Family history as a predictor of early menopause*

t

FERTILITY AND STERILITY Copyright

Vol. 64, No.4, October 1995

Printed on acid-free paper in U. S. A.

1995 American Society for Reproductive Medicine

Family history as a predictor of early menopause*

Daniel W. Cramer, M.D., Sc.D. t Huijuan Xu, M.P.H. Bernard L. Harlow, Ph.D. Obstetrics and Gynecology Epidemiology Center, Department of Obstetrics and Gynecology, Brigham and Women's Hospital, Boston, Massachusetts

Objective: To determine the relative importance of family history as a predictor of early menopause. Design: Case-control study. From a population-based survey of 10,606 women between 45 and 54 years of age, we selected 344 cases with early menopause (average age 42.2 years) and 344 age-matched controls who were still menstruating or who had a menopause after age 46 years. Subjects were interviewed about their medical and family history and blood was drawn for identification of women who were carriers for the classic or Duarte variant of galactosemia, a potential hereditary factor for early menopause. Logistic regression analysis was used to estimate the risk of an early menopause in women with and without a family history of early menopause. Results: Overall 129 (37.5%) of the early menopause cases reported a family history of menopause before age 46 years in a mother, sister, aunt, or grandmother compared to 31 (9.0%) of controls yielding an odds ratio (OR) of 6.1 (95% confidence interval [CI] of 3.9 to 9.4) after adjustment for smoking history, education, parity, and body mass index. Risk for early menopause associated with family history of same was greatest: for family history in a sister, OR = 9.1 (95% CI 3.1 to 26.5); multiple relatives, OR = 12.4 (95% CI 4.4 to 34.2); and cases menopausal before age 40 years, OR = 8.4 (95% CI 2.5 to 31.2). Cases with a family history of early menopause were not more likely to have errors of galactose metabolism compared with cases without a family history or to all controls, nor did they possess Turner's stigmata such as short stature, but they were less likely to have brothers in their sibships. Conclusions: Although preferential recall of family history by women with early menopause could contribute to the association between family history and early menopause observed in this study, a genetic factor is also plausible including partial deletions of the X chromosome compatible with the deficiency of male siblings in cases with family history of early menopause. Fertil Steril 1995;64:740-5 Key Words: Menopause, family history, galactose metabolism

Fewer than 10% of women undergo a menopause before age 46 years and only 1% before age 40 years (1). A possible genetic basis for early menopause is

Received December 9, 1994; revised and accepted April 13, 1995. * Supported by grant R01HD23661 from the National Institute of Child Health and Human Development, Bethesda, Maryland. t Reprint requests: Daniel W. Cramer, M.D., Sc.D., Obstetrics and Gynecology Epidemiology Center, Brigham and Women's Hospital, 221 Longwood Avenue, Boston, Massachusetts 02115 (FAX: 617-732-4899). 740

Cramer et al. Family history and early menopause

suggested by description of pedigrees in which early menopause occurred in two or more generations (27). However, such case reports do not provide a basis for appreciating the relative importance of family history as a predictor of early menopause. To identify potential predictors of early menopause, epidemiologic data, including family history, was collected for 344 women with an early age at menopause (average age 42.2 years) and was compared with similar data in 344 age-matched women who were still menstruating or had a more normal age at menopause. In this article, family history as a risk factor for early Fertility and Sterility

menopause is examined, potential biases related to this association are discussed, and the biologic interpretation of such data is considered. MATERIALS AND METHODS

Case-control methodology was used to define the risk for early menopause associated with family history. Subjects were selected from a cross-sectional survey of the population of greater Boston conducted between 1989 and 1992 which was described previously (8). Using town census lists to identify women aged 45 to 54 years, we sent 14,606 postal surveys requesting information about current menstrual status. After excluding 10% of subjects who had moved, were not in the target age range, or had language difficulties with the survey, and 17% of subjects who refused, 10,606 surveys were returned and used to select cases of early menopause for more intensive study. Women with an early menopause (cases) were selected from approximately 10% of survey respondents who had experienced menopause before age 47 years not as a consequence of surgery, radiation, or chemotherapy. Women with the earliest ages at menopause were oversampled preferentially. We selected all women with a natural menopause before age 40 years, a 75% random sample of women with a menopause between the ages of 40 and 43 years, and a 25% random sample of women with a menopause between the ages of 44 and 46 years. From 440 women identified as potential cases and invited to participate in the more detailed study, 344 (78.2%) agreed to participate including 64 women < 40 years at menopause, 140 between 40 and 43 years, and 140 between 44 and 46 years. The average age at menopause for the entire case group was 42.2 years. Potential controls were selected randomly from survey respondents who were the same current age as cases but who still were menstruating or had menopause after age 47 years. An equal number of controls was sought for cases in each of the three menopause categories described. We contacted a total of 464 women to obtain 344 controls for a response rate of 74.1 %. Approximately 30% of controls were themselves menopausal and their average age at menopause was 49.8. The study protocol involved the following: informed consent; a questionnaire administered inperson dealing with reproductive, medical, and family history; a self-administered food frequency questionnaire; and a venous blood specimen that was separated into plasma, buffy coat, and red cell components. From the red cell specimen, the activity and electrophoretic pattern of galactose-I-phosphate uridyltransferase (GALT) was analyzed at the ChilVol. 64, No.4, October 1995

dren's Hospital of Los Angeles to explore a possible connection between deficiency of GALT and early menopause (9). The GALT genotype has been inferred from the electrophoretic pattern and activity level and includes alleles for the classic, Duarte, and Los Angeles variants (see footnote to Table 3). Details from the medical history on the questionnaire were not verified independently from other medical records. In addition, because cases were interviewed :2: 1 years after their menopause had occurred, exposures occurring after or around the time of the menopause needed to be censored. A "reference age" was defined for the cases as 1 year before the age at menopause; and only those exposures occurring before this age were counted to avoid the possibility of including exposures that occurred as a consequence of perimenopausal events. For cases the average age at interview was 49.4 years, and the average reference age was 41.2 years. As previously described (8), a comparable reference age was defined for the age-matched controls such that their average age at interview was 49.6 years and their average reference age was 41.1 years. Regarding family history, subjects were asked about a variety of medical problems including cancer, infertility, and early menopause among their mother, sisters, aunts, or grandmothers. An early menopause in a relative was defined as a natural menopause before age 46 years. Data were analyzed first using standard 2 X K tables with calculation of crude exposure odds ratios (ORs) and then using multiple logistic regression analysis to compute adjusted ORs and 95% confidence intervals (CIs). RESULTS

No general tendency was observed for cases to report an excess of medical events among first-or second-degree relatives. For example, a family history of diabetes was reported by 11.3% of cases, heart disease and stroke by 33.1% of cases, and breast cancer by 26.5% of cases compared with 14.8%, 34.9%, and 27.9% of controls for the same diseases. There was a slight but not significant excess of cases who reported a family history of infertility (22.7% of cases versus 18.9% of controls, P = 0.22). However, there was a highly significant excess of cases who reported a family history of early menopause (Table 1). One hundred twenty-nine (37.5%) of cases reported they had at least one female relative who had experienced a menopause before age 46 years compared with 31 (9.0%) of controls. This translated into an OR of 6.1 (95% CI 4.0 to 9.3) (P < 0.001) for early menopause associated with family history of same. The frequency of a family history of early menopause among controls who were themselves Cramer et al. Family history and early menopause

741

,. Table 1 Odds Ratios for Early Menopause by Family History of Early Menopause Family history of early menopause None Anyt Single relative Paternal, 2nd degree Maternal, 2nd degree Mother Sister Multiple relatives

(n

Control (n = 344)

Odds ratio*

215 129

313 31

1.0 6.1 (4.0 to 9.3)

2

0

10 58 25 34

6 17 4 4

Case = 344)

-:j: 2.4 (0.9 5.0 (2.8 9.1 (3.1 12.4 (4.4

to to to to

6.8):j: 8.8)§ 26.5)§ 34.2)§

* 95% confidence interval in parentheses. t Defined as a natural menopause before age 46 in a mother, sister, aunt, or grandmother. :j: P (X 2 ) = 0.09. §P < 0.001.

menopausal was 13 (12.8%) out of 102; and, using this group as the referent, the OR for family history of early menopause as a predictor of same was 4.1 (95% CI 2.2 to 7.6) with statistical significance of P < 0.001. The nature of the family history of early menopause as it translated into risk for early menopause is further detailed in Table 1. Family history of early menopause only in a single aunt or grandmother did not lead to a statistically significant elevation in risk for early menopause in the proband. Greatest risk for early menopause was associated with a history of early menopause in a mother, sister, or multiple relatives. The ORs (and 95% CI) for early menopause associated with such histories were 5.0 (2.8 to 8.8), 9.1 (3.1 to 26.5), and 12.4 (4.4 to 34.2). The majority of the early menopause cases with multiple relatives had a mother and one or more sisters with early menopause but no instances occurred where the transmission appeared to have been exclusively along paternal lines, such as menopause occurring in a sister and a paternal aunt or grandmother. Table 2 examines the association between family history of early menopause and occurrence of same in the proband by categories of selected variables that might be considered potential confounding factors such as educational level, smoking history, body mass index (BMI), parity, and estimated number of ovulatory cycles (8). In all categories of these variables, we observed a significantly elevated risk for early menopause associated with a similar family history. The final two entries in Table 2, number of siblings and "earliness" of menopause, are of interest in identifying subgroups where family history may be a more dominant risk factor for early menopause. Family history was a very strong risk factor for early menopause among subjects who were an only child, 742

Cramer et al. Family history and early menopause

where the OR was 22.9, although the 95% CI was very wide (2.7 to 503.7). The association was also strong in the category of cases with menopause before age 40 years where the OR (and 95% CI) was 8.4 (2.5 to 31.2). Adjusting for education, smoking history, BMI, ovulatory cycles, and number of siblings, the OR (and 95% CI) for early menopause associated with a similar family history was 6.1 (3.9 to 9.4). Finally, Table 3 examines some characteristics of the cases with a family history of early menopause compared with those cases without a family history of early menopause and all controls. Features that notably did not distinguish cases with a family history of early menopause was GALT activity or the possession of a low activity variant of GALT (Duarte or classic galactosemia). Two features that distinguished the cases with a family history of early menopause from all controls included shorter cycle lengths and a greater frequency of never married subjects; but these were traits that, in general, occurred more frequently among women with early menopause. Cases with a family history of early menopause were less likely to have brothers in their

Table 2 Odds Ratios for Early Menopause by Family History Within Categories of Potential Confounding Factors Case, family history Confounder Education 22.5 Smoking Never Former Current Live birth None lor 2 .,,3 Ovulatory cycles <300 .,,300 Siblings None lor 2 .,,3 Category:j: <40y 40to43y 44to46y

Control, family history

No

Yes

No

Yes

Odds ratio*

99 116

56 73

102 211

14 17

4.1 (2.1 to 8.3) 7.8 (4.3 to 14.5)

111 104

68 61

172 141

14 17

7.5 (3.9 to 14.8) 4.9 (2.6 to 9.2)

77 64 74

46 42 41

141 85 87

7 13 11

12.0 (4.9 to 30.8) 4.3 (2.0 to 9.2) 4.4 (2.0 to 9.8)

57 99 59

35 58 36

54 160 99

6 12 13

5.5 (2.0 to 16.0) 7.8 (3.8 to 16.2) 4.6 (2.2 to 10.1)

78 137

49 80

155 158

15 16

6.5 (3.3 to 13.0) 5.8 (3.1 to 10.8)

19 109 87

15 66 48

29 163 121

1 16 14

22.9 (2.7 to 503.7) 6.2 (3.3 to 11.8) 4.8 (2.4 to 9.7)

41 81 93

23 59 47

60 124 129

4 16 11

8.4 (2.5 to 31.2) 5.6 (2.9 to 11.0) 5.9 (2.8 to 12.8)

* 95% confidence interval in parentheses. t Body mass index = weight (kg) divided by square of height (m). :j: Case and Control category for age at menopause.

Fertility and Sterility

Table 3 Characteristics of Cases With and Without a Family History of Early Menopause Compared With All Controls

Variables Age at menarche (y)* sl1 12 and 13 .",14 Cycle length (d)*t <28 28 and 29 .",30 Ever married*t No Yes Live birth*:/: None lor 2 .",3

Height (cm)* <160 160 to 169 .",170 GALT genotype*§ NN DN orGN LN or Unknown Transferase activitYIl Hypertension* No Yes Hypothyroid* No Yes No. of brothers*t~ None .",1

Case, with family history

Case, without family history

DISCUSSION All controls

34 (26.4) 66 (51.2) 29 (22.4)

60 (27.9) 109 (50.7) 46 (21.4)

70 (20.4) 200 (58.1) 74 (21.5)

28 (21.7) 73 (56.6) 28 (21.7)

35 (16.3) 132 (61.4) 48 (22.3)

54 (15.7) 172 (50.0) 118 (34.3)

20 (15.5) 109 (84.5)

25 (11.6) 190 (88.4)

32 (9.3) 312 (90.7)

15 (13.8) 58 (53.2) 36 (33.0)

34 (17.9) 98 (51.6) 58 (30.5)

29 (9.3) 171 (54.8) 112 (35.9)

27 (20.9) 72 (55.8) 30 (23.3)

33 (15.3) 130 (60.5) 52 (24.2)

89 (25.9) 175 (50.9) 80 (23.3)

110 (85.3) 14 (10.8) 5 (3.9)

174 (80.9) 26 (12.1) 15 (7.0)

283 (82.3) 48 (13.9) 13 (3.8)

22.4 ± 0.27

22.7 ± 0.25

22.3 ± 0.21

125 (96.9) 4 (3.1)

204 (94.9) 11 (5.1)

337 (98.0) 7 (2.0)

121 (93.8) 8 (6.2)

195 (90.7) 20 (9.3)

323 (93.9) 21 (6.1)

53 (41.1) 76 (58.9)

64 (29.8) 151 (70.2)

105 (30.5) 239 (69.5)

Brothers/siblings ratiotll~

0.38 ± 0.03

0.47 ± 0.03

0.48 ± 0.02

* Values are number of cases with percentage in parentheses. t p < 0.05 for difference between cases with a family history and all controls. :/: Restricted to married women. § GALT genotype inferred, biochemically, on the basis of the electrophoretic pattern and GALT activity level and commonly includes: normal-normal individuals with a single moving band and activity in the range of 17 to 37 micromoles of hexose conversion per hour per gram of hemoglobin; galactosemia-normal individuals with a single moving band at the normal position but activity in the range of 6 to 16; Duarte-normal variants with two faster moving bands in addition to a normal band and activity in the range of 12 to 24; and Los Angeles-normal individuals with three moving bands corresponding to the positions of the Duarte variant but activity in the range of 20 to 36. II Values are means ± SE. ~ P < 0.05 for difference between cases with a family history and cases without a family history.

sibships compared with either all controls or cases without a family history. Forty-one percent of the cases with a family history had no brothers compared with approximately 30% in both the control group and the case group without a family history. The ratio of brothers to all siblings was also signifiVol. 64, No.4, October 1995

cantly lower in the family history case group compared with all others.

Pedigrees have been described in which a premature menopause (before age 40 years) appeared in two or more generations, suggesting autosomal dominant or X-linked inheritance (2-3). The X-linked inheritance postulated has been demonstrated in other pedigrees describing early menopause in association with partial deletions of the X chromosome (4-7). There is no firm agreement on what proportion of women who present with premature ovarian failure will have a family history of early menopause. Starup and Sele (10) reported only 1 in 26 women (4%) had a family history of early menopause, but this series consisted of women in their teens or twenties with very early ovarian failure. In a more recent series (11) of patients whose age at menopause was in their late twenties and in whom known causes of ovarian failure such as chemotherapy, autoimmunity, and X monosomy were excluded, the frequency of a family history of early menopause was 12 of 40 (30%). This frequency more closely approximates the 37.5% found in the current study that focused on "idiopathic" early menopause in the late 30s and early 40s. Knowledge of the frequency of a family history of early menopause among women with early menopause does not indicate the importance offamily history as a predictor of early menopause without knowledge of the occurrence of early menopause in a general population or control group. In this study the 37.5% frequency of a family history of early menopause (i.e., before age 46 years) in the case group contrasted sharply with the 9% frequency observed in the control group and translated into an approximately sixfold increase in risk for early menopause associated with such a history. We were able to identify only one other epidemiologic study that has addressed the issue of family history as a predictor of menopause. In a cross-sectional study of women 45 to 49, Torgerson et al. (12) found that women were more likely to have made their menopausal transition during this interval iftheir mother had also gone through a menopause at age 45 to 49 years. There are some obvious potential biases to consider in the interpretation of this association including under-reporting by controls or over-reporting by cases of family medical events. Because the controls were still menstruating or had experienced a normal age at menopause, they may have been less aware and under-reported a family history of early menopause. However, the association persisted when the Cramer et al. Family history and early menopause

743

r

.

referent group was restricted to those controls who had themselves undergone menopause; and the overall control frequency of 9% who reported that a relative had a natural menopause before age 46 approximates the cumulative incidence of menopause before age 46 years in the general population (1). No general tendency for cases to over-report family medical events such as diabetes, female cancer, or heart disease was observed. Because we could not verify age at menopause directly from the relatives, the possibility remains that relatives of cases were more likely to recall preferentially (or erroneously) an early menopause than were the relatives of controls. It is also possible that family history as a predictor of risk for early menopause is mediated through habits that tend to cluster in families rather than genetic factors. Traits like smoking could have been learned from a mother or sister and caused early menopause (13) in both the proband and her relative, suggesting a familial tendency. However, Table 2 indicates that the association between early menopause and family history of same was stronger among nonsmokers. Treating educational status, smoking, parity, and BMI as potential confounders in a multivariate model, the OR for family history as a predictor of early menopause remained significant. Despite the possible biases of a retrospective study, it seems likely that the association between family history and early menopause has a genetic basis. Premature ovarian failure has been reported in women who have classic galactosemia (14) and an early menopause before age 48 years in carriers for galactosemia, including the Duarte variant (15). However, in this data set, there was no excess of women who carried the Duarte or classic variant of galactosemia observed among cases with a family history of early menopause compared with controls. A limitation of our study was the fact that GALT genotype was inferred biochemically from red cell enzyme characteristics rather than assessed by direct DNA testing. Only recently has the GALT gene been cloned and key mutations identified including the Q188R mutation associated with classic galactosemia and the N314D mutation associated with the Duarte variant (16). In a separate study, possession of an N314D or Q188R mutation was associated with increased early follicular phase FSH levels, a potential marker for ovarian dysfunction (17). There is considerable support for X chromosomal errors as a factor in early menopause. In the extreme, X monosomy is associated with primary amenorrhea, sexual infantilism, short stature, gonadal maldevelopment (dysgenesis), and other medical conditions such as hypertension, autoimmune disorders, or hypothyroidism (18). Partial deletions 744

Cramer et al. Family history and early menopause

of the long or short arm of the X chromosome may also occur and be associated with primary or secondary amenorrhea with or without Turner's stigmata. In a scholarly review, Therman and Susman (19) hypothesized that loss ofXp or major deletions ofXq that might somehow affect the activation region on Xp lead to gonadal dysgenesis with Turner's stigmata but that deletions distal to Xq25 may less severely affect the gonads and produce early menopause in women without Turner's stigmata. The early menopause cases in this study with a family history did not have a history of Turner characteristics such as short stature, hypertension, or thyroid disease, although it must be conceded that no standardized search for such features was conducted. Of potential interest, however, was a lower frequency of brothers in sibships of cases with a family history of early menopause compared with cases without a family history or all controls. Furthermore, instances of "transmission" of early menopause along paternal lines were rare in this series. As pointed out by Fitch et al. (5), no male subject with a terminal deletion of the long arm of the X chromosome has been described (possibly due to the presence of one or more genes essential for life including glucose-6-phosphate dehydrogenase). These observations invite speculation that defects of the X chromosome might, at least partially, mediate the importance offamily history as a risk factor for early menopause. Karyotyping was not performed on subjects in this study so no definite answer can be provided in this paper whether X chromosomal abnormalities accounted for the observed association between family history and early menopause. In summary, using a case-control format, we have observed that family history of early menopause (before age 46 years) increased the risk for an early menopause in the proband. The association was strongest in women who had a menopause before age 40 years and those with a sister or multiple relatives with early menopause and was not confounded by possible familial traits such as smoking or BM!. Although over-reporting of a family history by the cases is certainly possible, a genetic basis for the association is plausible. Decreased metabolism of galactose reported as a genetic factor in early menopause did not account for the association but the possibility of partial deletions of the X chromosome is suggested by a decreased frequency of males observed in the sibships of women with a family history of early menopause. Regardless of what genetic factors may be found eventually as the basis for a familial tendency for early menopause, women should make themselves aware of the age at menopause of their relatives because it may impact upon decisions about the timing of pregnancies or use of replacement therFertility and Sterility

apy. The strength of this association is such that history of menopause before age 46 years in a sister or mother increases the probability that a woman will have her menopause before age 46 years from approximately 5.0% to approximately 25%. REFERENCES 1. Coulam CB, Adamson SC, Annegers JF. Incidence of pre mature ovarian failure. Obstet Gynecol 1980;67:604-6. 2. Coulam CB, Stringfellow SB, Hoefnagel D. Evidence for a genetic factor in the etiology of premature ovarian failure. Fertil Steril 1983;40:693-5. 3. Mattison DR, Evan MI, Schwimmer WB, White BJ, Jensen B, Schulman JD. Familial premature ovarian failure. Am J Human Genet 1984;36:1341-8. 4. Veneman TF, Beverstock GC, Exalto N, Mollevanger P. Premature menopause because of an inherited deletion in the long arm of the X-chromosome. Fertil Steril1991;55:631-3. 5. Fitch N, Saint Victor J de, Richer CL, Pinsky L, Sitahal S. Premature menopause due to a small deletion in the long arm of the X chromosome: a report ofthree cases and a review. Am J Obstet Gynecol 1982; 142:968-72. 6. Krauss CM, Turksoy RN, Atkins L, McLaughlin C, Brown LG, Page DC. Familial premature ovarian failure due to an interstitial deletion of the long arm of the X chromosome. N Engl J Med 1987;317:125-31. 7. Bates A, Howard PJ. Distal long arm deletions ofthe X chromosome and ovarian failure. J Med Genet 1990;27:722-3. 8. Cramer DW, Xu H, Harlow BL. Does "incessant" ovulation increase the risk for early menopause. Am J Obstet Gynecol 1995; 172:568-73.

Vol. 64, No.4, October 1995

9. Lee JES, Ng WG. Semi-micro techniques for the genotyping of galactokinase and galactose-I-phosphate uridyl transferase. Clin Chern Acta 1982;124:351-6. 10. Starup J, Sele V. Premature ovarian failure. Acta Obstet Gynecol Scand 1973;52:259-68. 11. Eden JA. Menopause before 40-premature, but not always permanent. Aust NZ J Obstet Gynaecol 1993;33:201-3. 12. Torgerson DJ, Avenell A, Russell IT, Reid DM. Factors associated with the onset of menopause in women aged 45-49. Maturitas 1994; 19:83-92. 13. Jick H, Porter J. Relation between smoking age of natural menopause. Lancet 1977;i:1354-5. 14. Kaufman FR, Kogut MD, Donnell GN, Goebelsmann U, March C, Koch R, et al. Hypergonadotropic hypogonadism in female patients with galactosemia. N Engl J Med 1981; 304:994-8. 15. Cramer DW, Harlow BL, Barbieri RL, Ng WG. Galactose-1phosphate uridyl transferase activity associated with age at menopause and reproductive history. Fertil Steril 1989; 51:609-15. 16. Reichardt JKV, Woo SLC. Molecular basis of galactosemia mutations and polymorphisms in the gene encoding human galactose-I-phosphate uridyl transferase. Proc Natl Acad Sci USA 1991;88P:2633-7. 17. Cramer DW, Barbieri RL, Xu H, Reichardt JKY. Determinants of basal follicle stimulating hormone levels. J Clin Endocrinol Metab 1994; 79:1105-9. 18. Speroff L, Glass RH, Kase NG, editors. Clinical gynecologic endocrinology and infertility. Baltimore (MD): Williams and Wilkins, 1989. 19. Therman E, Susman B. The similarity of phenotypic effects caused by Xp and Xq deletions in the human female: a hypothesis. Hum Genet 1990;85:175-83.

Cramer et al. Family history and early menopause

745