Menstrual and reproductive factors and endometrial cancer risk: Results from a population-based case-control study in urban Shanghai

Authors


Abstract

The purpose of our study was to evaluate the association of menstrual and reproductive factors with the risk of endometrial cancer. In a population-based case-control study conducted in urban Shanghai, in-person interviews were completed for 833 women aged 30–69 years and an equal number of controls frequency-matched to cases by age. All cases were newly diagnosed with endometrial cancer between January 1, 1997 and December 31, 2001. The unconditional logistic regression model was employed to derive the adjusted odds ratios (ORs) of endometrial cancer and 95% confidence intervals (CIs) in relation to menstrual and reproductive factors. Earlier menarche age, particularly among premenopausal women, and later menopausal age were associated with an elevated risk of endometrial cancer. A clear dose-response relation between endometrial cancer risk and years of menstruation was observed (p for trend < 0.01). Compared to women ever having a pregnancy and women ever having had a live birth, respectively, nulligravity and nulliparity were both associated with a more than one-fold elevated risk of endometrial cancer. Both completed (OR = 3.02, 95% CI 1.10–8.32 for women never having a complete pregnancy) and incomplete pregnancy (OR = 0.69, 95%CI 0.55–0.87) conferred a protective effect against endometrial cancer, and the protective effect appeared to increase with total number of pregnancies (p for trend = 0.01). The effect of pregnancy on endometrial cancer remained unchanged with increasing time since the last pregnancy. Stillbirth and age at first pregnancy was unrelated to endometrial cancer risk. Our study suggests that prolonged menstruation was related to an increased risk of endometrial cancer while pregnancy, including induced abortion, reduced the risk of endometrial cancer. © 2003 Wiley-Liss, Inc.

Chinese women historically have a lower risk of endometrial cancer compared to their counterparts in the United States and other western countries. However, the incidence of endometrial cancer in Shanghai women has substantially increased over the past 2 decades. The age-adjusted endometrial cancer incidence rate increased 76% from 2.5/100,000 in 1972–74 to 4.4/100,000 in 1994, becoming 1 of the 10 most common cancers among Shanghai women.1 The increase was more pronounced among older women, with 132.1% and 2.08% for women 55–64 and 35–44, respectively, for the same time period. Menstrual and reproductive factors, such as age at menarche, age at menopause, nulligravity and live birth, have been associated with the risk of endometrial cancer in many previous studies, including a study conducted among Chinese women.2, 3, 4, 5 Prolonged estrogen stimulation without the protection of progesterone has been suggested as the possible mechanism for the association of menstrual and reproductive factors with the risk of endometrial cancer.6, 7, 8 However, results from earlier studies were not entirely consistent, particularly regarding the effect of induced abortion.4, 5, 9 The high frequency of receiving hormone replacement therapy and hysterectomy among western populations makes it difficult to study attribution of menstrual and reproductive factors on endometrial cancer and may have caused some of the inconsistency.

The major objective of our study was to test the hypothesis that prolonged menstruation increases and pregnancy, including induced abortion, reduces the risk of endometrial cancer among women in Shanghai. The populated-based Shanghai cancer registry, high rate of induced abortion, low rate of hysterectomy and low frequency of using hormonal replacement therapy among women in Shanghai provided us a unique opportunity to study the hypothesis.

MATERIAL AND METHODS

All cases were permanent residents of urban Shanghai between the ages of 30 and 69 years who were newly diagnosed with endometrial cancer during the period of January 1997 to December 2001. Through the population-based Shanghai Cancer Registry, 983 eligible endometrial cancer cases were identified, and in-person interviews were completed for 833 (84.7%) of them. Among 150 nonparticipants, there were 73 refusals (7.4%), and 36 deaths before interview (3.7%), while 22 could not be located (2.2%), 10 were absent during the study period (1.0%) and 9 were excluded for other miscellaneous reasons (0.9%).

Controls were randomly selected from female permanent residents in the Shanghai urban area and frequency-matched to cancer cases by age (5-year interval) in a 1:1 ratio. The random selection was performed at the Shanghai Resident Registry. Women who had a prior history of endometrial cancer or hysterectomy were ineligible for the study (5.1%). The number of controls in each age-specific stratum was determined in advance according to the age distribution of the incident endometrial cancer cases reported to the Shanghai Cancer Registry in 1996. In-person interviews were completed for 833 of the 1,165 eligible controls identified. The response rate for the controls was 71.5%. The reasons for nonparticipation were refusal (24.9%) and absence during the study period (3.6%).

Interviewers were trained nurses. The structured questionnaire used for the study took, on average, 70 min to complete. It covered demographic factors, menstrual and reproductive history, hormone use, usual dietary intake, disease history, tobacco and alcohol use, weight history and family history of cancer. Date of diagnosis (for cases) and interview (for controls) were used as the “reference date” during the interview. All exposure information was collected up to or before the reference date, often one year before the interview date. Data collected on reproductive history included outcome and duration of each pregnancy. Menopause was defined as cease of menstrual period, excluding those caused by pregnancy and nursing, for at least 12 months before the reference date. The median interval between diagnosis and interview (25th, 75th percentile) for cases was 5 months (3 months, 8 months). Years of menstruation were age at reference date minus age at menarche for premenopausal women, and age at menopause minus age at menarche for postmenopausal women. Each participant was also measured for current weight, circumferences of the waist and hip and sitting and standing heights. All interviews were tape-recorded and were reviewed by the field supervisor and quality control staff to monitor the quality of interview data. Additional data quality procedures included range and logic check and double entry.

Odds ratios (ORs) were used to measure the association of endometrial cancer risk with menstrual and reproductive characteristics. The unconditional logistic regression model was employed in the analysis to obtain maximum likelihood estimates of the ORs and their 95% confidence intervals (CIs) with an adjustment for potential confounders.10 Collinearity between variables was checked before they were included in the regression model. Categorical variables were treated as dummy variables in the model. Tests for trend were performed by entering the categorical variables as continuous parameters in the models.

RESULTS

Cases and controls were well matched on age distribution (Table I). The mean age was 55.2 years for cases and 55.1 years for controls. There were no major case-control differences with respect to marital status, per capita income one year before interview or in 1978 and total energy intake. Compared to controls, cases were more likely to be better educated (p = 0.05), have a higher body mass index (BMI) (p < 0.01) and waist-to-hip ratio (WHR) (p < 0.01), more likely to have a family history of cancer among first-degree relatives (p < 0.01), less likely to be a drinker (p < 0.01) and less likely to have used oral contraceptives (p < 0.01). All subsequent analyses included these variables and age as potential confounders.

Table I. Comparison of Cases and Controls on Demographics and Selected Endometrial Cancer Risk Factors1
 Cases (n = 833)Controls (n = 833)p-value
No.%No.%
  • 1

    Subjects with missing values were excluded from the analysis.

Age (mean ± SD)55.2 ± 8.7055.1 ± 8.560.83
Per capita income in last year (yuan)     
 ≤4166.723127.824028.8 
 4166.8∼5833.3688.2759.0 
 5833.4∼7500.020424.519323.2 
 7500.1∼8750.015118.116019.2 
 >8750.017821.416419.70.80
Per capita income in 1978 (yuan)     
 ≤187.515618.819423.5 
 187.6∼250.021125.420524.8 
 250.1∼312.512915.612815.5 
 312.6∼416.716920.415819.1 
 >416.716419.814217.20.19
Education     
 ≤Elementary20124.123227.9 
 Middle & high50660.750460.5 
 Prof. & College12615.19711.60.05
Material Status     
 Unmarried141.7111.3 
 Married or cohabit72486.973087.6 
 Seperation/divorce/bereft of spouse9511.49211.00.81
Oral contraceptive use     
 Ever14817.820724.9 
 Never68582.262675.2<0.01
Cancer among first degree relatives     
 No53564.860573.1 
 Yes29035.222326.9<0.01
Body Mass Index (by quartile)     
 ≤21.41112314.920925.2 
 21.412∼23.68116620.120825.0 
 23.682∼26.27121726.320624.8 
 >26.27132038.720825.0<0.01
Waist-to-Hip Ratio (by quartile)     
 ≤0.78210612.820724.9 
 0.783∼0.81616219.520925.1 
 0.817∼0.85523928.820724.9 
 >0.85532338.920925.1<0.01
Regular alcohol drinking     
 Ever202.4425.0 
 Never81397.679195.0<0.01
Total energy intake (Kcal, Mean ± SD)2213.1 ± 492.42201.9 ± 512.50.66

Table II presents ORs and 95% CIs for menstrual characteristics in association with endometrial cancer risk. There was an inverse association between age at menarche and endometrial cancer risk (p for trend = 0.02). Compared to women with menarchal age 12 or less, women with menarchal age 17 or more had a 45% (95% CI 0.36–0.83) lower risk of endometrial cancer. Cases were more likely than controls to be premenopausal (OR = 0.61, 95% CI 0.45–0.84). Older age at menopause was associated with an elevated risk of endometrial cancer (p for trend < 0.001). Compared to subjects whose menopause occurred before age 45, those whose menstrual periods continued until age 55 or later had a 5-fold (95% CI 2.48–10.69) increased risk of endometrial cancer. The risk of endometrial cancer increased with years of menstruation (p for trend < 0.01). Compared to women whose days of monthly menstrual bleeding was 5–7 days, women whose monthly bleeding was less than 5 days had a lower risk of endometrial risk (OR = 0.60, 95% CI 0.48–0.75). Length and regularity of menstrual cycle was unrelated to the risk of the disease.

Table II. Odds Ratios (ORs) of Endometrial Cancer Associated with Menstrual Characteristics1
 CasesControlsOR1 (95% CI)OR2 (95% CI)
  • 1

    OR1, Adjusted for age; OR2, Adjusted for age, education, oral contraceptives use, alcohol drinking, any cancer history among first degree relatives, body mass index, ever having had a live birth and ever having had an induce abortion.

  • 2

    Additionally adjusted for menstrual status.

  • 3

    Additinally adjusted for age at menarcheal.

  • 4

    Additionally adjusted for menstrual status, age at menarcheal.

Age at menarche (years)    
 ≤12109771.001.002
 131671470.80 (0.56–1.16)0.79 (0.54–1.15)
 141671520.78 (0.54–1.12)0.81 (0.55–1.18)
 151661770.66 (0.46–0.95)0.76 (0.52–1.12)
 161281410.64 (0.44–0.94)0.75 (0.51–1.13)
 ≥17941370.49 (0.33–0.72)0.55 (0.36–0.83)
   <0.010.02
Menstrual status    
 Premenopause3302941.001.003
 Natural menopause4915340.62 (0.46–0.84)0.61 (0.45–0.84)
 Induced menopause1251.84 (0.64–5.34)1.68 (0.56–5.07)
Age at menopause (among women with natural menopause)
 <4531661.001.003
 45∼491922471.65 (1.03–2.63)1.67 (1.03–2.70)
 50∼542202042.27 (1.42–3.63)2.38 (1.46–3.87)
 ≥5548175.80 (2.86–11.75)5.15 (2.48–10.69)
   <0.01<0.01
Years of menstruation    
 <301231731.001.002
 30∼2863371.34 (1.00–1.81)1.30 (0.95–1.78)
 35∼3272701.99 (1.46–2.72)1.93 (1.38–2.70)
 40∼95513.21 (2.06–5.02)2.71 (1.67–4.40)
   <0.01<0.01
Length of menstrual cycle    
 26∼305535401.001.004
 <261281490.84 (0.64–1.09)0.90 (0.68–1.19)
 >301521441.03 (0.80–1.33)1.00 (0.76–1.31)
Regularity of menstrual cycle    
 Always regular7707931.001.004
 Always irregular40261.58 (0.86–2.62)1.37 (0.80–2.35)
 Sometimes irregular22141.62 (0.82–3.19)1.61 (0.80–3.26)
Days of monthly menstrual bleeding    
 5∼75715091.001.004
 <52072890.64 (0.52–0.79)0.60 (0.48–0.75)
 >744271.45 (0.89–2.38)1.62 (0.96–2.75)

Nulligravity was associated with a significantly elevated risk of endometrial cancer (OR = 1.82, 95% CI 1.16–2.85) (Table III). Exclusion of unmarried women (14 cases and 11 controls) did not change the association. Among subjects who were involuntarily infertile (i.e., tried to become pregnant for 2 or more years without success), 17 cases and 7 controls reported to have a physician diagnosed fertility problem. The protective effect of pregnancy on the risk of endometrial cancer increased with the increasing number of pregnancies (p for trend = 0.01). Although women whose latest pregnancy ended with a miscarriage had a marginally elevated risk (OR = 2.27, 95% CI = 0.97–5.25), ever having had a miscarriage did not appear to be related to the risk of endometrial cancer (OR = 1.03, 95% CI = 0.67–1.58). Induced abortion only before first birth was rare (19 cases and 15 control) in the study population and was not related to the risk of the disease. Age at last pregnancy was not significantly associated with the risk of endometrial cancer (data not shown). Neither the number of live births nor the number of induced abortions conferred additional protection from endometrial cancer. The protective effort of pregnancy on endometrial cancer remained little changed for more than 30 years after the last pregnancy among all gravid women and women who had no induced abortion after their last birth. Excluding women who had ever used hormone replacement therapy (HRT) did not change the associations (data not shown).

Table III. Odds Ratios (ORs) of Endometrial Cancer Associated with Reproductive Characteristics1
 CasesControlsOR1 (95% CI)OR2 (95% CI)
  • 1

    OR1, Adjusted for age; OR2, Adjusted for age, education, oral contraceptives use, years of menstruation, menstrual status any cancer history among first degree relatives, body mass index and alcohol drinking.

  • 2

    Additionally adjusted for ever having had a live birth, ever having had an induce abortion, number of pregnancy.

  • 3

    Additionally adjusted for number of incomplete pregnancy.

  • 4

    Additionally adjusted for number of complete pregnancy.

  • 5

    Additionally adjusted for ever having had a live birth, number of pregnancy.

  • 6

    Conditions in footnote 2 plus excluding oral contraceptives use

Pregnancy    
 Ever7707991.001.00
 Never63341.93 (1.26–2.96)1.82 (1.16–2.85)
Reasons for nulligravity    
 Unmarried14111.001.00
 Infertile due to a female fertility problem1771.90 (0.58–6.20)2.64 (0.70–9.91)
 Infertile due to a male fertility problem1333.45 (0.78–15.29)4.50 (0.90–22.60)
 Others19131.16 (0.40–3.34)1.23 (0.38–3.97)
Number of pregnancies (among gravid women)
 11361061.001.00
 21972010.76 (0.55–1.05)0.75 (0.54–1.05)
 31962070.74 (0.54–1.02)0.66 (0.47–0.94)
 41411550.71 (0.50–1.00)0.63 (0.43–0.93)
 ≥51001300.60 (0.42–0.86)0.59 (0.39–0.92)
 p trend  <0.010.01
Outcome of most recent pregnancy (among gravid women)
 Live birth3983641.001.002
 Induced abortion3334150.73 (0.60–0.90)1.11 (0.76–1.62)
 Miscarriage2892.84 (1.32–6.11)2.27 (0.97–5.35)
 Others11110.91 (0.39–2.13)0.98 (0.39–2.50)
Years since last pregnancy    
 <202592761.001.002
 20∼2432611.03 (0.76–1.38)0.84 (0.60–1.16)
 30∼2642581.16 (0.77–1.74)0.98 (0.63–1.54)
 p trend  0.470.89
Years since last pregnancy (among women who have no induced abortion after their last birth)
 <20113901.001.002
 20∼1331260.77 (0.49–1.22)0.69 (0.41–1.17)
 30∼1911680.77 (0.41–1.44)0.86 (0.42–1.75)
   0.460.77
Stillbirth (among gravid women)    
 Never7607821.001.002
 Ever10170.60 (0.27–1.31)0.64 (0.27–1.48)
Livebirth (among gravid women)    
 Ever7547931.001.003
 Never1662.86 (1.11–7.36)3.02 (1.10–8.32)
Number of live births (among parious women)    
 13543551.001.003
 22032290.80 (0.60–1.05)0.82 (0.60–1.10)
 31181080.92 (0.63–1.34)1.00 (0.65–1.53)
 ≥4791010.64 (0.42–0.98)0.73 (0.45–1.18)
 p trend  0.090.30
Incomplete pregnancy (among gravid women)    
 Never3042461.001.004
 Ever4665530.68 (0.55–0.84)0.69 (0.55–0.87)
 Induced only3714580.65 (0.53–0.81)0.67 (0.53–0.84)
 Miscarriage only66491.09 (0.73–1.63)1.03 (0.67–1.58)
 Both induced and miscarriage29460.51 (0.31–0.84)0.53 (0.31–0.89)
Number of induced abortions (among gravid women)
 Never3702951.001.004
 12393080.62 (0.49–0.78)0.64 (0.50–0.82)
 21221540.63 (0.47–0.84)0.63 (0.46–0.85)
 ≥339420.74 (0.46–1.17)0.84 (0.52–1.38)
Time of the induced abortion (among gravid women)
 Never3702951.001.005
 Only before first birth19151.00 (0.50–2.03)1.02 (0.47–2.21)
 Only after last birth2903650.63 (0.51–0.79)0.68 (0.52–0.88)
 Both before and after first birth9190.38 (0.18–0.85)0.44 (0.18–1.05)
 Others821050.62 (0.45–0.86)0.62 (0.41–0.94)
Oral contraceptive use (among gravid women)    
 Never6255931.001.006
 Ever1452060.67 (0.52–0.85)0.64 (0.50–0.84)
Duration of oral contraceptive use (months)    
 Never6255931.001.006
 ≤24971110.81 (0.61–1.09)0.80 (0.58–1.10)
 >2445940.46 (0.31–0.67)0.44 (0.29–0.65)
 p trend  <0.01<0.01
Contraceptive injection use (among gravid women)
 Never7517761.001.002
 Ever19230.86 (0.46–1.59)1.04 (0.54–1.98)
HRT use    
 Never7988001.001.002
 Ever35331.06 (0.65–1.73)1.06 (0.63–1.78)

Additional analyses stratified by menopausal status were conducted and presented in Table IV. Later age at menarche was associated with a markedly decreased risk of endometrial cancer among premenopausal women (p for trend = 0.01) but the reduction in risk was much smaller among postmenopausal women. However, test for multiplicative interaction was not significant. Similarly, prolonged menstruation was associated with a slightly higher risk of endometrial cancer among premenopausal women than that of postmenopausal women, although test for interaction was not significant and the linear trend between years of menstruation and endometrial cancer risk was only observed for postmenopausal women. Ever having a pregnancy or a live birth appears to have a slightly stronger protective effect on postmenopausal endometrial cancer, while the effect of induced abortion was slightly more protective among premenopausal women. Again, the interaction did not reach statistical significance.

Table IV. Odds Ratios (ORs) of Endometrial Cancer Associated with Menstrual and Reproductive Factors by Menopausal Status1
 PremenopausePostmenopausep value for test of heterogeneity
Cases/controlsOR (95% CI)Cases/controlsOR (95% CI)
  • 1

    Adjusted for age, education, any cancer history among first degree relatives, body mass index, alcohol drinking.

  • 2

    Additionally adjusted for oral contraceptives use, ever had a live birth and an induced abortion, and for postmenopausal women, menopausal age.

  • 3

    Additionally adjusted for oral contraceptives use, ever had a live birth and an induced abortion.

  • 4

    Additionally adjusted for oral contraceptives use, years of menstruation, number of induced abortion.

  • 5

    Additional adjusted for oral contraceptives use, years of menstruation, number of livebirth.

  • 6

    Additionally adjusted for years of menstruation, ever had a live birth and an induced abortion, number of pregnancies.

Age at menarche (years)2     
 ≤1254/221.0055/551.00 
 1364/540.45 (0.24–0.87)103/931.07 (0.66–1.76) 
 1470/610.45 (0.24–0.86)97/911.21 (0.73–1.99) 
 1571/740.43 (0.23–0.80)95/1031.06 (0.64–1.75) 
 1650/480.47 (0.24–0.91)78/930.97 (0.58–1.62) 
 ≥1721/330.29 (0.13–0.63)73/1040.80 (0.47–1.35)p = 0.24
 p trend p = 0.01 p = 0.22 
Years of menstruation3     
 <3086/981.0037/751.00 
 30–117/1111.23 (0.70–2.17)169/2261.47 (0.93–2.33) 
 35–98/771.39 (0.64–3.00)229/1932.47 (1.56–3.92) 
 40–29/65.12 (1.29–20.44)66/452.65 (1.48–4.76)p = 0.52
  0.15 <0.01 
Pregnancy     
 Ever303/2791.00467/5201.00 
 Never27/151.63 (0.81–3.29)36/192.25 (1.25–4.05)p = 0.58
Livebirth (among gravid women)c     
 Never9/41.007/21.00 
 Ever294/2750.55 (0.15–2.02)460/5180.21 (0.04–1.16)p = 0.34
Number of livebirth (among parous women)4     
 1247/2351.00107/1201.00 
 238/370.73 (0.42–1.29)165/1920.93 (0.64–1.35) 
 39/31.06 (0.25–4.51)111/1061.08 (0.68–1.74) 
 ≥4  77/1000.79 (0.46–1.34)p = 0.41
  p = 0.41 p = 0.51 
Induced abortion (among gravid women)5     
 Never126/851.00244/2101.00 
 Ever177/1940.60 (0.42–0.86)223/3100.69 (0.53–0.91)p = 0.75
Number of induced abortion (among gravid women)5    
 Never126/851.00244/2101.00 
 1114/1170.66 (0.44–0.99)125/1910.63 (0.46–0.85) 
 246/590.46 (0.28–0.77)76/950.75 (0.51–1.10) 
 ≥317/180.64 (0.31–1.37)22/241.06 (0.55–2.08)p = 0.35
  p < 0.01 p = 0.18 
Oral contraceptive use (among gravid women)6     
 Never253/2341.00372/3591.00 
 Ever50/450.89 (0.55–1.45)95/1610.57 (0.42–0.78)p = 0.14
Duration of oral contraceptive use (months) (among gravid women)6   
 Never253/2341.00372/3591.00 
 ≤2428/250.93 (0.51–1.72)52/580.87 (0.57–1.35) 
 >2421/200.81 (0.41–1.62)41/1020.39 (0.26–0.59) 
 p trend 0.54 <0.01p = 0.09

Among gravid women, a lower proportion of cases than controls had ever used oral contraceptives (OR = 0.64, 95% CI 0.50–0.84). It was uncommon to use noncontraceptive hormones among Shanghai women and there were no case-control differences with respect to HRT use and injectable contraceptive use (Table III). Further stratified analysis by menstrual status showed that the difference and the trend of decreasing ORs with increasing duration of oral contraceptive use were observed only among postmenopausal women (Tables III and IV).

Finally, we ran a stratified analysis of number of induced abortions by number of live births (Table V). The inverse association of endometrial cancer with induced abortion appeared to be weakened among women with 3 or more live births. Similarly, the protective effect of live birth was also less apparent among women with 2 or more induced abortions. However, test for multiplicative interaction was not significant.

Table V. Odds Ratios of Endometrial Cancer Associated with Live Birth and Induced Abortion1
  Number of live birthsp value for test of heterogeneity
12–3>3
Cases/controlsOR (95% CI)Cases/controlsOR (95% CI)Cases/controlsOR (95% CI)
  • 1

    Adjusted for age, education, menopausal status, years of menstruation, any cancer history among first degree relatives, body mass index, alcohol drinking, and oral contraceptive use.

No. of induced abortion0165/1231.00137/1100.73 (0.50–1.06)55/590.57 (0.35–0.96) 
 1121/1490.56 (0.40–0.78)100/1280.44 (0.30–0.64)16/300.36 (0.18–0.75) 
 ≥268/830.52 (0.35–0.77)84/990.48 (0.32–0.70)8/120.49 (0.18–1.32)p = 0.35

DISCUSSION

Similar to many studies conducted in western countries,2, 4, 5, 11 this population-based case-control study of endometrial cancer, conducted in a low risk population, found that prolonged menstruation, nulligravity and nulliparity were related to an increased risk of endometrial cancer. Pregnancy, regardless of whether it was a complete pregnancy or induced abortion, was related to a reduced risk of endometrial cancer, and the protective effect lasted for more than 30 years after the last pregnancy. We also found that the effect of menstrual factors appeared to be more pronounced among premenopausal women, while the reproductive factors appear to be slightly more strongly linked to postmenopausal endometrial cancer. An interesting finding of the study was the demonstration that induced abortion was associated with a decreased risk of endometrial cancer.

Endometrium undergoes a cyclic change from proliferative to secretory phase and then to menstruation phase among menstruating women under the influence of the rise and fall of estrogen and progesterone levels. Excessive proliferation or inadequate shedding caused by prolonged estrogen stimulation, particularly unopposed by progesterone, has been long suggested as one of the major mechanisms of endometrial cancer development. In addition to contributing to the years of menstruation, early menarche and later menopause have been both linked to anovulatory cycles.12, 13 It has also been suggested that level of androstenedione, a precursor of estradiol, increased with anovulation at menopause. Our present study, like many previous studies,2, 3, 4 found that earlier age at menarche and later age at menopause are the major risk factors of endometrial cancer. The stronger association between years of menstruation and endometrial cancer among premenopausal women was similar to the report by La Vecchia et al.14 and may be the result of exposure to estrogen at an earlier age. The latter is supported by a stronger association between earlier menarche with premenopausal endometrial cancer found in our study.

Several decades ago it was observed that nulliparous women in England and Wales had higher deaths rates from endometrial cancer than parous women.15 U.S. investigators also reported higher incidence of the disease in single than in married women.16 Case-control and cohort studies have consistently shown that nulliparity is associated with a 2- to 3-fold increased risk of endometrial cancer.3, 17 In agreement with these studies, we found more than a 2-fold of risk among nulliparous women. In addition, we found that the risk of endometrial cancer decreased with increased number of pregnancies, and the effect was unaffected by pregnancy outcome.

Several studies showed that age at last live birth might decrease the risk of the cancer.5, 18, 19, 20 Kvale et al.21 hypothesized that the protective effect of late delivery age may relate to the mechanical exfoliation of the endometrium at each delivery, by which epithelial cells in early and later stages of malignant transformation might be removed. However, we found age at last pregnancy was not significantly associated with the risk of endometrial cancer after adjustment, suggesting uncontrolled confounding factors might have accounted for findings from the earlier report. In contrast, we found that years since last pregnancy was unrelated to risk of endometrial cancer, suggesting a sustained protective effect of pregnancy on endometrial cancer.

Association of abortion with the risk of endometrial cancer remains a subject of much debate. Results from McPherson et al.'s22 study showed a positive relationship between endometrial cancer risk and abortion, especially induced abortion, but 2 other studies reported no such association.12, 23 Three other studies revealed that abortion was inversely associated with the risk.3, 9, 14 The high frequency of induced abortion in Shanghai women and lack of social stigma associated with it makes our study particularly informative with respect to the possible effects of induced abortion on the risk of endometrial cancer. After adjustment for age, education, number of complete pregnancies and other traditional risk factors, ever having an induced abortion was related to a 30% reduced risk among gravid women. In China, the practice of induced abortion was performed with suction under negative pressure and/or apoxesis. Both procedures can remove endometrium epithelial cells, including cells that have undergone early stages of malignant transformation, therefore decreasing the risk of endometrial cancer. The low frequency of induced abortion before live birth in the study population prohibited a meaningful conclusion regarding its effect on endometrial cancer. It is noteworthy that increased number of abortions was not related to a further reduction in risk among women who had 3 or more live births. Similarly, the protective effect of live birth did not appear to increase with the number of live births among women who had 2 or more induced abortions. This observation supports the hypothesis that reproductive factors alter the risk of endometrial cancer not only by shedding the mutative endometrial cells but also by altering the level of hormones.21 For example, increased numbers of induced abortions may weaken the protective effect of pregnancy on the development of endometrial cancer by depriving exposure to high levels of progesterone during the pregnancy. However, we cannot rule out the possibility of a chance finding given the small number of subjects involved in the extreme strata. Future studies are needed to confirm these results.

As with most case-control studies, our study is limited by the recall nature of exposure information. Although reproduction and menstruation are important events in women's lives, the validity of self-reported reproductive and menstrual information is unknown to the study population. However, the exposure of misclassification, although magnitude is unknown, is most likely to be nondifferential and thus bias the results towards null. The case-control difference (84% vs. 71%) in study participation rates raised some concern about selection bias. Our study, however, has many strengths. The population-based study design avoided the referral bias and increased the generalizability of study results. The low rate of hysterectomy in our study population also minimized the potential selection bias encountered by studies from other populations with a high rate of hysterectomy. Furthermore, large sample size and extensive survey information allowed us to evaluate the effect of menstrual and reproductive factors on endometrial cancer with proper controlling of the multiple potential confounders.

In conclusion, in line with reports from western countries, we found that menstrual and reproductive factors are important risk factors for endometrial cancer in Shanghai women. Prolonged menstruation increases the risk of endometrial cancer, while pregnancies, including induced abortion, were related to a decreased risk, supporting the hypothesis that estrogens are important and perhaps essential mediators in the incidence of endometrial cancer. Mechanical shedding of malignant or premalignant cells at each delivery may play a role in the reduction of the risk of endometrial cancer.

Ancillary