OVARIAN CANCER

Incidence and Case-Control Study

JOHN PHD, H. STROM,MD, DAVIDG. DECKEK,
F. ANNEGEKS, MD,
MD,'k A N D W. MICHAEI.
MALXOLMB. DOCKEKIY, PHD
O'FAL.I.ON,

The incidence of ovarian cancer in Rochester, Minnesota over the 40-year

period 1935 through 1974 was determined; and risk factors for epithelial
ovarian cancer occurring in Rochester from 1945 to 1974 were examined in
116 patients and 464 controls. Among the characteristics studied, only nul-
liparity was found to be a significant risk factor-relative risk 1.8. Other
suspected risk factors-including hypertension, obesity, age at menopause,
prior therapeutic pelvic radiation, and prior exposure to exogenous estro-
gen-were found not to differ significantly between patients and controls.
The ovarian cancer patients were found to have a significantly lower fre-
quency of prior hysterectomy and of unilateral oophorectomy than the con-
trol group. Thus our data show that hysterectomy, even when one or both
ovaries are preserved, is associated with a lower risk of subsequent ovarian
cancer.
Cancer 43:723-729, 1979.

0 VARIAN C A N C E R is both common and

highly fatal. In the United States, it
causes some 12,000 deaths per year. Although
there are wide international variations in
ovarian cancer incidence rates, the rates are
very similar among the industrialized nations
of northern Europe and the United States.
Also, in these countries the incidence rates of
ovarian cancer appear to have been stable for
as long as there have been adequate tumor
registry data. The fact that incidence rates are
relatively low in southern Europe and South
America and very low in Japan suggests en-
vironmental factor^,^ although differences in
diagnosis and reporting cannot be entirely dis-
missed. In addition, migrant studies show that
the rates among Oriental immigrants to the
United States approach those of Caucasian
Americans after a generation or two. In
Israel, rates are high among immigrants from
Europe and low among immigrants from
Asiaz1 We are aware of only four case-con-
trol studies of ovarian cancer.6*9*'9~21 These
studies have not consistently found any
suspected risk factors to be of significance.
From the Mayo Clinic a n d Mayo Foundation,
Rochester, Minnesota.
Supported in part by N I H Grant GM 14231.
* Emeritus staff.
Address for reprints: J. F. Annegers, PhD, c/o Section
of Medical Statistics, Mayo Clinic, 200 First Street SW,
Rochester, MN 55901.
Accepted for publication April 28, 1978.
0008-543X1791020010723 $0.85 0 American Cancer Society
723
MATERIALSA N D METHODS
Our basic resources were the medical
records and retrieval system of the Rochester
Project at the Mayo Clinic, which have been
utilized in the study of many diseases.8
Cases
Records of all Rochester residents with a
diagnosis of ovarian cancer from 1935
through 1974 were reviewed. These included
records at the Mayo Clinic and at other medi-
cal facilities serving the local population. All
autopsy diagnoses of ovarian cancer were re-
viewed also; and three cases first diagnosed
at autopsy, satisfying the pathologic criteria,
were included in the series. It should be noted
that the autopsy rate in the Rochester popu-
lation is very high (65%).
For the incidence study, a total of 151 cases
met the residential, diagnostic, and pathologic
criteria. Sections of the pathologic material
were available from all but four cases and
were reviewed by the same pathologist
(M.B.D.). In the four cases without such ma-
terial, the recorded clinical indications of
ovarian cancer were strong; and these cases
were included in the series. The distribution
of cell types (Table 1) is similar to that
found in various reported series of ovarian
cancer cases.
T h e case-control study was limited to the
724 CANCER
February 1979
TABLE I . Histologic Types of Ovarian Cancer in
Incidence Study (Rochester MN, 1935- 1974)
Cell type No.
Epithelial
Serous 49
Mu ci n ou s 17
Mixed serous-mucinous 3 Statistical Methods
Endometroid 28
Mesonephric 4 The relative risks (RR) of ovarian cancer
Anaplastic 36
Adenocarcinoma unclassified 1 mated by the matched-quintuplets method of
Miscellaneous
Teratoma (malignant) 3
Granulosa theca 3
Teratoma with adenocarcinoma and
malignant melanoma and
fibrosarcoma 1 nored the matching characteristic of the con-
Carcinosarcoma 1 trols and estimated the RR by the odds
Ovarian remnant 1
(Clinical diagnosis only)
TOTAL
116 patients with a diagnosis of epithelial
ovarian cancer during the years 1945 through
1974.
Controls
For each of the 116 patients, 4 controls were
selected. These were women age-matched
(within 2 years) to the patient and resident
in Rochester at the time when the diagnosis
was made. Since every patient had at least
some ovarian tissue at the time of diagnosis
of ovarian cancer, the same requirement was
made of the controls. Review of the control
group revealed that the usual occasions for
their being seen at the medical facilities
from which they were selected were emer-
gency-room visits and general medical,
ophthalmologic, and dental examinations.
So far as we know, the control group was
free of a selection bias for any specific medi-
cal condition.
Previous Circumstances
For the period before the diagnosis of
ovarian cancer (index date), information was
collected on each patient and the controls
matched to her. These data included parity,
height, weight, blood pressure, use of exog-
enous estrogen, age at and type of meno-
pause, therapeutic pelvic radiation, and pelvic
surgery. An important aspect of the medical
records linkage system of the Rochester Proj-
ect is the period of medical-records coverage
of Rochester residents. For more than half of
the patients and controls, the interval from
Vol. 43
first medical record to the index date was
greater than 30 years; and for the vast ma-
cases jority of subjects the records encompassed at
least 10 years of prior medical care in the
138 community.
associated with various risk factors were esti-
9 Miettinen13 when the value of the factor was
known for all patients and controls. When two
risk factors were being studied simultaneously
or when some data were lacking, we ig-
(4)
ratio, using the correction factor discussed by
151 Flei~s.~
INCIDENCE RATES
The incidence rates of ovarian cancer in
Rochester over the 40-year period, 1935
through 1974, are presented in Fig. 1 and
Table 2. As has been demonstrated many
times, the incidence rates rise sharply until
about age 60 and then plateau in the older
age groups. Since women with a prior bilateral
oophorectomy are not at risk to ovarian can-
cer, the proportion of the population at risk
is less than one. The prevalence of such
previous oophorectomy was determined
among 820 women from Olmsted County
whose admissions to the local medical facilities
over the period 1945-1974 could be re-
viewed. In the first decade (1945-1954),
about 4% of the postmenopausal population
were without ovaries, and in the succeeding
two decades 10% and 8%, respectively. The
adjustment for this variable slightly increases
the incidence rate among those at risk to
ovarian cancer (Table 2).
When compared with results from other
studies, two aspects of the Rochester data are
of interest. First, the age-adjusted incidence
rates among the Rochester population are
somewhat higher than those reported else-
where, even when compared to the highest
rates from Connecticut (see Fig. 1) and
Scandinavia.3 It is not known to what ex-
tent this is due to completeness of diagnbsis
and case ascertainment or to characteristics of
the Rochester population. Second, the age-
'
adjusted incidence rate among the Rochester
population has declined in the last two
decades, even with correction for the popula-
tion at risk. The incidence rate in the first
No. 2 CANCER. Annegers et al.
OVARIAN 725

2. Incidence of Ovarian Cancer (Rochester M N , 1935-1974): Rates per 100,000

TABLE

1935-1944 1945-1954 1955-1964 1965-1974 1935-1974

Age (yr) No. Rate No. Rate No. Rate No. Rate No. Rate

15-34 3 5.1 2 3.5 2 2.9 4 3.9 11 3.8

35-44 3 15.6 4 18.9 5 21.1 4 14.3 16 17.4
45-54 6 41.8 11 58.0 7 32.2 10 39.5 34 42.3
54-64 7 70.2 13 90.2 13 69.1 11 47.6 44 66.4
65-74 5 79.2 6 63.0 8 60.0 6 32.8 25 52.7
75+ 2 68.7 2 38.0 7 79.4 10 64.9 21 64.8
Total 26 38 42 45 151
Age-adjusted 22.6 24.6 20.7 16.5 20.3
(95% CI*) (13.9-3 1.3) (17.0-32.2) (14.6-26.9) (1 1.6-21.4) (17.1-23.5)
Adjusted for
population at risk t 25.0 22.9 17.8

* Confidence interval. t Data on prior surgery were not collected for 1935-
1944 interval.

decade (1935- 1944) could be an under- Case-Control Study

estimate, in that it is possible that a number
of cases with ovarian cancer were not diag-
nosed and at death were recorded as "ab-
dominal carcinomatosis."
The Rochester population apparently dif-
fers from the national population in having
a high percentage of nulliparous women,
and this difference was greatest during the
first decades of our study. Of our con-
trol group the nulliparous women comprised
38% in 1945-1954, 34% in 1955-1964, and
21% in 1965-1974. In the United States,
24.7% of the white female population aged
45 and over were never married or nul-
liparous in 1960 and 21.1% in 19'70 (US
Census)." Thus, the high proportion of nullip-
arous women in the Rochester population
prior to 1965 could possibly account for both
the high incidence and the recent decline.
Association with Breast Cancer
It has been reported that women with
breast cancer are twice as likely as those
without to develop ovarian cancer later, and
that those with ovarian cancer are four times
as likely as those without to develop breast
cancer later.16 Of our 151 patients, 5 had de-
veloped breast cancer earlier. The expected
number of prior breast cancer cases, based
upon age-specific prevalence rates of breast
cancer in the local population, is 2.9. One
patient had subsequent breast cancer, for
which the expected number is 1.4. Thus, in
this series there is no significant excess of
association between breast cancer and ovarian
cancer-observed, 6; expected, 4.3. How-
ever, the sample size is too small to contradict
the relationship found by others.
Table 3 presents highlights of the four
previous ovarian cancer case-control studies.
It should be noted that only in the study by
Lau et aL9 was there a general-population con-
trol: the others used hospitalized patients,
and t w o of them used patients with specific
diseases. Reports of clinical series have indi-
cated that a high proportion of women with
ovarian cancer are nulliparous and also that
therapeutic pelvic radiation-particularly
radiation given to induce menopause-may
be a risk factor.
Further, many have considered the use of
exogenous estrogen to be a possible risk
factor in ovarian cancer.1° In a cohort of
908 women who had taken conjugated estro-
gen for at least 6 months, studied by Hoover
et u . L . , ~ the estimated relative risk of 2.4,
with a 95% confidence interval from 1.0 to
4.8, was barely significant. Indeed, if account
is taken of the estimated 10% of the popula-
tion without ovaries-as in the Rochester
population-the expected numbers are ad-
justed upward and the lower limit of the con-
fidence interval becomes 0.9, an even more
borderline situation. West,Ig Wynder et al.,"
and Lau et ~ 1 all. evaluated
~ the prior use of
exogenous estrogen and found no significant
differences in exposure between their patients
and controls.
The studies by Westlg and Lau et aLy did
not reveal association with prior pelvic radia-
tion. However, Lau et al. found a relative-
risk measurement of the association between
hypertension and ovarian cancer to be of
borderline significance. The risk associated
with obesity was studied by Wynder et al.,"
who found it only slightly greater among
726 February 1979
CANCER Vol. 43

FIG. 1. Ovarian cancer in-

cidence rates in Rochester MN
and in Connecticut, with 95%
confidence intervals of Rochester
rates.

I 1 I 1 I I I I

1940 1950 1960 1970

ovarian cancer patients than among controls.
A negative relationship between mumps and
ovarian cancer was reported by Westlg but
was not duplicated by Wynder et al. Wynder
et al. reported an increased risk with past
history of dysmenorrhea among his ovarian
cancer cases.
Parity, long thought to be strongly as-
sociated with ovarian cancer, was not found to
be significant by West,lg Wynder et al.,21 or
Lau et aL9 Joly et al.,6 however, found a low
but significant risk associated with nulliparity
3. Case Control Studies of Ovarian Cancer
TABLE
and with various other indices of low fertility.
An analysis of the Third National Cancer
Survey data, by Weiss et al.,ls revealed that
never-married women have an incidence rate
1.5 times that of ever-married women (95%
confidence interval, 1.4- 1.7).
RESULTS
Among our 116 patients with epithelial
ovarian cancer and 464 controls, nulliparity
was the only significant risk factor (Table 4).

Conditions at or before diagnosis

Exog-
No. Pelvic enous
No. con- radia- estro- Hyper- Other
Source pts trols Source of controls tion gen tension Obesity Parity associations

West,Ig 97 97 Benign ovarian NS NS 0 0 NS Negative for

1966 tumors mumps
Wynder 150 300 Hospitalized 0 NS 0 NS 0.8 Positive for
PL n1.,21 dysmenor-
1969 rhea

I
JO~Y 399 395 Non-neoplastic
et ~ l . , ~ disease
1974
362 Colon-rectal 0 0 0 0 1.5 Positive for
cancer (1.1- 1.9) various in-
dices of
low fertility
396 Breast cancer
Lau 149 149 General population NS NS 1.5 0 1.4
et al.,9 receiving x-rays ( 1.O- 2.4) (0.8- 2.2)
1977

0 = not studied; NS = not significant.

No. 2 CANCER
OVARIAN Annegers et al. 727

Parity and 164 (35%)of the controls. The difference,

again, was not of statistical significance.
We found nulliparity in 44% of our ovarian
cancer patients and in 30% of the controls. Prior Pelvic Radiation
The relative risk of 1.8 was significant. Among
the parous, 38% of the patients and 51% of There were 12 patients and 24 controls who
the controls were para three or more. had a history of prior therapeutic pelvic
radiation. Most of the irradiation was done
Age at Menopause between 1920 and 1945. These treatments
were usually for infertility, dysmenorrhea, or
Among those ovarian cancer patients who induction of menopause. The relative risk of
had a normal last menstrual period, it oc- 1.8 (95% confidence interval, 0.9-3.5) was not
curred at or before 52 years of age in 49 statistically significant.
and later in 18 women. Among the controls,
the last normal menstrual period occurred Prior Exogenous Estrogen
at or before age 52 in 196 and later in 65. All indications of estrogen use 6 months or
Thus there was essentially no difference be- more prior to the index date were recorded.
tween the two groups in frequency of early Such indications were found for 25 (22%) of
or late menopause. Both Wynder et aL2' and the 116 patients and 148 (32%) of the 464
Lau et al.,9 in their studies, found a slight controls. Thus estrogen exposure of any type
association between early menopause and and any duration was significantly less
ovarian cancer. frequent among the patients. This difference
Obesity probably is due to the greater parity of the
controls, because there was considerably more
Obesity was defined as weight 30% or more estrogen use associated with pregnancy and
above the upper limit of the ideal weight lactation suppression among the controls. Es-
for given height and medium frame12 at the trogen use of long duration-and specifically
time of diagnosis-or, for controls, the meas- use of conjugated estrogen-did not differ
urement nearest to that date. Among subjects significantly between patients and controls.
for whom data were available, obesity was This is in contrast to our finding in endo-
present in 15% of the patients and 12% of the metrial cancer of a strong relationship with use
controls. of estrogen-and especially use of conjugated
estrogen for more than 6 months."
Hypertension
Prior Surgery
Hypertension, defined as blood pressure
160/95 mm Hg at the time of initial diagnosis, The patients and controls were found to
was tabulated for 35 (30%) of the patients differ greatly in the frequency of prior hys-
4. Risk Factors for Ovarian Cancer (Rochester MN, 1945-1974)
TABLE

Patients Controls

KR 95% CI Pos Neg Pos Neg

Nulliparity 1.8 I .2-2.8 51 64 139 32 1
Menopause, normal at age
5 5 2 yr 1.1 0.6-2.0 49 196
53+ 18 65
Obesity 1.4 0.8-2.5 17 96 52 399
Hypertension 0.8 0.5- 1.2 35 81 164 296
Prior treatment
Pelvic radiation 1.8 0.9-3.5 12 104 24 440
Estrogen
Exogenous:
all 0.5 0.3-0.9 25 91 148 316
2 6 mo 1.o 0.5-1.9 13 103 33 43 1
Conjugated:
all 0.9 0.4- 1.8 9 107 40 424
2 6 mo 0.7 0.2-1.8 3 113 16 448

,,,,-odds ratio estimate of RR; CI-confidence interval.

728 CANCER
February 1979 Vol. 43

TABLE
5. Surgery Prior to Ovarian Cancer: ovarian cancer who had a prior hysterectomy
Genital Status at Diagnosis ranges from 3.6 to 11.7. Since the prevalence
Pa- Con- of intact uteri in that proportion of the popu-
tients trols KH (95% CI) lation at risk to ovarian cancer is not known,
it is not possible to state whether these
Rochester M N , frequencies of prior hysterectomy differ from
1945-1974
Uterus with
those in the general population. The largest
2 ovaries 110 358 1 of the studies is that by Counseller et al. ,2 who
1 ovary 2 26 0.3 (0.08-1.14) found that 4.5% of 1,500 patients with ovarian
N o uterus cancer diagnosed from 1930 through 1952 at
2 ovaries 4 53 0.36 (0.10-0.73) the Mayo Clinic had had prior hysterectomy.
1 ovary 0 27 0.06 (0.004-0.975)
Wynder rt al. This percentage is on the same order as ours,
(1 Y69)2' but most of the patients in Counseller's study
Prior hyster- were diagnosed at an earlier period, and also
ectomy 24 68 0.7 (0.4-1.0) there may be some bias related to referral.
Unilateral
oophorec-
There are few cohort studies wherein a
tomy 3 17 0.4 (0 1-1.2) series of patients who had hysterectomy or
'rOTAL 150 299* unilateral oophorectomy, or both, were
followed up to determine their experience
* One controlhad prior bilateral oophorectomy. with ovarian cancer. Randall et al.I5 studied a
RK-oddsratio estimate of RR; C1-confidence interval
cohort of 345 patients who had had one ovary
removed because of benign ovarian cyst. The
terectomy or unilateral oophorectomy, or mean age of the patients was 40 years and
both (Table 5): such prior treatment had been they were followed up for an average of 15
carried out in only 6 (5%)of the patients but years from the time of their unilateral oopho-
in 106 (23%) of the controls. These data show rectomy. During this period, ovarian cancer
a possible protective effect of prior hysterec- was observed in four patients. Although the
tomy and oophorectomy. The relationship be- number of person-years of follow-up is not
tween prior hysterectomy and ovarian cancer known, the number of ovarian cancers is prob-
was independent of that of parity (Table 6). ably on the order of what one would expect
The study by Wynder et ~ 1 also . supplied
~ ~ in such a group of women in a period of
data on prior surgery. In their report it is that length.
not possible to distinguish subgroups with In another cohort study, Randall14followed
prior hysterectomy, unilateral oophorectomy , 288 women who had had hysterectomy and
and both. However, both prior hysterectomy unilateral oophorectomy and 627 who had
and prior oophorectomy were more frequent had hysterectomy with both ovaries pre-
(though not significantly so) in the controls served. The mean follow-up of both groups
than in the patients (Table 5 ) . was 20 years. He observed only one case of
Over the years there has been considerable subsequent ovarian cancer in each group.
interest in the risk of ovarian cancer sub- Thus, in the hysterectomized group, the
sequent to a hysterectomy or unilateral frequency of subsequent ovarian cancer ap-
oophorectomy, or both. The numerous re- pears to be low, but since the age-specific
ports of the frequency of prior hysterectomy years of follow-up is not known, it is not
among patients with ovarian cancer have been possible to determine the expected number.
summarized by Christ et al.' and Kofler.' In In a similar study, WhitelawZ0identified
these studies the percentage of women with 1,215 women who had had hysterectomy with
I'ABLF
6. Relative Risk of Ovarian Cancer: Prior Hysterectomy vs. Parity (Rochester M N , 1945- 1974)

Cases Controls RR 95% CI

Parous
Without prior hysterectomy 61 257 1
With prior hysterectomy 3 64 0.2 (0.02-0.6)
Nulliparous
Without prior hysterectomy 50 123 1.7 (1.1-2.6)
With prior hysterectomy 1 16 0.4 (0.1-2.1)
RR-OddSratio estimate of relative risk; CI-confidence interval.
No. 2 OVARIAN
CANCER * Annegers et al. 729

one or both ovaries left intact. He observed no was not found to be associated with ovarian
cases of subsequent ovarian cancer; however, carcinoma in a population where a study
the duration and extent of follow-up are not utilizing the same methodology did find a
known. Thus, although the data are far from highly significant association between endo-
satisfactory, available information seems to metrial carcinoma and long-term use of
support the results of our case-control study, conjugated estrogen. Prior hysterectomy with
in that women who have had a hysterectomy or without unilateral oophorectorny was
but retain at least one intact ovary appear, forfound to be far less common in our cases than
some reason, to be at lower risk to subsequent in our controls. Our data and those of others
ovarian cancer than those who have not had suggest that hysterectomy, even with preser-
hysterectomy. vation of one or both ovaries, may reduce the
risk of subsequent ovarian cancer.
DISCUSSION T h e incidence rates of ovarian cancer in
Rochester were found to be higher than those
In this, as in other case-control studies of reported from Connecticut and other parts of
ovarian cancer, no major risk factors have the world with high rates of ovarian cancer.
been identified. With the exception of a slight It is also noted that, at least for the last
risk increment associated with nulliparity, three decades, there has been a slight decline
there were no positive associations between in the incidence of ovarian cancer in the
ovarian cancer and suspected risk factors. Rochester population. It is possible that both
Prior therapeutic radiation, obesity, age at of these phenomena are due to the high pro-
menopause, and prior use of exogenous estro- portion of Rochester women, especially in the
gen were not found to be significant risk fac- earlier decades of our study, who were
tors. Long-term use of conjugated estrogen nulliparous.

REFERENCES
1. Christ, J. E., and Lotze, E. C.: T h e residual ovary
syndrome. Obstet. Gynecol. 46:55 1-556, 1975.
2. Counseller, V. S., Hunt, W., and Haigler, F. H., Jr.:
Carcinoma of the ovary following hysterectomy. Am. J.
Obstrt. Gynecol. 69:538-546, 1955.
3. Fathalla, M. F.: Factors in the causation and
incidence of ovarian cancer. Ob.\tet. Gynecol. Surv. 27:
751-768, 1972.
4. Fleiss, J. L.: Statistical Methods for Rates and Pro-
portions. New York, John Wiley, 1973; p. 46.
5. Hoover, R., Gray, L. A,, Sr., and Fraumeni, J. F.,
Jr.: Stilboestrol (diethylstilbestrol) and the risk of ovarian
cancer. Lancet 2:533-534, 1977.
6. Joly, D. J., Lilienfeld, A. M., Diamond, E. L., and
Bross, I. D. J.: An epidemiologic study of the relation-
ship of reproductive experience to cancer of the ovary.
Am. J. Epidemiol. -99: 190-209, 1974.
7. Kofler, E.: Uber die Haufigkeit vorheriger Hysterek-
tomein undioder unilateraler Ovarektomien bei Frauen
mit malignen Ovarialtumoren. Geburt.shil&e Fruuenhoilkd.
32~873-881, 1972.
8. Kurland, L. T., Elveback, L. R., and Nobrega, F. T.:
Mayo Clinic records linkage system in the Rochester-
Olmsted epidemiology program project. I n International
Epidemiological Association: Uses of Epidemiology in
Planning Health Services, A. M. Davies, Ed. Proceedings
of the Sixth International Scientific Meeting, Belgrade,
Savremena Administracija, 1973; pp. 164- 176.
9. Lau, H.-U., Petschelt, E., Poehls, H., Pollex, G.,
Unger, H.-H., and Zegenhagen, V.: Zur epidemiologie
des Ovarialkarzinoms. Arch. Ge.w,hzuulstfor.\rh. 47:57-66,
1977.
10. Lingeman, C. H.: Etiology of cancer of the human
ovary: A review.J. Natl. CancrrInst. 53: 1603-1618, 1974.
11. McDonald, T . W., Annegers, J. F., O’Fallon, W. M.,
Dockerty, M . B., Malkasian, G. D.,Jr., and Kurland, I,. T . :
Exogenous estrogen and endometrial carcinoma: Case-
control and incidence study. Am. J. 0b5trt. Gynecd.
127:572-580, 1977.
12. Metropolitan Life Insurance Company: Fifty years
of life conservation. Stat. Bull. MPtropol. L f e Ins. Co. 40:
1-4, 1959.
13. Miettinen, 0. S.: Estimation of relative risk from
individually matched series. Biometrirs 26:75-86, 1970.
14. Randall, C. L.: Ovarian conservation. In Progress
in Gynecology, J. V. Meigs and S. H . Sturgis, Eds. New
York, Grune & Stratton, 1963; pp. 457-464.
15. Randall, C. L., Hall, D. W., and Armenia, C. S.:
Pathology in the preserved ovary after unilateral
oophorectomy. Am. J . Obstet. GynrCol. 84: 1233- 1240,
1962.
16. Schottenfeld, D., and Berg, J.: Incidence of mul-
tiple primary cancers. IV. Cancers of the female breast
and genital organs.,\. Natl. Cancer Inst. 46: 161- 170, 197 1.
17. United States Department of Commerce, Bureau
of the Census: Subject reports, women by number of
children ever born, 1960 and 1970, Washington, D. C.,
Government Printing Office, 1973.
18. Weiss, N. S., Young, J. L., Jr., and Roth, G. J.:
Marital status and incidence of ovarian cancer: The U. S.
Third National Cancer Survey, 1969- 197 I , ,J. ,\‘utl.
Cancer Inst. 58:913-915, 1977.
19. West, R. 0.:Epidemiologic study of malignancies
of the ovaries. Cancrr 19:lOOl-1007, 1966.
20. Whitelaw, R. G.: Pathology and the conserved
ovary. J. Obstet. Gynaecol. Br. E m f . 66:413-416, 1959.
21. Wynder. E. L., Dodo, H., and Barber, H. R. K.:
Epidemiology of cancer of the ovary. Cancer 23:352-370,
1969.