Incidence and mortality in epithelial ovarian cancer by family history of any cancer


  • Kari Hemminki MD, PhD,

    Corresponding author
    1. Division of Molecular Genetic Epidemiology, German Cancer Research Center, Heidelberg, Germany
    2. Center for Primary Care Research, Lund University, Malmö, Sweden
    • Division of Molecular Genetic Epidemiology, German Cancer Research Center (DKFZ), Heidelberg, Germany
    Search for more papers by this author
  • Jan Sundquist MD, PhD,

    1. Center for Primary Care Research, Lund University, Malmö, Sweden
    2. Stanford Prevention Research Center, Stanford University School of Medicine, Stanford, California
    Search for more papers by this author
  • Andreas Brandt PhD

    1. Division of Molecular Genetic Epidemiology, German Cancer Research Center, Heidelberg, Germany
    Search for more papers by this author



Practically all data on familial risk in ovarian and other cancers are based on incident cancer, whereas familiality in cancer mortality is largely unknown. If fatal forms of cancer are a highly familial subtype, then familial risk for mortality may exceed that of incidence, which is relevant for clinical decision making and counseling.


Ovarian cancer patients in the nationwide Swedish Family Cancer Database were classified according to fatal and nonfatal (incident) family history. Familial risks for incident and fatal ovarian cancer were calculated for offspring based on their parental or sibling family history of any cancer using standardized incidence ratios (SIRs) for incidence and standardized mortality ratios (SMRs) for mortality. Offspring without family history were referents.


The database included 24,757 mothers and 8138 daughters with ovarian cancer. When a mother had ovarian cancer, the SIR for incident ovarian cancer in daughters was 2.69, and when a sister had ovarian cancer it was 3.49. The SMRs for fatal cancer by fatal cancer in probands were 3.39 and 5.80, respectively. For fatal serous cancers among siblings, the SMR was 6.16, compared with 10.01 for the endometrioid type. Ovarian cancer was associated with maternal (SIR, 1.22; SMR, 1.56) and sororal breast cancer (SIR, 1.27). Another discordant association was between ovarian and paternal prostate cancer (SIR, 1.12; SMR, 1.66).


Fatal familial risks were higher for concordant ovarian, ovarian-breast, and ovarian-prostate cancers than the corresponding incident risks. This may suggest that highly fatal subtypes exist for these cancers, calling for genetic dissection. Cancer 2011. © 2011 American Cancer Society.

The incidence in ovarian cancers has decreased by about 40% in Sweden from the peak years in the mid-1970s, reaching a rate of 15 in 100,000 and ranking 10th among all female cancers.1 Sweden is still internationally a high-risk country, and no immigrant group matches native Swedes in ovarian cancer incidence.2 Because ovarian cancer is a relatively fatal cancer, it ranks fifth in cancer mortality, accounting for some 6% of cancer deaths ( Epithelial tumors present in 3 main histological forms: serous, mucinous, and endometrioid carcinomas.3 In all these forms, the TNM or International Federation of Gynecology and Obstetrics stage is a critical predictor of survival.3 The serous histology tends to be more advanced at presentation than the other histologies, and consequently survival is poorer for serous cancers.4 Known risk/protective factors of ovarian cancer are reproductive and hormonal factors and family history.5-9 A previous Swedish study estimated population-attributable fractions for epithelial ovarian cancer and found the highest contribution of 22.3% for reproductive factors, well above 2.6% for family history.10

In population-based studies with medically verified diagnosis, the familial risk of invasive ovarian cancer has ranged between 2.0 and 3.0.10-17 Survival in ovarian cancer has been worse for familial compared with sporadic patients.4 Ovarian cancer is a manifestation in families with BRCA1/2 mutations and in hereditary nonpolyposis colorectal cancer (HNPCC).3 The prevalence of BRCA1/2 mutations, 11.7% to 15.3%, in unselected cases of invasive ovarian cancer has been high in a Canadian and a US study.18, 19 Many of the mutation carriers had no family history of ovarian cancer. The survival in BRCA1/2 mutation carriers and HNPCC patients has been better than that of sporadic patients.20-22 Familial clustering of ovarian cancer with cancers at other, discordant sites has also been reported, particularly involving endometrial and breast cancers.17, 23 The association with endometrial cancer has been strongest with ovarian cancer of endometrioid histology.24

Most family studies of ovarian cancer to date have been based on data on cancer incidence, and only a few studies have taken the outcome into account.4 If a random proportion of ovarian cancer patients died from this cancer, then the familial relative risk of death from ovarian cancer should be equal to or smaller than familial risks of incident cancer. If fatal forms of ovarian cancer are associated with fatal forms of ovarian or any discordant cancers in relatives, data on familial mortality alone might be able to detect them. This would be a particularly relevant observation for clinical decision making and counseling, where fatal outcomes might deserve special attention. Finding fatal subtypes of concordant or discordant familial cancers would also stimulate mechanistic and gene finding studies.

In the present study, we used the nationwide Swedish Family Cancer Database, updated in 2010, to explore familial risks for incident and fatal ovarian cancers with any discordant cancers in parent-offspring and sibling pairs. To exclude chance findings because of multiple comparisons, we applied 3 independent tests for each pair of cancer sites: offspring ovarian cancer by parental cancer X, offspring cancer X by parental ovarian cancer, and sibling ovarian cancer by cancer X. For internal validity, we require at least 2 of these tests to be positive. In addition to showing associations between incident ovarian cancer and other female cancers and prostate cancer, we demonstrate for the first time that these clusters also include fatal cancers, some with higher familial risks than with incident cancers.


The Swedish Family Cancer Database is an anonymous dataset constructed from the MigMed database located at the Center for Primary Health Care at Lund University in Malmö. The Family Cancer Database contains information from the Multigeneration Register, national censuses, Swedish Cancer Registry, and death notifications.25 Data on family relationships were obtained from the Multigeneration Register, where children born in 1932 and later are registered with their biological parents as families. Thus, the individuals in the database can be divided into offspring generation (individuals born 1932 and later) and parental generation. The Swedish Cancer Registry is based on compulsory reports of diagnosed cases, with coverage of cancer registration close to 100%.1 Cases are reported separately by clinicians and pathologists/cytologists. Information on cancers based on death certificates is not used. The underlying cause of death was available from the Swedish Causes of Death Register. This register annually adds some 2000 notifications of cancer as an underlying cause of death that are not on file in the Cancer Registry; these cases were included in the analysis of familial mortality. The 2010 update of the database includes >12.1 million individuals and their cancers from years 1958 to 2008. Our study population constituted 7.8 million individuals from the offspring generation of the database with linkage to both parents. Most individuals without identified parents (∼1.8 million) were immigrants. The age structure of the database (offspring born after 1932) implicates that the maximum age of diagnosis in the offspring generation was 76 years. The age at diagnosis in the parental generation was not limited. The Systematized Nomenclature of Human Medicine histology was available since year 1993.

The database also includes borderline ovarian tumors, but these were not included in the present analysis, because we compared familial incidence and mortality. Patients with borderline tumors have excellent survival, and analysis of familiar risks in fatal forms would not have been possible.4

Individuals entered the risk period at birth, immigration date, or January 1, 1961, whichever came last, and they were followed until the first diagnosis of cancer, death, out-migration, or end of study on December 31, 2008. Standardized incidence ratios (SIRs) were calculated for diagnosis with ovarian cancer when parent or sibling were diagnosed by cancer X. In addition, we carried out analysis by reverse order, ie, SIR for cancer X by diagnosis of maternal ovarian cancer. This was a completely independent analysis. Standardized mortality ratios (SMRs) were used to compare cause-specific mortalities in individuals with a deceased parent or sibling with the cause-specific mortality rates in the general population. The number of individuals with both an affected or deceased parent and an affected sibling was too small for a separate analysis; these individuals were excluded. SIRs and SMRs were standardized for sex, age (5-year bands), calendar year (5-year groups between 1961-2000, 2001-2008), socioeconomic index (categories: blue collar worker, white collar worker, farmer, private, professional, other/unknown), and region (categories big cities, South Sweden, North Sweden, unknown). Confidence intervals (CIs) were calculated assuming a Poisson distribution.

Cancer studies in the MigMed Database was approved by the Lund regional ethical committee on August 12, 2008 (No. 409 of 2008) and with complementary approvals dated September 1, 2009 and January 22, 2010.


In the database, the offspring population has reached a maximal age of 76 years; siblings can only be identified in the offspring population. Ages in parents are not limited. The current update of the Family Cancer Database (FCD2008) covered years 1958 to 2008 from the Swedish Cancer Registry, and included 24,757 mothers and 8138 daughters with ovarian cancer. Familial risks for incident (SIR) and fatal (SMR) ovarian cancer in daughters are shown in Table 1 by familial incident and fatal cancers in parents or siblings. The data were adjusted for age, period, socioeconomic index, and region; for sex-specific female cancer, parity and age at first birth were also adjusted for. When a mother had ovarian cancer, the SIR for incident ovarian cancer in daughters was 2.69, and when a sister had ovarian cancer it was 3.49. The SMRs for fatal cancer by fatal cancer in probands were 3.39 and 5.80, respectively; for siblings the CIs of the SIR and the SMR did not overlap. Ovarian cancer was associated with maternal (SIR, 1.22; SMR, 1.56) and sororal breast cancer (SIR, 1.27); association with endometrial cancer was observed only with maternal probands (SIR, 1.24; SMR, 1.84).

Table 1. SIR for Diagnosis of Ovarian Cancer for Individuals With a Parent/Sibling Who Was Affected With Cancer; SMR for Death From Ovarian Cancer for Individuals With a Parent/Sibling Who Died From Cancer
SiteParent Affected/DiedSibling Affected/Died
  • O indicates observed number of cases; SMR, standardized mortality ratio; SIR, standardized incidence ratio; CI, confidence interval.

  • a

    Adjusted for age, calendar period, socioeconomic status, geographical region, number of children, and age at first birth.

  • b

    95% CI does not include 1.00.

Upper aerodigestive tract        
 Incident cases1101.140.941.37280.920.611.33
 Incident cases381.100.781.5080.800.351.58
 Incident cases2101.000.871.14361.55b1.08b2.14b
Small intestine        
 Incident cases191.060.641.6581.260.552.49
 Incident cases5351.050.971.151281.090.911.29
 Incident cases1361.030.861.22351.74b1.21b2.42b
 Incident cases1461.130.951.33271.160.761.68
 Incident cases3041.000.891.121031.130.921.36
 Incident cases5341.22b1.11b1.32b3191.27b1.13b1.42b
 Incident cases931.120.911.38250.880.571.30
 Incident cases1421.24b1.05b1.46b451.080.791.44
 Incident cases2482.69b2.36b3.04b1273.49b2.91b4.16b
 Incident cases6511.050.971.142181.17b1.02b1.33b
 Incident cases1361.020.861.21351.070.741.48
Urinary bladder        
 Incident cases1970.960.831.11541.060.791.38
 Incident cases1031.100.901.34931.24b1.00b1.52b
Nervous system        
 Incident cases1101.050.871.27641.090.841.40
Connective tissue        
 Incident cases210.810.501.24181.78b1.06b2.82b
Non-Hodgkin lymphoma        
 Incident cases1291.100.921.31491.060.781.40
Hodgkin disease        
 Incident cases221.230.771.86111.200.602.14
 Incident cases761.160.921.45201.350.822.08
 Incident cases1040.870.711.06431.210.871.62

Data for the reverse analysis (risk for any cancer in male and female offspring by ovarian cancer in mothers or siblings) are shown in Table 2. For ovarian cancer, the affected mother-daughter pairs were the same as in Table 1, but for any discordant sites, the analysis in Table 2 was completely independent from that in Table 1. However, for siblings the affected pairs were the same in Tables 1 and 2. Daughters' risk for breast cancer was increased when mother (SIR, 1.35; SMR, 1.47) or sister (SIR, 1.26) was diagnosed with ovarian cancer. Additional discordant associations were found with paternal prostate (SIR, 1.12; SMR, 1.66; nearly nonoverlapping CIs) and kidney cancers (SIR, 1.29; SMR, 1.48). The association of prostate-ovary cancer among siblings was not significant in Table 2, but it was significant in Table 1 (SIR, 1.15). Other significant sibling SIRs in Table 2—stomach (1.71), liver (1.63), and connective tissue (1.88) cancers—were also found in Table 1.

Table 2. SIR for Diagnosis of Cancer for Individuals With a Parent/Sibling Who Was Affected With Ovarian Cancer; SMR for Death From Cancer for Individuals With a Parent/Sibling Who Died From Ovarian Cancer
SiteParent Affected/DiedSibling Affected/Died
  • SMR indicates standardized mortality ratio; SIR, standardized incidence ratio; CI, confidence interval.

  • a

    Adjusted for age, calendar period, socioeconomic status, and geographical region.

  • b

    95% CI does not include 1.00.

  • c

    Also adjusted for number of children and age at first birth.

Upper aerodigestive tract        
 Incident cases660.890.691.14290.950.641.37
 Incident cases150.670.371.1080.790.341.56
 Incident cases511.070.801.41361.71b1.20b2.37b
Small intestine        
 Incident cases171.150.671.8581.340.582.63
 Incident cases3091.100.981.231331.090.911.29
 Incident cases461.040.761.38321.63b1.12b2.30b
 Incident cases470.900.661.19271.150.761.67
 Incident cases2131.080.941.241031.160.951.41
 Incident cases9371.35b1.26b1.43b3311.26b1.13b1.40b
 Incident cases901.060.861.31260.910.591.33
 Incident cases1121.140.941.37471.080.801.44
 Incident cases2482.69b2.36b3.04b1273.49b2.91b4.16b
 Incident cases4971.12b1.02b1.22b2241.090.951.24
 Incident cases981.29b1.05b1.58b331.010.701.42
Urinary bladder        
 Incident cases1161.010.831.21541.060.791.38
 Incident cases2571.090.961.23931.150.931.41
Nervous system        
 Incident cases1720.980.841.14651.090.841.39
Connective tissue        
 Incident cases270.920.611.34191.88b1.13b2.93b
Non-Hodgkin lymphoma        
 Incident cases1231.060.881.27541.180.881.53
Hodgkin disease        
 Incident cases391.180.841.61121.300.672.28
 Incident cases391.190.841.62201.420.862.19
 Incident cases880.890.711.10441.260.911.69

The histology-specific case numbers in Table 3 give crude data on the survival in each histological type; among serous patients, 41% (43 + 16 divided by 99 + 46) had died; for endometrioid cases, 26% had died, compared with 29% for mucinous and 23% for clear cell cases. Analysis of incident familial risk for specific histologies when probands were diagnosed with any ovarian cancer revealed that serous and endometrioid cancers were most familial, and the risks were higher for siblings than for daughter-mother pairs; the sibling SIRs were 4.30 and 4.74, respectively. For fatal serous cancers among sisters, the SMR was 6.16 compared with 10.01 for the endometrioid type and 9.64 for the clear cell type. These specific histologies dominated also the discordant associations with breast and endometrial cancers. The SMR for endometrioid ovarian cancer mortality was 4.15 when mothers were diagnosed with breast cancer. The “unspecified and other” group included mainly unspecified adenocarcinomas.

Table 3. Histology-Specific SIR of Ovarian Cancer for Women With a Relative Affected by Cancer and Histology-Specific SMR of Ovarian Cancer for Women With a Relative Who Died From Cancer
Family HistorySNOMEDParent Affected/DiedSibling Affected/Died
  • SIR indicates standardized incidence ratio; SMR, standardized mortality ratio; SNOMED, Systematized Nomenclature of Human Medicine; CI, confidence interval.

  • a

    Adjusted for age, calendar period, socioeconomic status, geographical region, number of children, and age at first birth.

  • b

    95% CI does not include 1.00.

 Clear cell92.56b1.17b4.87b42.970.817.62
 Unspecified and other472.42b1.78b3.22b273.65b2.41b5.31b
 Clear cell11.100.036.1229.64b1.17b34.83b
 Unspecified and other212.14b1.33b3.28b72.79b1.12b5.75b
 Clear cell211.220.751.86121.330.682.32
 Unspecified and other1231.32b1.10b1.57b661.30b1.01b1.66b
 Clear cell52.420.795.6511.800.0510.05
 Unspecified and other321.46b1.00b2.06b91.500.682.84
 Clear cell91.940.893.6910.660.023.69
 Unspecified and other241.010.641.5080.940.411.85
 Clear cell00.000.0011.0300.000135.89
 Unspecified and other30.840.172.4500.00011.43


The main novelty of the present study was the systematic analysis of association of fatal ovarian cancer with any fatal familial cancers and comparison of the findings with familial incident cancers. The salient finding was that not only were fatal familial risks increased for many cancers, but that the fatal risks tended to exceed the incident risks for concordant ovarian cancer and some discordant cancers. For ovarian cancer, the implication is that the high fatality may be related to yet unknown gene defects. BRCA1/2 or mismatch repair gene mutations are an unlikely explanation, because these defects are associated with favorable survival.20-22 The data are also clinically relevant. Sibling risks for serous and endometrioid cancers were 4.30 and 4.74, respectively, which are high in comparison to familial risks of most other cancers.26 Moreover, the fatal familial risks for these histologies were as high as 6.16 and 10.01, respectively, suggesting that in taking family histories it may be useful to record also the outcome of relatives' ovarian cancer. The present data should also encourage a molecular search for genes modifying survival, supported by recent evidence on such a locus at chromosome 19p13.11.27 Serous histology was associated most strongly with the recently detected low-risk germline loci for ovarian cancer.27, 28 Transcriptome sequencing has recently revealed common somatic mutations in the ARID1A gene, encoding a subunit for the chromatin remodeling complex, in clear cell and endometrioid tumors, reinforcing the evidence that these types of ovarian cancers arise from endometriosis.29

For incident discordant cancers, at least 2 independent associations were observed only for ovary-breast (3 independent associations) and ovary-prostate (2 associations). For ovary-endometrium, only a single incident association was significant, but this was supported by a borderline reverse association in Table 2 and a high fatal association of 1.84 in Table 1. For the other single significant associations, an Icelandic study reported an association between ovarian and stomach cancers.17 The association of ovarian cancer with breast cancer could be explained by common hormonal mechanisms and BRCA1/2. However, the data on breast cancer survival for BRCA1/2 mutation carriers are not essentially different from those of patients with sporadic breast cancer.30, 31 For ovarian cancer, mutation carriers have a favorable prognosis, as pointed out earlier.20-22 Thus, the higher fatal associations between ovarian and breast cancers compared with incident associations cannot be explained by BRCA1/2 alone. Hormonal mechanisms could also underlie the association of ovarian and prostate cancers, similar to breast and prostate cancers, which have extensive biological similarities in hormone-dependent pathways and also in familial associations.32, 33 However, again the high fatal risk of 1.66 compared with the incident risk of 1.12 may suggest the presence of aggressive genetic variants.

The present study is the largest on familial ovarian cancer and discordant cancers. Our previous study on incident cases is next in size.23 However, in comparison to that previous study, we have doubled the case numbers for main cancers among parents and offspring and tripled those among siblings; ovarian-prostate sibling pairs have increased 10-fold, because for late onset cancers, the aging of the offspring (sibling) population vastly increases case numbers. However, is spite of the incomparable size of the present study, the numbers of familial fatal cases were small whereby both type 1 and type 2 statistical errors were possible. We have tried to avoid type 1 errors by requiring support by 2 independent analyses.

In conclusion, fatal familial risks were higher for concordant ovarian, ovarian-breast, and ovarian-prostate cancers than the corresponding incident familial risks. For ovarian cancer, familial risks for incident cancer were high for endometrioid and serous histologies, but they were even higher for the fatal cancers of these histologies. The present data suggest that highly fatal subtypes may exist for these cancers. These may be amenable to genetic identification and molecular dissection.


The study was supported by Deutsche Krebshilfe, Swedish Council for Working Life and Social Research, and Swedish Cancer Society.