Publicationof the lnternaiionalUnion Against Cancer Publication de I Union Internationale Contre le Cancer
lnt. J. Cancer: 52,35 1-354 (1992) 0 1992 Wiley-Liss, lnc.
REPRODUCTIVE FACTORS AND THE RISK OF HEPATOCELLULAR CARCINOMA IN WOMEN Carlo LA vECCHIA’,2,4,Eva NEGRI’,Silvia FRANCESCHI~ and Barbara D’AVANZO’ IIstiticto di Ricerclie Famiacologiche “Mario Negn”’, 20157 Milan, Italy; %tstitute of Social and Preventive Medicine, University of Luusatiiie, 1005 Laiisanne, Switzerland; and 3Aviano Cancer Center, 33081 Aviano (Pordenone), Italy. The relationship between reproductive factors and the risk of primary liver cancer was analyzed using data of a case-control study conducted in Northern Italy between I984 and I99 I on 79 women with histologicallyor serologically confirmed hepatocelM a r carcinoma and 344 controls in hospital for a wide spectrum of acute, non-neoplastic diseases. The multivariate relative risk (RR) for parous vs. nulliparous women was 2.6 (95% confidence interval (CI) I .2 to 5.8), and the risk increased with parity from 2. I for I, to 2.6 for 2, to 3.2 for 3, to 3.5 for 4 or more births trend = 6.49, p = 0.01). The relative risks were above unity, though not significantly, in women reporting spontaneous (RR = 1.3) and induced (RR = 1.6) abortions, and there was a significant trend in risk with total number of abortions. An apparent inverse trend in risk with age and first birth was accounted for by parity. No relationship emerged with age at menarche, at menopause or other menstrual factors. The association between parity and hepatocellular carcinoma was, if anything, more marked at older ages, since the RR was I .6 (95% CI 0.5 to 4.6) below age 60, and 4.8 (95% CI I .3 to 18. I) at age 60 or over. This observation has relevant public-health implications, since in developed countries primary liver cancer is extremely rare among young women, but not at older ages. The association between parity and hepatocellular carcinoma is similar to that described for combined oral contraceptives, again confirming that the impact of contraceptiveson the risk of several neoplasms is similar to that of pregnancy. 0 199-3 wi/ey-Liss, Inc.
There is convincing evidence that exogenous steroid hormones, including artdrogenic-anabolic steroids and oestrogensprogestins combined in oral contraceptives, are associated with elevated risk of hepatocellular carcinoma (Henderson et al., 1983; Neuberger et al., 1986; Forman et al., 1986; Palmer et al., 1989; La Vecchia et al., 1989, 1990; Hsing et al., 1992). A few studies suggested that reproductive factors, which modify endogenous levels of steroid hormones, are also associated with primary liver cancer. These include 2 cancer-registrybased studies of parity and cancer incidence (Miller et al., 1980; Plesko et al., 1985), which found higher liver cancer rates in parous than in nulliparous women, but no significant and consistent trend in risk. Further, the WHO Collaborative Study of Neoplasia and Steroid Contraceptives (Stanford et al., 1992), a case-control study conducted in 4 developing countries, found a direct association of primary liver cancer with parity, and a significant trend in risk with number of births. To further explore the issue of reproductive pattern and hepatocellular carcinoma, this report considers the data from a case-control study of primary liver cancer conducted in Northern Italy. SUBJECTS AND METHODS
The data were derived from an ongoing case-control study of several digestive tract neoplasms, whose general design has been previously described (La Vecchia et al., 1988, 1989). Briefly, trained interviewers identified and questioned eases of primary liver cancer and patients admitted for a wide spectrum of acute, non-neoplastic diseases (controls) to a network including the major teaching and general hospitals in the Greater Milan area. On the average, less than 2% of eligible subjects (cases and controls) refused to be interviewed.
The cases included in the present analysis were 79 women with histologically (n = 61) or serologically (elevated alphafoetoprotein levels) (n = 18) confirmed hepatocellular carcinoma diagnosed between January 1984 and September 1991, after specific exclusion, on the basis of the above criteria of all metastatic or undefined liver neoplasms. The age range was 28 to 74, and median age was 57 years. The controls were 344 women, admitted over the same calendar period to the same network of hospitals for traumas (33%), other orthopaedic conditions (20%), acute surgical diseases (28%), and other miscellaneous illnesses (19%). Controls were not individually matched with cases but their age distribution was comparable. The age range of the comparison group was 25 to 74, and median age was 58 years. The structured questionnaire included information on sociodemographic factors and lifestyle habits, a few selected dietary indicators and medical history; further, data were elicited on menstrual and reproductive factors and use of oral contraceptives and female hormones for other purposes. Statistical analyses were based on standard methods for case-control studies (Breslow and Day, 1980), including stratification and the ManteLHaenszel(l959) procedure and unconditional multiple logistic regression to obtain odds ratios as estimates of relative risk (RR) and the corresponding 95% confidence intervals (CI) (Baker and Nelder, 1978). Included in the regression equations were terms for age, education, alcohol consumption, history of hepatitis, and oral contraceptive use. RESULTS
Table I gives the distribution of cases of hepatocellular carcinoma and the comparison group according to age and selected co-variates. Cases were slightly younger than controls, less frequently unmarried and significantly less educated. There was no significant difference in alcohol consumption, but cases reported more frequently a history of clinically evident hepatitis and oral contraceptive use. Table I1 includes selected reproductive variables. Parity was significantly associated with hepatocellular carcinoma; the RR for parae vs. nulliparae was 2.5 (95% CI 1.2-5.2). Further, there was a significant trend in risk with number of births: relative to nulliparae, the RRs were 1.9 for 1,2.2 for 2,2.7 for 3 and 4.0 for 4 or more births (x: trend = 8 . 5 9 , ~< 0.01). The relative risks were above unity, though not significantly, in women reporting spontaneous (RR = 1.3) and induced (RR = 1.7) abortions, and there was a trend in risk of borderline significance with total number of abortions (x: = 3.96, p < 0.05). None of these risk estimates was materially modified by allowance for several identified potential confounding factors, using multivariate logistic regression. Further, there 4To whom corres ondence and reprint requests should be adRicerche Farmacologiche “Mario Negri”, Via dressed, at Istituto Eritrea, 62, 20157 Milan, Italy. Fax: 0213536277.
8
Received: March 27,1992.
TABLE I - DISTRIBUTION OF 79 CASES OF HEPATOCELLULAR CARCINOMA AND 344 CONTROLS ACCORDING TO AGE AND SELECTED COVARIATES. MILAN, ITALY, 1984-91
Hepatocellular carcinoma
Age group (years) < 45 45-54 5564 65-74 Marital status Never married Married Education (wars) 2-4 >4 Unknown Clinical diagnosis or hepatitis No Yes Oral contraceDtive use Never Ever
Controls
Number
%
Number
%
18 15 23 23
22.8 19.0 29.1 29.1
59 72 113 100
17.2 20.9 32.8 29.1
2 77
2.5 97.5
40 304
11.6 88.4
54 14 10 1
68.4 17.7 12.6 1.3
179 97 66 2
52.0 28.2 19.2 0.6
33 25 16 5
41.8 31.6 20.3 6.3
-
-
138 86 102 17 1
40.1 25.0 29.7 4.9 0.3
68 11
86.1 13.9
327 17
95.1 4.9
71 8
89.9 10.1
325 19
94.5 5.5
TABLE I1 - DISTRIBUTION OF 79 CASES OF HEPATOCELLULAR CARCINOMA AND 344 CONTROLS ACCORDING TO SELECTED REPRODUCTIVE VARIABLES, MILAN, ITALY, 1984-91 Relative risk estimates (9S%, CI)
Hepaloceliuidr carcinoma
Controls
8 71
75 269
1
19
94
2
26
100
3
13
44
24
13
31
60 19
275 69
12 1.3 (0.7-2.3)
Number of induced abortions 0 21
68 11
314 30
12 1.7 (0.8-3.5)
1.6 (0.7-3.6)
Total number of abortions 0 1
51 15
248 64
13
32
12 1.1 (0.6-2.2) 2.0 (1.040) 3.96"
1.0 (0.5-2.0) 2.1 (1.0-4.3) 3.86"
9 31
23 106
25-29
22
97
2 30
9
43
Nulliparae
8
75
Number of births Nulliparae Parae
xi (trend) Number of spontaneous abortions 0 21
22
x: (trend) Age at first birth (years) < 20 20-24
x: (trend), nulliparae excluded
-
MH'
13
2.5 (1.2-5.2) 1.9 (0.84.5) 2.2 ( 1.O-5.1) 2.7 (1 .l-6.8) 4.0 (1.5-10.4) 8.59"
1'
0.8 (0.3-1.8) 0.6 (0.2-1.4) 0.5 (0.2-1.4) 0.3 (0.1-0.9) 2.04 (ns.)
m1r2 13
2.6 (1.2-5.8) 2.1 (0.8-5.2) 2.6 (1.1-6.3) 3.2 (1.2-8.7) 3.5 (1.2-9.9) 6.49* 12
1.3 (0.7-2.3) 11
12
12
0.7 (0.2-1.6) 0.5 (0.2-1.4) 0.7 (0.2-2.1) -
1.28 (ns.)
'Mantel-Haenszel estimates adjusted for age.-2Estimates from multiple logistic regression equations, including terms for age, education, alcohol consumption, history of he atitis and oral-contraceptive use; estimates for age of first birth included also terms for parity.- PReference category.-*,p < 0.05; * * , p < 0.01.
REPRODUCTIVE FACTORS A N D LIVER CANCER RISK TABLE 111- DISTRIBUTION OF 79 CASES OF HEPATOCELLULAR CARCINOMA AND 344 CONTROLS ACCORDING TO 2 SEPARATE STRATA OF AGE. MILAN, ITALY. 1984-91
Hepatocellular carcinoma
Age
lnt. J. Cancer: 52,35 1-354 (1992) 0 1992 Wiley-Liss, lnc.
REPRODUCTIVE FACTORS AND THE RISK OF HEPATOCELLULAR CARCINOMA IN WOMEN Carlo LA vECCHIA’,2,4,Eva NEGRI’,Silvia FRANCESCHI~ and Barbara D’AVANZO’ IIstiticto di Ricerclie Famiacologiche “Mario Negn”’, 20157 Milan, Italy; %tstitute of Social and Preventive Medicine, University of Luusatiiie, 1005 Laiisanne, Switzerland; and 3Aviano Cancer Center, 33081 Aviano (Pordenone), Italy. The relationship between reproductive factors and the risk of primary liver cancer was analyzed using data of a case-control study conducted in Northern Italy between I984 and I99 I on 79 women with histologicallyor serologically confirmed hepatocelM a r carcinoma and 344 controls in hospital for a wide spectrum of acute, non-neoplastic diseases. The multivariate relative risk (RR) for parous vs. nulliparous women was 2.6 (95% confidence interval (CI) I .2 to 5.8), and the risk increased with parity from 2. I for I, to 2.6 for 2, to 3.2 for 3, to 3.5 for 4 or more births trend = 6.49, p = 0.01). The relative risks were above unity, though not significantly, in women reporting spontaneous (RR = 1.3) and induced (RR = 1.6) abortions, and there was a significant trend in risk with total number of abortions. An apparent inverse trend in risk with age and first birth was accounted for by parity. No relationship emerged with age at menarche, at menopause or other menstrual factors. The association between parity and hepatocellular carcinoma was, if anything, more marked at older ages, since the RR was I .6 (95% CI 0.5 to 4.6) below age 60, and 4.8 (95% CI I .3 to 18. I) at age 60 or over. This observation has relevant public-health implications, since in developed countries primary liver cancer is extremely rare among young women, but not at older ages. The association between parity and hepatocellular carcinoma is similar to that described for combined oral contraceptives, again confirming that the impact of contraceptiveson the risk of several neoplasms is similar to that of pregnancy. 0 199-3 wi/ey-Liss, Inc.
There is convincing evidence that exogenous steroid hormones, including artdrogenic-anabolic steroids and oestrogensprogestins combined in oral contraceptives, are associated with elevated risk of hepatocellular carcinoma (Henderson et al., 1983; Neuberger et al., 1986; Forman et al., 1986; Palmer et al., 1989; La Vecchia et al., 1989, 1990; Hsing et al., 1992). A few studies suggested that reproductive factors, which modify endogenous levels of steroid hormones, are also associated with primary liver cancer. These include 2 cancer-registrybased studies of parity and cancer incidence (Miller et al., 1980; Plesko et al., 1985), which found higher liver cancer rates in parous than in nulliparous women, but no significant and consistent trend in risk. Further, the WHO Collaborative Study of Neoplasia and Steroid Contraceptives (Stanford et al., 1992), a case-control study conducted in 4 developing countries, found a direct association of primary liver cancer with parity, and a significant trend in risk with number of births. To further explore the issue of reproductive pattern and hepatocellular carcinoma, this report considers the data from a case-control study of primary liver cancer conducted in Northern Italy. SUBJECTS AND METHODS
The data were derived from an ongoing case-control study of several digestive tract neoplasms, whose general design has been previously described (La Vecchia et al., 1988, 1989). Briefly, trained interviewers identified and questioned eases of primary liver cancer and patients admitted for a wide spectrum of acute, non-neoplastic diseases (controls) to a network including the major teaching and general hospitals in the Greater Milan area. On the average, less than 2% of eligible subjects (cases and controls) refused to be interviewed.
The cases included in the present analysis were 79 women with histologically (n = 61) or serologically (elevated alphafoetoprotein levels) (n = 18) confirmed hepatocellular carcinoma diagnosed between January 1984 and September 1991, after specific exclusion, on the basis of the above criteria of all metastatic or undefined liver neoplasms. The age range was 28 to 74, and median age was 57 years. The controls were 344 women, admitted over the same calendar period to the same network of hospitals for traumas (33%), other orthopaedic conditions (20%), acute surgical diseases (28%), and other miscellaneous illnesses (19%). Controls were not individually matched with cases but their age distribution was comparable. The age range of the comparison group was 25 to 74, and median age was 58 years. The structured questionnaire included information on sociodemographic factors and lifestyle habits, a few selected dietary indicators and medical history; further, data were elicited on menstrual and reproductive factors and use of oral contraceptives and female hormones for other purposes. Statistical analyses were based on standard methods for case-control studies (Breslow and Day, 1980), including stratification and the ManteLHaenszel(l959) procedure and unconditional multiple logistic regression to obtain odds ratios as estimates of relative risk (RR) and the corresponding 95% confidence intervals (CI) (Baker and Nelder, 1978). Included in the regression equations were terms for age, education, alcohol consumption, history of hepatitis, and oral contraceptive use. RESULTS
Table I gives the distribution of cases of hepatocellular carcinoma and the comparison group according to age and selected co-variates. Cases were slightly younger than controls, less frequently unmarried and significantly less educated. There was no significant difference in alcohol consumption, but cases reported more frequently a history of clinically evident hepatitis and oral contraceptive use. Table I1 includes selected reproductive variables. Parity was significantly associated with hepatocellular carcinoma; the RR for parae vs. nulliparae was 2.5 (95% CI 1.2-5.2). Further, there was a significant trend in risk with number of births: relative to nulliparae, the RRs were 1.9 for 1,2.2 for 2,2.7 for 3 and 4.0 for 4 or more births (x: trend = 8 . 5 9 , ~< 0.01). The relative risks were above unity, though not significantly, in women reporting spontaneous (RR = 1.3) and induced (RR = 1.7) abortions, and there was a trend in risk of borderline significance with total number of abortions (x: = 3.96, p < 0.05). None of these risk estimates was materially modified by allowance for several identified potential confounding factors, using multivariate logistic regression. Further, there 4To whom corres ondence and reprint requests should be adRicerche Farmacologiche “Mario Negri”, Via dressed, at Istituto Eritrea, 62, 20157 Milan, Italy. Fax: 0213536277.
8
Received: March 27,1992.
TABLE I - DISTRIBUTION OF 79 CASES OF HEPATOCELLULAR CARCINOMA AND 344 CONTROLS ACCORDING TO AGE AND SELECTED COVARIATES. MILAN, ITALY, 1984-91
Hepatocellular carcinoma
Age group (years) < 45 45-54 5564 65-74 Marital status Never married Married Education (wars) 2-4 >4 Unknown Clinical diagnosis or hepatitis No Yes Oral contraceDtive use Never Ever
Controls
Number
%
Number
%
18 15 23 23
22.8 19.0 29.1 29.1
59 72 113 100
17.2 20.9 32.8 29.1
2 77
2.5 97.5
40 304
11.6 88.4
54 14 10 1
68.4 17.7 12.6 1.3
179 97 66 2
52.0 28.2 19.2 0.6
33 25 16 5
41.8 31.6 20.3 6.3
-
-
138 86 102 17 1
40.1 25.0 29.7 4.9 0.3
68 11
86.1 13.9
327 17
95.1 4.9
71 8
89.9 10.1
325 19
94.5 5.5
TABLE I1 - DISTRIBUTION OF 79 CASES OF HEPATOCELLULAR CARCINOMA AND 344 CONTROLS ACCORDING TO SELECTED REPRODUCTIVE VARIABLES, MILAN, ITALY, 1984-91 Relative risk estimates (9S%, CI)
Hepaloceliuidr carcinoma
Controls
8 71
75 269
1
19
94
2
26
100
3
13
44
24
13
31
60 19
275 69
12 1.3 (0.7-2.3)
Number of induced abortions 0 21
68 11
314 30
12 1.7 (0.8-3.5)
1.6 (0.7-3.6)
Total number of abortions 0 1
51 15
248 64
13
32
12 1.1 (0.6-2.2) 2.0 (1.040) 3.96"
1.0 (0.5-2.0) 2.1 (1.0-4.3) 3.86"
9 31
23 106
25-29
22
97
2 30
9
43
Nulliparae
8
75
Number of births Nulliparae Parae
xi (trend) Number of spontaneous abortions 0 21
22
x: (trend) Age at first birth (years) < 20 20-24
x: (trend), nulliparae excluded
-
MH'
13
2.5 (1.2-5.2) 1.9 (0.84.5) 2.2 ( 1.O-5.1) 2.7 (1 .l-6.8) 4.0 (1.5-10.4) 8.59"
1'
0.8 (0.3-1.8) 0.6 (0.2-1.4) 0.5 (0.2-1.4) 0.3 (0.1-0.9) 2.04 (ns.)
m1r2 13
2.6 (1.2-5.8) 2.1 (0.8-5.2) 2.6 (1.1-6.3) 3.2 (1.2-8.7) 3.5 (1.2-9.9) 6.49* 12
1.3 (0.7-2.3) 11
12
12
0.7 (0.2-1.6) 0.5 (0.2-1.4) 0.7 (0.2-2.1) -
1.28 (ns.)
'Mantel-Haenszel estimates adjusted for age.-2Estimates from multiple logistic regression equations, including terms for age, education, alcohol consumption, history of he atitis and oral-contraceptive use; estimates for age of first birth included also terms for parity.- PReference category.-*,p < 0.05; * * , p < 0.01.
REPRODUCTIVE FACTORS A N D LIVER CANCER RISK TABLE 111- DISTRIBUTION OF 79 CASES OF HEPATOCELLULAR CARCINOMA AND 344 CONTROLS ACCORDING TO 2 SEPARATE STRATA OF AGE. MILAN, ITALY. 1984-91
Hepatocellular carcinoma
Age
