Risk factors for complete molar pregnancy from a case-control study

Risk factors for complete molar pregnancy from a case-control study

Risk factors for complete molar pregnancy from a case-control study Ross S. Berkowitz, M.D., Daniel W. Cramer, M.D., Sc.D., Marilyn R. Bernstein, B.F...

462KB Sizes 2 Downloads 41 Views

Risk factors for complete molar pregnancy from a case-control study Ross S. Berkowitz, M.D., Daniel W. Cramer, M.D., Sc.D., Marilyn R. Bernstein, B.F.A., Sally Cassells, B.S., Shirley G. Driscoll, M.D., and Donald P. Goldstein, M.D. Boston, Massachusetts Demographic, reproductive, and dietary histories for 90 white women with complete molar pregnancy were compared in a multivariate analysis with those of 90 parous controls matched to cases by residence, birth year, and race. Women with molar pregnancy were more likely to have been born outside North America (relative risk = 1.9, p = 0.05), were more likely to have been past age 30 at time of their molar pregnancy (relative risk = 1.6, p = 0.05), and were more likely to have diets deficient in the vitamin A precursor carotene. Women with dietary scores for carotene above the control median had a relative risk for molar pregnancy of 0.6 (p = 0.02). In addition, there was a significant trend for decreasing risk for molar pregnancy with increasing consumption of carotene. Although other nutritional deficiencies in patients with complete molar pregnancy may exist, carotene is a biologically plausible candidate for a nutritional risk factor that could explain the geographic distribution of molar pregnancy. (AM J OssTET GvNECOL 1985; 152:1016-20.)

Key words: Molar pregnancy, epidemiology, diet, carotene, fat

Description of the epidemiologic features of molar pregnancy has advanced little beyond the definition of broad demographic risk factors such as age, race, and geographic origin. Essentially no personal risk factors, such as contraceptive experiences or dietary habits, have been defined. Thus we undertook a case-control study of molar pregnancy and present the findings in this article.

Methods The cases were white women with complete molar pregnancy referred to the New England Trophoblastic Disease Center for evaluation and therapy between 1975 and 1980 and interviewed between December, 1981, and June 1982. Only patients with a confirmed histologic diagnosis of complete molar pregnancy who were residents of Massachusetts were chosen for study. Out of 150 potential cases, 32 (21%) had moved from the state, 15 (10%) could not be reached by phone or letter, and three (2%) declined to participate. Another 10 patients were interviewed but could not be matched

From the New England Trophoblastic Disease Center, Division of Gynecologic Oncology, Brigham and Women's Hospital, and the Departments of Obstetrics and Gynecology and Pathology, Harvard Medical School. This study was funded in part through a grant from theM assachusetts Division of the American Cancer Society. Ross S. Berkowitz, M.D., is a junior Faculty Clinical Fellow of the American Cancer Society. ReceivedforpublicationNovember 16, 1984; revisedApril12, 1985; accepted April29, 1985. Reprint requests: Ross S. Berkowitz, M.D., Brigham and Women's Hospital, 75 Francis St., Boston, MA 02115.

1016

to control subjects because they lived in an area of Massachusetts without Town Books (see below); they were excluded in this analysis based on matched pairs. Characteristics of the molar pregnancy patients who could not be interviewed or matched were similar to those of the interviewed cases (about 55% in both groups had had a pregnancy with a live birth prior to the molar pregnancy). Cases that could not be reached because they had moved from their last known address were younger than those who were reached. However, this was also true for the controls that had moved. The controls were selected from the Massachusetts Town Books, annual publications that list residents by name, age, and address. The controls matched the cases by year of birth (within 2 years), race, and precinct of residence, as a marker for socioeconomic status. It was also required that controls must have been delivered of a live-born infant. Of 187 women approached by introductory letter and telephone call, 44 (23%) had moved from that address, 14 (7%) could not be reached, 20 (11 %) had not had a live-born infant, 15 (8%) declined to participate, and four (2%) were ineligible because of race or language. These exclusions left 90 casecontrol pairs as the basis for analysis. Telephone interviews were conducted to assess a variety of factors including demographic history, pregnancies, contraception, prior gynecologic surgical procedures, smoking, coffee use, and dietary preferences. Exposure information such as contraceptive use, gy· necologic surgical procedures, prior pregnancies, and smoking history was censored after the date of an index pregnancy defined for cases and controls. The index

Risk factors for complete molar pregnancy

Volume 152 Number 8

1017

Table I. Distribution of case-control pairs and associated risk for molar pregnancy for various exposures, classified dichotomously Status of case-control pairs (N = 90) No discordancy

Factor

Subject's age ;.30 at index pregnancy Spouse's age ;.30 at index pregnancy Subject's birthplace outside North America Spouse's birthplace outside North America Catholic religion Age ;.J4 at menarche Oral contraceptive failure in index pregnancy Spermicide failure in index pregnancy IUD failure in index pregnancy Smoking during or prior to index pregnancy Induced abortion prior to index pregnancy Oral contraceptive use prior to index pregnancy Spermicide use prior to index pregnancy IUD use prior to index pregnancy Protein intake above control median Animal fat intake above control median Carotene intake above control median

Discordancy

Absent in both

Present in case, absent in control

9

56

20

19

37

0

Present in both

I

I

Present in control, absent in case

Relative

risk*

p Value

5

4.0

0.005

24

10

2.4

0.03

77

II

2

5.5

0.03

80

9

0

ll.lt

0.008

42 5 0

12 47 83

16 24 4

20 14 3

0.8 1.7 1.3

0.6 0.1 1.0

0

80

6

4

1.5

0.8

0

88

1.0

1.0

9

39

20

0.9

0.9

0

81

8

8.0

0.05

38

15

25

12

2.1

0.05

9

48

19

14

1.4

0.8

73

12

4

3.0

0.08

15

24

21

30

0.7

0.3

17

32

13

28

0.5

0.03

17

34

II

28

0.4

0.02

22

See text for definition of index pregnancy. *The relative risk is adjusted for the matching variables (year of birth, race, and residence). tUnpaired analysis used to estimate crude exposure odds ratio.

pregnancy was defined as the first molar pregnancy for the case and as the pregnancy with the same sequence number for the matched control. This method of censoring the exposure information in the controls is necessary because the occurrence of the molar pregnancy predated our interview by several years in some instances. Other methods of censoring the exposure information such as using the calendar date of the case's molar pregnancy in the matched control would not have allowed age at pregnancy to be studied as an "exposure" variable since case and control were already matched on year of birth. Each subject was asked about general diet including use of whole milk, cheese, butter, eggs, cabbage, cauliflower, brussel sprouts or broccoli, carrots, leafy greens, chicken, turkey, processed meats (hot dogs, salami, bologna, sausage), beef, pork, lamb, fish, regular coffee, decaffeinated coffee, tea, sodas or tonic, wine,

beer or liquor. The frequency of intake was categorized as daily use, weekly use, or never or less than weekly use. The consumption of coffee, tea, and sodas was measured in cups or cans per day. Crude intake scores for various nutrients were calculated based upon the subject's frequency of use of food items times the nutrient content of a typical portion. I. 2 The median value for any particular nutrient was identified for the control population, and members of each case-control pair were classified as "exposed" on the basis of whether their score fell above or below that control median. A paired analysis is appropriate since we matched for age, race, and residence. The effect of an exposure, controlling only for the matching factors, was tested with McNemar's 3 procedure based on dichotomous pairs. The effect of an exposure, controlling for other confounders, was examined with the use of multivariate logistic regression for paired data as described by Bres-

1018

Berkowitz et al.

August 15, 1985 Am J Obstet Gynecol

Table II. Matched multivariate model with greatest predictive value of risk for molar pregnancy* Factor

Carotene consumption above control median Subject's birthplace outside North America Subject's age ""30 at index pregnancy IUD use prior to index pregnancy

Adjusted relative riskt

95% Confidence limit

p Value

0.6

0.4-0.9

0.02

1.9

l.0-3.6

0.05

1.6

l.0-2.4

0.05

1.7

0.9-3.1

0.10

*x 2 Statistic to test all coefficients = 0 is 13.46 (4 df, p = 0.009). tThe relative risk for any factor is adjusted for the remaining three factors as well as for the matching variables (year of birth, race, residence).

low et a!! Finally, the presence of a dose response for a particular variable was tested by a matched pair technique for three exposure levels developed by Pike et al. 5 Results

The average age at interview for cases was 31.9 years and for controls 32.2 years. The index pregnancy (first molar pregnancy) for cases was the first pregnancy in 29 (32%) instances, the second pregnancy in 20 (22%) instances, and the third or later pregnancy in 4I (46%) instances. Forty-eight (53%) subjects had had a liveborn pregnancy before their molar pregnancy. No instances of twin pregnancies prior to or subsequent to the molar pregnancy occurred in this series. In all but one case, the index pregnancy prompted the first referral to the New England Trophoblastic Disease Center. Two of the ninety cases have had a subsequent molar pregnancy and one a subsequent postterm choriocarcinoma. Table I shows the distribution of case-control pairs and associated risk for molar pregnancy for various exposures, classified dichotomously. Exposures found to be associated with a significantly increased risk (p ~ 0.05) for molar pregnancy included subject's or spouse's age ;.30 years at index pregnancy, subject's or spouse's birthplace outside North America, induced abortion, and use of oral contraceptives prior to the index pregnancy. Exposures found to be associated with a significantly decreased risk (p ~ 0.05) for molar pregnancy included animal fat intake above the control median and carotene intake above the control median. Exposures found to be associated with an increased risk for molar pregnancy of borderline statistical significance (p :::: O.I) included age ;.I4 years at menarche and use of the intrauterine contraceptive device (IUD) prior to the index pregnancy. Exposures notable for their lack of association with molar pregnancy included religion, prior spermicide use, contraceptive failure during the index pregnancy, smoking prior to or dur-

ing the index pregnancy, and protein intake above the control median. Other exposures found not to be significantly related to molar pregnancy but not reported in Table I include: consanguinity, use of prescription medications (especially analgesics for dysmenorrhea) prior to the index pregnancy, regular use of alcohol, coffee consumption or other caffeinated beverage use, medical illness or reproductive losses prior to the index pregnancy, gynecologic surgical procedures prior to the index pregnancy, and occupation (professional, clerical, service, or factory). Some variables indicated to be significant in Table I are interrelated. To examine confounding among these variables, multivariate logistic regression was performed. Beginning with all the variables listed in Table I, step-down regression was performed until the 13-coefficient of each remaining variable had a statistical probability of~ I 0% of being equal to zero. This yielded the four-parameter model shown in Table II. Carotene consumption above the control median was associated with a relative risk of 0.6 (p = 0.02) and 95% confidence limits of 0.4 to 0.9 adjusted for age at index pregnancy, birthplace, and IUD use. Subject's birthplace outside North America was associated with an adjusted risk of 1.9 (p = 0.05) and confidence limits of 1.0 to 3.6. Subject's age ;.30 years at index pregnancy was associated with an adjusted risk of 1.6 (p = 0.05) and 95% confidence limits of 1.0 to 2.4. Use of the IUD prior to the index pregnancy was associated with an adjusted risk of I. 7 which, although not statistically significant, was included because it yielded the multivariate model which best accounted for the data. Spouse's age and spouse's birthplace were interchangeable with subject's age and birthplace in the model without materially affecting risks associated with carotene consumption. Distributions of carotene intake scores are shown in Table IliA, and trend in risk associated with level of carotene consumption is examined in Table IIIB. A significant trend (p = 0.01) for decreasing risk for mo-

Risk factors for complete molar pregnancy

Volume 152 Number 8

Table IliA. Distribution of carotene intake

Table IVA. Distribution of animal fat intake

scores in case-control pairs

scores in case-control pairs

Distribution in controls Distribution zn cases

Lowest third Middle third Upper third

Lowest third

Distribution in controls Upper third

I5

II

I8

IO

I3

7

5

6

5

Distribution zn cases

Lowest third Middle third Upper third

Lowest third

Upper third

15

14

13

8

9

7

5

7

12

Table IIIB. Associated trend of carotene

Table IVB. Associated trend of animal fat

intake in relative risk for molar pregnancy

intake in relative risk for molar pregnancy

Category of carotene consumption*

Relative riskt

Lowest third Middle third Upper third

0.6 0.4

1019

95% Confidence limit

Category of animal fat consumption*

Relative riskt

95% Confidence limit

0.2-1.2 0.1-0.8

Lowest third Middle third Upper third

1 0.9 0.4

0.3-2.0 0.2-0.9

I

*Based on thirty-third and sixty-sixth percentile intake scores for controls, the Mantel extension x for linear trend is -2.49 (p = 0.01). tThe relative risk is adjusted only for the matching variables.

*Based on thirty-third and sixty-sixth percentile intake scores for controls, the Mantel extension x for linear trend is -2.12 (p = 0.03). tThe relative risk is adjusted only for the matching variables.

lar pregnancy was associated with increasing carotene consumption such that women in the upper third for carotene consumption had about one third the risk for molar pregnancy of women in the lowest third. Distributions of animal fat intake scores are shown in Table IVA. Although the significant association between low dietary animal fat and molar pregnancy identified in the simple paired analysis did not persist in the multivariate analysis, it is of interest that there was a significant trend (Table IVB). Women in the upper third for animal fat consumption had about one third the risk for molar pregnancy of women in the lowest third (Table IVB).

among individuals with similar demographic characteristics. To meet this deficiency we performed a case-control study based on personal interview of 90 white women with complete hydatidiform mole and 90 parous controls matched to cases by birth date, race, and residence. Our findings concerning birthplace and age confirm the studies cited above. Women born outside North America had a 1.9-fold increased risk for molar pregnancy as compared to those born in North America. Spouse's birthplace was also significantly related to risk for molar pregnancy. In our analysis of age as a risk factor for molar pregnancy, we controlled for parity by comparing age at the first molar pregnancy with age at the pregnancy of the same sequence number in the age-matched control. Women pregnant past the age of 30 had a 1.6-fold increased risk for having a molar pregnancy compared to those pregnant before age 30. Spouse's age at index pregnancy was also a risk factor for molar pregnancy but not as strong as maternal age. Pregnancy at an older age may more likely be preceded by contraception or abortion and this may confound the association between prior oral contraceptive use or abortion and molar pregnancy found in a simple paired analysis of the data in Table I. In a multivariate analysis adjusted for age at pregnancy, induced abortion and oral contraceptive use did not persist as statistically significant risk factors. It should also be con-

Comment

Geographic origin and age appear to be important determinants of the frequency of hydatidiform mole and other forms of trophoblastic disease. The incidence is low in the United States" and Europe and greater in Asia." Increasing age at pregnancy is also associated with increased risk for hydatidiform mole,''· 7 although some studies have also suggested greater rates in adolescents.9 Virtually all of the studies cited are based on hospital or registry data in which the demographic profile of patients diagnosed as having a molar pregnancy is compared with that of women delivered of a live-born infant. Such studies are limited in that they do not permit the identifir<~tion of personal risk factors

1020

Berkowitz et al.

sidered that there may be preferential recall of such exposures by cases. However, it is noteworthy that contraceptive failure of the IUD, oral contraceptives, or spermicides in the index pregnancy was not associated with increased risk in this small study. Geographic differences in the incidence of hydatidiform mole have been attributed to nutritional factors with protein deficiency cited as a possible factor." Our study of diet in relation to molar pregnancy compared intake scores for protein, fat, and carotene for cases and controls. Although no differences were noted between cases and controls for protein intake scores, cases differed from controls in animal fat and carotene consumption with the difference favoring a protective effect for these nutrients. Carotene consumption was the variable of greatest predictive value for molar pregnancy risk in the multivariate analysis of the data. It should be emphasized that the food questionnaire used in this study was not exhaustive in quantifying all nutrients. In particular, the omission of breads, cereals, fruits, and liver from the questionnaire did not permit scoring of total carbohydrates, water-soluble vitamins, and preformed vitamin A. Since levels of one nutrient are often correlated with levels of other nutrients, the deficiencies in carotene and animal fat consumption observed for cases may simply correlate with deficiency of other nutrients. Furthermore the validity of these dietary data in relation to molar pregnancy depends upon the assumption that dietary patterns in individuals are relatively constant over time. Nevertheless, the suggested associations between molar pregnancy and deficiency of carotene (and possibly fat) do have biologic credibility. Both carotene, as a precursor to vitamin A, and fat have effects on reproduction. Vitamin A deficiency in the male rat and rhesus monkey produced degeneration of the seminiferous epithelium with consequent production of primitive spermatogonia and spermatocytes, while deficiency in the female rat and rhesus monkey produced fetal resorption or abortion. HI.'' Deficiency of fat is also associated with degeneration of the seminiferous epithelium in male rats and fetal resorption and impaired ovulation in female rats and mice. 12 · '" However, the influence of dietary fat deficiency on reproduction could be mediated in part through its effect on carotene absorption. When the diet is deficient in fat, the absorption of carotene is limited.'' Geographic areas with a high incidence of vitamin A deficiency" correspond to regions with a high incidence of molar pregnancy. Carotene and vitamin A deficiency have general effects on epithelial differentiation in addition to their specific effects on reproduction already

August 15, 1985 Am J Obstet Gynecol

discussed. Epidemiologic evidence linking deficiency of carotene and human cancer has recently been reviewed.'6 However, it is not clear precisely how deficiency of fat or carotene could cause molar pregnancy or how they might relate to the androgenetic origin of hydatidiform moles. Given the exploratory nature of this study it is premature to conclude there is a causal relationship between molar pregnancy and carotene or fat deficiency, but our findings do indicate the need for further studies of nutrition and molar pregnancy with emphasis on carotene or other fat-soluble vitamins and animal fat. REFERENCES I. Church CF, Church HN, eds. Food values of portions

2. 3. 4. 5.

6. 7. 8.

9.

10.

commonly used. 12th rev. ed. Philadelphia:JB Lippincott, 1975. Adams CF: Nutritive value of American foods. Washington DC: United States Department of Agriculture, 1975; no. 456. McNemar Q. Note on the sampling of the difference between corrected proportions on percentages. Psychometrika 1947;12:153. Breslow NE, Day NE, Halvorsen KT, et al. Estimation of multiple relative risk functions in matched case-control studies. Am J Epidemiol 1978; I 08:299. Pike MC, Casagrande J, Smith PG. Statistical analysis of individually matched case-control studies in epidemiology: factor under study a discrete variable taking multiple values. Br J Prev Soc Med 1975;29:196. Hayashi K, Bracken MB, Freeman DH Jr, et al. Hydatidiform mole in the United States ( 1970-1977): a statistical and theoretical analysis. Am J Epidemiol 1982; 115:67. Ringertz N. Hydatidiform mole, invasive mole and choriocarcinoma in Sweden, 1958-1965. Acta Obstet Gynecol Scand 1970;49: 195. Acosta-Sison H. Observations which may indicate the etiology of hydatidiform mole and explain its high incidence in the Philippines and Asiatic countries. Philipp J Surg Special 1959;14:290. Jacobs PA, Hunt PA, Matsuura JS, et al. Complete and partial hydatidiform mole in Hawaii: cytogenetics, morphology, and epidemiology. Br J Obstet Gynaecol 1982; 89:258. Thompson JN. The role of vitamin A in reproduction. In: DeLuca HF, Suttie JW, eds. The fat soluble vitamins. Madison: The University of Wisconsin Press, 1969:26781.

11. O'Toole BA, Fradkin R, Warkany J, et al. Vitamin A deficiency and reproduction in rhesus monkeys. J Nutr 1974;104:1513. 12. Evans HM, Lepkovsky S, Murphy EA. Vital need of the body for certain unsaturated fatty acids. VI. Male sterility on fat-free diets. J Bioi Chern 1934;106:445. 13. Kim HL, Picciano MF, O'Brien W. Influence of maternal dietary protein and fat levels on fetal growth in mice. Growth 1981;45:8. 14. Roels OA, Trout M, Dujacquier R. Carotene balances on boys in Ruanda where vitamin A deficiency is prevalent. J Nutr 1958;65: 115. 15. McLaren DS. Present knowledge of the role of vitamin A in health and disease. Trans R Soc Trop Med Hyg 1966;60:436. 16. Peto R, Doll R, Buckley JD, et al. Can dietary beta-carotene materially reduce human cancer rates? Nature 1981 ;290:20 I.