = severe maternal morbidity. *Number of pregnancies in propensity score–rematched cohorts were slightly lower than for the full model (total no. for age was 57 007: n = 11 513 with fertility treatment and n = 47 494 without fertility treatment; and total no. for singleton/multiple gestation was 58 837: n = 11 448 with fertility treatment and n = 47 389 without fertility treatment). Table 3: Odds of a woman having 1, 2 or ≥ 3 indicators of severe maternal morbidity between 20 weeks’ gestation and 42 days after the index birth, by infertility treatment within the index pregnancy versus no infertility treatment, and by treatment subtype versus no infertility treatment Exposure Women with 1 SMM indicator Women with 2 SMM indicators Women with ≥ 3 SMM indicators No. (rate per 1000) OR* (95% CI) No. (rate per 1000) OR* (95% CI) No. (rate per 1000) OR* (95% CI) Infertility treatment versus no infertility treatment No infertility treatment (n = 47 553) 768 (16.2) 1.00 (ref.) 164 (3.4) 1.00 (ref.) 121 (2.5) 1.00 (ref.) Infertility treatment (n = 11 546) 247 (21.4) 1.34 (1.16–1.54) 61 (5.3) 1.55 (1.15–2.08) 48 (4.2) 1.65 (1.18–2.30) Infertility treatment subtype versus no infertility treatment No infertility treatment (n = 47 553) 768 (16.2) 1.00 (ref.) 164 (3.4) 1.00 (ref.) 121 (2.5) 1.00 (ref.) Noninvasive infertility treatment† (n = 5572) 84 (15.1) 0.93 (0.74–1.17) 23 (4.1) 1.20 (0.77–1.85) 14 (2.5) 0.99 (0.57–1.72) Invasive infertility treatment‡ (n = 5974) 163 (27.3) 1.72 (1.45–2.04) 38 (6.4) 1.88 (1.32–2.68) 34 (5.7) 2.28 (1.56–3.33) Note: CI = confidence interval, OR = odds ratio, ref. = reference, SMM = severe maternal morbidity. *Odds ratios were calculated using multinomial logistic regression analysis, accounting for matching. †Includes intrauterine insemination and ovulation induction. ‡Includes in vitro fertilization and intracytoplasmic sperm injection. RESEARCH E126 CMAJ | FEBRUARY 4, 2019 | VOLUME 191 | ISSUE 5

Study limitations included possible nondifferential misclassifi- cation of exposure, despite prior high data accuracy for this data source. 30 Such misclassification is likely to be minor and would result in an attenuation of our risk estimates. Subfertil- ity, as measured by fertility-related conditions, may have also been undercaptured, as shown by lower than expected rates of endometriosis and polycystic ovary syndrome (Table  1). Because women who received noninvasive infertility treatment had a similar rate of severe maternal morbidity and risk to those who became pregnant without infertility treatment sug- gests that women with noninvasive infertility treatment might prove to be a useful subfertile comparison group for future studies. However, severity of infertility — not captured with our approach — may also explain differences in adverse pregnancy outcomes between invasive and noninvasive treatment sub- groups. 31 Similarly, we lacked detailed information on the superovulation protocols used. Aggressive superovulation pro- tocols have been linked with mortality in in vitro fertilization pregnancies; this may be a discriminating feature to include in future studies. 32 We included diagnostic codes for obesity in our propensity score because high body mass index (BMI) is an independent risk factor for severe maternal morbidity. 32 Body mass index was only available for 47 939 matched pregnancies; therefore, the true proportion of obesity in our cohort is unknown. However, the interaction between prepregnancy BMI and infertility treatment may not be as influential on severe maternal morbidity as previously thought. 33 Finally, we did not stratify our results by maternal ethnicity, which may modify the effect of infertility treatment.

Although severe maternal morbidity and maternal death are uncommon outcomes, our study suggests that women under- going noninvasive infertility treatment, such as intrauterine insemination, are not at increased risk of these outcomes, whereas those undergoing in vitro fertilization are. Whether spe- cific components of treatment using in vitro fertilization, such as the dose of ovarian hyperstimulation or fresh versus frozen embryo transfer, worsen maternal health, or whether the increased risk is a reflection of those who require or choose in vitro fertilization, remains to be determined. Nevertheless, studies comparing invasive with less invasive infertility treat- ment should extend their focus beyond rates of live births to encompass maternal health outcomes. Women whose pregnancy was achieved by infertility treat- ment, especially in vitro fertilization, are at a somewhat higher risk of severe maternal morbidity or death. Further research should identify patient- and treatment-specific factors that might be modified to mitigate excess maternal risks.

1. Bushnik T, Cook JL, Yuzpe AA, et al. Estimating the prevalence of infertility in Canada. Hum Reprod 2012;27:738-46. 2. Talaulikar VS, Arulkumaran S. Reproductive outcomes after assisted concep - tion. Obstet Gynecol Surv 2012;67:566-83. 3. Gunby J, Bissonnette F, Librach C, et al. Group of the Canadian Fertility Androl- ogy Society. Assisted reproductive technologies (ART) in Canada: 2007 results from the Canadian ART Register. Fertil Steril 2011;95:542-7.e1-10 4. Messerlian C, Platt RW, Tan SL, et al. Low-tech assisted reproduction and the risk of pre-term birth in a hospital-based cohort. Fertil Steril 2015;103:81-8.e2 5. Sazonova A, Kallen K, Thurin-Kjellberg A, et al. Obstetric outcome after in vitro fertilization with single or double embryo transfer. Hum Reprod 2011;26: 442-50. 6. Wennerholm UB, Sazonova A, Kallen K, et al. Obstetric outcome in singletons after in-vitro fertilization with frozen-thawed embryos. Am J Obstet Gynecol 2012;1:S42. 7. Quebec Assisted Procreation Services. Québec: Gouvernement du Québec; 2017. Available: www.quebec.ca/en/family-and-support-for-individuals/ pregnancy -and-parenthood/assisted-procreation-services/ (accessed 2017 Dec. 11). 8. Ontario offering 50 government funded fertility treatment clinics. CBC News Toronto 2015 Dec. 21. Available: www.cbc.ca/news/canada/toronto/ontario -fertility-clinics-1.3374634 (accessed 2017 Sept. 6) . 9. Tulandi T, King L. Public funding and access to in-vitro fertilization. N Engl J Med 2013;368:1948-9. 10. Lisonkova S, Potts J, Muraca G, et al. Maternal age and severe maternal mor- bidity: a population-based retrospective cohort study. PLoS Med 2017;14: e1002307. 11. Canadian Perinatal Surveillance System. Ottawa: Public Health Agency of Canada; 2004. Available: www.canada.ca/en/public-health/services/injury-prevention / health-surveillance-epidemiology-division/maternal-infant-health/canadian -perinatal -surveillance-system.html (accessed 2018 June 5). 12. Martin AS, Monsour M, Kissin DM, et al. Trends in severe maternal morbidity after assisted reproductive technology in the United States, 2008–2012. Obstet Gynecol 2016;127:59-66. 13. Wang ET, Ozimek J, Greene N, et al. Impact of fertility treatment on severe maternal morbidity. Fertil Steril 2016;106:423-6. 14. Belanoff C, Declercq ER, Diop H, et al. Severe maternal morbidity and the use of assisted reproductive technology in Massachusetts. Obstet Gynecol 2016; 127:527-34. 15. Vandenbroucke JP, von Elm E, Altman D, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE): explanation and elabora - tion. Epidemiology 2007;18:805-35. 16. Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol 2004;159:702-6. 17. BORN’s data quality report: 2012-2014 — executive summary. Ottawa: BORN [Bet- ter Outcomes Registry & Network] Ontario; 2016. Available: http://datadictionary. bornontario.ca/assets/documents/Data%20Quality/BORN%20Data%20Quality %20Report%202012-2014%20-%20Executive%20Summary.pdf (accessed 2016 July 20). 18. Joseph KS, Liu S, Rouleau J, et al. Severe maternal morbidity in Canada, 2003 to 2007: surveillance using routine hospitalization data and ICD-10-CA codes. J Obstet Gynaecol Can 2010;32:837-46. 19. Austin PC. Using the standardized difference to compare the prevalence of a binary variable between two groups in observational research. Commun Stat Simul Comput 2009;38:1228-34. 20. Urquia ML, Wanigaratne S, Ray JG, et al. Severe maternal morbidity associated with maternal birthplace: a population-based register study. J Obstet Gynaecol Can 2017;39:978-87. 21. Medcalf KE, Park AL, Vermeulen MJ, et al. Maternal origin and risk of neonatal and maternal ICU admission. Crit Care Med 2016;44:1314-26. 22. Joseph KS, Fahey J; Canadian Perinatal Surveillance System. Validation of perinatal data in the Discharge Abstract Database of the Canadian Institute of Health Information. Chronic Dis Can 2009;29:96-100. 23. Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res 2011; 46:399-424. 24. Lazariu V, Nguyen T, McNutt LA, et al. Severe maternal morbidity: a population- based study of an expanded measure and associated factors. PLoS One 2017;12:e0182343. RESEARCH CMAJ | FEBRUARY 4, 2019 | VOLUME 191 | ISSUE 5 E127 25. Ray JG, Urquia ML, Berger H, et al. Maternal and neonatal separation and mor- tality associated with concurrent admissions to intensive care units. CMAJ 2012;184:E956-62. 26. Ethics Committee of the American Society for Reproductive Medicine. Oocyte or embryo donation to women at advanced age: a committee opinion. Fertil Steril 2013;100:337-40. 27. Craig LB, Peck JD, Bullock S, et al. How large is too large: association of obesity with fecundability in women undergoing intrauterine insemination. Fertil Steril 2011;96(Suppl 1):S89. 28. Live births, by age of mother. Ottawa: Statistics Canada; 2016. Available: www150.statcan.gc.ca/t1/tbl1/en/tv.action?pid=1310041601 (accessed 2018 Aug. 23). 29. Cleary-Goldman J, Malone FD, Vidaver J, et al. Impact of maternal age on obstetric outcome. Obstet Gynecol 2005;105:983-90. 30. Baldur-Felskov B, Kjaer SK, Albieri V, et al. Psychiatric disorders in women with fertility problems: results from a large Danish register-based cohort study. Hum Reprod 2013;28:683-90. 31. Basso O, Weinberg CR, Baird DD, et al. Subfecundity as a correlate of preeclamp- sia: a study within the Danish National Birth Cohort. Am J Epidemiol 2003;157:195-202. 32. Braat DDM, Schutte JM, Bernardus RE, et al. Maternal death related to IVF in the Netherlands 1984–2008. Hum Reprod 2010;25:1782-6. 33. Dayan N, Fell DB, Guo Y, et al. Severe maternal morbidity in women with high BMI in IVF and unassisted singleton pregnancies. Hum Reprod 2018;33:1548-56. Competing interests: None declared. This article has been peer reviewed. Affiliations: Department of Medicine and Research Institute (Dayan), Department of Obstetrics and Gynecology (Basso), McGill University Health Centre; Department of Epidemiology, Biostatistics and Occupa- tional Health (Dayan, Basso), McGill University, Montréal, Que.; Depart- ment of Obstetrics and Gynaecology, and School of Population and Public Health (Joseph), University of British Columbia, Vancouver, BC; Children’s Hospital of Eastern Ontario Research Institute (Fell); School of Epidemiology and Public Health (Fell), University of Ottawa, Ottawa, Ont.; Department of Medicine and Obstetrics and Gynecology (Laskin), University of Toronto; TRIO Fertility (Laskin); ICES (Park, Luo, Guan, Ray, Fell); Department of Medicine (Ray), St. Michael’s Hospital, Toronto, Ont. Contributors: All of the authors contributed substantially to this study. Natalie Dayan designed the study, interpreted analyses, and wrote and revised the manuscript. Joel Ray designed the study with Natalie Dayan and interpreted analyses. K.S. Joseph, Alison Park, Olga Basso, Deshayne Fell and Carl Laskin helped with study design and data interpretation. Jin Luo and Jun Guan helped with study design and cohort creation, and performed data analyses. All of the authors revised the manuscript criti- cally for important intellectual content, approved the version to be pub- lished and agreed to be accountable for all aspects of the work. Funding: This study was supported by funding from the Canadian Insti- tutes of Health Research (grant no. 15139). Natalie Dayan is supported by a Fondation de Recherche du Québec – Santé (FRQS) Chercheur Boursier Clinicien award. K.S. Joseph is supported by an Investigator award from the BC Children’s Hospital Research Institute and an Applied Chair in Reproductive, Child & Youth Health Services and Policy Research from the Canadian Institutes of Health Research (APR-126338). This study was also supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC). Data sharing: The data set from this study is held securely in coded form at ICES. Although data-sharing agreements prohibit ICES from making the data set publicly available, access may be granted to those who meet pre- specified criteria for confidential access, available at www.ices.on.ca/DAS. The full data set, creation plan and underlying analytic code are available from the authors upon request, with the understanding that the computer programs may rely upon coding templates or macros that are unique to ICES and are therefore either inaccessible or may require modification.

The authors thank Dr. Janet Takefman for review of the manuscript from the perspective of the infertile patient, and Dr. Gabriel Shapiro and Hiba Jekki for their help with figure and manuscript formatting. Disclaimer: Parts of this material are based on data or information com- piled and provided by Canadian Institute for Health Information (CIHI). However, the analyses, conclusions, opinions and statements expressed in the material are those of the authors, and not necessarily those of CIHI. This study is based, in part, on data provided by the Better Outcomes Reg- istry & Network (BORN), part of the Children’s Hospital of Eastern Ontario. The interpretation and conclusions contained herein do not necessarily represent those of BORN Ontario. This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health and Long- Term Care (MOHLTC). The opinions, results and conclusions reported in this paper are those of the authors and are independent from the funding sources. No endorsement by ICES or the Ontario MOHLTC is intended or should be inferred. Accepted: Dec. 19, 2018 Correspondence to: Natalie Dayan, [email protected]