Int. J. Cancer: 50,736-739 (1992) 0 1992 Wiley-Liss, Inc.
Publication of the InternationalUnion Against Cancer Publicationde I‘Union lnternationaleContre le Cancer
NITRATE AND N-NITROSOPROLINE EXCRETION IN TWO ITALIAN REGIONS WITH CONTRASTING RATES OF GASTRIC CANCER: THE ROLE OF NITRATE AND OTHER FACTORS IN ENDOGENOUS NITROSATION T. KNIGHT’, R. PIRASTU’, D. PALL?,P. c O C C 0 4 , s. LEACH5,P. PACKER’,R. IANNARILLI‘, P. MANCA4,H. MBLLER’ and D. FORM AN'.^ ’ICRFEpidemiology Unit, Gibson Building, Radcliffe Irijirmary, onford, UK; ‘Department of Animal and Human Biology, University of Rome, “L.a Sapienza ”, Rome, Italy; ’Epidemiology Unit, Centro per lo Studio e la Prevenzione Oncologica (CSPO), Florence, Italy; 41stitutodi Medicina del Lavoro, Universita di Cagliari, Cagliari, Sardinia, Italy; ’Bacterial Metabolism Group, Pathology Division, PHLS-CAMR, Porton Down, Salisbuiy, UK; ‘Istituto Superiore di Sanita, Rome, Italy; and ’Danish Cancer Registry, Institute of Cancer Epidemiology, Danish Cancer SocieQ, Copenhagen, Denmark. Exposure to nitrate and propensity for endogenous nitrosation were examined in 80 healthy males, aged 2 5 4 0 years, residing in areas of Italy with long-standing high (Florence) and low (Cagliari) rates of gastric cancer. Nitrate exposure was assessed by measurement of urinary nitrate excretion over I 2 hr, and endogenous nitrosationwas assessed usingthe N-nitrosoproline test (NPRO-test). Our hypothesis was whether the geographic variation in cancer rate correlated with nitrate exposure or nitrosating ability. Exposure to background sources of NPRO was significantly higher in the high-risk subjects (p = 0.04) whereas no differences were found in exposure to nitrate or in urinary NPRO levels after L-proline loading (test NPRO levels). The regional difference in test NPRO was almost completely accounted for by background NPRO exposure. Examination of individual rather than grouped data revealed that exposure to nitrate was a major factor in NPRO formation. No other factors studied (age, dietary-questionnaire-assessed intake of anti-oxidant vitamins) had a significant effect. Geographical variation in gastric cancer risk did not, therefore, correlate with either nitrate exposure or propensity for endogenous nitrosation of L-proline.
The most widely accepted hypothesis for gastric carcinogenesis was put forward by Correa (1988) in which, inter alia, N-nitroso compounds (NOC) formed endogenously from dietary nitrate (via bacterial reduction to nitrite) are proposed as a source of genotoxic carcinogens acting in the stomach. Exposure to exogenous nitrate has therefore been considered a major risk factor for gastric cancer via the endogenous formation of NOC. The epidemiological evidence for an aetiological association between nitrate/nitrite exposure and gastric cancer is, however, conflicting and remains the subject of debate (Forman, 1989). A major advance in the study of nitrosation was made with the introduction of the N-nitrosoproline technique (or the “NPRO-test”) (Ohshima and Bartsch, 1981). A loading dose of the amino acid L-proline is ingested together with nitrate (usually from dietary sources). The resulting compound, N-nitrosoproline (NPRO) is, atypically, neither mutagenic nor carcinogenic and is not metabolized in vivo but is excreted quantitatively in the urine (Bartsch et al., 1989). The measurement of 12-hr or 24-hr urinary excretion of NPRO can, therefore, be used as a non-invasive marker for endogenous nitrosation in humans (Bartsch et al., 1989). The NPRO method has been applied in several countries in correlational studies to assess nitrosating ability in populations resident in regions with varying rates of gastric cancer (Moller and Forman, 1991). The age-groups included in these previous studies were such that an increased prevalence of gastric hypochlorhydria among the subjects was likely (Walters et al., 1982). L-proline is most readily nitrosated under acidic conditions (Mirvish etal., 1973) and NPRO formation may therefore only be relevant as a marker for acid-catalysed endogenous nitrosation in normochlorhydric individuals. Certainly studies with NPRO in individuals with gastric disease and reduced
acidity have demonstrated reduced levels of L-proline nitrosation (Bartsch et al., 1989). We report here the results of a study conducted in healthy young men (reducing the risk of hypochlorhydria) in 2 regions of Italy which show long-standing contrasts in rates of mortality from gastric cancer (DeCarli et al., 1986). Florence, in Tuscany, has a rate comparable to some of the highest in the world. In contrast, Cagliari on the island of Sardinia has low rates. The purpose of our study was to examine the association between nitrosating ability in healthy men and risk of gastric cancer in these 2 Italian populations. At the same time we have examined the inter-relationship between endogenous NPRO synthesis and nitrate intake. SUBJECTS AND METHODS
Subjects In each of the 2 study regions, 40 men aged 2 5 4 0 years, resident in local urban areas, were invited by their family doctors to participate in the study. In none of the study participants had any clinical gastric disease ever been diagnosed. Urine collections On 2 consecutive days, 12-hr urine samples were collected from each subject, commencing 1 hr after the evening meal. Thirty minutes prior to the second sample, subjects ingested 500 mg of L-proline dissolved in water. Five grams of sodium hydroxide were placed inside the urine containers as a preservative agent (Ohshirna and Bartsch, 1981) and, within 12 hr of completing collection, sample volumes were recorded, and 2 100-ml aliquots were extracted and stored. One aliquot (-2oOC) was analysed for NPRO (Ohshima and Bartsch, 1981) within 4 months of collection and the other aliquot (PC), was analysed for nitrate (Packer et al., 1989) within 3 months of collection. Questionnaires Questionnaires were used to collect details of age, current smoking habits and dietary practices. The questionnaire used was adapted from that developed for a case-control study of gastric cancer and diet in Italy (Buiatti et al., 1989). Questionnaires were administered by trained interviewers. A full description of the methodology has been reported (Knight, 1988). Statistics Data for urinary nitrate and NPRO were normalized by logarithmic transformation; means are therefore geometric. In general, 2 approaches have been used for comparison of groups: firstly, a conventional comparison of the geometric means using t-tests and linear regression and, secondly, logistic ‘To whom correspondence and reprint requests should be sent. Received: October 28,1991.
737
ENDOGENOUS NITROSATION AND GASTRIC CANCER TAR1.E I - (;ASTKIC ('ANCtK MOR 1'AI.ITY KATE5 A?JD GEOMETRIC M F A N URINAKY KITRATE ANI) NPKO 1:XC'REI'ION (Y5'i ('ONFIDEKCL LSTKKi'ALSI IN TWO AKEhS OF ITALY ~~
1980-1982,0-74 years. Gastric cancer standardized mortality rate (number of deaths in parentheses)' Males Females Urinary nitrate (mg/12 hr) Background
Test Urinary NPRO (p,g/12 hr) Background Test
Florence
Cagliari
27.2 (205 16.7 (1421 n 40 37.1 (29.7,46.3) 39 36.0 (29.7,43.6)
15.6 (49) 7.1 (22)
n 39 33.7 (27.0,42.0) 39 34.4 (28.8,41.0)
40 0.94 (0.72, 1.24) 40 1.59 (1.21,2.08)
40 0.58 (0.41,0.83) 40 1.44 (1.10, 1.87)
P'
0.55 0.73 0.04 0.61
'Comparison between populations (t-test).-'1980-1982 mortality rates per 100,000, 0-74 years, for the cities of Florence and Cagliari, age-standardized to the entire Italian population (1981). (lstituto Centrale di Statistica, Rome, 1988).
regression analysis. For the latter, subjects were categorized dichotomously according to whether or not they excreted 1 pg or more NPRO in 12 hr. Odds ratios (OR) for excreting 1pg or more NPRO were then calculated using univariate or multivariate logistic regression (Breslow and Day, 1980) with the CATMOD procedure in the SAS statistical package (SAS Institute, 1985). Due to the extremely skewed distribution of urinary NPRO, this method of analysis is most appropriate for statistical modelling of NPRO excretion data (Moiler et at., 1989). RESULTS
Urinary levels of nitrate and NPRO before (background) and after (test) L-proline loading are shown in Table I, together with gastric cancer mortality rates for the 2 study regions. Urinary nitrate excretion was constant over the 2 study days (background and test) and there was no significant difference in nitrate excretion between the 2 populations. Background levels of urinary NPRO were significantly greater in the subjects from Florence ( p = 0.04). Urinary excretion of NPRO approximately doubled under test conditions, i.e. after the L-proline loading dose, in both populations (increasing by 69% in Florence and 148% in Cagliari). There was no significant difference between the populations in mean test NPRO levels. In order to analyse the regional differences in NPRO excretion more fully, a logistic regression approach was used. Subjects were categorized dichotomously according to whether or not they excreted 1 pg or more NPRO in 12 hr, under test conditions. Odds ratios (OR) for excreting 1 pg or more NPRO were then calculated for the Florence subjects in comparison to the Cagliari subjects before and after adjustment for a number of co-variates. Results are shown in Table 11.
Subjects from Florence were more likely to excrete 1 pg or more NPRO (OR = 1.7) than residents of Cagliari. This excess was, however, not statistically significant. Two covariates independently improved the fit of the regression model. These were background NPRO excretion (for which subjects were also categorized dichotomously according to whether or not they excreted 1pg or more in 12 hr) and urinary nitrate excretion (for which subjects were categorized according to the tertile of nitrate excretion). Background NPRO adjustment acted to significantly reduce the OR for Florence subjects to 1.1, whereas urinary nitrate adjustment marginally reduced the OR (to 1.6). Fitting both co-variates simulta-
TABLE I1 -ODDS RATIOS FOR RESIDENCE IN FLORENCE COMPARED WITH RESIDENCE IN CAGLIARI FOR E L E V A T E D ~1 p j i 1 2 HR)
LEVELS O F TEST URINARY NPRO EXC E T
N
~~
OR' (95% CI)
Not adjusted Adjusted for: Background NPRO ( < 1 p,g vs. 2 1 pgil2 hr) Test urinary nitrate (tertiles) Background NPRO and test urinarv nitrate
1.7 (0.6, 4.8) 1.1 (0.3,3.4) 1.6 (0.5,4.8) 1.0 10.3.3.6)
'Odds ratio for residence in Florence (residence in Cagliari = 1.0) for excreting 1 kg or more urinary NPROi 12 hr after proline loading dose. TABLE 111-EFFECT OF URINARY NITRATE EXCRETION ON BACKGROUND URINARY NPRO EXCRETION IN FLORENCE AND CAGLIARI SUBJECTS COMBINED
n
Background urinary NPRO excretion (Pgi1.2 hr) geometriclmean (95% CI)
OR' (95% CI)
Background urina nitrate excretion7 LOW
Medium High All D-value 'Odds ratio for excreting 1 pg or more background urinary NPROil2 hr.-'low =
Publication of the InternationalUnion Against Cancer Publicationde I‘Union lnternationaleContre le Cancer
NITRATE AND N-NITROSOPROLINE EXCRETION IN TWO ITALIAN REGIONS WITH CONTRASTING RATES OF GASTRIC CANCER: THE ROLE OF NITRATE AND OTHER FACTORS IN ENDOGENOUS NITROSATION T. KNIGHT’, R. PIRASTU’, D. PALL?,P. c O C C 0 4 , s. LEACH5,P. PACKER’,R. IANNARILLI‘, P. MANCA4,H. MBLLER’ and D. FORM AN'.^ ’ICRFEpidemiology Unit, Gibson Building, Radcliffe Irijirmary, onford, UK; ‘Department of Animal and Human Biology, University of Rome, “L.a Sapienza ”, Rome, Italy; ’Epidemiology Unit, Centro per lo Studio e la Prevenzione Oncologica (CSPO), Florence, Italy; 41stitutodi Medicina del Lavoro, Universita di Cagliari, Cagliari, Sardinia, Italy; ’Bacterial Metabolism Group, Pathology Division, PHLS-CAMR, Porton Down, Salisbuiy, UK; ‘Istituto Superiore di Sanita, Rome, Italy; and ’Danish Cancer Registry, Institute of Cancer Epidemiology, Danish Cancer SocieQ, Copenhagen, Denmark. Exposure to nitrate and propensity for endogenous nitrosation were examined in 80 healthy males, aged 2 5 4 0 years, residing in areas of Italy with long-standing high (Florence) and low (Cagliari) rates of gastric cancer. Nitrate exposure was assessed by measurement of urinary nitrate excretion over I 2 hr, and endogenous nitrosationwas assessed usingthe N-nitrosoproline test (NPRO-test). Our hypothesis was whether the geographic variation in cancer rate correlated with nitrate exposure or nitrosating ability. Exposure to background sources of NPRO was significantly higher in the high-risk subjects (p = 0.04) whereas no differences were found in exposure to nitrate or in urinary NPRO levels after L-proline loading (test NPRO levels). The regional difference in test NPRO was almost completely accounted for by background NPRO exposure. Examination of individual rather than grouped data revealed that exposure to nitrate was a major factor in NPRO formation. No other factors studied (age, dietary-questionnaire-assessed intake of anti-oxidant vitamins) had a significant effect. Geographical variation in gastric cancer risk did not, therefore, correlate with either nitrate exposure or propensity for endogenous nitrosation of L-proline.
The most widely accepted hypothesis for gastric carcinogenesis was put forward by Correa (1988) in which, inter alia, N-nitroso compounds (NOC) formed endogenously from dietary nitrate (via bacterial reduction to nitrite) are proposed as a source of genotoxic carcinogens acting in the stomach. Exposure to exogenous nitrate has therefore been considered a major risk factor for gastric cancer via the endogenous formation of NOC. The epidemiological evidence for an aetiological association between nitrate/nitrite exposure and gastric cancer is, however, conflicting and remains the subject of debate (Forman, 1989). A major advance in the study of nitrosation was made with the introduction of the N-nitrosoproline technique (or the “NPRO-test”) (Ohshima and Bartsch, 1981). A loading dose of the amino acid L-proline is ingested together with nitrate (usually from dietary sources). The resulting compound, N-nitrosoproline (NPRO) is, atypically, neither mutagenic nor carcinogenic and is not metabolized in vivo but is excreted quantitatively in the urine (Bartsch et al., 1989). The measurement of 12-hr or 24-hr urinary excretion of NPRO can, therefore, be used as a non-invasive marker for endogenous nitrosation in humans (Bartsch et al., 1989). The NPRO method has been applied in several countries in correlational studies to assess nitrosating ability in populations resident in regions with varying rates of gastric cancer (Moller and Forman, 1991). The age-groups included in these previous studies were such that an increased prevalence of gastric hypochlorhydria among the subjects was likely (Walters et al., 1982). L-proline is most readily nitrosated under acidic conditions (Mirvish etal., 1973) and NPRO formation may therefore only be relevant as a marker for acid-catalysed endogenous nitrosation in normochlorhydric individuals. Certainly studies with NPRO in individuals with gastric disease and reduced
acidity have demonstrated reduced levels of L-proline nitrosation (Bartsch et al., 1989). We report here the results of a study conducted in healthy young men (reducing the risk of hypochlorhydria) in 2 regions of Italy which show long-standing contrasts in rates of mortality from gastric cancer (DeCarli et al., 1986). Florence, in Tuscany, has a rate comparable to some of the highest in the world. In contrast, Cagliari on the island of Sardinia has low rates. The purpose of our study was to examine the association between nitrosating ability in healthy men and risk of gastric cancer in these 2 Italian populations. At the same time we have examined the inter-relationship between endogenous NPRO synthesis and nitrate intake. SUBJECTS AND METHODS
Subjects In each of the 2 study regions, 40 men aged 2 5 4 0 years, resident in local urban areas, were invited by their family doctors to participate in the study. In none of the study participants had any clinical gastric disease ever been diagnosed. Urine collections On 2 consecutive days, 12-hr urine samples were collected from each subject, commencing 1 hr after the evening meal. Thirty minutes prior to the second sample, subjects ingested 500 mg of L-proline dissolved in water. Five grams of sodium hydroxide were placed inside the urine containers as a preservative agent (Ohshirna and Bartsch, 1981) and, within 12 hr of completing collection, sample volumes were recorded, and 2 100-ml aliquots were extracted and stored. One aliquot (-2oOC) was analysed for NPRO (Ohshima and Bartsch, 1981) within 4 months of collection and the other aliquot (PC), was analysed for nitrate (Packer et al., 1989) within 3 months of collection. Questionnaires Questionnaires were used to collect details of age, current smoking habits and dietary practices. The questionnaire used was adapted from that developed for a case-control study of gastric cancer and diet in Italy (Buiatti et al., 1989). Questionnaires were administered by trained interviewers. A full description of the methodology has been reported (Knight, 1988). Statistics Data for urinary nitrate and NPRO were normalized by logarithmic transformation; means are therefore geometric. In general, 2 approaches have been used for comparison of groups: firstly, a conventional comparison of the geometric means using t-tests and linear regression and, secondly, logistic ‘To whom correspondence and reprint requests should be sent. Received: October 28,1991.
737
ENDOGENOUS NITROSATION AND GASTRIC CANCER TAR1.E I - (;ASTKIC ('ANCtK MOR 1'AI.ITY KATE5 A?JD GEOMETRIC M F A N URINAKY KITRATE ANI) NPKO 1:XC'REI'ION (Y5'i ('ONFIDEKCL LSTKKi'ALSI IN TWO AKEhS OF ITALY ~~
1980-1982,0-74 years. Gastric cancer standardized mortality rate (number of deaths in parentheses)' Males Females Urinary nitrate (mg/12 hr) Background
Test Urinary NPRO (p,g/12 hr) Background Test
Florence
Cagliari
27.2 (205 16.7 (1421 n 40 37.1 (29.7,46.3) 39 36.0 (29.7,43.6)
15.6 (49) 7.1 (22)
n 39 33.7 (27.0,42.0) 39 34.4 (28.8,41.0)
40 0.94 (0.72, 1.24) 40 1.59 (1.21,2.08)
40 0.58 (0.41,0.83) 40 1.44 (1.10, 1.87)
P'
0.55 0.73 0.04 0.61
'Comparison between populations (t-test).-'1980-1982 mortality rates per 100,000, 0-74 years, for the cities of Florence and Cagliari, age-standardized to the entire Italian population (1981). (lstituto Centrale di Statistica, Rome, 1988).
regression analysis. For the latter, subjects were categorized dichotomously according to whether or not they excreted 1 pg or more NPRO in 12 hr. Odds ratios (OR) for excreting 1pg or more NPRO were then calculated using univariate or multivariate logistic regression (Breslow and Day, 1980) with the CATMOD procedure in the SAS statistical package (SAS Institute, 1985). Due to the extremely skewed distribution of urinary NPRO, this method of analysis is most appropriate for statistical modelling of NPRO excretion data (Moiler et at., 1989). RESULTS
Urinary levels of nitrate and NPRO before (background) and after (test) L-proline loading are shown in Table I, together with gastric cancer mortality rates for the 2 study regions. Urinary nitrate excretion was constant over the 2 study days (background and test) and there was no significant difference in nitrate excretion between the 2 populations. Background levels of urinary NPRO were significantly greater in the subjects from Florence ( p = 0.04). Urinary excretion of NPRO approximately doubled under test conditions, i.e. after the L-proline loading dose, in both populations (increasing by 69% in Florence and 148% in Cagliari). There was no significant difference between the populations in mean test NPRO levels. In order to analyse the regional differences in NPRO excretion more fully, a logistic regression approach was used. Subjects were categorized dichotomously according to whether or not they excreted 1 pg or more NPRO in 12 hr, under test conditions. Odds ratios (OR) for excreting 1 pg or more NPRO were then calculated for the Florence subjects in comparison to the Cagliari subjects before and after adjustment for a number of co-variates. Results are shown in Table 11.
Subjects from Florence were more likely to excrete 1 pg or more NPRO (OR = 1.7) than residents of Cagliari. This excess was, however, not statistically significant. Two covariates independently improved the fit of the regression model. These were background NPRO excretion (for which subjects were also categorized dichotomously according to whether or not they excreted 1pg or more in 12 hr) and urinary nitrate excretion (for which subjects were categorized according to the tertile of nitrate excretion). Background NPRO adjustment acted to significantly reduce the OR for Florence subjects to 1.1, whereas urinary nitrate adjustment marginally reduced the OR (to 1.6). Fitting both co-variates simulta-
TABLE I1 -ODDS RATIOS FOR RESIDENCE IN FLORENCE COMPARED WITH RESIDENCE IN CAGLIARI FOR E L E V A T E D ~1 p j i 1 2 HR)
LEVELS O F TEST URINARY NPRO EXC E T
N
~~
OR' (95% CI)
Not adjusted Adjusted for: Background NPRO ( < 1 p,g vs. 2 1 pgil2 hr) Test urinary nitrate (tertiles) Background NPRO and test urinarv nitrate
1.7 (0.6, 4.8) 1.1 (0.3,3.4) 1.6 (0.5,4.8) 1.0 10.3.3.6)
'Odds ratio for residence in Florence (residence in Cagliari = 1.0) for excreting 1 kg or more urinary NPROi 12 hr after proline loading dose. TABLE 111-EFFECT OF URINARY NITRATE EXCRETION ON BACKGROUND URINARY NPRO EXCRETION IN FLORENCE AND CAGLIARI SUBJECTS COMBINED
n
Background urinary NPRO excretion (Pgi1.2 hr) geometriclmean (95% CI)
OR' (95% CI)
Background urina nitrate excretion7 LOW
Medium High All D-value 'Odds ratio for excreting 1 pg or more background urinary NPROil2 hr.-'low =
