Cancer Causes & Control

, Volume 17, Issue 6, pp 821–829

Association of Reproductive Factors, Oral Contraceptive Use and Selected Lifestyle Factors with the Risk of Ovarian Borderline Tumors: A Danish Case-control Study

Authors

  • Lene Drasbek Huusom
    • Institute of Cancer EpidemiologyDanish Cancer Society/Rigshospitalet
  • Kirsten Frederiksen
    • Institute of Cancer EpidemiologyDanish Cancer Society/Rigshospitalet
  • Estrid V. S. Høgdall
    • Institute of Cancer EpidemiologyDanish Cancer Society/Rigshospitalet
  • Eva Glud
    • Institute of Cancer EpidemiologyDanish Cancer Society/Rigshospitalet
  • Lise Christensen
    • Department of Pathology RigshospitaletUniversity of Copenhagen
  • Claus K. Høgdall
    • The Gynecologic Clinic, The Juliane Marie Centre, RigshospitaletUniversity of Copenhagen
  • Jan Blaakaer
    • Department of Gynecology and ObstetricsAarhus University Hospital Skejby
    • Institute of Cancer EpidemiologyDanish Cancer Society/Rigshospitalet
Original Paper

DOI: 10.1007/s10552-006-0022-x

Cite this article as:
Huusom, L.D., Frederiksen, K., Høgdall, E.V.S. et al. Cancer Causes Control (2006) 17: 821. doi:10.1007/s10552-006-0022-x

Abstract

Objective

The aim was to examine risk factors for ovarian borderline tumors overall, and according to histological subtype (serous vs. mucinous), in a large Danish population-based case-control study.

Methods

Ovarian borderline cases and controls were recruited from 1995 to 1999, and personal interviews were conducted. In all, 202 cases and 1,564 randomly selected controls were included. The analysis was performed using multiple logistic regression models.

Results

The risk of ovarian borderline disease decreased with increasing parity (OR=0.79 per birth, 95% CI: 0.63–0.98) and older age at first birth (OR=0.67 per 5 years, 95% CI: 0.53–0.84). Both a history of breastfeeding and use of oral contraceptives reduced the risk of borderline tumor, the effect being most pronounced for serous tumors. Increasing body mass index (BMI) was associated with elevated risk of serous borderline tumor (OR=1.05 per BMI unit; 95% CI: 1.00–1.10), whereas current smoking was a strong risk factor only for mucinous tumors (OR=2.10; 95% CI: 1.22–3.60). Finally, increasing consumption of milk (all types) was found to increase the risk of borderline disease (OR=1.04 per glass milk per week; 95% CI: 1.02–1.06), and increasing intake of total lactose also increased the risk significantly (OR=1.16 per 50 gram lactose per week; 95% CI: 1.06–1.26).

Conclusion

The risk profile of ovarian borderline tumors is similar to that of ovarian carcinomas, and we observed significant etiological differences between serous and mucinous borderline tumors.

Keywords

Ovarian borderline tumorsRisk factorsEpidemiologyMilk intakeLactose

Introduction

Ovarian borderline tumors are a subgroup of ovarian tumors of low malignant potential accounting for 10–20% of all ovarian tumors [1, 2]. Compared to invasive carcinomas, ovarian borderline tumors occur at younger age (35–55 years), and the majority of patients presents with early stage disease. In general, the prognosis of ovarian borderline tumors is much more favorable compared to invasive tumors [24], but a small fraction is associated with poor prognosis and mortality rates of 20–50% [4, 5].

Several studies on epidemiological risk factors for ovarian cancers do not distinguish between borderline and invasive tumors, and relatively few studies have explored the epidemiology of ovarian borderline tumors separately. In these, the risk profile has shown some similarities with that of ovarian cancer [6, 7]. A consistent finding is the protective effect of parity, although the protection seems weaker for ovarian borderline tumors [7]. The protective effect of parity has been shown to include all histological types of ovarian borderline tumors, except mucinous borderline tumors where results have been controversial [7, 8]. It has been proposed that mucinous tumors are etiologically without relation to other epithelial tumors [9]. Use of oral contraceptives (OC) has also consistently been found to lower the risk of ovarian cancer [7], but the effect related to ovarian borderline disease is less clear.

The incidence of ovarian cancer varies widely between countries, and the influence of diet and specific lifestyle have been suggested as important factors in the development of ovarian cancer, even though conflicting results have been shown. High intake of the milk sugar lactose has in some studies been associated with increased risk of ovarian cancer, through a possible direct toxic effect of its metabolites (galactose) on the ovary or by compensatory gonadotropin stimulation [1012]. A recent meta-analysis of 21 studies investigating the link between ovarian cancer and consumption of milk products and lactose has found support for the hypothesis that high intakes of dairy foods and lactose may increase the risk of ovarian cancer [13]. Only two studies have examined milk and lactose in relation to histological subtypes of ovarian cancer, and they found the relation confined to serous ovarian cancer [14, 15]. To our knowledge, no studies have investigated the effect of milk and lactose on ovarian borderline disease separately.

The aim of the present study was to examine risk factors for ovarian borderline tumors, overall and according to histological subtype, in a large Danish population-based case-control study on ovarian malignancy. We have focused on the effects of reproductive factors including the role of pregnancy timing, use of oral contraceptives and different measures of lifestyle, such as smoking, body mass index and intake of milk and total lactose.

Material and methods

Cases

In this population-based case-control study of ovarian malignancies, we included women 35–79 years of age. The study, named the MALOVA study, was conducted from January 1995 to May 1999 and cases were recruited from 16 gynecological departments in a well-defined study area in Denmark (municipalities of Copenhagen and Frederiksberg as well as the counties of Copenhagen, Frederiksberg, Roskilde, Western Zealand, Funen, and Southern- 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. On the basis of the peroperative findings (microscopy of peroperative biopsy or macroscopic impression), women with ovarian borderline tumor were allocated to a personal interview at the hospital. In order to ensure that all eligible cases in the study area were included, the study database was linked to the Danish Cancer Registry every second month. If a woman was registered in the Cancer Registry with ovarian borderline, but had not primarily been included in the study, she was contacted by letter and asked to participate with an interview. The interview included information on social-, reproductive-, medical- and gynecological history and lifestyle factors such as smoking, body mass index (BMI) and intake of lactose. A life event calendar was used to obtain detailed information on pregnancies, births, abortions, lactation, hormone and contraceptive use and infertility periods. Intake of lactose was assessed from open-ended questions on average intake of milk (all types), curdled milk products, yoghurt, cheese, sour cream and double cream in the past year. Consumption of lactose was calculated by multiplying the frequency of intake by the nutrient content of specified portions. One cup of milk contains 9.44 grams of lactose, and the content of lactose per portion of the following nutrients is: junket=8.43 grams, yoghurt=5.90 grams, hard cheese=0.05 grams, and double cream=0.44 grams [16]. The measurements of BMI related to the average BMI within the last 5 years before the time of ovarian borderline diagnosis.

Pathology reports were collected and the histological diagnoses of ovarian borderline tumors were categorized into five groups (serous, mucinous, endometroid, papillary, clear cell). Pathology reports and tissue specimens were reviewed in a blinded fashion by a pathologist specialized in gynecologic tumors. In case of disagreement with the original diagnosis, the histological material was evaluated by another pathologist, who was unaware of any of the diagnoses, and a consensus histological diagnosis was obtained. Through this procedure, six ovarian borderline tumors changed category to invasive ovarian tumors, and 15 invasive ovarian tumors changed category to ovarian borderline disease.

A total of 276 ovarian borderline cases were identified, of these eight women were too ill to participate, and one woman died before being contacted, leaving 267 women as eligible cases. In all, 235 ovarian borderline cases (88%) were enrolled in the main study. In the present study we excluded 12 women with non-epithelial ovarian borderline tumor, and 21 women in whom only a blood sample was available. A final number of 202 women with ovarian borderline tumor remained: 104 serous cystadenomas borderline (52%), 88 mucinous cystadenomas borderline (44%), 4 endometrioid cystadenomas borderline (2%), 4 papillary cystadenomas borderline (2%), and 2 clear cell adenomas borderline (1%). Further details about the MALOVA study are described elsewhere [17].

Controls

In Denmark all inhabitants have a unique personal identification number, which is used universally in the central administration. These identification numbers, which comprise information on date of birth and sex, are registered in the computerized Central Population Register. By means of the Central Population Register, controls were drawn as a random sample from the general female population (35–79 years) in the study area. Controls were recruited between September, 1995 and August, 1999; i.e. simultaneously with the cases. They were frequency-matched in 5-year intervals by using the age distribution of women with ovarian cancer (1987–92) registered in the Danish Cancer Registry, which contains information on all persons in Denmark diagnosed with cancer since 1942. The controls were invited by letter to participate in the study with a personal interview and a blood sample. Those, who did not give notice about their participation, were contacted by telephone or a second letter. In all, 3,839 women were invited as controls in the study. Of these, contact could not be achieved with 301 women, and 269 were excluded due to bilateral ooforectomy, while 6 women had moved out of the study area, and 126 women were too ill to participate, leaving 3,137 women as eligible controls. We enrolled a total of 2,116 women (67.5%), 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 the present study is based exclusively on personal interviews.

Data analysis

The influence of reproductive and lifestyle factors on the risk of ovarian borderline disease was analyzed using multiple logistic regression analysis. All analyses included age categorized in 5-year age groups in accordance with the grouping used in sampling the controls. Additionally, all the reported risk estimates were adjusted for those variables showing the strongest effect when mutually adjusted, that is childbirth (ever/never), number of births, age at first birth, breastfeeding, oral contraception (ever/never), duration of OC, smoking (ever/never) and intake of milk. Separate estimates were calculated for serous and mucinous borderline tumors, adjusted for the same variables as in the overall analysis. The relationship between age at first birth and age at last birth was supplementary analyzed in a model restricted to women with at least two childbirths.

For all quantitative variables except age, linearity of the effect of the variable was tested by comparison with a model with a linear spline effect [18] with knots placed at the tertiles. No significant deviations from linearity were found for any of the variables. Tests of statistical significance were performed using the likelihood ratio test statistic, and 95% confidence intervals (CI) were based on the Wald test on the log-odds scale.

Results

In general, the cases and controls were nearly identical concerning marital status and length of education, and the age distribution in the case group was also almost identical to that in the control group. In all, 30% of the women in the case group were below the age 50 compared to 29% in the control group; 57% of the cases were 50–69 years compared to 55% in the control group, while 13% of the case women were older than 70 years compared to 16% in the control group (data not shown).

Table 1 presents odds ratios for ovarian borderline tumors in relation to reproductive factors. We found a lower risk of ovarian borderline tumor among parous women compared to nulliparous. The protection increased significantly with number of childbirths, the risk reduction per additional birth being 21% (OR=0.79; 95% CI: 0.63–0.98). The effect of the first birth was greater than that of succeeding births (data not shown). The risk reduction was present for both serous and mucinous borderline tumors, but the association did not reach statistical significance among the serous tumors.
Table 1

Odds ratios (OR) and 95% confidence intervals (CI) for the relationship of reproductive factors with ovarian borderline risk

Characteristic

All

Histologic subgroup

Cases

Serous Adjusted ORa (95% CI)

Mucinous Adjusted ORa (95% CI)

Cases (n=202)

Controls (n=1564)

Adjusted ORa (95% CI)

Serous (n=104)

Mucinous (n=88)

Childbirthb

Never

38

141

1.00

19

16

1.00

1.00

ever

164

1423

0.55(0.35–0.87)

85

72

0.60(0.33–1.11)

0.56(0.29–1.07)

Number of births

1

45

236

1.00

23

20

1.00

1.00

2

71

680

0.51(0.33–0.79)

34

36

0.44(0.24–0.81)

0.62(0.34–1.16)

3

33

368

0.41(0.23–0.72)

19

10

0.43(0.21–0.91)

0.32(0.13–0.76)

≥4

15

139

0.51(0.24–1.08)

9

6

0.62(0.24–1.61)

0.42(0.13–1.35)

Per birth among parous

  

0.79(0.63–0.98)

  

0.87(0.66–1.15)

0.67(0.48–0.95)

Age at first birth

≤19

32

185

1.38(0.86–2.20)

16

12

1.31(0.70–2.44)

1.04(0.50–2.18)

20–24

84

666

1.00

45

38

1.00

1.00

25–29

38

431

0.65(0.43–1.00)

20

17

0.67(0.37–1.18)

0.60(0.33–1.10)

≥ 30

10

141

0.52(0.25–1.06)

4

5

0.47(0.16–1.36)

0.47(0.17–1.27)

Per 5 years among parous

  

0.67(0.53–0.84)

  

0.68(0.49–0.95)

0.66(0.48–0.93)

Age at last birth

≤24

51

252

1.16(0.70–1.92)

27

21

1.44(0.73–2.85)

0.89(0.43–1.83)

25–29

61

538

1.00

26

33

1.00

1.00

30–34

37

439

0.99(0.61–1.61)

23

13

1.52(0.80–2.90)

0.67(0.32–1.38)

≥35

15

194

1.14(0.55–2.34)

9

5

1.59(0.61–4.15)

0.74(0.25–2.25)

Per 5 years among parous

  

0.91(0.67–1.23)

  

0.96(0.65–1.42)

0.87(0.55–1.38)

Years since last birth

≤ 9

8

108

0.87(0.31–2.50)

3

5

1.07(0.22–5.09)

0.95(0.23–3.90)

10–19

21

229

1.01(0.49–2.10)

7

12

0.86(0.30–2.51)

1.10(0.39–3.09)

20–29

67

385

1.60(0.95–2.68)

39

27

1.86(0.95–3.62)

1.43(0.65–3.13)

≥30

68

701

1.00

36

28

1.00

1.00

Per 5 years among parous

  

0.92(0.80–1.06)

  

0.90(0.73–1.09)

0.94(0.77–1.14)

Duration of breastfeeding (months)

Never breastfed

14

99

0.97(0.50–1.86)

5

9

0.71(0.27–1.90)

1.28(0.56–2.96)

1–5

65

431

1.00

35

28

1.00

1.00

6–11

43

437

0.73(0.48–1.13)

25

15

0.79(0.46–1.38)

0.62(0.32–1.20)

12–24

31

294

0.93(0.57–1.50)

14

15

0.75(0.39–1.45)

1.22(0.62–2.39)

≥ 25

10

155

0.32(0.11–0.95)

5

5

0.29(0.07–1.31)

0.42(0.09–1.89)

Per 5 months among women who breastfed

  

0.90(0.80–1.00)

  

0.84(0.71–1.00)

0.96(0.83–1.11)

aAdjusted for age (in categories), childbirth (ever/never), number of additional births (linear), age at first birth (linear), breastfeeding (linear), duration of oral contraceptives (linear), smoking (ever/never), intake of milk (linear)

bOR for childbirth (ever/never) adjusted for age (in categories), breastfeeding (linear), duration of oral contraceptives (linear), smoking(ever/never), and intake of milk (linear)

Overall, the risk decreased significantly with an older age at first birth, corresponding to a 33% risk reduction per 5 years (OR=0.67; 95% CI: 0.53–0.84). The same pattern was seen for both serous and mucinous borderline tumors. There was no clear association between age at last birth and risk of ovarian borderline tumor when adjusted for age at first birth, but if age at first birth was omitted from the adjustment, older age at last birth reduced the risk of ovarian borderline with 26% per 5 years (OR=0.74; 95% CI: 0.60–0.90) (data not shown). Since women with only one birth have the same age for first and last birth, we also conducted the analysis restricted to women with a least two births, and found the same trend with age at first birth being more strongly associated with the risk of ovarian borderline tumor than age at last birth. When both age at first birth and age at last birth were included in the model, age at first birth was still the strongest factor (data not shown). No clear association between time since last birth and risk of ovarian borderline disease was found, when adjusted for age at first birth.

A history of breastfeeding implied a decreased risk of developing a borderline tumor even after adjustment for confounding factors, including ever children, number of births and age at first birth, with a reduction of 10% per 5 months of lactation (95% CI: 0.80–1.00). Neither abortions, nor menopausal status or infertility were significantly associated with the risk of borderline tumor (data not shown).

The relationship between borderline tumor risk and OC use is presented in Table 2. We found a reduced risk of ovarian borderline tumors in relation to ever use of OC, but the association did not reach statistical significance. The association appeared weaker for mucinous tumors (OR=0.91; 95% CI: 0.53–1.55) than for serous tumors (OR=0.66; 95% CI: 0.41–1.06), which was also reflected in the association with duration of OC use, where a borderline significant reduction of 5% (95% CI: 0.91–1.00) was observed for serous borderline tumors for each year of OC use. Furthermore, a marginally significantly increased risk of serous borderline tumors was observed with increasing number of years since last OC use, with an increase of 5% for each additional year since last OC use (95% CI: 0.99–1.12). Tubal ligation was not associated with the risk of ovarian borderline tumor; however, the pattern differed between serous and mucinous tumors (data not shown). In relation to serous borderline tumors, a decreased risk was observed for women who reported a previous tubal ligation (OR=0.49; 95% CI: 0.20–1.16), whereas tubal ligation seemed to increase the risk of mucinous borderline tumors (OR=1.61; 95% CI: 0.85–3.06), although none of the associations reached statistical significance (data not shown). Hysterectomy was not a statistically significant risk determinant, and there was no relation observed for any of the histological subgroups (data not shown).
Table 2

Odds ratios (OR) and 95% confidence intervals (CI) for the relationship of oral contraceptive use

Characteristic

All

Histologic subgroup

Cases

Serous Adjusted ORa (95% CI)

Mucinous Adjusted ORa (95% CI)

Cases (n=202)

Controls (n=1564)

Adjusted ORa (95% CI)

Serous (n=104)

Mucinous (n=88)

Use of oral contraceptives

never

89

704

1.00

52

34

1.00

1.00

ever

113

860

0.81 (0.56–1.16)

52

53

0.66 (0.41–1.06)

0.91 (0.53–1.55)

Duration of oral contraceptive use (years)

<1

22

138

1.39 (0.77–2.54)

11

7

1.48 (0.64–3.40)

0.91 (0.36–2.30)

1–4

32

261

1.00

14

17

1.00

1.00

5–9

28

187

1.23 (0.70–2.16)

17

11

1.68 (0.78–3.60)

0.93 (0.42–2.06)

≥10

30

268

0.77 (0.45–1.34)

10

18

0.54 (0.23–1.27)

0.91 (0.44–1.85)

Per year among women who used OC

  

0.97 (0.94–1.00)

  

0.95 (0.91–1.00)

0.99 (0.95–1.04)

Years since oral contraceptive use

0–10

17

187

1.00

7

6

1.00

1.00

11–20

36

249

1.59 (0.80–3.16)

10

23

1.05 (0.34–3.20)

2.04 (0.84–4.91)

≥ 21

55

410

1.63 (0.72–3.70)

32

20

2.17 (0.62–7.67)

1.42 (0.47–4.33)

Per year among women who used OC

  

1.01 (0.97–1.05)

  

1.05 (0.99–1.12)

0.99 (0.95–1.04)

aAdjusted for age (in categories), childbirth (ever/never), number of additional births (linear), age at first birth (linear), breastfeeding (linear), duration of oral contraceptives (linear), smoking (ever/never), intake of milk (linear)

In Table 3, the relationship between selected lifestyle factors and ovarian borderline tumor risk is presented. The risk of serous borderline tumors was associated with increasing body mass index, with a 5% increase in risk per kg/m2 (OR=1.05; 95% CI: 1.00–1.10), whereas no association was observed for mucinous tumors. Current smokers had a more than 2-fold increased significant risk of mucinous borderline tumors (CI: 1.22–3.60) compared to never smokers, and there was a trend of increased risk among women who had smoked for a long time, with an increase of 2% per additional year of smoking, although the estimate was only marginally significant (95% CI: 0.99–1.04). No clear association between smoking and serous tumors was found.
Table 3

Odds ratios (OR) and 95% confidence intervals (CI) for the relationship of body mass index (BMI), smoking habits and intake of milk and lactose with ovarian borderline risk

Characteristic

Cases (n=202)

Controls (n=1564)

Adjusted ORa (95% CI)

Histologic subgroup

Cases

Serous Adjusted ORa (95% CI)

Mucinous Adjusted ORa (95% CI)

Serous (n=104)

Mucinous (n=88)

Body mass index (kg/m2)

<22

67

484

1.00

27

36

1.00

1.00

22–24

52

520

0.76 (0.51–1.14)

24

25

0.83 (0.46–1.50)

0.76 (0.44–1.31)

25–26

29

218

1.06 (0.64–1.74)

16

13

1.22 (0.61–2.44)

1.06 (0.53–2.08)

27–29

29

172

1.33 (0.80–2.19)

21

7

2.27 (1.21–4.27)

0.56 (0.23–1.39)

≥ 30

24

156

1.09 (0.64–1.84)

15

7

1.58 (0.80–3.15)

0.63 (0.27–1.49)

Per kg/m2

  

1.01 (0.98–1.05)

  

1.05 (1.00–1.10)

0.96 (0.90–1.02)

Smoking status

Never

67

660

1.00

41

22

1.00

1.00

Former

53

391

1.34 (0.90–1.99)

30

20

1.32 (0.79–2.18)

1.49 (0.79–2.81)

Current

81

512

1.24 (0.86–1.79)

32

46

0.83 (0.50–1.37)

2.10 (1.22–3.60)

Duration of smoking (years)

never

67

660

1.00

41

22

1.00

1.00

≤10

11

109

1.08 (0.54–2.13)

5

5

0.85 (0.32–2.23)

1.35 (0.49–3.68)

11–20

13

138

0.84 (0.44–1.61)

7

6

0.88 (0.38–2.05)

1.05 (0.40–2.71)

21–30

50

212

1.86 (1.20–2.88)

21

27

1.44 (0.79–2.64)

2.70 (1.44–5.08)

31–40

39

210

1.45 (0.92–2.27)

21

16

1.26 (0.71–2.25)

1.76 (0.87–3.55)

≥41

20

233

0.83 (0.46–1.49)

8

12

0.50 (0.21–1.19)

1.80 (0.80–4.07)

Per year among smokers

  

1.00 (0.99–1.02)

  

1.00 (0.97–1.02)

1.02 (0.99–1.04)

Intake of milk (all types) in the last year (glasses of milk per week)

0

49

379

1.00

24

23

1.00

1.00

≤5

42

435

0.78 (0.50–1.23)

22

18

0.83 (0.45–1.52)

0.75 (0.39–1.43)

6–10

46

411

0.94 (0.61–1.46)

26

19

1.10 (0.61–1.98)

0.82 (0.43–1.55)

11–20

34

195

1.48 (0.90–2.42)

20

11

1.79 (0.93–3.41)

1.02 (0.48–2.19)

≥ 20

27

115

2.07 (1.21–3.55)

9

16

1.64 (0.73–3.71)

2.35 (1.17–4.74)

Per glass per week among women who drink milk

  

1.04 (1.02–1.06)

  

1.03 (1.01–1.06)

1.04 (1.02–1.07)

Total lactose intake in the last year (g/week)

0–49.9

70

523

1.00

38

29

1.00

1.00

50–149.9

72

725

0.76 (0.53–1.09)

37

34

0.70 (0.43–1.13)

0.88 (0.53–1.49)

150–249.9

34

176

1.57 (0.99–2.49)

16

14

1.30 (0.69–2.44)

1.58 (0.80–3.11)

≥250

18

59

2.52 (1.35–4.68)

8

9

2.42 (1.04–5.65)

2.70 (1.18–6.17)

Per 50 gram lactose per week among women who have intake of lactose

  

1.16 (1.06–1.26)

  

1.16 (1.03–1.30)

1.15 (1.02–1.29)

aAdjusted for age (in categories), childbirth (ever/never), number of additional births (linear), age at first birth (linear), breastfeeding (linear), duration of oral contraceptives (linear), smoking (ever/never), intake of milk (linear)

We also assessed the association between intake of milk (all types) and the risk of borderline tumor and found that an increasing consumption of milk increased the risk of both serous and mucinous ovarian borderline tumor, the risk per glass of milk per week being respectively 3% (95% CI: 1.01–1.06) and 4% (95% CI: 1.02–1.07) for serous and mucinous tumors. When the risk of ovarian borderline tumor according to lactose intake in milk was assessed, we found the same trend, with significantly increased risk of borderline tumor following increasing consumption of lactose in milk (OR=1.11 per 25 gram per week; 95% CI: 1.05–1.17) (data not shown). The same pattern was seen for both serous and mucinous tumors. No significant association between intake of yoghurt, junket or cheese with borderline tumors was seen, but for double cream we found that increasing consumption increased the risk of borderline tumor, as the risk per gram double cream per week was 18% (95% CI: 1.08–1.29), and similar patterns were seen for both serous and mucinous tumors (data not shown). In addition, intake of total lactose was assessed, and an increasing risk of both serous and mucinous borderline tumors with increasing consumption of total lactose was seen, the risk more than doubled for women in the highest category of lactose intake compared to the lowest (OR=2.52; 95% CI: 1.35–4.68).

Discussion

The results of this study confirm the protective effect of increasing parity on ovarian borderline tumor risk, and in addition an association between pregnancy timing and ovarian borderline disease was found, with a decrease in risk with late age at first birth. Both duration of breastfeeding and duration of oral contraceptive use were associated with a reduced risk of borderline tumors, the effect being most pronounced for serous tumors. We found an elevated risk of serous borderline tumor among women with increasing body mass index, whereas current smoking was a strong risk factor only for mucinous tumors. Another interesting finding of this study was an elevated risk of ovarian borderline tumors with high intake of milk and total lactose.

The biological mechanism of the protective effect of parity on ovarian borderline disease as well as invasive ovarian cancer has not yet been identified, but several hypotheses have been suggested, including the incessant ovulation hypothesis [19], and the raised gonadotropin hypothesis [20]. Pregnancy leads to as well anovulation as reduced gonadotropin production and the protective effect of parity can therefore be explained by both of these hypotheses. A more recent hypothesis has suggested that pregnancy induces clearance of precancerous cells from the ovaries [21], possibly through apoptosis due to high levels of progesterone during pregnancy [22].

In previous studies the association of timing of pregnancy with the risk of invasive ovarian cancer has been investigated and controversial results have been demonstrated [2226], but some studies have reported reduced risk of invasive ovarian tumors by pregnancy at older ages [2225]. The effect of the timing of pregnancy on the risk of ovarian borderline tumors is unclear, as some studies report a reduced risk related to older age at first birth [21] while others find no association [8, 9, 27]. In our study we found that the risk of ovarian borderline decreased significantly with older age at first birth, and the same association was observed when looking at serous and mucinous borderline tumors separately. A protective effect of a later first pregnancy fits the hypothesis of pregnancy clearing cells from the ovaries that have undergone malignant transformation, as the risk of having accumulated pre-malignant ovarian epithelial cells would be expected to increase with age [21]. To our knowledge, only one study has investigated age at last birth in relation to ovarian borderline disease, and no clear association was found [8]. In the present study no significant association between age at last birth and ovarian borderline risk was found after adjustment was performed. In this study, as in most previous reports [7], the risk of borderline tumor in the ovaries decreases with duration of breastfeeding. This is consistent with hypotheses regarding incessant ovulation.

It is widely accepted that OC use reduces the risk of invasive ovarian tumors, and the risk decreases with increasing duration of use. For ovarian borderline tumors the results have been less consistent, as some studies report a reduced risk of ovarian borderline tumor among OC users [9, 28], while others find no association [8]. In our study we found that women who had used OC had a lower risk of borderline tumor compared to never users. The protective effect was strongest in relation to serous borderline tumors, and virtually no effect was seen for mucinous tumors. This is in line with a study by Risch et al. [9], who also found that use of OC protected against ovarian borderline of all histological types, except mucinous. However, the protective effect of OC on ovarian borderline risk is in general weaker than the effect related to invasive ovarian cancer. It has been suggested that differences between the relation of OC and respectively invasive ovarian carcinomas and ovarian borderline could be due to differences in hormone doses, since it could be assumed that the relatively younger women at risk of ovarian borderline disease may have used the low-dose pill types, whereas the earlier high-dose pills may have been more widespread in the older group of women at risk of invasive ovarian cancer [7]. Most of the case-control studies concerning the protection of oral OC against invasive ovarian cancer most likely have enrolled women using the high-dose pills. It has not yet been determined whether the low-dose pill types also prevent ovarian malignancies [7].

In the present study we found an increased BMI to be associated with increased risk of serous borderline tumors, but no clear association with mucinous tumors was found. This corroborates the results from a recent Swedish study by Riman et al. [8]. Adipose women have more fat tissue producing estrogen, and it has been suggested that estrogens act as promoting agents that increase epithelial cell proliferation [29]. Also in relation to smoking we found a difference between serous and mucinous tumors as cigarette smoking was a significant risk factor for mucinous tumors, but not for serous tumors. Our findings are in agreement with two previous studies investigating serous and mucinous ovarian borderline tumors separately [8, 30]. The relationship between smoking and invasive mucinous ovarian tumor has also been reported in several studies [30, 31]. There have been speculations whether the relationship between cigarette smoking and mucinous tumors is due to the resemblance of mucinous tumors with gastrointestinal mucosa, since cigarette smoking has been shown to increase the risk of colon cancer, and affect the production of mucus in the large bowel [31].

In studies by Cramer and colleagues, milk products have been found to be a risk factor for invasive ovarian carcinomas [1012]. Two recent meta-analyses of epidemiological studies concerning milk and lactose consumption found conflicting results, since Qin et al. reported no association between milk/dairy products and invasive ovarian cancer risk [32], while Larsson et al. found that high intake of milk and lactose was associated with increased risk of ovarian cancer [13]. A possible explanation of the different results of these meta-analyses could be that Larsson et al. included more prospective cohort studies. In the present study we observed that milk and total lactose intake were strong risk factors for ovarian borderline tumor with a more than 2-fold increase in risk among women with high intake of milk and total lactose. To our knowledge, no previous studies have examined milk in relation to ovarian borderline tumors, except for one study examining the relation between galactose (originating from the hydrolysis of lactose, which is found naturally in milk products) consumption and galactose-1-phosphate uridyl transferase (GALT) genotypes and the risk of borderline tumors [33], with the conclusion that galactose intake may play a role in the development of ovarian borderline among women carrying an uncommon GALT N314D polymorphism. It has been proposed that high levels of galactose may cause direct toxicity to the ovary, or may cause a breakdown in the feedback control of gonadotropins [34], leading to increases in gonadotropin levels. Milk contains several nutrients as fat, lactose and calcium. Some studies have found no association between calcium intake and ovarian cancer risk [35, 36], while other studies have found that calcium intake reduces the risk of invasive ovarian cancer [37]. We did not examine other nutrients in milk than lactose, but future studies may have the possibility to elucidate whether it is the fat content rather than lactose, galactose or calcium that is related to ovarian borderline risk. In the present study we were not able to adjust lactose intake for total energy intake, which may influence the results, however, when adjusting the results for BMI no changes in the estimates were found (data not shown).

The risk profile of serous and mucinous borderline tumors seems to be different in some aspects. For the reproductive factors we found no differences between histological subtypes, but for BMI, smoking and use of OC divergence between serous and mucinous tumors was found. There have been speculations whether a possible explanation for the difference between non-mucinous and mucinous tumors could be, that some mucinous tumors may occur as a result of intestinal metaplasia of the surface epithelium or originate from gastrointestinal teratomas of germ-cell origin, as the mucinous tumors resemble gastrointestinal mucosa [9].

The strengths of this study include the relatively high number of epithelial ovarian borderline cases compared to other studies. Furthermore, the review procedure of ovarian borderline diagnoses by a specialized pathologist provides additional strength to the study. The participation rates among cases and controls were reasonably high, although 18% of eligible controls participated with a telephone interview only. Those who had telephone interview did not differ from the remaining controls with respect to number of pregnancies, length of OC use, menopausal status or hysterectomy. We have no information about BMI and smoking status on the telephone interviewed women. Recall biases can be a source of error in case-control studies, but by interviewing the case-group personally shortly after their operation, and by using life-event calendars, the recall biases were most likely reduced. Although the present study is among the largest case-control studies reported on ovarian borderline tumors, it is still limited by small numbers of cases in some of the histological categories, leading to low statistical power, especially for relatively infrequent exposures.

In conclusion, this study shows etiological differences between serous and mucinous borderline tumors, and it confirms that the risk factors of ovarian borderline disease share similarities with those of invasive ovarian tumors. These data strengthen the hypothesis that invasive epithelial ovarian cancers and epithelial ovarian borderline tumors share the same etiologies, and may suggest that ovarian borderline disease is not a separate entity.

Acknowledgments

We are grateful to the nurses and physicians from the departments of Gynecology and Pathology for their contribution, and to the interviewers for their great work.

Copyright information

© Springer Science+Business Media B.V. 2006