Reproductive history and mortality from cardiovascular disease among women textile workers in Shanghai, China

Abstract

Background Few studies have examined the possible effects of reproductive factors on cardiovascular disease (CVD) risks in Asian women.

Methods A cohort of 267 400 female textile workers in Shanghai, China, was administered a questionnaire at enrolment (1989–91) and followed for mortality through 2000. Relative risks (hazard ratios) for ischaemic heart disease (IHD), ischaemic stroke and haemorrhagic stroke were calculated using Cox proportional hazards modelling, adjusting for relevant co-variates.

Results Risks were not consistently associated with age at menopause, parity, stillbirths, miscarriages or duration of lactation. An increasing trend in IHD mortality risk, but not stroke, was observed with decreasing age at menarche. There was no evidence of increased CVD mortality risk by oral or injectable contraceptive use or induced abortions. As expected, greater mortality rates from CVD and increased CVD risks were also observed with smoking.

Conclusions Use of steroid contraceptives, induced abortions and reduced parity from China's one-child-per-family policy has not had an adverse effect on risk of CVD mortality in this cohort.

Introduction

Cardiovascular diseases (CVDs) are among the leading causes of morbidity and mortality worldwide.¹ Mortality from ischaemic heart disease (IHD) is most common, followed by stroke in Western countries, although in China, stroke is more common.^1,2 Established CVD risk factors in addition to age and male sex include hypertension, obesity and smoking.^1,3 Recent surveys have shown the prevalence of the latter three factors to be increasing in China.²

For women in Asia, the health consequences of reproductive factors on CVD risks have not been well quantified. The effects are of particular interest for Chinese women, a population with unique reproductive histories from the country's ‘One Child Policy’ introduced in 1979, which set rules and regulations for family size, affecting parity, contraceptive use and number of induced abortions.⁴ Reproductive factors may play a role in CVD risk through hormone levels, which vary over a woman's lifetime. Pre-menopausal women have lower risk of CVD morbidity and mortality compared with men and post-menopausal women of the same age, indicating a possible protective effect of endogenous oestrogen and progesterone.⁵ Also, low testosterone is associated with elevated CVD risk.³ CVD risk may depend on dose and type of hormonal preparation or presence of other risk factors. Observational studies have found some coronary heart disease (CHD) benefit with use of oestrogen alone as post-menopausal hormone therapy (PHT), although the Women's Health Initiative (WHI) trials found no change in risk.⁶ Elevated CHD risk was found in WHI users of oestrogen–progestin combination PHT, and increased risk of ischaemic, but not haemorrhagic stroke, was seen with both types of PHT.⁶ Oral contraceptive (OC) use has been associated with CVD, especially in smokers and hypertensive women.⁷

The majority of studies on reproductive history and CVD have been conducted among women from North America and Europe; thus, findings may not be applicable to women in other parts of the world. Studies of Asian women, in particular, have been limited in scope. Studies by the World Health Organization (WHO) on CVD outcomes and OCs grouped women from Asia, Africa and Latin America together for many analyses and found evidence for increased risk of stroke in current OC users in this group.^8–10 An early study in Taiwan of OC use and stroke mortality¹¹ was the only study in Asia in a recent meta-analysis¹² and one of the few that found no association, rather than increased risk. More recently, studies in Asian women have shown increased risk of stroke to be associated with increased number of pregnancies or births and later age of menarche and increased risk of CHD to be associated with early age at menopause.^13–15

The current study assessed possible etiologic associations of a wide range of reproductive factors with CVD mortality in a prospective study of a large cohort of female workers employed in the textile industry in Shanghai, China.

Methods

Study population

Female workers in 526 textile factories in Shanghai were enrolled in a randomized trial of breast self-examination in 1989–91.¹⁶ Women were actively followed for cancer incidence and vital status through 2000. The cohort included 267 400 active and retired women born between 1 January 1925 and 31 December 1958. At enrolment, pre-coded questionnaires were administered to subjects by trained medical workers to elicit information on reproductive and contraceptive history, cigarette smoking and other risk factors. The study was approved by the Institutional Review Boards at Fred Hutchinson Cancer Research Center, the University of Washington and the Station for the Prevention and Treatment of Cancer of the Shanghai Textile Industry Bureau (STIB).

Outcome definition

The STIB periodically received information on all cancers and all deaths from each factory, and maintained a cancer and death registry. Cause of death information was collected from this registry and classified according to the International Classification of Diseases, ninth revision (ICD-9 code). End-points include deaths from IHD (ICD-9 codes 410–414), ischaemic stroke (434) and intra-cerebral haemorrhage (431). Each member of the cohort contributed person-years from the date of enrolment to date of death, lost to follow-up (severed ties with the STIB) or end of vital status follow-up period (31 December 2000).

Data analysis

Mortality rates were computed for each CVD cause of death, and age-adjusted by the direct method using the 2000 World Standard Population in 5-year age categories as the standard population.¹⁷ These rates were also calculated separately by smoking status (ever/never).

Multivariable analyses were conducted using Cox proportional hazards modelling to estimate hazard ratios (HRs) and 95% confidence intervals (95% CIs) for reproductive factors associated with each CVD outcome. Smoking is a known risk for CVD,¹ and as the prevalence of smoking was exceedingly low in this cohort (2.9%), there was a rare opportunity to examine the effect of reproductive factors on CVD mortality in non-smoking women by restricting the analysis to non-smokers. Analyses on the entire cohort were also performed to evaluate the expected association between CVD and smoking.

Information for all women was only available at baseline; thus, all factors were analysed as fixed variables. Potential confounding variables were chosen for preliminary models a priori based on established associations and/or plausible biological relations with the exposures of interest. Co-variates were retained if they changed the relative risk estimates for one or more outcomes by >10%. The same co-variate models were applied to all three CVD outcomes for comparability. Median ages at baseline were compared among categories of CVD risk factors to assess possible birth cohort effects, as women's reproductive practices have changed substantially over time in China. To ensure complete adjustment for age at enrolment, we also included an age-squared term in the model if it changed the relative risk estimates for one or more outcomes by >10%.

Trends of relative risk were estimated treating reproductive factors as continuous variables in final regression models. Multiplicative interactions between smoking and contraceptive use were assessed with the likelihood ratio test. All statistical analyses were completed in SAS 9.2 (SAS Institute, Cary, NC, USA).

Results

Women in the cohort contributed 2 565 433 person-years of follow-up and ranged in age at enrolment from 30 to >60 years old (Table 1). The majority of women had ever been pregnant and never smoked, but prevalence of many factors of interest varied by birth cohort. Over 80% of the singleton births were born to the younger women, who were in their early reproductive years when China's one-child-per-family policy was introduced. In contrast, 97% of the women who had five or more births were born prior to 1936. These older women also constituted 82% of those who ever smoked.

Table 1.

Distribution of selected lifestyle and reproductive characteristics of female textile workers in Shanghai, China (1989–91)

Characteristic	Number of women (N = 267 400) n (%)	Median age (years)
Age group at study entry
All women (age in years)		43
30–39	106 855 (40.0)
40–49	51 527 (19.3)
50–59	65 385 (24.5)
≥60	43 633 (16.3)
Smoking history
Never	259 494 (97.0)	42
Ever	7868 (2.9)	59
Nulligravid women	11 375 (4.3)	40
Livebirth historya
0	1907 (0.7)	39
1	129 145 (48.3)	36
2	42 317 (15.8)	49
3	33 726 (12.6)	57
4	24 914 (9.3)	59
≥5	24 014 (9.0)	61
Age at first livebirth (years)
≤19	12 460 (4.7)	58
20–24	72 570 (27.1)	57
25–29	124 044 (46.4)	38
≥30	45 041 (16.8)	38
Oral contraceptive use
Never	229 327 (85.8)	42
Ever	38 072 (14.2)	44
Menopause
Natural	98 180 (36.7)	50
Surgical	11 052 (4.1)	43

^aIncludes only women with previous pregnancies.

Overall in the cohort, there were 4349 deaths from CVD. There were 2776 deaths from stroke, including 699 from ischaemic stroke and 1815 from haemorrhagic stroke. There were 494 deaths from IHD. The age-standardized mortality rates were 6.9/100 000 person-years for IHD and 37.9/100 000 for all strokes (8.7 for ischaemic stroke and 24.9 for haemorrhagic stroke).

Smoking

The rates per 100 000 person-years for non-smokers and smokers were, respectively, 6.5 and 20.6 for IHD (IHD: HR = 1.60, 95% CI 1.20–2.14), 8.3 and 15.9 for ischaemic stroke (HR = 1.79, 95% CI 1.43–2.25) and 23.9 and 92.9 for haemorrhagic stroke (HR = 1.20, 95% CI 1.01–1.43). Among smokers, risks of IHD and ischaemic stroke mortality tended to increase with years of smoking determined at baseline (data not shown). We also assessed risk in relation to duration of smoking over follow-up, assuming that current smokers at baseline did not quit and never and former smokers did not start smoking. Data available for a subset of the cohort with smoking information from several years after the baseline questionnaire indicated that these assumptions were valid. The results were very similar to years of smoking at baseline. Results for analysis of the reproductive factors presented below are for women who reported at baseline that they had never smoked. They are similar to results for the entire cohort when adjusted for smoking (data not shown).

Pregnancy

No associations were observed with number of livebirths, age at first livebirth, spontaneous abortions or induced abortions, although the power was low to detect relations with intentionally aborted first pregnancies. There was some suggestion of slightly increased risk of CVD mortality associated with more than five births (Table 2). A slightly increased risk of haemorrhagic stroke mortality was seen for women whose first birth ended in a stillbirth, but risk was not increased in women with multiple stillbirths. Conversely, no association was seen with ischaemic stroke mortality and first pregnancy ending in a stillbirth, but increased risk was observed in women who had multiple stillbirths.

Table 2.

Adjusted HRs and 95% CIs for cardiovascular disease mortality in relation to reproductive history among non-smoking female textile workers in Shanghai, China (1989–2000)

		Ischaemic heart disease			Ischaemic stroke			Haemorrhagic stroke
Reproductive characteristic	Person- years at risk	Numbera	HR (95% CI)	P trend	Numbera	HR (95% CI)	P trend	Numbera	HR (95% CI)	P trend
Number of livebirthsb,c,d
0	120 932	22	1.17 (0.66–2.10)	0.20	37	1.15 (0.72–1.82)	0.01	87	1.31 (0.98–1.76)	0.05
1	1 193 771	33	1.00 (referent)		45	1.00 (referent)		133	1.00 (referent)
2	412 279	67	0.98 (0.62–1.57)		69	0.70 (0.46–1.04)		278	1.07 (0.84–1.35)
3	319 960	100	0.99 (0.62–1.57)		124	0.78 (0.53–1.15)		385	1.03 (0.82–1.31)
4	230 629	89	0.97 (0.60–1.55)		121	0.78 (0.53–1.15)		349	1.07 (0.84–1.36)
≥5	211 391	131	1.23 (0.78–1.95)		216	1.10 (0.76–1.59)		438	1.22 (0.96–1.55)
Age at first livebirth (years)b,c
≤19	110 194	43	0.93 (0.67–1.28)	0.11	75	1.23 (0.96–1.59)	0.06	178	1.09 (0.92–1.28)	0.0002
20–24	676 422	239	1.00 (referent)		309	1.00 (referent)		850	1.00 (referent)
25–29	1 160 345	101	0.75 (0.59–0.95)		141	0.88 (0.72–1.07)		423	0.84 (0.74–0.94)
≥30	421 047	37	0.85 (0.60–1.20)		50	0.95 (0.71–1.29)		132	0.81 (0.67–0.97)
First pregnancy outcomec,e
Livebirth	2 191 094	394	1.00 (referent)		554	1.00 (referent)		1496	1.00 (referent)
Spontaneous abortion	124 332	22	1.23 (0.78–1.94)		15	0.58 (0.33–1.01)		61	0.90 (0.68–1.17)
Stillbirth	27 523	6	0.89 (0.40–2.01)		8	0.80 (0.40–1.61)		29	1.17 (0.81–1.69)
Induced abortion	38 767	1	0.61 (0.08–4.44)		2	0.94 (0.22–3.97)		7	1.04 (0.49–2.22)
Other	3094	2	2.04 (0.45–9.26)		0	–		3	0.86 (0.26–2.82)
History of stillbirthc,f
No	2 321 098	408	1.00 (referent)	0.78	550	1.00 (referent)	0.25	1517	1.00 (referent)	0.34
Single	55 934	13	0.76 (0.43–1.31)		21	0.86 (0.55–1.33)		68	1.09 (0.86–1.40)
Multiple	7744	4	1.02 (0.38–2.73)		8	1.36 (0.68–2.75)		11	0.80 (0.44–1.44)
History of spontaneous abortionc,f
No	2 092 348	350	1.00 (referent)	0.16	482	1.00 (referent)	0.90	1303	1.00 (referent)	0.69
Single	233 896	51	0.88 (0.66–1.19)		72	0.87 (0.68–1.12)		224	1.06 (0.92–1.22)
Multiple	58 565	24	1.27 (0.84–1.92)		25	0.90 (0.60–1.35)		69	1.01 (0.80–1.29)
History of induced abortionc,f
No	1 114 138	270	1.00 (referent)	0.14	385	1.00 (referent)	0.52	1009	1.00 (referent)	0.04
Single	889 671	120	1.00 (0.80–1.24)		138	0.86 (0.71–1.05)		400	0.86 (0.76–0.96)
Multiple	381 000	35	0.76 (0.53–1.08)		56	0.96 (0.72–1.27)		187	1.03 (0.88–1.20)
Breastfeeding history (months)c,e,g
Never (with livebirth)	368 108	54	1.00 (referent)	0.23	47	1.00 (referent)	0.08	143	1.00 (referent)	0.35
<6	340 530	21	0.70 (0.42–1.16)		25	1.02 (0.63–1.66)		69	0.84 (0.63–1.12)
7–12	702 203	37	0.50 (0.33–0.76)		64	1.05 (0.72–1.54)		198	0.98 (0.79–1.22)
13–24	372 338	74	0.67 (0.46–0.97)		76	0.90 (0.62–1.31)		302	1.01 (0.82–1.24)
25–36	234 664	60	0.53 (0.36–0.79)		98	1.15 (0.79–1.67)		257	0.88 (0.71–1.09)
37–48	155 294	61	0.71 (0.48–1.06)		88	1.21 (0.83–1.77)		234	1.02 (0.82–1.28)
≥49	194 892	113	0.78 (0.53–1.14)		177	1.20 (0.84–1.72)		380	1.05 (0.84–1.30)
Age at menarche (years)c,h
≤13	256 455	47	1.44 (1.00–2.05)	0.007	50	1.05 (0.75–1.45)	0.49	130	0.97 (0.79–1.19)	0.06
14	459 130	63	1.06 (0.76–1.47)		71	0.82 (0.61–1.09)		251	1.03 (0.88–1.22)
15	556 680	83	1.00 (referent)		123	1.00 (referent)		336	1.00 (referent)
16	531 148	107	1.09 (0.82–1.45)		143	0.97 (0.76–1.23)		361	0.92 (0.79–1.07)
≥17	684 948	142	0.85 (0.65–1.12)		225	0.87 (0.70–1.09)		592	0.90 (0.79–1.03)
Age at natural menopause (years)c,h
<40	16 029	7	1.10 (0.52–2.34)	0.48	13	1.18 (0.68–2.05)	0.31	30	1.21 (0.84–1.75)	0.55
40–44	46 200	20	1.12 (0.70–1.77)		24	0.78 (0.52–1.19)		63	0.90 (0.69–1.16)
45–49	347 760	156	1.18 (0.95–1.46)		182	0.82 (0.68–0.99)		524	1.00 (0.89–1.12)
50–54	450 359	175	1.00 (referent)		295	1.00 (referent)		691	1.00 (referent)
≥55	44 349	23	1.24 (0.81–1.92)		26	0.83 (0.55–1.23)		85	1.17 (0.94–1.47)
Cause of menopausec,d
Pre-menopausal	1 479 020	32	1.00 (referent)		25	1.00 (referent)		157	1.00 (referent)
Natural	906 697	381	1.35 (0.73–2.50)		542	1.97 (0.99–3.93)		1397	1.14 (0.86–1.51)
Surgical	101 652	27	1.23 (0.61–2.48)		44	2.08 (1.00–4.33)		113	1.18 (0.85–1.63)

^aNumber of mortalities.

^bTrend includes only women who have had one or more livebirths.

^cAdjusted for age.

^dAdjusted for age-squared.

^eAdjusted for number of livebirths.

^fTrend from single to multiple outcome.

^gTrend includes only women who have breastfed.

^hTrend includes women in all age categories.

Breastfeeding history

No consistent trends in risk of any of the three causes of death were observed in relation to duration of breastfeeding, although parous women who did not breastfeed may have been at slightly increased risk of IHD.

Age at menarche

Risk of death from IHD was increased in women with menarche at the age of ≤13 years. No consistent associations with age at menarche were found for deaths from ischaemic or haemorrhagic stroke.

Menopause

Compared with women of the same age, women who reported natural or surgical menopause at baseline were at increased risk of mortality from IHD and haemorrhagic and ischaemic stroke, but there were no trends in risk with age at natural menopause. There were too few deaths to assess trends in relation to age at surgical menopause.

Exogenous hormones

Women who had ever used OCs were at decreased risk of mortality from all three conditions considered, with the weakest association seen for haemorrhagic stroke, and there were no trends in risk with duration of use (Table 3). Most women used Chinese Pill No.1. Use of monthly injectable contraceptives was associated with a decreased risk of mortality from IHD, and possibly also ischaemic stroke, but not from haemorrhagic stroke. Women who were given progesterone injections to prevent miscarriage were at a decreased risk of ischaemic stroke mortality, but there was no trend in risk with the number of pregnancies during which treatment was received.

Table 3.

Adjusted HRs and 95% CIs for cardiovascular disease mortality in relation to exogenous reproductive factors among non-smoking female textile workers in Shanghai, China (1989–2000)

		Ischaemic heart disease			Ischaemic stroke			Haemorrhagic stroke
Reproductive characteristic	Person- years at risk	Numbera	HR (95% CI)	P trend	Numbera	HR (95% CI)	P trend	Numbera	HR (95% CI)	P trend
OC useb,c
Never	2 122 083	407	1.00 (referent)		576	1.00 (referent)		1503	1.00 (referent)
Ever	366 890	35	0.79 (0.56–1.12)		36	0.65 (0.46–0.91)		167	0.97 (0.82–1.14)
Current	27 311	2	2.38 (0.58–9.76)		0	–		5	1.28 (0.53–3.10)
Past	339 579	33	0.76 (0.53–1.09)		36	0.66 (0.47–0.92)		162	0.96 (0.82–1.13)
OC typec
Never	2 122 083	407	1.00 (referent)		576	1.00 (referent)		1503	1.00 (referent)
Pill no.1	319 983	29	0.74 (0.50–1.08)		33	0.67 (0.47–0.96)		143	0.93 (0.78–1.11)
Other	36 133	6	1.55 (0.69–3.49)		3	0.62 (0.20–1.93)		21	1.40 (0.91–2.15)
Duration of OC use (years)c
Never	2 122 083	407	1.00 (referent)	0.86	576	1.00 (referent)	0.06	1503	1.00 (referent)	0.28
<5	277 834	24	0.82 (0.54–1.24)		18	0.50 (0.31–0.80)		106	0.92 (0.75–1.12)
≥5	88 553	11	0.74 (0.41–1.35)		18	0.94 (0.59–1.51)		61	1.07 (0.83–1.38)
Injectable contraceptive useb,c
Never	2 367 632	437	1.00 (referent)		606	1.00 (referent)		1620	1.00 (referent)
Ever	121 352	5	0.44 (0.18–1.06)		6	0.45 (0.20–1.00)		50	1.10 (0.83–1.46)
Current	7877	0	–		0	–		0	–
Past	113 475	5	0.45 (0.18–1.08)		6	0.45 (0.20–1.02)		50	1.13 (0.85–1.50)
Type of injectable contraceptivec
Never	2 367 632	437	1.00 (referent)		606	1.00 (referent)		1620	1.00 (referent)
Monthly	114 880	4	0.37 (0.14–1.00)		5	0.40 (0.17–0.96)		46	1.09 (0.81–1.46)
Every 3 months	4174	1	1.96 (0.28–13.98)		0	–		2	1.00 (0.25–4.01)
Miscarriage injectionsc
Never	2 124 972	401	1.00 (referent)		560	1.00 (referent)		1501	1.00 (referent)
Ever	259 838	24	0.88 (0.58–1.34)		19	0.55 (0.35–0.87)		95	0.89 (0.72–1.09)
Number of pregnancies receivedc
Never	2 124 972	401	1.00 (referent)	0.26	560	1.00 (referent)	0.69	1501	1.00 (referent)	0.82
1	221 381	21	0.99 (0.64–1.54)		14	0.52 (0.31–0.89)		75	0.89 (0.71–1.13)
≥2	38 457	3	0.50 (0.16–1.56)		5	0.63 (0.26–1.53)		20	0.87 (0.56–1.35)

^aNumber of mortalities.

^bEver use defined as current or past use reported at baseline.

^cAdjusted for age.

Interactions

No meaningful interactions were seen between smoking (ever/never) and use of OCs, injectable contraceptives or injections to prevent miscarriage (data not shown), but the power to detect such interactions was low because of the small number of smokers. Residual confounding by age was not evident, as there were only slight differences in results for use of contraceptives or injections for women aged <50 and ≥50 years. However, the power to detect such differences by age was limited due to the small numbers of deaths for women <50 years old.

Discussion

In general, the reproductive history of Chinese women in this cohort did not appear to impact risk of CVD mortality. The current study did not confirm prior reports of associations between risks of death from CVD and pregnancy-related factors. Parity has been associated with increased risk of CHD and stroke in other cohorts in the USA, Europe and Asia,^{14,15,18–21}, although primarily with four or more births, and no associations were shown in two large US cohorts.¹⁸ Further study of CVD risk in women with very high parity is warranted, given the slight increased risk for IHD for five or more births observed in this study and some others.^19,20 Some previous studies have shown an increased CVD risk with pregnancy loss, but factors such as pregnancy-induced hypertension may cause the loss and also increase CVD risk.^18,22 Trends in decreasing risk with increasing duration of breastfeeding found in other studies were not observed here, but results were consistent in finding women who never lactated may have increased risk of IHD mortality.²³ Lactation may reduce risk by mobilizing fat stores, or a lack of breastfeeding may spuriously appear to increase risk because lactation failure is associated with obesity and insulin resistance, which in turn are CVD risk factors.²³

A few studies have shown results similar to ours for increased risk of IHD in women with an early age at menarche^24,25, although a study of Japanese women found no such association.¹³ Confirmation in other populations and further exploration of the biological mechanisms is warranted, including examining age at menarche as a marker for other CVD risk factors. Post-menopausal women at baseline were at increased risk of all three end-points compared with pre-menopausal women of the same age. No trends in risk with age at menopause were observed. This is contrary to other observations.^5,13,26 A true trend could have been obscured in our data because many women entered menopause over the 10–12-year follow-up period but were classified as pre-menopausal.

Increased CVD risk in OC users was observed in early studies, but primarily for use of preparations with high doses of oestrogen (≥50 µg), and mainly attributed to synergistic effects with smoking and hypertension.⁷ Interactions with smoking were not observed here, although the low frequency of smoking limited our power to detect such an interaction. Chinese Pill No.1 was the most commonly used OC and contained 0.625 mg norethisterone and 35 µg ethinyl estradiol until 1974 and 0.300 mg norethisterone and 30 µg ethinyl estradiol thereafter. For other similar low-dose OCs, only current use, not assessed in this study, has been associated with increased risk of myocardial infarction and possibly ischaemic stroke.^12,27 With the exception of a WHO study,⁹ past low-dose OC use has primarily been associated with decreased risk or no change in risk for myocardial infarction^28,29 and ischaemic stroke.^8,30–32 Our results primarily reflected past OC use and were consistent with this evidence. There may be a prothrombotic effect, potentially exerted by affecting coagulation factors, which is reversed when use is stopped, or there may be a persistent benefit, similar to the anti-atherosclerotic effect shown in animal models for oestrogen.³³ No increased risk of CVD mortality was observed for injections of progesterone during pregnancy to prevent miscarriage or for use of monthly combined injectable contraceptives. Although the power to detect associations was limited by small numbers, the result for injectable contraceptives is consistent with that of a similar WHO study.³⁴

Important study strengths are the wide variation in values of the major pregnancy-related variables of interest and large number of person-years in each category, providing sufficient power to detect true associations. Women were relatively homogeneous for socio-economic factors (income, education), which are potential confounders linked to parity and behavioural patterns that can influence CVD risk, such as diet and physical activity level.^1,35 Residual confounding by such factors could explain the increased risk observed in some other studies. This study also had the unique ability to examine risks in relation to use of one type of OC (in two doses), because most women used the same formulation.

Available data indicate that CVD mortality in the cohort was likely to be similar to or lower than that for women in Shanghai. Using the same standardizing population, a surveillance project reported annual mortality rates for coronary events from 1.0 to 3.0/100 000 for women aged 35–64 years old in Shanghai, and from 29.5 to 43.3/100 000 for stroke.³⁶ In women of the same age, the rate of IHD (4.3/100 000) and stroke (26.3/100 000) is similar in our cohort. Women working in the textile industry appear to have CVD mortality rates comparable with the general population in Shanghai, although there are indications of a possible healthy worker effect. These similarities indicate vital status follow-up was likely successful in identifying most CVD deaths that occurred in the cohort.

This study has some potential limitations. Only fatal outcomes were assessed, unlike many other studies that included both fatal and non-fatal outcomes. Steroid contraceptive use in the cohort was infrequent and for short durations, limiting power to detect altered risks, especially in relation to long-term use. Thrombotic events are more common during the first year of use;⁷ thus, users who died from CVD before enrolment may have been missed.

Reproductive history was collected by questionnaire and subject to misclassification from inaccurate self-reporting. This is unlikely to be related to CVD mortality and would be non-differential, tending to bias results towards the null. Other analyses in the cohort have found lung cancer to be associated with smoking and breast cancer to be associated with reproductive factors.^37,38 These expected results provide evidence that the questionnaire provided valid information. Report validity is also evident by differences in the cohort after institution of the one-child-per-family policy, with younger women representing >70% of the induced abortions and OC use reported.

Fixed baseline variables may be misclassified for factors that can change over time, particularly duration of smoking, menopausal status and contraceptive use. Adjusting for follow-up time did not change the association with duration of smoking seen using years smoked as reported at baseline. Pre-menopausal women would be expected to become menopausal over the follow-up, and given the one-child-per-family policy, non-OC users may have become users, but this is not likely to happen since use of OCs is typically short term to delay the first pregnancy and women then choose childbearing sterilization. Both types of misclassification would have tended to bias the results towards the null value. Variables related to childbearing were unlikely to change since very few women were nulliparous at the beginning of the study and few women would be expected to have more children due to the one-child-per-family policy. Unlike most studies of induced abortion, women in China are likely to report having had the procedure because it is associated with little or no social stigma.

Cause of death codes could not be validated, but a study in Chinese cities, including Shanghai, found reasonable sensitivity for cerebrovascular disease (82%) and IHD (69%) for death registration compared with clinical diagnoses.³⁹ Misclassification of stroke subtype may have occurred since the use of brain imaging for diagnostic evaluation appears to have increased during this time period, as indicated by a recent Chinese study.⁴⁰

Information on blood cholesterol levels, coagulation factors, blood pressure, blood lipids, diabetes and obesity were not available. Although these might be regarded as potential confounders, it is more likely that they are largely intermediates on the causal pathway, in which case, statistical control would not be appropriate. For example, oestrogen use can affect lipoproteins and blood pressure, which in turn affects subsequent development of CVD.³³ Other factors such as prevalent diabetes or CVD are generally rare or unrecognized in women of reproductive age and unlikely to have affected decisions surrounding pregnancy and contraceptive use and hence confound the results. Information on family history of CVD was also unavailable, but would unlikely have been associated with any independent variables considered.

This study was able to examine a large cohort of Asian women for CVD mortality in relation to a wide range of reproductive factors. Induced abortions and hormonal contraceptives and the one-child-per-family policy are unlikely to have played a role in CVD mortality in this cohort.

Funding

The National Cancer Institute at the National Institutes of Health (grant number R01CA80180).

Conflict of interest: None declared.

KEY MESSAGES.

• Possible effects of reproductive factors on CVD risks in a cohort of Asian women were examined.

• For most reproductive factors, there were no consistent associations with CVD. In particular, there was no evidence of increased CVD mortality risk by oral or injectable contraceptive use, induced abortions and reduced parity.

• The results suggest that China's one-child-per-family policy has not had an adverse effect on risk of CVD mortality in this cohort.