Familial Cancer

, Volume 6, Issue 1, pp 121–129

The interval between cancer diagnosis among mothers and offspring in a population-based cohort

Authors

    • Braun School of Public HealthHadassah-Hebrew University
    • School of Public Health
  • Yehiel Friedlander
    • Braun School of Public HealthHadassah-Hebrew University
  • Lisa Deutsch
    • Braun School of Public HealthHadassah-Hebrew University
  • Rebecca Yanetz
    • Braun School of Public HealthHadassah-Hebrew University
  • Ronit Calderon-Margalit
    • Braun School of Public HealthHadassah-Hebrew University
  • Efrat Tiram
    • Braun School of Public HealthHadassah-Hebrew University
  • Hagit Hochner
    • Braun School of Public HealthHadassah-Hebrew University
  • Micha Barchana
    • Israel Cancer RegistryMinistry of Health
  • Susan Harlap
    • Mailman School of Public HealthColumbia University
  • Orly Manor
    • Braun School of Public HealthHadassah-Hebrew University
Original Paper

DOI: 10.1007/s10689-006-9113-9

Cite this article as:
Paltiel, O., Friedlander, Y., Deutsch, L. et al. Familial Cancer (2007) 6: 121. doi:10.1007/s10689-006-9113-9
  • 83 Views

Abstract

Background

Familial cancers may be due to shared genes or environment, or chance aggregation. We explored the possibility that ascertainment bias influences cancer detection in families, bearing upon the time interval between diagnosis of affected mothers and offspring.

Methods

The Jerusalem Perinatal Study (JPS) comprises all mothers (n = 39,734) from Western Jerusalem who gave birth 1964 –1976 and their offspring (n = 88,829). After linking identification numbers with Israel’s Cancer Registry we measured the absolute time interval between initial cancer diagnoses in affected mother-offspring pairs. We tested the probability of obtaining intervals as short as those observed by chance alone, using a permutation test on the median interval.

Results

By June 2003 cancer had developed in 105 mother-offspring pairs within the cohort. Common sites among mothers were breast (47%), colorectal (9%), non-Hodgkin lymphoma (NHL) (8%) and cervix (7%), while for offspring in affected pairs common cancers were leukemia (12.4%), thyroid (13.3%), NHL (10.5%), breast (10.5%) and melanoma (7.6%). The median interval between diagnoses was 5.9 years, but for 33% of affected pairs the interval was ≤3 years. The probability of this occurring by chance alone was 0.03. This held true whether the offspring’s or mother’s diagnosis was first (P < 0.01).

Conclusions

In a population-based cohort followed for three decades, the absolute interval between the diagnosis of cancer in mothers and their offspring is shorter than expected by chance. Explanations include shared environmental exposures or the possibility that cancer ascertainment in one pair member affects health behaviors in the other resulting in early diagnosis. The latter may bias the estimation of anticipation and survival in familial cancers.

Keywords

Hereditary cancerTime to diagnosisAscertainment biasCohort study

Introduction

Cancers occurring in first degree relatives of cancer patients may be due to rare highly penetrant genetic syndromes (reviewed in [1]), more frequent low-penetrance genetic variants, shared environment [2], or chance occurrences within a family. Cancers occurring in carriers of the mutations in cancer-predisposing genes such as BRCA1 and 2, TP53, APC, mismatch repair genes and others are often characterized by a younger age at onset than sporadic cancers [1]. Often these syndromes are typified by tumors at different sites among family members (such as sarcoma in children and breast cancer in mothers in the Li-Fraumeni syndrome), rather than identical sites [3]. Evidence of “anticipation”, that is, younger age at onset or more aggressive disease in successive generations has been observed in hematologic malignancies [46] and has been suggested in hereditary colon [7] and breast cancer [8]. Despite the fact that a biological explanation for anticipation has been developed [9] controversies remain regarding the extent and validity of this phenomenon [10, 11]. The estimation of anticipation is determined by the time interval between cancer diagnoses in subsequent generations.

It is estimated that only 5–10% of cancers in the population are due to hereditary single gene mutations [1]. Thus many cancers occurring among multiple family members are multifactorial in origin and may be attributed to gene-environment interactions [12], environmental exposures or shared polygenic inheritance. Stiller has proposed that the excess cancer risk observed in first degree family members of children diagnosed with cancer is due to known hereditary syndromes [13]. However, studies in large population-based registries report only modestly elevated overall excess risks for maternal cancers in families where a child has experienced childhood cancer, with standardized incidence ratios (SIRs) of 1–1.1 [14, 15]. Cancers at specific sites, such as retinoblastoma, sarcomas and lymphomas have been found to be elevated in parents of affected children [15]. When the risk for children with affected mothers is assessed, increased relative risks have been observed for those whose mothers were diagnosed with breast, gynecological , thyroid, endocrine, hematopoietic and nervous system cancers [16]. Cancers at concordant sites are associated with high SIRs within families [1719].

The time interval between the diagnosis of mothers and offspring with cancer has not been the subject of intensive investigation. Familial cancers which do not occur at similar ages, but rather are closely spaced in calendar time suggest a shared environmental exposure in susceptible family members [20], or health behaviors which promote the diagnosis of clinically silent cancers or both.

We examined patterns of diagnosis of cancer within mother-offspring pairs in a population-based cohort, the Jerusalem Perinatal Study (JPS), in order to obtain clues to the contribution of biological and behavioral factors to the timing of detection of familial cancers.

Methods

Study population

The Jerusalem Perinatal Study is a database containing information on all births to residents of Western Jerusalem between 1964 and 1976. The study was initially established for research on pre-eclampsia and was later expanded to include other pregnancy complications and birth outcomes. Details of the study cohort have been previously published [21, 22]. Briefly, the database consists of 42,957 mothers and 91,459 live-born offspring. Information regarding ethnicity (including mother’s father’s (maternal grandfather’s) place of birth), socioeconomic status graded into six categories by father’s occupation, and years of education are available on virtually all mothers. Information on all offspring includes birth weight, birth order, type of delivery and singleton vs multiple birth. Information on maternal health and obstetric characteristics is available on 95% of the traced mothers. Smoking status (ever/never) is known for 54% of the mothers. The median age of living mothers in the cohort at last follow-up was 59 years (June 2004) while that of the offspring was 33 years.

Data Linkage

In Israel all residents have a unique identity number. Using this number we linked the JPS file with the Israel Population Registry, tracing vital status on 39,734 (92.5%) mothers and 90,078 (98.5%) offspring in January 2004. We then linked this file to the Israel Cancer Registry, updating cancer incidence in the cohort to June 30, 2003. The Cancer Registry has existed since 1960 and notification of all malignant tumors (except non-melanoma skin cancer), as well as benign brain tumors, has been obligatory by law since 1981. Even before that, registration for most tumor sites was considered >90% complete [23]. This study focuses on the mother-offspring pairs in which both received a diagnosis of cancer.

Statistical methods

We examined the frequencies of sociodemographic characteristics in the affected mother-offspring pairs, comparing them with the entire JPS cohort. Mother’s education was recorded as the maximal number of years noted in any birth in the cohort. Mother’s age and father’s age were assessed for the first birth in the cohort. We then compared cancer types, which occurred in the pairs with those ascertained among cohort members who did not have an affected child or mother according to ICD-O topography and morphology codes [24]. For several cancers we calculated age-standardized morbidity ratios comparing the observed numbers of cancers with those expected given the age distribution of the entire cohort. 95% confidence intervals were calculated for these ratios.

We estimated the correlation between the age at diagnosis of cancer for the mother and the offspring as well as year of diagnosis in the pairs. We examined the time interval, in absolute terms, between the mothers’ and offsprings’ cancer diagnoses in terms of mean and median. We then performed a permutation test to examine the probability of occurrence of time intervals as short as or shorter than those observed in our cohort occurring by chance alone (one-sided test). The permutation test was based on randomly matching pairs of mothers and offspring and computing the median of the time interval elapsed between the diagnosis of the first and second pair members. For every configuration 50 × 104 permutations were generated. We then obtained the P value by counting the percentage of permutations in which the median was smaller or equal to that observed. All permutations were programmed using the C language.

We examined these permutations in different subgroups: occurrence of mother’s cancer first or second, mother with breast cancer or other cancer, mother’s age at cancer diagnosis <50 years or ≥50 years, and offspring’s age at cancer diagnosis <15 years or 15+ years. We repeated the analysis excluding non-invasive cancers (such as in-situ carcinoma of the cervix) where the diagnosis is more likely to have been influenced by ascertainment bias. Out of concern that short follow-up may cause artefacts in the assessment of the time interval between cancer detection across generations, we also analyzed pairs in which the offspring were born in the 1960s and 1970s separately.

As an additional procedure to determine the probability of obtaining the observed median survival by chance alone we examined the occurrence of cancer in offspring and mother within the entire cohort using random combinations as follows: from the entire cohort of mothers and children diagnosed with cancer we excluded those cases where both mothers and offspring were affected, yielding 3684 mothers and 785 offspring with cancer. From this subgroup of non-familial cancers we randomly created 100 hypothetical pairs of mothers and offspring and calculated the absolute time interval elapsed between the diagnoses of the generated pair members. We repeated this procedure 1,000 times and determined the P value by calculating the proportion of replications in which the median was shorter than that observed.

Finally for mother-offspring pairs with mother’s cancer occurring first, we fitted a Cox proportional hazards model to the time to diagnosis of offspring, starting at the time of mother’s diagnosis and adjusting for offspring age. We then plotted the estimated survival function at the mean offspring age. In a similar way we have plotted the estimated survival function for mother’s time to diagnosis among pairs with offspring cancer occurring first.

The study received ethical approval from the Institutional Review Boards of the Hadassah University Hospital and Columbia University.

Results

There were 3784 mothers and 890 offspring in the JPS cohort who were reported to the cancer registry with a diagnosis of a first primary cancer until June 2003. Of these, 105 were mother-offspring pairs. Five mothers had two offspring with cancer and each was treated as an independent event. The demographic characteristics of mothers are shown in Table 1. Compared to women in the entire cohort, women with cancer were more likely to be of Western (European, North and South American, Australian) origin, and had a higher mean age at first birth. Women who had cancer and had a child with cancer were less likely to be of high socioeconomic status, more likely to be educated beyond 12 years and less likely to be uniparous, smokers, or to have given birth to a male child as their firstborn in the cohort compared with women with cancer and no affected offspring in the cohort.
Table 1

Sociodemographic characteristics of mothers in the cohort and mothers with cancer

Variable

Entire cohort (n = 39,734)

%

Women in the cohort with cancer (n = 3,784)

%

Women with cancer whose offspring had cancer (n = 100)

%

Birthplace of mother

Israel

18,362

46.2

1,796

47.5

50

50

Other

21,372

53.8

1,988

52.5

50

50

Mother’s father’s birthplace

Israel

5,401

13.6

520

13.7

11

11

West (Europe, America etc.)

14,098

35.5

1,532

40.5

45

45

North Africa

8,480

21.3

685

18.1

16

16

Western Asia

11,164

28.1

1,020

27

28

28

Religion

Jewish

39,129

98.5

3,755

99.2

100

100

Other

605

1.5

29

0.8

  

Socioeconomic Status (based on father’s occupation)

1–2 High

13,998

35.2

1,344

35.5

28

28

3–4

15,545

39.2

1,513

40

47

47

5–6 Low

10,191

25.6

927

24.5

25

25

Education

0–8 year

11,069

28.7

1,068

29.1

26

26

9–12 year

14,047

36.9

1,306

34.5

36

36

13+ years

12,833

32.3

1,294

34.2

38

38

Missing

 

3

 

3.1

  

Mean age at first birth (SD)

26.3 (5.7)

 

28.1 (5.9)

 

27.0 (5.1)

 

Sibship size in cohort

1

15,576

39.2

1,519

40.1

16

16

1`2–3

18,650

47

1,816

48

63

63

4–5

4,400

11.1

367

9.7

18

18

6+

1,108

2.8

82

2.2

3

3

Father’s mean age at first birth (SD)

33.0 (7.2)

 

34.6 (7.1)

 

31.7 (6.5)

 

Gender of first child

Male

20,601

51.8

2,003

52.9

45

45

Female

19,133

48.2

1,781

47.1

55

55

Smoking

Ever

8,688

39

732

37.6

18

29

Never

13,563

61

1,214

62.4

44

71

The tumor characteristics of mother and offspring cancer pairs are shown in Table 2. The most prevalent types of cancer in mothers without an affected child were breast (41.8%), colon and rectum (9%), melanoma (5.8%) and thyroid (4.9%), whereas the most common diagnoses of mothers in mother-offspring cancer pairs were breast (47%), colon (9%), non-Hodgkin lymphoma (NHL) (8 %) and cervical cancer (7%). Non-Hodgkin lymphomas (observed/expected 2.2, 95% confidence interval 1.1–4.3) were significantly over-represented in mothers with affected offspring. As for the offspring, the most prevalent sites for those in the cohort without an affected mother were Hodgkin Disease (12.2%), leukemia (8.7%), breast cancer (8.8%), NHL (7.5%), cervical cancer (7.6%), brain (6%) and melanoma (6.4%), whereas in the mother-offspring pairs thyroid tumors (13.3%), leukemia (12.4%), breast cancers (10.5%) and NHL (10.5%) were most common. Rare childhood tumors such as medulloblastoma (observed/expected 5.9, 95% confidence interval 1.6–15.11) were over-represented in the children whose mothers also had cancer, whereas Hodgkin Disease (observed/expected = 0.39, 95% confidence interval 0.13–0.9) was significantly under-represented. The mean and median age of cancer diagnosis for all mothers in the cohort was 52 years, similar to that of mothers with an affected offspring. Likewise, the corresponding ages for offspring with cancer in the entire cohort were similar to those with an affected mother (mean 22.6 and 22.7 years respectively).
Table 2

Tumor characteristics of mothers and offspring with and without affected family members

 

Mothers without affected offspring (= 3684)

Mothers with affected offspring n = 100

Offspring without affected mothers(n = 785)

Offspring with affected mother (n = 105)

Cancer site

n (%)

n &%

n (%)

n (%)

Breast

1539 (41.8)

47

69 (8.9)

11 (10.5)

Colon/Rectum

332 (9)

9

12 (1.5)

3 (2.9)

Ovary

136 (3.7)

1

19 (2.4)

Uterus

146 (4)

2

6 (0.8)

2 (1.9)

Melanoma

212 (5.8)

3

50 (6.3)

8 (7.6)

Lung

103 (2.8)

3

4 (0.5)

Thyroid

180 (4.9)

4

66 (8.4)

14 (13.3)

Cervix

150 (4.1)

7

60 (7.6)

7 (6.7)

Non-Hodgkin Lymphoma

137 (3.7)

8

59 (7.5)

11 (10.5)

Leukemia

53 (1.4)

68 (8.7)

13 (12.4)

Hodgkin Disease

35 (1.0)

2

96(12.2)

5 (4.8)

Brain

49 (1.3)

55 (7)

4 (3.8)

Stomach

64 (1.7)

4

5 (0.6)

Pancreas

50 (1.4)

1 (0.1)

Kidney

47 (1.3)

1

5 (0.6)

Bladder

47 (1.3)

7 (0.9)

Sarcoma

68 (1.8)

4

39 (5)

7 (6.7)

Liver/ biliary tract

32 (0.9)

2

1 (0.1)

Head and neck

66 (1.7)

12 (1.5)

2 (1.9)

Unknown primary

70 (1.9)

7 (0.9)

1 (1)

Testis

50 (6.4)

4 (3.8)

Retinoblastoma

6 (0.8)

1 (1.0)

Neuroblastoma

8 (1)

3 (2.9)

Nephroblastoma

4 (0.5)

1 (1.0)

Medulloblastoma

5 (0.6)

4 (3.8)

Other

169 (4.5)

1

72 (9,2)

4 (3.8)

% Males

44.4

43.8

Age at diagnosis Mean (SD)

51.7 [10.6]

52 [10]

22.6 [9.6]

22.7 [9.8]

Median

52

52

24.5

24.8

0–4

  

75 (8.4)

10 (9.5)

5–14

2 (0.1)

 

95 (10.7)

11 (10.5)

15–29

11 (4.3)

1

502 (56.4)

60 (57.1)

30–44

809 (21.4)

25

216 (24.3)

24 (22.9)

45–59

2029 (53.6)

51

 

60+

826 (21.8)

23

 

Missing

4 (0.1)

 

2 (0.2)

 

Breast cancer in the mother was co-observed with the same tumor in a daughter in seven families. Other concordant sites were lymphoma in three pairs, melanoma and colorectal cancers in two pairs each, and thyroid cancer in one pair. The combination of sarcoma and breast cancer occurred in four pairs, suggesting the Li- Fraumeni syndrome. There were no cases of the combination of endometrial and colon cancer in pairs. Thyroid cancer in the offspring was associated with breast cancer in the mother in six cases.

The interval between cancer diagnosis in the first and second pair members was evaluable in 103 pairs in which the date of diagnosis of both members was documented. The correlation between the child’s and mother’s age at diagnosis was moderate in the pairs (r = 0.199, r= 0.04, P = 0.042), while the correlation between calendar year of diagnosis among the pairs was weaker (r = 0.175, r= 0.03, P = 0.07). The mean and median time between cancer discovery in one member of a mother-child pair and the second member were relatively short, that is 8.6 and 5.9 years respectively (Table 3). The probability of an interval this short or shorter occurring by chance is 0.03. The findings were similar whether the mother’s or child’s diagnosis came first, however the median was shorter in the latter.
Table 3

Absolute Interval between diagnoses of malignancies in affected mother and offspring pairs

Interval

All pairs

Mother diagnosed first

Offspring diagnosed first

Mother with breast cancer

Mother other cancer

Mother <50 years of age at diagnosis

Mother ≥50 years of age at diagnosis

Offspring diagnosed <15 years of age

Offspring diagnosed ≥15years of age

Child DOB < 1970

Child DOB ≥1970

n

103

46

57

48

55

42

61

21

82

57

46

% ≤3 yearrs

33

30

35

29

36

26

38

14

38

32

35

Mean years

8.6

8.7

8.5

8.6

8.6

10.3

7.4

15.2

6.9

8.6

8.5

SD

8.2

8.3

8.2

7.4

8.9

8.7

7.7

9.1

7.1

8.3

8.2

Median years

5.9

6.4

5.9

7.3

5.2

8.1

5.1

15.2

4.7

5.9

6.3

P value

0.03

0.0067

0.0015

>0.1

0.06

0.075

>0.1

>0.1

0.026

0.026

>0.1

P value, for a permutation test on the median

The median interval was particularly short (4.7 years) in the subgroup of offspring diagnosed as adolescents and young adults ≥15 years of age. The finding was not observed when only mothers with breast cancer were considered. Those pairs in which the offspring was born in the 1960s showed a shorter median interval (5.9 years) compared with those born in the 1970s (median 6.3 years).

Restricting our analysis to invasive cancers (ie excluding 13 cases of in situ tumours) yielded a similar overall median and P value (0.02). When we based our calculations on the entire cohort of mothers and offspring diagnosed with cancer generating hypothetical pairs, our results were similar, with only 49 of 1000 replications yielding medians as short or shorter than the observed median. Moreover the lower quartile observed among the 103 pairs in our study was 2.1 years, whereas in only 15 of 1000 replications in the whole cohort the lower quartile was as short or shorter than the observed.

Figures 1 and 2 show the time to cancer detection (based on a Cox proportional hazards model adjusted for age ) after proband detection in children of affected mothers and in mothers of affected children, respectively. In both instances we observed a 50% probability that cancers detected among pair members would be diagnosed within five years of the proband’s diagnosis.
https://static-content.springer.com/image/art%3A10.1007%2Fs10689-006-9113-9/MediaObjects/10689_2006_9113_Fig1_HTML.gif
Fig. 1

Cox Model for time to offspring’s diagnosis of cancer. X-axis; Years from mother’s diagnosis Y-axis: Percent not diagnosed

https://static-content.springer.com/image/art%3A10.1007%2Fs10689-006-9113-9/MediaObjects/10689_2006_9113_Fig2_HTML.gif
Fig. 2

Cox model for time to mother’s diagnosis of cancer. X-axis: Years from offspring’s diagnosis Y-axis: Percent not diagnosed

We specifically examined the 34 (33%) cases in which both members of the pairs were diagnosed within 3 years of each other (Table 4) We found that in 18 (53%) the offspring was diagnosed first, both diagnoses occurred in the same month in 2 (6%) and the mother’s diagnosis came first in 14 (41%) pairs. In six (18%) of these cases cancer sites were concordant. Breast cancers in the mother represented 42% of cancers in this subgroup, similar to their proportion in the cohort, whereas in the offspring, breast cancers were somewhat over-represented, comprising 15% of tumors in the group. All cancers but two (one case of NHL and one of lung cancer) detected in mothers within one year of their offspring could have been detected by early diagnosis procedures (breast, colon, cervix-in situ, and uterus). On the other hand, when the offspring was diagnosed soon after the mother the cancer sites included sarcomas, lymphoma and brain cancer which are less likely to be subject to over-diagnosis or lead time bias.
Table 4

Year, age and site of tumors in pairs diagnosed within 3 years of each other

Interval between offspring’s and mother’s diagnosis

Year of diagnosis

Age at diagnosis

Cancer site/type

Offspring

Mother

Offspring

Mother

Offspring

Mother

25–36 months offspring first n = 5

1989

1992

26

60

Hodgkin

Breast

1999

2002

32

66

Cervix in situ

Colon

1998

2000

33

69

Thyroid

Lung

1988

1990

22

49

Melanoma

Melanoma

1999

2002

31

58

Breast

Osteosarcoma

13–24 months offspring first n = 7

1970

1972

5

40

NHL

Breast

1997

1999

28

60

NHL

Breast

2000

2002

30

62

Osteosarcoma

Cutaneous NHL

1993

1995

20

56

NHL

Breast

1997

1999

31

55

Breast

Colon

1996

1997

20

60

NHL

Breast

1971

1972

7

42

AML

Breast

0–12 months n = 13

2002

2002

34

57

NHL

Breast

2002

2003

33

67

Breast

Lung

2001

2002

31

68

Unknown 1˚

NHL

1993

1993

24

45

Rectum

Colon

2002

2002

29

67

Breast

Breast

1990

1990

17

46

Soft tissue sarcoma

Cervix in situ

1991

1990

16

49

Sarcoma

Breast

1996

1997

28

50

Thyroid

Breast

2000

2000

29

53

Breast

Breast

2002

2002

31

55

ALL

Cervix in situ

1995

1995

23

50

Hodgkin

Uterus

2000

2000

25

51

AML

Colon

2002

2002

36

65

Sinus

Breast

13–24 months mother first n = 5

1997

1996

26

60

Thyroid

Thyroid

2000

1998

33

58

Brain

Cervix

1994

1992

28

57

Colon

Breast

1998

1996

27

46

Thyroid

Breast

1988

1986

11

45

Osteosarcoma

Uterus

25–36 months mother first n = 4

1993

1991

24

46

Astrocytoma (brain)

Stomach

1997

1995

24

55

Melanoma in situ

Melanoma

1998

1995

24

47

Oral sqamous Cell

Lung

1994

1991

27

60

NHL

NHL

Total n = 34

Discussion

In our study population we were able to discern several families which could be suspected (but not confirmed, due to the limited extent of the pedigrees) and lack of analysis of the appropriate genes to have hereditary cancer syndromes such as Li-Fraumeni syndrome, hereditary breast cancer, FAP, or familial melanoma. Apart from these we observed a relatively large number of families in which the offspring was diagnosed with non-medullary thyroid cancer and the mother suffered from a variety of tumors.

Family studies of cancer have been fraught with bias, such as biased ascertainment of more severely affected family members or biased participation rates among family members with a higher perceived risk of cancer [25]. Studies based on cancer registries are generally thought to largely mitigate these biases. In this study, neither the mothers nor the offspring have been followed up sufficiently in order to ascertain all potential cases of familial cancer. Furthermore to our knowledge none of the mothers in the cohort were survivors of childhood cancer although three were diagnosed before the birth of their affected offspring. The short follow-up (maximum 40 years) in the offspring could spuriously result in a false inference of anticipation in these families. In our cohort, however, pairs followed-up from the 1960’s had a shorter interval than those followed from the 1970’s rendering this artefact unlikely. Other causes of spurious anticipation have included cohort effects [10], secular trends in earlier exposures to carcinogens such as cigarette smoking [26], curtailment of fertility in women with early onset hereditary breast cancer [27], gene-environment interactions such as obesity and lack of physical activity in younger generations of BRCA mutation carriers [28], or methodologic issues such as use of mean ages at cancer diagnosis instead of life table approaches [11].

In this study, the striking finding is the temporal proximity of the diagnosis of mothers and offspring. One third of the pairs were diagnosed within three years of each other, mainly at discordant sites. This clustering in time may suggest common exposures to carcinogenic factors, as was suggested by Grossman in a study of familial brain cancer [20]. Furthermore, inheritance of low penetrance modifying genes by the offspring may contribute to the short interval between diagnoses.

Alternatively or additionally, the large proportion of tumors in this group which are detectable by screening techniques suggests that health behaviors in one of the members of the pair may have been modified by the detection of cancer in the other pair member. Bermejo and Hemminki [29] have recently reported results from the Swedish Family Cancer database which corroborate our findings. They found that daughters of women with breast cancer and melanoma were particularly likely to have these tumors detected within a year of the mother’s diagnosis. This suggests that lead time bias may operate preferentially in families where one family member has already been diagnosed with cancer. We do not have data on stages of cancer at the time of detection, but removing the few in-situ cancers from our analysis did not substantially alter the results. While daughters of breast cancer patients have been found to have greater feelings of vulnerability to breast cancer there are no differences in their mammography practices compared to women without a maternal history [30]. Older women with a family history of breast cancer have not been found to be compliant with recommendations for yearly mammography screening [31], however the short-term response to having had a family member diagnosed with cancer is not known and may vary among populations. Most studies have focused on coping strategies of mothers with young children or adolescents with cancer [32] and not on their health behaviors. In our study, most of the offspring from pairs in whom cancer was diagnosed within three years of their mother, were in their third or fourth decades of life. We are not aware of any published data regarding health behaviors, especially screening practices, among mothers of young adult cancer patients. Further follow up of this cohort, as well as an assessment of cancer detection among fathers will shed more light on the robustness of our findings.

In conclusion, there is an unexpectedly short time interval between the diagnosis of cancer among mothers and their young adult children in the Jerusalem Perinatal Study cohort. Apart from the obvious psychosocial implications for families coping with two generations with cancer in a short interval, this finding has implications for research exploring anticipation as well as survival in familial cancers. Furthermore, studies examining health behaviors of parents of young adult cancer patients are warranted.

Copyright information

© Springer Science + Business Media B.V. 2007