Parity and Risk of Subarachnoid Hemorrhage in Women
A Nested Case-Control Study Based on National Swedish Registries
Background and Purpose— Subarachnoid hemorrhage (SAH) is the only type of stroke with female predominance, suggesting that reproductive factors may play a role in the etiology. We conducted a population-based study to examine the influence of parity on the risk of SAH in women.
Methods— We linked data from 3 national Swedish registries to identify first-ever hospitalizations for SAH in a cohort of women followed up since first childbirth during 1973–1997. Within this cohort, we conducted a nested case-control study and estimated the odds ratio (OR) and 95% CI of SAH by parity adjusted for age, calendar period, and length of follow-up. Information on smoking habits before the subject’s first childbirth was available in a subset of the data (women with first childbirths during 1982–1997).
Results— Of the 887 cases identified, 70% had suffered from SAH ≥5 years after giving birth to their last child. The OR declined with increasing parity (1 child: reference; 2: OR=0.83 [95% CI, 0.70 to 0.99]; 3: OR=0.72 [95% CI, 0.58 to 0.91]; 4: OR=0.72 [95% CI, 0.48 to 1.08]; ≥5: OR=0.67 [95% CI, 0.32 to 1.41]). Adjusting for daily cigarette consumption before first childbirth in the subsample in which this information was available reduced but did not eliminate the association of the disorder with parity.
Conclusions— Parity may confer a moderate long-term protective effect on the risk of SAH. The biological mechanism underlying this association is currently unknown.
Subarachnoid hemorrhage (SAH) is caused, in 85% of cases, by a ruptured aneurysm. The disorder primarily affects individuals aged <60 years, carries a high morbidity and mortality rate, and is the only type of stroke that afflicts women more than men.1,2 The gender gap may provide clues about the pathology of this disorder.3 Efforts to explain the female susceptibility to SAH have hitherto focused mainly on exogenous factors (use of oral contraceptive and hormone replacement therapy) and endogenous hormonal (mainly menopausal) factors.4,5 The influence of reproductive factors such as parity has received relatively little attention. Three studies examining the associations of parity and gravidity with stroke did not provide separate estimates for SAH.6–8 Two recently published case-control studies with well-characterized cases of SAH yielded conflicting results concerning parity,9,10 possibly because of limitations of methodology and sample size.
To investigate further the effect of parity on SAH, we performed the present nationwide population-based study in which we followed a large cohort of parous women for SAH through registries with decades of national coverage.
See Editorial Comment, page 32
Subjects and Methods
The present case-control study, nested within a population-based cohort, was based on data from 3 nationwide Swedish registries: the Inpatient Register, Birth Register, and Cause of Death Register.
The coverage of the Inpatient Register of in-hospital admissions in Sweden increased over time, from 60% in 1973 at the time the register was founded, to 80% in 1980, and to >99% since 1987.11 Discharge diagnoses are recorded according to the International Classification of Diseases, Eighth Revision (ICD-8) (1969–1986), International Classification of Diseases, Ninth Revision (ICD-9) (1987–1996), and International Classification of Diseases, 10th Revision (ICD-10) (since 1997). Recorded data include the dates of admission and discharge, up to 6 discharge diagnoses (8 discharge diagnoses in 1997), and up to 6 codes for surgical procedures performed (12 codes in 1997).11
The Birth Register covers >99% of all deliveries in Sweden.12 Since 1982, the Birth Register also includes data on the mother’s daily consumption of tobacco (none, 1 to 9 cigarettes daily, and ≥10 cigarettes daily), which was recorded at the time of the women’s first visit to the antenatal clinic.
The Cause of Death Register, which has been computerized since 1961, includes person-specific information about date and cause of death of all deaths in Sweden.
Information from these registers was linked by means of the unique, national registration number assigned to each resident in Sweden. All Swedish residents have access to tax-supported medical care.
We identified a cohort comprising all women with a record of a first childbirth in the Birth Register from January 1973 through December 1997, a total of 1 082 658 primiparous women. Information on any subsequent births was also retrieved from the Birth Register. The dates of death for deceased women were available through previous linkage of the Birth Register with the Cause of Death Register, but only for subjects giving birth in 1987–1997. Data on admissions to hospitals in Sweden were retrieved from the Inpatient Register from the earliest date available (1973) and until 1997.
Women entered the cohort at the time of their first childbirth and were followed up until 1 of the following events occurred: a first-ever admission for SAH, death, or end of the study period (December 31, 1997), whichever came first.
Definition of Cases and Controls
For a nested case-control study within the cohort, cases were defined as women with a discharge diagnosis of SAH (ICD-8 or ICD-9: 430; ICD-10: I60) at any time during follow-up. The date of first admission for SAH was considered the index date.
Excluded from the study were the following: (1) women who had an Inpatient Register diagnosis of SAH dated before their first childbirth; (2) cases with concurrent diagnoses of potential nonaneurysmal causes of SAH on their index admission (ie, trauma, arteriovenous malformation, or cerebral neoplasm) (n=39); (3) cases with any registered diagnoses of a number of rare disorders reported to predispose to SAH (eg, polycystic kidney disease, Ehlers-Danlos disorder, congenital heart disease) (n=3); and (4) cases of SAH that arose during the puerperium (up to 6 weeks after delivery) (n=23), since this is a period with excess risk of not only ischemic stroke and intracerebral hemorrhage13 but probably also SAH14 because of short-term postpartum changes in hemodynamics, coagulation, or vessel walls.
For each case, we identified all women in the cohort who were still at risk on the index date of the case, were of the same age as the case (year of birth), and had been followed up for the same length of time (year and month of first childbirth). As with the cases, controls were ineligible if identified within 6 weeks of their latest childbirth. We randomly selected 10 eligible controls for each case. Fewer than 10 controls were matched in 7.7% of cases, but only 1.1% of case-control sets included fewer than 5 controls.
Validity of Discharge Diagnoses
Since we did not have access to hospital records, we relied on the validity of the registry diagnosis of SAH. In a previous Danish study in a highly comparable setting,15 a neurologist with access to medical records validated the inpatient registry diagnosis of SAH (ICD-8 and ICD-10 codes) according to strict predefined criteria. The degree of misclassification was found to be <20%. Among the validated cases in that study, 25 had also received codes for operative procedures for aneurysms, all of whom were classified as definite cases of aneurysmal SAH (D. Gaist, MD, PhD, unpublished data, 2000). Guided by these findings, we identified all cases in the present study with a diagnostic code for SAH who had also received a code for an aneurysm operation at any admission within 21 days of the index admission. The 21-day time limit allowed for transfers between departments. We termed this subsample the aneurysmal surgery group.
Definition of Parity and Potential Confounders
Parity was defined as the number of children recorded in the last childbirth record of each woman registered during follow-up, ie, before the index date. Parity was classified into 5 groups: 1 child (reference group), 2, 3, 4, and ≥5 children. Multiple births (eg, twins, triplets) contributed to parity with the number of children born. The time gap between last recorded delivery and date of index date was also determined (<1 year, 1 to 4 years, 5 to 9 years, ≥10 years).
Women with any cesarean deliveries were identified because this mode of delivery might be a proxy for pregnancies with problematic conditions that carry a high short-term risk of SAH. Women with multiple births, who might have been exposed to exogenous hormone stimulation, were likewise identified.
We used conditional logistic regression analysis to estimate odds ratios (ORs) and 95% CI of SAH by parity. With the matched design, these ORs took into account age at index date, follow-up period, and calendar date. Cesarean delivery and multiple births were not included in the conditional logistic regression since they changed the crude OR for parity by <10%.16 In nested case-control studies, the OR is an unbiased estimate of the relative risk.
To further minimize issues of bias, confounding, and censoring, we repeated the analyses in 2 subgroups: the aneurysmal surgery group and women aged ≥40 years at the time of diagnosis.
We adjusted for the effect of smoking in the subset of data that included this information, ie, women with first births during 1982–1997. Subjects from this subset with missing information on smoking (4% of cases and controls) were included in the reference group because a separate analysis indicated that the OR of SAH in the group with missing information (OR=1.07) was highly comparable with that of the reference group (OR=1.0).
Trends in ORs over parity were tested by means of the χ2 for trend test. The data were analyzed with the use of the SAS version 8.02 statistical package (SAS Institute Inc).
After application of the exclusion criteria, a total of 887 cases of SAH were identified, matched to 8512 controls. The median age of cases at the time of diagnosis was 38.1 years (range, 17.7 to 62.1 years), and they had been followed up for a median of 12.9 years.
Most cases suffered from SAH >1 year after their last registered childbirth (93%); in 70% of cases, ≥5 years had elapsed (Table 1). Twin births were equally frequent in cases and controls, but cesarean deliveries were more frequent among cases. The parity of cases was less than that of controls (Table 1).
The OR of SAH declined with increasing parity (1 child: reference; 2: OR=0.83 [95% CI, 0.70 to 0.99]; 3: OR=0.72 [95% CI, 0.58 to 0.91]; 4: OR=0.72 [95% CI, 0.48 to 1.08]; ≥5: OR=0.67 [95% CI, 0.32 to 1.41]) (χ2 for trend: 8.2; P=0.004). The point estimates were not materially different in the restricted samples, but, probably because of the lower number of cases, parity was no longer significantly associated with risk of subarachnoidal bleeding (Table 2).
We separately analyzed the subset of cases with first births in 1982–1997 in which information was available on smoking habits at the time of the first visit to the antenatal clinic. Cases belonging to this subset were considerably younger (median age at diagnosis, 33.3 years [range, 19.9 to 55.7 years]) and had been followed up for a shorter period (median follow-up, 7.1 years). The distribution of parity in cases and controls in the subset is presented in Table 3. Adjusting for the effect of smoking in the subset reduced but did not eliminate the association with parity in all strata (Table 3). Daily cigarette smoking recorded before first childbirth was clearly linked in a dose-response fashion to later-life relative risk of SAH (0 cigarettes: reference; 1 to 9 cigarettes: OR=2.31 [95% CI, 1.65 to 3.24]; ≥10 cigarettes: OR=3.68 [95% CI, 2.58 to 5.24]). This relation was even more pronounced in the subcohort with cases from the aneurysmal surgery group (0 cigarettes daily: reference; 1 to 9 cigarettes: OR=3.24 [95% CI, 1.80 to 5.80]; ≥10 cigarettes: OR=4.39 [95% CI, 2.25 to 8.59]).
In this large, population-based study, we found that high parity was associated with a moderately reduced risk of SAH in women several years after their last childbirth. Controlling for smoking habits before first childbirth in a subset of the data in which this information was available somewhat reduced but did not eliminate the association.
The relation between childbearing and SAH has received little attention. Three studies6–8 have focused on the risk of stroke in general or of ischemic stroke and found moderately elevated risks linked to gravidity and parity. Only 1 of these studies8 considered intracerebral hemorrhage. While it found a 2-fold to 3-fold increase in risk for parous compared with nulliparous women, the study comprised only 33 cases, and the CIs were wide.
Two recently published studies9,10 focused on the relation between childbearing and the risk of a first-ever SAH. In a case-control study involving 268 women with SAH, Mhurchu et al9 found that parity (≥1 child) was not associated with the disorder (relative risk=0.93 [95% CI, 0.58 to 1.53]). In a smaller study of 124 cases of SAH, Okamoto et al10 found that nulliparity conferred additional risk in Japanese women (relative risk=1.82 [95% CI, 0.76 to 17.5]). The conflicting results of these 2 studies may be due, at least in part, to methodological limitations. Proxy responses were used in 29%10 and 65%9 of cases, and in the study by Mhurchu et al,9 data on parity were missing in 9% of cases but in none of the controls. Furthermore, parous women were compared only with nulliparous women, and nulliparity may reflect an inability to conceive or complete a pregnancy because of health factors with unknown influences on the risk of suffering from a SAH. Possibly as a result of sample size limitations, the effect of the number of children was not explored in either of these studies.
The present population-based study was conducted within a uniformly organized health system with free access to medical care. The use of prospectively collected registry information safeguarded the study against potential bias introduced by recall and proxy-reported data, which may have influenced previous studies.
Nonetheless, this study has several other potential weaknesses. Both our inpatient data and vital status information were incomplete before 1987. The incompleteness of inpatient data may result in misclassification with regard to case status. However, this problem is unlikely to have biased our findings: the rarity of this disorder (<1 case per 10 000 person-years2) makes it highly unlikely that more than an extremely small fraction of controls was affected. The incomplete data on vital status would affect only controls and may have led to an overestimation of the time at risk and hence an underestimation of parity in this group. The direction of this potential bias is therefore conservative, and its magnitude is probably negligible because of the relatively low mortality rates in the age groups studied.
As a result of the short length of follow-up, not all women in our sample will have completed their reproductive history. The effect of this potential lack of information appears to have little impact on our findings, as illustrated by the analyses restricted to women aged ≥40 years at index date, at which point a more complete reproductive history can be assumed. Although the birth register covered births, including stillbirths, it did not include information on abortions. We were therefore unable to explore the possible role of gravidity on the risk of SAH.
Use of an inpatient register to capture cases necessitates that patients are admitted to the hospital and have correctly coded diagnoses. Studies of intensive monitoring of stroke in Sweden have shown that approximately 90% of all cases of SAH are hospitalized.17 Most of our cases occurred during a time period that had wide availability of neuroimaging facilities, a factor known to influence positively the validity of diagnosis.18 Restricting the sample to patients who also had codes for aneurysmal operative procedures yielded results that were highly comparable to the main analysis, indicating that misclassification of diagnosis is not of major concern. Furthermore, we have no reason to believe that the completeness or validity of the data on SAH depended on the patient’s reproductive history.
Hypertension and alcohol consumption are recognized risk factors of SAH.2 The association of gravidity with hypertension and systolic blood pressure was examined in a large cross-sectional study in the United States,19 which reported modest associations of increasing gravidity with lower systolic blood pressure and hypertension. A cross-sectional survey of 3576 Norwegian women aged 40 to 42 years found high parity to be correlated with low intake of alcoholic beverages.20 Thus, an important limitation of our study was the lack of data on accepted risk factors for SAH such as hypertension and alcohol abuse as well as more contentious ones (eg, oral contraceptives and hormone replacement therapy).
In the subset of data in which information on cigarette smoking at the time of the first antenatal visit was available, a clear dose-response relationship was established for this known risk factor, which was further enhanced when the sample was restricted to women who had undergone aneurysmal surgery. However, although residual confounding cannot be ruled out, our data indicate that the association between parity and SAH is not due to tobacco consumption since adjusting for this effect did not substantially change the results.
Our findings regarding parity can have only 3 likely explanations. Because of the nature of our study and the relative dearth of information on potential risk factors available, we cannot rule out that the observed associations are due to confounding. The size of the study makes random error a less likely explanation. A final intriguing possibility is that a woman’s reproductive history is causally linked to her risk of SAH.
In conclusion, our population-based data show that increasing parity, at least for women of a relatively young age, may protect against SAH. Our results suggest that reproductive factors and smoking play a role in the etiology of SAH.
- Received March 10, 2003.
- Revision received September 3, 2003.
- Accepted September 5, 2003.
van Gijn J, Rinkel GJE. Subarachnoid hemorrhage: diagnosis, causes and management. Brain. 2001; 124: 249–278.
Longstreth WT, Nelson LM. The gender gap in aneurysmal subarachnoid hemorrhage. Stroke. 2001; 32: 611–612.
Teunissen LL, Rinkel GJE, Algra A, van Gijn J. Risk factors for subarachnoid hemorrhage: a systematic review. Stroke. 1996; 27: 544–549.
Beral V. Long term effects of childbearing on health. J Epidemiol Community Health. 1985; 39: 343–346.
Mhurchu CN, Anderson C, Jamrozic K, Hankey G, Dunbabin D, for the Australasian Cooperative Research on Subarachnoid Hemorrhage Group. Hormonal factors and risk of aneurysmal subarachnoid hemorrhage: an international population-based case-control study. Stroke. 2001; 32: 606–612.
Okamoto K, Horisawa R, Kawamura T, Asai A, Ogino M, Takagi T, et al. Menstrual and reproductive factors for subarachnoid hemorrhage risk in women: a case-control study in Nagoya, Japan. Stroke. 2001; 32: 2841–2844.
Epidemiologisk Centrum, Socialstyrelsen. Patientregistret: Utskrivningar från sluten vård 1964–1997: Kvalitet och innehåll. Stockholm, Sweden: Socialstyrelsen; 2000.
Kittner SJ, Stern BJ, Feeser BR, Hebel R, Nagey DA, Buchholz DW, et al. Pregnancy and the risk of stroke. N Engl J Med. 1997; 335: 768–774.
Gaist D, Væth M, Tsiropoulos I, Christensen K, Corder E, Olsen J, et al. Risk of subarachnoid haemorrhage in first degree relatives of patients with subarachnoid haemorrhage: follow up study based on national registries in Denmark. BMJ. 2000; 320: 141–145.
Greenland S, Rohtman KJ. Introduction to stratified analysis. In: Rothman KJ, Greenland S, eds. Modern Epidemiology. 2nd ed. Philadelphia, Pa: Lippincott-Raven: 1998: 253–279.
Ingall T, Asplund K, Markku M, Bonita R, for the WHO MONICA Project. A multinational comparison of subarachnoid hemorrhage epidemiology in the WHO MONICA stroke study. Stroke. 2000; 31: 1054–1061.
Linn FHH, Rinkel GJE, Algra A, van Gijn J. Incidence of subarachnoid hemorrhage: role of region, year, and rate of computed tomography: a meta-analysis. Stroke. 1996; 27: 625–629.