Different Risk Factor Profiles for Mucinous and Nonmucinous Ovarian Cancer: Results from the Danish MALOVA Study

Ovarian cancer is the most lethal gynecologic cancer. The overall 5-year survival is ∼40%, varying from ∼10% for generalized disease to ∼80% for localized disease (1, 2). The age-standardized incidence of ovarian cancer shows a marked geographic variation with the highest incidence in Western countries (10.2/100.000; ref. 3).

Tumors of the ovarian surface epithelium account for ∼90% of all malignant ovarian tumors. The group can be divided into five major histologic subtypes: (a) serous tumors, (b) mucinous tumors, (c) endometrioid tumors, (d) clear cell tumors, and (e) transitional cell (Brenner) tumors representing, respectively, 50%, 5% to 14%, 10% to 25%, 4% to 5%, and 1% of all malignant ovarian tumors (4-7). Included in the 90% of malignant ovarian tumors arising from the surface epithelium are highly malignant epithelial-stromal tumors lacking any specific differentiation and classified as undifferentiated (representing ∼5% of all malignant ovarian tumors), and those not designated a specific subtype commonly recorded as adenocarcinomas not otherwise specified (5).

The close relationship between the ovarian epithelium and epithelia elsewhere in the genital tract causes similarities between different types of ovarian tumors and other genital tract tumors. Mucinous tumors resemble either the endocervical epithelium or, more frequently, the intestinal epithelium, whereas serous tumors resemble the interior of the fallopian tube. Endometrioid tumors resemble the internal lining of the uterus, clear cell tumors are formed by clear, peglike cells and Brenner tumors resemble the internal lining of the urinary bladder (5, 6, 8, 9). It is known that different genital tract tumors have different risk profiles, and it has been suggested that this also applies to the different histologic types of ovarian cancer. In particular, this distinction is apparent between mucinous and nonmucinous tumors (10-14). The hypothesis about different etiologies for mucinous and nonmucinous ovarian tumors was first given by Risch et al. (14).

Several epidemiologic studies have examined the association between reproductive factors and the risk of ovarian cancer. The majority of studies have shown that reproductive factors are associated with the risk of developing nonmucinous tumors, whereas the corresponding results for mucinous tumors have been less clear. Some studies have found an association for mucinous tumors with reproductive variables, although less pronounced than for nonmucinous tumors (11, 14-16), whereas other studies did not find any association (17, 18).

Several studies have reported that the use of oral contraceptives lowers the risk of ovarian cancer (19, 20). Recently, a large pooled analysis of 10 case-control studies found this effect to be independent of the histologic type (21); however, the results are conflicting, as other studies found the effect of oral contraceptives to be most pronounced for nonmucinous tumors (10, 22).

The association between hormone replacement therapy (HRT) and risk of ovarian cancer is still equivocal in spite of several observational studies (14, 21, 23, 24). Whether there is a difference with regard to histologic subtype is also controversial, as Riman et al. (24) found an increased risk of epithelial ovarian cancer after ever use of HRT for serous, mucinous, and endometrioid subtypes, whereas Risch (25) only found the risk increased for nonmucinous tumors.

Studies on life-style factors and ovarian cancer have reported different risk profiles for mucinous and nonmucinous tumors with regard to smoking, in which tobacco smoking was found to increase the risk for mucinous tumors, but not for nonmucinous tumors (21, 26-29).

It is still controversial whether body mass index (BMI) has an effect on the risk of ovarian cancer. Some studies have documented that increasing BMI increases the risk of epithelial ovarian cancer especially for mucinous and clear cell tumors (30, 31). In contrast, one study found that BMI ≥ 24 had a protective effect on the risk of serous tumors (21), whereas other studies did not find any association between BMI and the risk of developing ovarian cancer (32, 33).

The aim of this study was to further examine risk factors for mucinous and nonmucinous ovarian tumors, using data from a large Danish population–based case-control study. We focused on the effects of reproductive factors such as pregnancy, length of ovulation, use of oral contraceptives, and of life-style factors such as smoking and BMI.

The present article is based on data from a population-based case-control study (“the MALOVA study”). A detailed description of the study population and data collected has been provided previously (34).

Briefly, the case group consisted of women 35 to 79 years of age. From January 1995 to May 1999, cases were recruited from 16 gynecologic departments in Denmark (municipalities of Copenhagen and Frederiksberg as well as the counties of Copenhagen, Frederiksberg, Roskilde, Western Zealand, Funen, Southern Jutland, and Northern Jutland). Women, scheduled for an explorative laparotomy or laparoscopy on the suspicion of an ovarian tumor, were requested to participate in the study with blood and tissue samples and a personal interview. In order to ensure that all eligible cases in the study area were included, we also linked the study database to the Danish Cancer Registry every second month. Women registered in the Cancer Registry with ovarian cancer, but not primarily included in the study, were contacted by letter and asked to participate with an interview. The interview included information on social, reproductive, medical, and gynecologic history and life-style factors. A life event calendar was used to obtain detailed information on pregnancies, births, abortions, lactation, hormone and contraceptive use, and infertility periods.

Pathology reports and tissue specimens were collected, and in 30% of cases, reviewed in a blinded fashion by one pathologist specializing in gynecologic tumors. In case of discrepancy from the original diagnosis, the evaluation was done by a third pathologist specializing in gynecologic tumors, who was unaware of any of the diagnoses, and a consensus histologic diagnosis was obtained. For cases in which the histopathologic slides were not reviewed (70%), all histologic descriptions were scrutinized for consistency between the hospital description and the resulting diagnosis.

A total of 681 women (79.9%) with ovarian cancer were enrolled in the study, including 653 women with epithelial ovarian cancer, of whom 554 women had completed the personal interview making them eligible for the analysis in this study. This covered 50 women with mucinous tumors, 343 women with serous tumors, 75 women with endometrioid tumors, and 86 women with other types of epithelial ovarian cancer (8 undifferentiated carcinomas, 34 adenocarcinomas, and 44 clear cell neoplasms). Clinically and pathologically, most of these undifferentiated or poorly differentiated tumors are serous tumors that have lost their differentiation, and it is reasonable to include them among the serous tumors. However, after reviewing the histologic diagnosis resulting in reclassification of a majority of the originally classified undifferentiated carcinomas and the adenocarcinomas to serous tumors, we chose to include the ones still classified as undifferentiated carcinomas and adenocarcinomas after the review in the group designated as other tumor types. Pseudomyxomas of the peritoneum and ovary were not included in the study.

A random sample from the general female population, 35 to 79 years of age in the study area, was drawn by means of the computerized Civil Registration System (all inhabitants in Denmark have a unique personal identification number, which is registered in the Civil Registration System). Controls were recruited simultaneously with the cases, and from the area which generated the cases. They were frequency-matched in 5-year intervals by using the age distribution of women with ovarian cancer (1987-1992) registered in the Danish Cancer Registry. The controls were invited by letter to participate in the study with a personal interview and a blood sample. In the case of no response, they were contacted by telephone or by a second letter. A total of 3,839 women were invited to participate in the study as controls. Of these, contact could not be achieved with 301 women, 269 women were excluded due to bilateral oophorectomy, 6 women had moved out of the study area, and 126 women were too ill to participate, leaving 3,137 women as eligible controls. A total of 2,116 women were enrolled of whom 1,564 participated with a personal interview and 552 participated with a telephone interview. The telephone interview was less comprehensive than the personal interview, so this study is based exclusively on personal interviews.

To analyze the influence of reproductive factors, oral contraceptives, HRT, smoking, and BMI on epithelial ovarian cancer risk, we did multiple logistic regression analysis. Odds ratios (OR) and 95% confidence intervals (CI) were estimated for ovarian cancer overall, and separately for each of the four largest histologic groups. In all analyses, we adjusted for age in 5-year categories corresponding to the sampling of controls. Variables were retained in the model if it was significant and/or influenced the other estimates. We tested linearity for the quantitative variables for spline effects with knots placed at the tertiles. No significant deviations from linearity were found.

On the basis of information obtained at the interview, a variable was constructed to estimate the woman's lifetime years of ovulation, i.e., the ovulation time. The total number of years with ovulation was calculated by subtracting the age at menarche from age at menopause (for premenopausal women; age at last menstrual period) and then further subtracting total time pregnant, total time breast-feeding, and total time taking oral contraceptives (excluding gestagen-only pills). This variable only exists for a subgroup of the study population because of missing values for one or more of the variables necessary for the calculation.

The association between reproductive factors and epithelial ovarian cancer risk is presented in Table 1. Ever pregnancy significantly decreased the risk of ovarian cancer (OR, 0.40; 95% CI, 0.30-0.55). Similar significantly decreased risk estimates were observed for the different nonmucinous histologic types (serous: OR, 0.42; 95% CI, 0.29-0.61; endometrioid: OR, 0.37; 95% CI, 0.19-0.73; other ovarian cancers: OR, 0.27; 95% CI, 0.15-0.48), whereas the risk of mucinous tumors was not affected by ever being pregnant (OR, 1.13; 95% CI, 0.34-3.73; P = 0.035 for the difference between the mucinous and the nonmucinous tumors). The number of additional pregnancies significantly decreased the risk of ovarian cancer, and the same pattern was seen for all the types of nonmucinous tumors, whereas the risk for mucinous carcinomas was much less strongly related with the increasing number of pregnancies. Both older age at first pregnancy and at last pregnancy reduced the risk of epithelial ovarian cancer. The effect was mostly pronounced for serous ovarian carcinomas in which the risk decreased by 19% and 13% per 5 years older age for, respectively, first pregnancy (95% CI, 0.70-0.95) and last pregnancy (95% CI, 0.77-0.99; data not shown). The risk estimates for the other histologic types were similar, but did not reach statistical significance.

Finally, we assessed the combined effect of the variables related to ovulation. Increasing number of years with ovulation significantly increased the risk for all nonmucinous tumors. The risk for, e.g., serous carcinomas was more than four times higher for women with ≥36 years of ovulation compared with women with ≤24 years of ovulation (OR, 4.22; 95% CI, 2.39-7.45). In general, we found a 7% to 8% increase in risk per year of ovulation for nonmucinous tumors (data not shown), whereas the association with mucinous tumors was much weaker (P = 0.10 for the difference between the mucinous and the nonmucinous group; Table 1).

In Table 2, the relationship between ovarian cancer risk and use of oral contraceptives and HRT is presented. Ever use of oral contraceptives significantly decreased the overall risk of ovarian cancer (OR, 0.67; 95% CI, 0.53-0.85) and the same was found when the different histologic types were analyzed separately. Duration of oral contraceptive use decreased the risk for all nonmucinous tumors with a 5% to 7% decrease per year of use. There was no apparent effect on the risk of mucinous tumors with increasing duration of oral contraceptive use, but because none of the duration subgroups contain more than seven mucinous cases, it is not possible to make a firm conclusion (P = 0.17 for the difference between the mucinous and the nonmucinous groups). We found little evidence of declining protection with time since last use of oral contraceptive and we observed no association between age at start of oral contraceptive use and risk of ovarian cancer when adjustment included duration of oral contraceptive use (data not shown).

Ever users of HRT compared with never users had an increased risk of serous tumors (OR, 1.30; 95% CI, 1.00-1.68), endometrioid tumors (OR, 1.75; 95% CI, 1.07-2.84), and all ovarian cancers combined (OR, 1.30; 95% CI, 1.05-1.61). For the group classified as other types of ovarian cancer, an increased but nonsignificant risk was also seen, whereas no evident association was found between HRT use and risk of mucinous ovarian cancer (P = 0.048 for the difference between the mucinous and the nonmucinous group).

Current smoking significantly increased the risk of mucinous tumors (OR, 1.78; 95% CI, 1.01-3.15), but did not affect the risk of nonmucinous ovarian cancers (P = 0.0055 for the difference between the mucinous and the nonmucinous group; Table 3). The same picture emerged for increasing number of years with smoking and maximum number of cigarettes smoked per day, in which only the risk estimates for mucinous tumors were increased, although not being statistically significant.

We also examined the association between BMI and the risk of ovarian cancer (Table 3) and found an increased risk with increasing BMI for mucinous and endometrioid tumors (P = 0.17 for the difference between mucinous and nonmucinous tumors). The estimates for BMI at age 30 to 39 years are shown in Table 3 (mucinous: OR, 2.23; 95% CI, 0.95-5.21; endometrioid: OR, 1.63; 95% CI, 0.79-3.35), and a similar risk pattern was observed for BMI at age 20 to 29, 40 to 49, and 50 to 59 years (data not shown). In addition, we looked at the effect of ever having had a BMI of 25 or higher during adult life, and found a nonsignificantly increased risk for only mucinous and endometrioid carcinomas. By stratifying BMI for pariety (never/ever), an increasing risk of ovarian cancer was seen for parous women compared with nulliparous, when adjusted for age, duration of oral contraceptive use and additional births (data not shown).

Finally, we found that neither duration of breast-feeding, sterilization, nor hysterectomy were statistically significant determinants for ovarian cancer risk overall or for any of the histologic subtypes (data not shown).

The strength of our study is the population-based design, the size of the study population, the thorough pathologic examination of the tissue specimens, the extensive personal interview shortly after operation, reducing the likelihood of recall bias. However, our study also has limitations. Even though it includes a large number of cases, it is limited by a relatively low number of cases with mucinous tumors, which is the case for most studies of this kind. This implies that there is a risk of overlooking associations for mucinous tumors. Among the 30% that were reviewed by an independent pathologist, overall, 17.8% had some kind of discrepancy if categorized in the four histologic categories (mucinous, serous, endometrioid, or others). If categorized as mucinous/nonmucinous, 3% had discrepancies. Of these, some originally diagnosed as mucinous tumors were reclassified as nonmucinous tumors. If the same applies to those not reviewed, it may potentially have biased the results towards lower observed differences between mucinous and nonmucinous tumors. Those who did not participate in the study might have differed and this may potentially have biased the results. However, after enrollment of the 1,564 controls who participated with a personal interview, we had 552 additional controls participating with a telephone interview. Those who had telephone interview did not differ from the remaining controls with respect to number of pregnancies, length of oral contraceptive use, menopausal status, or hysterectomy. We have no information about BMI and smoking status on the telephone-interviewed women.

We found that reproductive factors were strongly associated with the risk of contracting nonmucinous tumors, whereas the associations with mucinous tumors were absent or much weaker. Ever pregnancy and number of pregnancies statistically significantly decreased the risk of nonmucinous tumors, but not of mucinous tumors. Similar results have been found in other studies (11, 13, 14, 32). When examining whether age at first and last pregnancy had an effect on ovarian cancer, we found a significantly reduced risk of serous tumors with older age at both first and last pregnancies, respectively. The risk estimates for the other histologic types were similar, but did not reach statistical significance. This is in line with previous studies of the effect of age at first/last pregnancy and birth on the risk of epithelial ovarian cancer (15, 23, 35-38). Except for two of the studies, the histologic subtypes were not analyzed separately (23, 30, 36). Riman et al. (30) only found a significant association when examining all tumor types together, whereas Albrektsen et al. (36) found the protective effect to differ only slightly between the different histologic groups. However, the results are equivocal, as Tung et al. (16) did not find any association between age at last birth and years since last pregnancy for invasive ovarian cancer when examining subtypes, except for a nonsignificantly decreased risk of mucinous tumors for women having their last delivery after the age of 34 years. Even though the overall picture is not clear, it seems that most studies point to a decreased risk of epithelial ovarian cancer with older age at last pregnancy/birth. One reason for this might be that a larger accumulation of damaged or precancerous cells, the number of which increases with age, are eliminated by a more recent (later) pregnancy.

When examining the risk related to lifetime number of years with ovulation, we found, in agreement with Purdie et al. (39) and Tung et al. (16), a 7% to 8% increase in risk of nonmucinous tumors for each year of ovulation, but no significantly increased risk for mucinous tumors. Purdie et al. (39) examined ovulation and risk of ovarian cancer, taking age into account, and found that ovulations during the ages of 20 may be those most associated with disease risk.

Use of oral contraceptives showed a protective effect on the risk of ovarian cancer for all histologic subtypes in line with several previous studies (13, 16-18, 21, 32); however, duration of oral contraceptive use was less pronounced for mucinous tumors. This observation is also in agreement with previous studies (13, 14, 16).

Risch (40) examined whether oral contraceptive use has additional independent effects on ovarian cancer risk other than suppressing ovulation by adjusting for length of ovulatory life. He found that oral contraceptives seem to have a protective influence on epithelial ovarian cancer beyond their anovulatory action, so they may not only act by suppressing ovulation, but also by somehow altering the tumor-promoting milieu.

The increased risk of nonmucinous ovarian cancer after ever use of HRT found in this study has also been observed in some studies (14, 21, 23, 24), although others found no effect of HRT (21). No association was seen between HRT and mucinous tumors supporting the findings of Risch (25) and Riman et al. (23). We found a statistically significant difference between nonmucinous and mucinous tumors. The data on HRT and the effect of estrogen intake on the risk of ovarian cancer in the MALOVA cohort has previously been published (34).

It has been speculated whether the histologic similarities between mucinous tumors and tumors of intestinal and cervical origin could explain the association of the mucinous subtype with cigarette smoking, as the cervix and colon are both known to be susceptible to tobacco carcinogens (41-45). In line with this, we found an increased risk of ovarian cancer related to current smoking only for mucinous tumors (statistically significant difference between mucinous and nonmucinous tumors). Our results are in agreement with those of an Australian study and two American studies (25-27). However, Goodman and Tung (46) did not find an association between tobacco smoking and risk of invasive ovarian cancer according to histologic subtype. We found a nonsignificant increase in risk of mucinous tumors with increasing duration of smoking and increasing number of cigarettes per day among ever smokers. The association with smoking has also been found for borderline mucinous tumors, where it seems to be even stronger than the association between smoking and invasive epithelial carcinoma of mucinous type, and has been found for both current and former smokers (26, 47).

We found an increasing risk for epithelial ovarian cancer with increasing BMI especially for mucinous and endometrioid tumors. Previously, Riman et al. (30) and Purdie et al. (31) had the same findings for mucinous tumors and clear-cell tumors, and Farrow et al. (48) have reported an increased risk for endometrioid tumors with increasing Quetelet index (five categories of BMI). They also found, that among women with serous tumors, only those in the highest Quetelet index category were at a greater than 2-fold excess risk, but the risk was not increased in the intermediate categories. For both serous and endometrioid tumors, the excess risk was largely confined to premenopausal women, which is in line with the finding by Fairfield et al. (49).

We observed no significant effects of breast-feeding, tubal sterilization, and hysterectomy on the risk of epithelial ovarian cancer, including the different histologic subtypes. This is in line with some previous findings (14, 15, 17, 23), whereas others (13, 21) find these factors to be protective, especially for nonmucinous tumors.

Carcinomas with mucinous histologies are rarely observed in BRCA1/2 mutation carriers (breast and ovarian cancer susceptibility genes; ref. 50) supporting the theory of different etiologies between mucinous and nonmucinous tumors.

Although many theories have been suggested to explain ovarian carcinogenesis, the etiology of ovarian cancer still remains unknown. One theory suggested by Fathalla (51) is that the “incessant” ovulation causes repeated ruptures of the epithelial surface of the ovary followed by repair and exposure to estrogen-rich follicular fluid, thereby increasing the risk of malignant transformation of the cells. Another theory is the “apoptosis theory” formulated by Adami et al. (35). They suggested that high levels of circulating hormones during pregnancy eliminate cells from the ovaries that have already undergone malignant transformation. As apoptosis is one of the most protective mechanisms against neoplasia (52), it is plausible that exposure to agents that promote apoptosis might reduce the risk of cancer. Rodriguez et al. (53) have suggested that it is possible that pregnancy, in which the progestagen level is known to be high and ovulation is inhibited, confers further protection against ovarian cancer through a mechanism that involves apoptotic effects on the ovarian epithelium. In support of this progesterone/apoptosis theory is the result of a randomized controlled trial conducted on primates. This trial showed that the frequency of apoptotic cells was substantially higher (24.9%) in the group randomized to levonorgestrol, which is a synthetic progestagen, than in the control-group (3.8%) or in the group randomized to ethinyl estradiol (1.8%) or in the group receiving a combined preparation of ethinyl estradiol and levonorgestrol (14.5%; ref. 53). In 1983, Cramer and Welch (54) published an article on the pathogenesis of ovarian cancer launching the gonadotropin-estrogen hypothesis. This hypothesis proposed that ovarian cancer was primarily caused by high levels of gonadotropins stimulating the ovarian stroma and subsequently leading to high levels of estrogen or estrogen precursors, which may cause the ovarian surface epithelia to undergo proliferation and malignant transformation.

In conclusion, we have confirmed already known risk factors for ovarian cancer, and further observed differences in the risk profile between mucinous and nonmucinous tumors, supporting the hypothesis that mucinous tumors may have a different etiology than nonmucinous tumors.