Dietary Vitamins A And C And Lung Cancer Risk In Louisiana

Dietary Vitamins A and C and Lung Cancer Risk in Louisiana ELIZABETH T. H. FONTHAM,' LINDA WILLIAMS PICKLE,* WILLIAM HAENSZEL,* PELAYO CORREA,* YOUPING LIN,' AND RON1 T. FALKt

The authors describe the results of a hospital-based incident case-control study of lung cancer conducted in a high-risk region of southern Louisiana from January 1979 through April 1982. Dietary intake of carotene, retinol, and vitamin C was estimated from food frequency questionnaires administered to 1253 cases and 1274 controls. An inverse association was fouhd between level of carotene intake and lung cancer risk, and this protective effect was specific for squamous and small cell carcinoma (odds ratio [OR] = 0.84, 95%confidence interval: 0.64-1.09, high intake). A stronger protective effect for these tumors was associated with dietary vitamin C intake (OR = 0.65, b.50-0.87, high intake). A significant inverse gradient in risk with retidol intake was limited to adenocarcinoma (OR = 0.64, 0.44-0.94, high intake) and more pronounced among blacks. Cancer 622267-2273, 1988.

T

HE FIRST REPORT that dietary vitamin A intake was

inversely associated with lung cancer risk was published by Bjelke. This observation was reproduced in subsequent studies that calculated an index of vitamin A or based their findings on a high consumption of green-yellow vegetables rich in car~tene.'?~ Whether a protective effect of vitamin A prevails for all types of lung cancer or is limited to some histologic subtypes and whether protection is conferred by the carotene or retinol component of total vitamin A remained topics for investigation. A Norwegian cohort study4 described a protective effect of vitamin A for squamous cell, and possibly small cell, carcinomas but not for adenocarcinoma. A case-control study of white males in Buffalo, New York, reported the same findings specific for squamous and small cell carcinoma.' Shekelle et al.' calculated separate indices of carotene and retinol intake and found carotene to be related inversely to lung cancer risk; they observed no such relationship for retinol. The issue of effective components of vitamin A has also been addressed by Samet et aL9 who, in their New Mexico study, found that the odds ratio for lung cancer risk increased as the intake of total vitamin A and carotene declined. Neither Shekelle nor Samet reported information on tumor histology. The study of New

'

From the Department of *Pathology, LSU Medical Center, New Orleans, Louisiana, the ?Population Studies Section, Environmental Epidemiology Branch, National Cancer Institute, Bethesda, Maryland, and the $Illinois Cancer Council, Chicago, Illinois. Supported by National Cancer Institute Grant NO1 CP91023. Address for reprints: Elizabeth T. H. Fontham, MD, Department of Pathology, Louisiana State University Medical Center, 1901 Perdido Street, New Orleans, LA 701 12. Accepted for publication May 5, 1988.

Jersey white males by Ziegler et a1." was the first to report on both vitamin A components and lung tumor histology. The authors saw no association between lung cancer risk and dietary retinol, The inverse relationship with carotene was most pronounced for squamous cell carcinomas; the risk of lung adenocarcinomas was unrelated to carotene intake. Wu et al.," in a study limited to white females, also failed to find an association between lung cancer and retinol, but they found that increased risk of both squamous cell carcinoma and adenocarcinoma of the lung was associated with decreased consumption of carotene. reported an inverse association Most recently Byers et between carotene and the risk of squamous cell but not small cell carcinoma; there was a suggestive inverse association with adenocarcinoma risk. We present the results of a study of 1253 lung cancer patients and 1274 controls conducted in southern Louisiana that collected data on histologic type and estimated the intake of the carotene and retinol components of vitamin A. Since fruits and vegetables are common sources of both vitamin C and carotene, we also report results for vitamin C. The consistency of our findings with those reported by other investigators is reviewed. Materials and Methods

Our study was begun in 1979 and was part of a casecontrol study of cancers of the lung, stomach, and pancreas in 26 parishes (counties) of southern Louisiana. These studies were carried out concurrentlyusing identical questionnaires and study designs which have been described previously.13314 Cases of primary lung carcinoma diagnosed from January 1979 through April 1982 were rapidly ascertained

2267

CANCERNovember 15 1988

2268

TABLE 1. Distribution of Lung Cancer Cases and Controls bv Selected Variables

Total” Race and sex WM WF BM BF Age c40 40-49 50-59 60-69 70+ Smoking history Nonsmokers Exsmoker (quit 3+ yr) Current smoker 5 2 0 cigaretteslday 2 1 + cigaretteslday Respondent status Self-respondent Next-of-kin Residence Urban Triparish area (Orleans, Jefferson, St. Bernard) Rural Family income <$1 O,OOO/year $ I O,OOO+ /year Cajun background Yes No

Cases

Controls

No. (90)

No. (96)

1253 (100)

1274 (100)

655 (52.3) 212 (16.9) 31 1 (24.8) 75 (6.0)

665 (52.2) 217 (17.0) 3 17 (24.9) 75 (5.9)

16 (1.3) 99 (7.9) 367 (29.3) 491 (39.2) 280 (22.4)

17 (1.3) 106 (8.3) 364 (28.6) 490 (38.5) 297 (23.3)

51 (4.3) 258 (21.6)

388 (31.6) 315 (25.7)

371 (31.1) 514 (43.0)

329 (26.8) 195 (15.9)

919 (73.3) 334 (26.7)

1128 (88.5) 146 ( 1 I .5)

778 (62.1)

842 (66.1)

277 (22.1) 475 (37.9)

293 (23.0) 432 (33.9)

652 (57.0) 491 (43.0)

630 (55.2) 51 I (44.8)

594 (48.2) 639 (5 I .8)

575 (45.4) 690 (54.6)

WM: white males; WF: white females; BM: black males; BF: black females. * Numbers do not add to total because of missing values.

from pathology records and admission diagnoses. All major hospitals were included in the study in the 23 parishes outside of metropolitan New Orleans and a sample representing public and private hospitals in the triparish city of New Orleans was included. These latter hospitals account for approximately one half of the lung cancer cases in the city based on incidence data from the Louisiana Tumor Registry. Controls were subjects admitted to the same hospitals and matched by race, sex, and age within 5 years. Patients whose admitting diagnosis was emphysema, chronic bronchitis, chronic obstructive pulmonary disease, or cancer of the larynx, oral cavity, esophagus, or bladder were excluded from selection as controls. These smoking-related disorders were infrequent admission diagnoses and resulted in relatively few exclusions. The distribution of controls by primary diagnosis included disorders of the (1) circulatory system, 26.6%; (2) gastrointestinal system, 16.6%; (3) genitourinary system, 10.3%; (4) musculoskeletal system, 10.6%;( 5 ) endocrine system, 5.7%;(6) trauma, 4.8%; (7) other tumors, 4.9%;(8) nervous

Vol. 62

system, 4.5%; (9) skin, 2.7%;(10)respiratory system, 2.2%; (1 1) infections, 1.2%; and (12) other, 9.9%. There was good correspondence in the distribution of cases and controls over a wide range of demographic and other variables, except for smoking history (Table 1). Interviews were completed for 1253 incident cases (76%)and 1274 of the eligible controls (85%).A next-ofkin respondent was interviewed when the subject was not available because of illness or death (26.7% of cases and 11.5% of controls). Priority for choice of next-of-kin respondent was given to spouse, then adult relatives living in the subject’s household, and then adult offspring. Methodologic studies indicate that surrogate dietary histones obtained from spouses correlate reasonably well with direct interviews of s~bjects.’~-’’ Initial analyses were conducted separately for next-of-kin and self respondents. There were no significant differences noted in vitamin consumption by respondent status and analyses with and without surrogate respondents yielded similar results. Histologic confirmation was obtained for 98% of the cases, and classification was done by local hospital pathologists. In an earlier study which validated accuracy of histologic typing of lung cancer in southern Louisiana by means of a slide review, diagnosis of lung cancer for the main categories (squamous cell, small cell, adenocarcinoma, and “other”) was found to be quite reliable.” Of 264 cases reviewed there was only one false positive, and subclassification reached approximately 90% agreement. No histologic review was considered necessary for this study. The 2% of cases lacking histologic confirmation had strong clinical evidence with radioscopic or surgical visualization of a lung mass without biopsy and were included in analyses only in the category “all lung cancers.” Table 2 shows the sex and race distribution by histologic type. An extensive questionnaire obtained information concerning diet and tobacco use, and occupational, residential, medical, and family histories. Dietary questions referred to usual adult frequency of consumption of 59 food items on a monthly basis before onset of illness or sympt o m ~ .Indices ’~ for retinol, carotene, and vitamin C intake were created by summing the frequency of consumption of each food item times the median nutrient content of a typical s e r ~ i n g . ’ ~The , * ~food items and weights used to determine the indices have been detailed p~viously.’~ The unit of measurement for carotene and retinol indices was retinol equivalents (RE) and for vitamin C was milligrams. Approximately 5% of the respondents did not know the frequency of consumption of any given food item. Subjects with missing values were excluded from the pertinent analyses. Nutrient indices were stratified into tertiles of consumption based on the control distribution, with level 1 indicating the highest tertile. The values of cutpoints for these categories are as follows: carotene in RE/month,

DIETARY VITAMINS A AND C AND LUNGCANCER

No. 10

*

Fontham et al.

2269

TABLE 2. Distribution of Lung Cancer Cases by Race, Sex, and Histologic Type

Cases [No. (%)I White

Black

Histologic type

Male

Female

Male

Female

Squamous cell carcinoma Small cell (oat cell) carcinoma Adenocarcinoma Other and carcinoma not otherwise specified No histologic confirmation

270 (41.2) 101 (15.4) 124 (18.9)

55 (25.9) 52 (24.5) 48 (22.7)

161 (51.8) 40 (12.9) 51 (16.4)

24 (32.0) 18 (24.0) 16 (21.3)

5 10 (40.7) 211 (16.8) 239 (19.1)

146 (22.3) 14 (2.2)

50 (23.6) 7 (3.3)

52 (16.7) 7 (2.2)

17 (22.7)

265 (21.2) 28 (2.2)

Total

655 (100)

212 (100)

31 I (100)

75 (100)

1253 (100)

( 1 ) 8339 and above, (2) 5237 to 8338, and (3) 0 to 5236; retinol in RE/month, (1) 32,838 and above, (2) 17,028 to 32,837, and (3) 0 to 17,027; and vitamin C in mg/month, (1) 4180 and above, (2) 2666 to 4179, and (3) 0 to 2665. The odds ratio (OR) was used to estimate the relative risk. Mantel-Haenszel” summary OR were calculated to adjust for confounders. The extension test was used to test for trend. Logistic multiple regression analysis of the data was performed using the SAS LOGIST proc e d u r e ~Confidence .~~ intervals (CI) (95%)were calculated from the logistic model. Variables examined as potential confounders included age (<40, 40-49, 50-59, 60-69, 70+), race (white, black), sex, cigarette use (pack-years), respondent status, family annual income (<$10,000, $10,000+), urban residence, and Cajun ethnicity (French spoken in home as a child).

Results The Louisiana data display the typical high risks for cigarette smokers (Table 3) and thus can provide infor-

- _

Total

mation on effects within groups at varying exposures to tobacco. An analysis of the separate data for squamous and small cell carcinomas revealed very similar patterns of relative risks for the two histologies for both dietary intake and smoking, and they have been combined for presentation here. The relative risk of lung cancer associated with dietary intake was first determined for the above-median consumption of selected food groups. No significant protective or adverse effects were associated with consump tion of dairy products, pork products, nitrite-preserved meats, seafood, all meats, or breads, grains, and cereals. Fruits and vegetables, which are major sources of carotene and vitamin C , discriminated best between the cases, in particular those with squamous and small carcinoma, and controls as seen in Table 4. For these analyses the controls were ordered into high, middle, and low tertiles of use and the cases assigned to tertiles as defined by the control series. Fruit consumption adjusted for vegetable intake remains a protective factor for all lung cancers combined (relative risk = 0.66,95%confidence interval: 0.54-0.82,

TABLE3. Relative Risk* of Lung Cancer According to History of Cigarette Use by Histologic Type

All lung cancers

Squamous and small cell carcinomas

Adenocarcinomas

No. of cases

No. of controls

OR (95%CI)

OR 195%CI)

OR (95%CI)

Never smoked Ever smoked?

51 1202

388 886

Exsmoker

258

315

Current smoker

943

57 1

1-20 cigarettes/day

37 1

329

2 1 + cigarettes/day

514

195

1.O 11.4 (8.7- 1 5.0) 6.5 (4.7-9.1) 14.2 (10.8-18.7) 9.3 (6.8-12.7) 25.3 (18.5-34.6)

1 .o 28.3 (18.643.2) 15.5 (9.3-26.0) 34.6 (22.8-52.4) 23.2 (14.6-37.0) 54.8 (35.6-89.2)

1 .o 5.6 (3.6-8.8) 3.7 (2.1-6.5) 6.7 (4.3-10.6) 4.3 (2.6-7.2) 12.0 (7.3-19.7)

* Adjusted for sex.

t Numbers by type of ever smoked do not sum to total because of missing values for amount.

CANCERNovember 15 1988

2270

TABLE4. Relative Risk* of Lung Cancer According to Level of Dietarv Intake of Fruits and Vegetables Tertiles of Consumption [OR (95% CIN Food Fruits All lung cancers Squamous and small cell carcinoma Adenocarcinoma Vegetables All lung cancers Squamous and small cell carcinoma Adenocarcinoma Fruits and vegetables All lung cancers Squamous and small cell carcinoma Adenocarcinoma

High

Moderate

Low

Pvalue

0.67 (0.52-0.86) 0.70 (0.52-0.94) 0.89 (0.58-1.36)

0.84 (0.67-1.07) 0.79 (0.60- I .04) 0.99 (0.66-1.49)

1.o

0.002

1.o

0.0 1

1.o

0.46

0.80 (0.62- 1.04) 0.80 (0.60-1.08) 0.77 (0.50-1.19)

0.78 (0.62-0.98) 0.78 (0.89-1.02) 0.64 (0.41-0.98)

1 .o

0.09

1.o

0.12

1.o

0.25

0.70 (0.55-0.91) 0.65 (0.48-0.87) 0.77 (0.49-1.21)

0.75 (0.59-0.95) 0.73 (0.55-0.96) 0.93 (0.62-1.41)

1.o

0.004

1.o

0.003

1.o

0.31

* Adjusted in a logistic regression model for race, sex, age ( t 4 0 , 4049, 50-59,60-69, 70+), pack year of cigarette use, family income, ethnic group, and respondent status.

for high consumption),but adjustment of vegetable intake for fruit lessens its protective effect (0.90, 0.74-1.1 1). Tables 5 and 6 describe the gradients in lung cancer risk according to the estimated dietary intake of carotene, retinol, and vitamin C. Relative risks have been adjusted for race, sex, age, and pack years of cigarette use; however, the effect of these adjustments was minimal. A weak inverse gradient in risk by carotene intake was seen for the total series, but this association was confined to the subset of squamous and small cell carcinomas. No relationship between carotene intake and adenocarcinomas was observed. The Louisiana findings on risk gradients for carotene were reproduced and expressed in an even stronger manner in the data for vitamin C. The vitamin C results pinpointed the association with squamous and small cell carcinomas (P< 0.001) and revealed no significantrelationship with adenocarcinomas. For all lung cancers combined and for the squamous/small cell carcinoma group, the risks associated with retinol intake were less than one; however, these estimates were not significant, and no trends in risk were apparent. When adenocarcinomas were examined separately, a statistically significant inverse gradient (P = 0.02) in risk with dietary retinol was seen. The inverse gradients by vitamin C and carotene intake were evident for both whites and blacks (Table 6). The inverse relationship between dietary intake of retinol and

Vol. 62

risk of adenocarcinoma of the lung was expressed best among blacks, and race was a significant effect modifier. A significant trend (P < 0.05) was observed for blacks while the trend for whites did not achieve statistical significance. When the data were examined by gender, an inverse trend by level of intake of carotene and vitamin C (P< 0.05) was noted for men. Among women the effect of carotene and vitamin C was modified by smoking. The inverse trend in risk associated with intake of these vitamins was not apparent for women who were heavy smokers, but it was similar for men and women in the lighter smoking categories. The retinol effects associated with adenocarcinoma were also similar in men and women. The similar carotene and vitamin C findings in this study are not surprising, given the correlation ( r = +0.64) in the carotene and vitamin C values recorded for individual cases and controls. Individual foods containinghigh levels of vitamin C but not carotene (e.g., strawberries, OR = 0.70, 95% CI 0.58-0.84), high levels of carotene but not vitamin C (e.g., carrots, OR = 0.58, 0.48-0.70), or high levels of both (e.g., broccoli, OR = 0.64, 0.540.78) showed similar patterns of risk for above median intake. Because of high correlations among the nutrients we examined, the pattern of risk for each nutrient adjusted for the others. Table 7 displays the combined risk of squamous and small cell carcinomas associated with dietary levels of carotene and vitamin C. Adjustment for vitamin C eliminates the protective effect of carotene, but adjustTABLE 5. Relative Risk* of Lung Cancer by Level of Dietary Intake of Carotene, Vitamin C, and Retinol Tertiles of Consumption [OR (95% CI)] Vitamin Carotene All lung cancers Squamous and small cell carcinoma Adenocarcinoma Vitamin C All lung cancers Squamous and small cell carcinoma Adenocarcinoma Retinol All lung cancers Squamous and small cell carcinoma Adenocarcinoma

High

Moderate

Low

Pvalue

0.88 (0.70-1.1 1) 0.84 (0.64-1.09) 0.94 (0.63-1.39)

0.96 (0.76-1.20) 0.92 (0.71-1.20) 1.oo (0.68-1.47)

1.o

0.29

1 .o

0.19

1.o

0.75

0.67 (0.53-0.84) 0.65 (0.50-0.87) 0.83 (0.55- 1.24)

0.87 (0.70- 1.09) 0.89 (0.69-1.16) 0.98 (0.67-1.44)

1.o


1 .o

<0.001

1.o

0.36

0.89 (0.72-1.12) 0.90 (0.69-1.17) 0.64 (0.44-0.94)

0.85 (0.68- 1.06) 0.86 (0.66-1.12) 0.67 (0.46-0.97)

1.o

0.32

1.o

0.42

1.o

0.02

* Adjusted in logistic regression model for age, race, sex, and pack years of cigarette use.

DIETARY VITAMINS A AND c AND LUNGCANCER

No. 10

Fontham el d.

227 1

TABLE 6. Relative Risk of Lung Cancer by Level of Dietary Intake of Carotene, Vitamin C, and Retinol for Selected Subgroups Squamous and Small Cell Carcinoma [OR (95% CI)] Carotene

Vitamin C

Females Race? Whites Blacks

Retinol

High

Moderate

Low

High

Moderate

Low

High

Moderate

Low

0.80 (0.60-1.09) 1.04 (0.50-2.06)

0.85 (0.64-1.13) 1.36 (0.69-2.70)

1.o

0.6 1 (0.45-0.83) 1.16 (0.56-2.40)

0.8 1 (0.6 1-1.08) 1.80 (0.90-3.57

1.O$

0.67 (0.44-1.04) 0.57 (0.26-1.22)

0.70 (0.45-1.08) 0.59 (0.29-1.23)

1.o

0.88 (0.63-1.23) 0.73 (0.46-1.16)

0.95 (0.69-1.30) 0.82 (0.51-1.33)

0.70 (0.50-0.98) 0.54 (0.33-0.89)

0.93 (0.68-1.28) 0.79 (0.50-1.26)

1.O$

0.73 (0.47- I . 13) 0.45 (0.21-0.95)

0.68 (0.44-1.05) 0.6 1 (0.29-1.27)

Sex*

Males

Adenocarcinoma [OR (95% CI)]

1.o 1.o 1.o

* Adjusted for race, age, pack years. t Adjusted for sex, age, pack years.

1.o

1.O$

1.o

1.o 1.O$

$ P value for trend <0.05.

ment for carotene has virtually no effect on the strong protective effect associated with vitamin C intake, a finding which parallels that for fruits and vegetables. Because the effects of total vitamin A (and carotene) intake in an earlier study were strongest among exsmokers, suggesting that dietary vitamin A might be most effective against weaker carcinogens, we examined this point? Logistic modeling detected no significant differences in the protective effects of carotene or vitamin C for squamous or small cell carcinomas by smoking category (ever/never, current/ex, light/heavy) although the protective effects appeared stronger for light smokers. Similarly lower risks for high retinol intake were seen for adenocarcinoma for all smoking categories. Overall no protective effects of carotene or vitamin C were seen for adenocarcinoma, although a decreasing trend in risk with increasing carotene intake was seen among light smokers. Discussion

The Louisiana lung cancer study produced several findings of note. There was a suggestive inverse relation-

ship of dietary carotene intake with lung cancer risk specific for squamous and small cell carcinomas. An inverse gradient in risk of squamous and small cell carcinomas was also associated with dietary vitamin C intake, and this association was stronger than that described for carotene. An inverse gradient in risk with dietary retinol was limited to adenocarcinomas. This significant inverse association of retinol and adenocarcinoma and the protective effects of carotene and vitamin C were apparent for all categories of smokers. Ziegler et al.lo were the first to collect information on intake of both vitamin A components and histologic type and to report an effect for carotene limited to squamous cell carcinoma. The results of our study, which also collected data on vitamin A components and histology, appear consistent with theirs. Both studies show a statistically significant reduced risk for squamous cell carcinoma in the high intake category on the order of 25% to 30%.The protective effect of carotene for white females noted by Wu et al. I was not confined to squamous cell carcinoma. Other papers have reported protective effects limited to squamous and small cell carcinomas when cases and con-

TABLE7. Relative Risk of Squamous and Small Cell Carcinoma of the Lung According to Level of Dietary Vitamin C and Carotene Intake Vitamin C intake [OR (95% CI)] Carotene intake High Moderate LOW

Crude Adjusted for carotene

High

Moderate

LOW

Crude

Adjusted for Vitamin C

0.60 (0.44-0.81) 0.58 (0.39-0.86) 0.38 (0.14-1.04) 0.56 (0.44-0.72) 0.52 (0.38-0.72)

0.75 (0.50-1.12) 0.82 (0.60-1.13) 0.83 (0.59-1.16) 0.78 (0.63-0.98) 0.78 (0.62-0.99)

1.07 (0.64-1.80) 1.08 (0.77-1.5 1) 1.o

0.75 (0.59-0.95) 0.9 1 (0.72-1.14) 1 .o

1.03 (0.75-1.42) 1.06 (0.83-1.36) 1.o

1.o 1.o

2272

CANCERNovember 15 1988

trols were classified by total vitamin A intake.4,7The New Jersey and Louisiana results suggest that the protective effects of total vitamin A for epidermoid cancers described by Kvale4 and Byers' reflect the contribution of the carotene component. Other studies have described an inverse relationship between carotene intake and all lung cancers irrespective of histologic type.8,99'3324 Since squamous and small cell carcinomas constitute more than half of all lung cancers in American men, the results for carotene in these studies probably reflect protection against these two histologic types of carcinoma. The Louisiana description of an inverse gradient in risk linked to vitamin C is surprising in view of the failure of several earlier studies by Hinds and K0lone1~~ and Byers et aL7,I2to detect a vitamin C effect. More recently, Ziegler et aL2' commented that their New Jersey study did not reveal strong inverse associations with vitamin C-containing foods; however, a major contributor to vitamin C intake, orange juice, was not included in their questionnaire. Lower levels of vitamin C in the Louisiana population relative to other American populations may account for regional differences in findings. In a nationwide food consumption survey conducted in 1977-78, food and nutrient intakes were determined for four broad regions of the United States.26Average fruit intake in the South was approximately 30%lower than in the other three regions, and vitamin C intake was also reduced compared with the other regions. Serum levels of vitamin C were determined for a sample of Louisiana residents in the TenState Nutrition Survey.27For low-income Louisiana residents, 15% were found to be deficient and low in vitamin C, and 9% with higher incomes were low or deficient in vitamin C. A strong protective effect was found for fruit consumption and dietary vitamin C on gastric cancer risk in the southern Louisiana p~pulation.'~ Low fruit intake was also associated with an increased risk of pancreatic cancer among these people.28These findings suggest that the protective effect associated with vitamin C may only be expressed in populations with relatively low intake. The Louisiana results for vitamin C should be viewed with reserve given the absence of confirmatory data from other studies and the equivocal findings in animal studies. Our finding, of course, might be the spurious result of some unique, nondietary factor operating in the study population. No evidence of a confounding factor was apparent, however, when respondent status, urban/rural residence, Cajun ethnicity, income, or other variables were examined. The magnitude and internal consistency of the Louisiana results for vitamin C and the histologic specificity and presence of plausible reasons as to why an effect may appear in some, but not all, studies suggests that efforts to elaborate the result should continue.

Vol. 62

The association of retinol with a decreased risk of adenocarcinoma was another unanticipated finding. The New Jersey study observed no such association between adenocarcinoma and retinol; however, only 16% of the 763 cases in this study were classified as adenocarcinomas." Risk gradients for retinol intake were similar for categories of carotene and vitamin C intake, indicating that the retinol effect is not dependent on consumption of the other nutrients. The retinol index in Louisiana was dominated by the contribution of organ meats (beef, calf, and chicken liver). The correlation between the retinol index and the retinol provided by consumption of organ meats in this study, as measured by Spearman's r, was near unity, r = 0.98, P < 0.000 1. The association of retinol with decreased lung cancer risk was most pronounced in blacks, and the black study subjects, on average, ate organ meats twice as frequently as whites. It is possible that the dietary pattern in this black population provided the opportunity for expression of a retinol effect which would not be noted in a population with low organ-meat consumption, resulting in a relatively low retinol index with a narrower range of values. It may be also that some component of organ meats other than retinol was responsible for the effect noted; these include riboflavin, niacin, folic acid, or other nutrients. Before retinol can be considered a candidate as a protective factor, replication of the Louisiana results in other data sets is needed. The results of our study support earlier findings of a protective effect of dietary carotene specific for squamous and small cell carcinoma. A stronger protective effect for these histologic types was found for dietary vitamin C in our population, and there was a protective effect for retinol in adenocarcinoma of the lung. This study in southern Louisiana may have provided a unique opportunity to examine the protective effects of vitamins A and C in a population with a wider variation of intake than is generally seen in the more homogenous American diet. REFERENCES 1 . Bjelke E. Dietary vitamin A and human lung cancer. Int J Cancer 1975; 15:561-565. 2. Mettlin C, Graham S, Swanson M. Vitamin A and lung cancer. JNCI 1979; 62:1435-1438. 3. Gregor A, Lee PN, Roe FJC, Wilson MJ, Melton A. Comparison of dietary histories in lung cancer cases and controls with special reference to vitamin A. Nutr Cancer 1980; 2:93-97. 4. Kvale G, Bjelke E, Cart JJ. Dietary habits and lung cancer risk. Int JCancer 1983; 31:397-405. 5. MacLennan R, DaCosta J, Day NE, Law CH, Ng YK, Shanmugaratnam K. Risk factors for lung cancer in Singapore Chinese, a pop ulation with high female incidence rates. Znt J Cancer 1977; 20854860. 6. Hirayama T. Diet and cancer. Nutr Cancer 1979; 1:67-8 I . 7. Byers T, Vena J, Mettlin C, Swanson M, Graham S. Dietary vitamin A and lung cancer risk An analysis by histologic subtypes. Am J Epidemiol 1984; 120~769-76.

No. 10

DIETARY VITAMINS A AND c AND LUNGCANCER

8. Shekelle RB, Leppler M, Liu S ef al. Dietary vitamin A and risk of cancer in the Western Electric Study. Lancet 1981; 2: 1185-1 190. 9. Samet JM, Skipper BJ, Humble CG, Pathak DR. Lung cancer risk and vitamin A consumption in New Mexico. Am Rev Respir Dis 1985; 13 1 198-202. 10. Ziegler RG, Mason TJ, Stemhagen A ef al. Dietary carotene and vitamin A and risk of lung cancer among white men in New Jersey. JNCI 1984; 73:1429-1435. 11. Wu AH, Henderson BE, Pike ME, Yu MC. Smoking and other risk factors for lung cancer in women. JNCZ 1985; 74747-751. 12. Byers TE, Graham S, Haughey BP, Marshall JR, Swanson MK. Diet and lung cancer risk Findings from the western New York diet study. Am J Epidemiol 1987; 125:351-363. 13. Correa P, Pickle LW, Fontham E et al. The causes of lung cancer in Louisiana. In: Mizell M, Correa P, eds. Lung Cancer: Causes and Prevention. Deerfield Beach, F L Verlag Chemie International, 1984; 73-82. 14. Correa P, Fontham E, Pickle LW, Chen V, Lin Y, Haenszel W. Dietary determinants of gastric cancer in south Louisiana. JNCZ 1985; 75~645-653. 15. Kolonel LN, Hirohata T, Nomura AMY. Adequacy of survey data collected from substitute respondents. Am J Epidemiol 1977; 106: 476-484. 16. Marshall J, Wore R, Haughey B, Rzeska T, Graham S. Spousesubiect interviews and the reliability of diet studies. Am J Epidemiol 1980; 112:675-683. 17. Himble CG. Samet JM. SkiDoer BE. ComDarison of self- and surrogate-reported dietary info&atib;l. Am J Epid&ol 1984; 1 1923698.

-

Fontham d d.

2273

18. Rothschild H, Buechner H, Welsh R, Vial U,Weinberg R. Histologic typing of lung cancer in Louisiana. Cancer 1982; 49: 1874-1877. 19. Pickle LW, Hartman AM. Indicator foods for vitamin A assessment. Nutr Cancer 1985; 7:3-23. 20. Composition of Foods. Agriculture Handbook No. 456. Washington, DC:US Government Printing Office, 1975. 2 1. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. JNCI 1959; 22:7 19-748. 22. Mantel N. Chi-square tests with one degree of freedom: Extensions of the Mantel-Haenszel procedure. J Am Stat Assoc 1963; 58:690-700. 23. Harrell F. The LOGIST procedure. In: Reinhardt PS, ed. SAS Supplemental Library User’s Guide. Cary, NC: SAS Institute, 1980 83102. 24. Hinds MW, Kolonel LN. Dietary vitamin A, carotene, vitamin C and risk of lung cancer in Hawaii. Am J Epidemiol 1984; 119:227237. 25. Ziegler RG, Mason TJ, Stemhagen A et al. Carotenoid intake, vegetables and the risk of lung cancer among white men in New Jersey. Am J Epidemiol 1986; 123:1080-1093. 26. Food and Nutrient Intakes: Individuals in Four Regions, 197778. NFCS 1977-78. Report No. 1-3. Hyattsville, MD, U. S. Department of Agriculture, 1985; 18-3 I . 27. Ten-State Nutrition Survey 1968-70, U. S . Department of Health, Education and Welfare Publ. (HSM), 72-8 129,72-8130,72-8 13 I, 728132,72-8133. 28. Falk RT, Pickle LW, Fontham E, Correa P, Fraumeni J. Lifestyle and pancreatic cancer in Louisiana: A case-control study. Am JEpidemiol 1988; 324-336.