A Case-Control Assessment of Risk Factors for Gallbladder Carcinoma

Digestive Diseases and Sciences, Vol. 44, No. 8 (August 1999), pp. 1619 ± 1625

A Case± Control Assessment of Risk Factors for Gallbladder Carcinoma THAYER E. SCO TT, MPH, MITCHELL CARROLL, MD, FRANCIS D. CO GLIANO, MD, BERNARD F. SMITH, MD, and WAYNE W. LAMO RTE, MD, PhD, MPH

Gallbladde r carcinom a is an uncommon, but highly fatal disease . Its symptoms fre que ntly mirror those of gallstone dise ase , and in most instance s, diagnosis is an incide ntal ® nding at surge ry. While risk factors have be e n sugge sted for this cancer, many may in re ality simply be a conseque nce of the olde r age of the population. This study is one of the fe w to approach this que stion by using a case± control study design comparing gallbladde r carcinoma patie nts with a gallstone population, couple d with multivariate analysis to de te rmine age -inde pe nde nt risk factors. Univariate analyse s showe d gallbladde r carcinoma patie nts to be olde r than gallstone patie nts and to have many age -associate d dise ase s. Following multiple re gre ssion adjustme nt for age , this disease was associate d with fe male gende r and with a pre vious history of gallstone symptoms. Carcinoma patie nts were le ss like ly to have chole ste rol gallstone s in the ir gallbladde rs at surge ry. A previous history of smoking was a substantial risk but of borde rline statistical signi® cance . Previous studie s report associations that may be due to the olde r age of the gallbladde r carcinoma patie nt. O ur re sults show that afte r adjusting for age with multivariate analysis, gallbladde r cance r subje cts were predominantly female , more like ly to re port pre vious gallstone symptom ology, and to smoke . While gallstone s were not unive rsally isolate d from carcinoma patie nts at chole cyste ctomy, when pre sent, they were less freque ntly classi® ed as chole sterol gallstone s base d on visual inspe ction. Furthe r cohort studie s which targe t these populations will allow us to gain a more solid consensus on the risk factors for this dise ase . KEY WORDS: gallbladde r carcinoma; risk factors; gallstone s; case ± control study; multivariate analysis.

Gallbladde r carcinoma is an uncom mon but highly fatal disease that has be e n linke d to gallstone s. Le ss than 0.5% of the population with gallstone s will be af¯ icte d with it (1), but the 5-ye ar survival rate is le ss Manuscript re ceive d July 2, 1997; re vised manuscript re ce ived May 21, 1998; accepte d February 10, 1999. From the Departmen ts of Surgery and Gastroe nterology, Boston Unive rsity School of Me dicine, Boston, Massachusetts; Carne y Hospital, Dorcheste r, Massachuse tts; Brockton Hospital, Brockton, Massachuse tts; and South Shore Hospital, We ymouth, Massachusetts. Institutional review board approval was obtained from the following institutions: Unive rsity HospitalÐ Nove mber 23, 1993; Carney HospitalÐ March 3, 1994; Brockton HospitalÐ April 15, 1994; Cape Cod HospitalÐ October 6, 1994; South Shore HospitalÐ May 26, 1994; and Quincy City HospitalÐ De ce mber 23, 1994. Address for reprint requests: Dr. Wayne W. LaMorte , Surgical Re se arch Se ction, Boston University Medical School, 80 East Concord St., Boston, Massachuse tts 02118.

than 5% (2). The only inte rvention known to have a de ® nitive impact on gallbladde r carcinoma is prophylactic re moval of the gallbladde rÐ an approach that would not be feasible for a population of this size, e ven with the adve nt of laparoscopic chole cyste ctomy. Yet, de spite over two centurie s of scrutiny, the causative factors re main elusive and it is not possible to identify individuals at high risk (3, 4). Most studie s sugge st that the individuals with the greate st risk are olde r, obe se female s; howeve r, these characte ristics also de scribe the population with gallstone s, the vast majority of whom do not progre ss to deve lop gallbladde r carcinom a. A pre fe rable approach for ide ntifying risk factors is to compare carcinoma patie nts with a suitable control

1619

Digestive Diseases and Sciences, Vol. 44, No. 8 (August 1999) 0163-2116/99/0800-1619$16.00/0 Ñ

1999 Plenum Publishing Corporation

SCOTT ET AL

population, such as those with gallstone s, in a case ± control study. Seve ral such studie s have be en pe rforme d to date ; howeve r, fe w have use d multivariate analysis to e xamine the risk re lationships inde pe nde nt of age . We conducte d a matche d hospital-base d re trospe ctive case ± control study design in an atte mpt to furthe r de ® ne the risk pro® le of the individual af¯ icte d with gallbladde r carcinoma. MATERIALS AND METHODS We selecte d our study population from hospitals in southeaste rn Massachusetts located betwe en Boston and Cape Cod. This network included the following hospitals: Unive rsity Hospital, Carney Hospital, Brockton Hospital, Cape Cod Hospital, South Shore Hospital, and Quincy Hospital. Following approval from the respective institutional review boards, either the medical records department or the cancer registry was used to identify all patie nts discharged with a diagnosis of primary gallbladder carcinoma (ICD code, 9th revision: 156.0) ove r the 12-ye ar period of January 1983 to Decembe r 1994. All primary gallbladder carcinomas we re con® rmed by surgical specimen pathology report. While all subjects e ntered surgery with the intention of undergoing a cholecystectomy, in a fe w situations the carcinoma proved to be so invasive that only palliative surgery was possible. As no gallbladder specimen was available for the se patients, a pathology report based on e ither a biopsy of the gallbladder or anothe r organ such as the liver was used to con® rm the diagnosis. Patients with a pathology report indicating primary gallbladder carcinoma or probable primary gallbladder carcinoma we re included in the study, but questionable cases were e xcluded. The control population was de® ned as patients with cholelithiasis (ICD code, 9th revision: 574.0 ± 574.5) who underwent either a traditional cholecystectomy (ICD code, 9th revision: 51.22) or a laparoscopic cholecyste ctomy (ICD code, 9th revision: 51.23) for uncomplicate d cholelithiasis. All patients who me t the se criteria and atte nded the same hospital ove r the identical 12-ye ar time frame as the cases were e ligible. To control for any improvements in the rapy or dete ction that may have occurred ove r the 12 ye ars, each case of gallbladder cancer was matched to the four patie nts without carcinoma who had the neare st admission date . In all instances, the se patie nts ente red the same hospital within three months of the ir corresponding case. No atT ABLE 1. D EMOGRAPHIC

AND

tempts we re made to match in any other manne r, such as by age or gender, so that these relationships could be addressed as potential risk factors in the analysis stage . Anticipating that some control records would be unusable because of incomplete information, we initially matched e ach case with the 10 neare st controls by admission date. Grossly unusable records, eg, those lacking a recorded history, physical e xam, or a pathology report we re e xcluded. From the usable records, we randomly selected four controls for e ach case. Each patie nt’s me dical record was extracte d for information on patient demographics, past and present me dical histories, smoking and alcohol intake, and pathology reports for the gallbladder, including presence, number and composition of gallstone. Medical histories we re obtained from the preoperative history, which for cases preceded cancer diagnosis. Gallstone composition was based on visual e xamination by the various hospital pathologists without further interpretation or chemical analysis. Admissions for reasons other than the cholecyste ctomy we re also e xtracte d to obtain information such as blood type or parity, which may have not bee n reported in the cholecyste ctomy record. Case s and controls we re initially compared using t te st, Mante l-Hae nszel chi square, or logit e stimator te sts using 1:4 matching where appropriate . To simultaneously assess the impact of various risk factors, a matched multivariate analysis was performed e mploying the conditional logistic regression me thod for 1:4 matching. The initial multivariate model was a simple age -adjusted relationship that individually te ste d all variable s. Subsequent models used forward stepwise multivariate model te chniques that included age and gender and te ste d all variables separate ly (Tables 1± 4 below) until e very signi® cant factor had been identi® e d. This ® nal model was rete sted by individually forcing all nonsigni® cant variables into the model to assure comprehensiveness.

RESULTS Univariate Analyses . O ur record re vie ws e ncompasse d a 12-ye ar time frame and accrue d 68 case s with gallbladde r carcinoma (gallbladde r carcinoma case s). Following matching by hospital and date of admission, 272 controls with chole lithiasis who had unde rgone a chole cyste ctomy were use d for comparison (gallstone controls) . As de monstrate d in Table 1,

SOCIOECONOMIC FACTORS A SSOCIATED WITH G ALLBLADDER C ARCINOMA D ISEASE: U NIV ARIATE AND A GE -ADJUSTED A NALYSES Odds ratio

Age (years 6 SE) Female gender Income ($) Insured White race

1620

G allbladder carcinom a (N 5 68)

G allstones (N 5 272)

Crude

74.4 6 1.3 51 (75% ) 33K 6 1K 68 (100% ) 64 (94% )

55.2 6 1.1 189 (69% ) 34K 6 0.5K 257 (94% ) 250 (92% )

1.09/yr 1.3 1.000/$ 1.6 1.5

95% con® dence interval 1.06± 0.7± 1.000± 0.5± 0.4±

1.1 2.4 1.000 4.5 5.8

OR

G ALLSTONE Odds ratio

P 0.0001 0.4 0.7 0.045 0.5

Age-adjusted

P

1.4 1.000

0.4 0.3 1.0 0.8

1.2

Digestive Diseases and Sciences, Vol. 44, No. 8 (August 1999)

RISK FACTORS FOR GALLBLADDER CARCINOMA T ABLE 2. FACTORS A SSOCIATED

G ALLBLADDER C ARCINOMA AND G ALLSTONE D ISEASE U NIVARIATE AND A GE -ADJUSTED A NALYSES

WITH

FROM

MEDICAL H ISTORY :

Odds ratio G allbladder carcinom a (N 5 68) 2

Obesity (kg/m ) Hype rte nsion Diabe tes He art disease In¯ ammatory bowel disease He patitis Pe ptic ulce r Hiatal hernia Cancer other than gallbladde r Past radiation or chemothe rapy Past history of gallstone s

28 6 1.0 38 (56% ) 11 (16% ) 22 (32% ) 2 (3% ) 7 (10% ) 10 (15% ) 7 (10% ) 19 (28% ) 7 (11% ) 3 (4% )

G allstones (N 5 272) 28 116 22 45 9 18 49 52 34 11 21

6 0.4 (43% ) (8% ) (17% ) (3% ) (7% ) (18% ) (19% ) (13% ) (4% ) (8% )

the mean age for our gallbladde r carcinoma patie nts was 74 ye ars, almost 20 years olde r than our controls (55 years of age ). While both patie nts and controls were for the most part female , the female pre dominance was some what greate r among the cance r cases, but not signi® cantly gre ater than in controls. Access to the health care syste m, as asse ssed by re port of medical insurance cove rage , was unive rsal in the carcinoma patie nts and almost as fre que nt for the gallstone group. No differe nce s were detected in racial makeup betwe en the two groups, with both the carcinoma patie nts and the controls be ing pre dominantly white . Differe nce s in the fre que ncy of othe r dise ase s are pre sented in Table 2. While dieting within the sixmonth pe riod prior to admission was more pre vale nt in the gallstone group (odds ratio 5 0.8, P 5 0.08) , obe sity was comparable be twee n the two groups. Not T ABLE 3. D IFFERENCES

IN

Crude

B. Smoking and ethanol e xposures History of smoking Curre nt smoke rs Smoked in past Ye ars smoke d 6 SE Ye ars since quit 6 SE History of more than 3 drinks/day

95% con® dence interval

0.99 1.7 2.2 2.4 0.9 1.6 0.8 0.5 3.0 2.9 0.6

0.96± 1.008± 1.01± 1.3± 0.2± 0.7± 0.4± 0.2± 1.5± 1.1± 0.2±

1.04 3.0 4.6 4.3 4.1 3.7 1.6 1.1 5.8 7.9 1.8

AND

P

0.9 0.047 0.046 0.004 0.9 0.3 0.5 0.07 0.001 0.04 0.3

1.03 0.7 1.3 0.96 1.2 1.5 0.6 0.4 1.08 1.5 0.6

0.2 0.3 0.6 0.9 0.8 0.5 0.2 0.4 0.9 0.5 0.5

Odds ratio G allstones (N 5 187)

Crude

95% con® dence interval

2.7 6 0.4 0 (0% ) 5 (10% )

2.3 6 0.2 7 (4% ) 21 (11% )

1.1/birth 0.5 0.9

0.94± 1.3 0.1± 2.3 0.3± 2.6

N 5 64 41 (62% ) 12 (18% ) 41 (87% ) 33 6 2.7 7.9 6 1.5

N 5 271 158 (59% ) 90 (33% ) 161 (77% ) 22 6 1.3 4.4 6 0.6

1.1 0.4 1.9 1.04/yr 1.03/yr

0.6± 0.2± 0.7± 1.02± 0.99±

Digestive Diseases and Sciences, Vol. 44, No. 8 (August 1999)

Age-adjusted

G ALLSTONE D ISEASE : U NIVARIATE

G allbladder carcinom a (N 5 51)

7 (10% )

P

surprisingly, dise ase s of the elde rly such as hype rte nsion, diabe te s mellitus, and heart dise ase were more common among the carcinoma case s. The cance r patie nts were also three times more like ly to re port a pre vious cance r othe r than their curre nt gallbladde r cance r. This ® nding is supporte d by the ir freque nt re port of past chemotherapy and radiation therapy. Similar freque ncies of dise ase s of the gastrointe stinal tract emerge d, with the exception of a borde rline pre dominance of hiatal he rnias in the gallstone patie nts. While past historie s of weight loss, gallstone s (re corded as a history of gallstone s or gallstone symptoms in the medical history) , and pancre atitis were not statistically diffe rent, the gallbladde r carcinoma patie nts were three times more like ly to re late a past attack of jaundice . No associations betwee n gallbladde r carcinoma and blood type were found. Gende r re late d risk factors were also e xamine d. As

E XPOSURES B ETWEEN G ALLBLADDER CARCINOMA A GE -A DJUSTED A NALYSES

A. Fe male e xposures Births 6 SE Oral contraceptive use Hormone replacem ent

Odds ratio

34 (13% )

0.8

1.9 0.8 4.9 1.06 1.07

0.3± 1.9

AND

Odds ratio P

Age-adjusted

P

0.3 0.1 0.8

1.0

0.9

1.2

0.8

0.7 0.01 0.2 0.001 0.09

1.4 0.9 1.9 1.01 1.0

0.3 0.8 0.2 0.4 0.4

0.6

0.8

0.6

1621

SCOTT ET AL T ABLE 4. FACTORS A SSOCIATED

WITH

G ALLBLADDER C ARCINOMA AND G ALLSTONE D ISEASE A GE -A DJUSTED A NALYSES

FROM

P ATHOLOGY R EPORT: U NIV ARIATE

Odds ratio G allbladder carcinom a (N 5 57) A. Gallbladde r pathology He morrhagic Gangrenous or ulcerous Porcelain Choleste rolosis B. Gallstone pathology Choleste rol Pigme nt Brown Mixed C. Stone number and size of largest stone 0 gallstones 1 gallstone 2± 25 gallstones . 25 gallstone s Size (cm 6 SE)

(11% ) (7% ) (1% ) (9% ) 249 (53% ) (14% ) (4% ) (29% )

Crude

5 (9% ) 3 (5% ) 0 (0% ) 0 (0% ) N 5 45 14 (31% ) 8 (18% ) 2 (4% ) 21 (47% )

30 19 2 24 N5 133 34 9 73

7 (12% ) 14 (24% ) 18 (31% ) 20 (34% ) 1.9 6 0.2

4 (1% ) 13.7 50 (19% ) 1.5 81 (30% ) 1.1 135 (50% ) 0.4 1.2 6 0.06 1.6/cm

Table 3 illustrate s, our study groups reporte d similar mean birth rate s, with 2.3 births re porte d by patie nts and 2.7 births for controls. We were also unable to de te ct any in¯ ue nce for oral intake of fe male hormone s, e ithe r as oral contrace ptive s or as hormone re place ment the rapy. Tobacco usage was assesse d with se veral diffe rent approache s. Any e xposure to tobacco smoking, indicated as eithe r past or prese nt use , was similar in case s and controls. Be cause tobacco risk may play two role s in this diseaseÐ eithe r as an initiator in the past, or as a promote r more current to the diagnosisÐ we e xamine d these re lationships more close ly. Among those subje cts whose prese nt smoking status was re corde d (N 5 327) , the gallstone controls were significantly more like ly to be curre nt smokers. In contrast, among smokers who had smoked in the past and had quit, a nonsigni® cant tre nd towards more pre vious e xposure was note d in carcinoma subje cts (N 5 187) . Gallbladde r carcinoma was signi® cantly associate d with longe r smoking e xposure s, and borde rline associate d with more re mote cessation date s (N 5 160) . Elevate d ethanol intake , de® ned as inge stion of thre e or more drinks pe r day or a re port of alcohol abuse , was similar betwe en the two groups (N 5 327). Possible pre dictors for gallbladde r carcinoma, such as patie nt symptomology at admission and indications for surge ry, were compare d betwe en the two groups. The carcinoma patie nts were 4.5 time s more to have complaine d of anore xia and also more like ly to have had fatigue and jaundice . Pre sence of an abdominal mass occurred e xclusive ly in the gallbladde r carci-

1622

G allstones (N 5 272)

95% con® dence interval

AND

Odds ratio P

Age-adjusted

P

0.6 0.7 0.8 0.6

0.2± 0.2± 0.06± 0.3±

1.7 2.9 11 1.6

0.3 0.7 0.5 0.02

0.7 0.6

0.5 0.4

0.4 1.5 0.9 2.5

0.2± 0.6± 0.2± 1.2±

0.8 3.4 4.1 5.3

0.008 0.4 0.9 0.02

0.4 0.5 1.1 1.3

0.02 0.2 0.9 0.5

4.0± 0.7± 0.6± 0.2± 1.2±

47 3.1 2.1 0.8 2.1

0.001 0.3 0.9 0.01 0.002

7.6 1.1 0.8 0.4 1.2

0.008 0.8 0.6 0.02 0.2

noma cases. O the r symptoms such as nause a, pancre atitis, and re cent weight loss occurre d with similar fre que ncie s in the two groups. Gallstone patie nts re porte d that the ir prese nt onse t of symptoms pe rsiste d for an ave rage of 11 months, longe r than the carcinom a patie nt’s experience of 5 months. The primary indication for surge ry was the alle viation of symptoms attribute d to gallstone s. Almost all of the gallstone control patie nts (92% ) unde rwent surge ry for gallstone s; howe ver, only 64% of the carcinoma patie nts were ope rated on for that reason. O f the re maining carcinoma patie nts, 23% of the surge rie s were for suspicion of cance r and 11% were to re lieve an obstruction. Pathology ® ndings are summarized in Table 4. Porcelain gallbladde rs, fre que ntly sugge ste d as an important risk factor for gallbladde r carcinoma, were found e xclusive ly in the control patie nts. Chole sterolosis, he morrhagic gallbladde rs, and gangre nous or ulce rous gallbladde rs were also more freque nt in the gallstone controls. Not all gallbladde r specime ns submitte d to pathology include d gallstone s. Gallstone s were isolate d from 88% of the carcinoma gallbladde rs at surge ry, and in 98% of the control population. In the re maining 2% of controls, gallstone s were visualize d pre ope rative ly, but gallstone count, size , and composition were not re corded in the pathology re port. For those spe cimens containing gallstone s, visual inspe ction sugge sted the pre sence of chole sterol gallstone s in 53% of the gallstone controls, and in 31% of the carcinoma patie nts. In contrast, mixed gallstone s were more pre vale nt in the gallbladde rs with carciDigestive Diseases and Sciences, Vol. 44, No. 8 (August 1999)

RISK FACTORS FOR GALLBLADDER CARCINOMA T ABLE 5. D IFFERENCES B ETWEEN G ALLBLADDER C ARCINOMA AND G ALLSTONE D ISEASE: M ULTIPLE LOGISTIC R EGRESSION A NALYSIS (N 5 228) Variable

Odds ratio

Age (yr) Fe male ge nde r Cholesterol gallstones History of gallstone symptoms Smoker in past

1.1/yr 8.3 0.25 17.2 4.8

95% con® dence interval 1.05± 1.8± 0.07± 1.5± 0.94±

1.16 39 0.85 190 25

P 0.0002 0.007 0.03 0.02 0.059

noma. These patie nts’ gallstone s were not as nume rous; howe ver, the ir large st gallstone was of gre ate r size than those of the gallstone controls. Mu ltivar iate An alys es. Many of the appare nt associations ide nti® e d in the univariate (crude) analyse s could be age -re late d. Inde e d, as the far right side s of Table s 1± 4 demonstrate , adjusting for age differe nce s re nde red many of the associations nonsigni® cant. Spe ci® cally, chronic dise ase s, smoking historie s, cholesterolosis, size of large st gallstone , and insurance status were no longe r associate d with gallbladde r cance r. Furthe r multivariate relationships were e xamined using stepwise multiple logistic re gre ssion analysis, which include d age and prese nce or abse nce of gallstone s, and tested individually all othe r pote ntial risk factors. Table 5 summarize s the results of a multivariate regression mode l comparing the gallbladde r carcinoma cases with the gallstone controls. This analysis indicate s that the gallbladde r carcinoma patie nts were olde r female s whe n compare d to the subje cts without carcinoma. Afte r adjusting for age , cance r patie nts were more like ly to have bee n symptomatic. The se patie nts were le ss incline d to have chole sterol gallstone s isolate d from the ir gallbladde rs at surge ry. No othe r gallstone type de monstrate d a statistically signi® cant re lationship, eithe r positive or ne gative , to this disease. Size of the large st gallstone re covere d at surge ry also did not reach statistical signi® cance in any tested model. Two othe r factors were of borde rline signi® cance. Gallbladde r carcinoma patie nts were approximate ly ® ve time s more like ly to report a pre vious smoking history than was the gallstone population. Hiatal he rnia was more fre que nt in the gallstone controls. There was no evide nce of interaction be twee n age and gende r, smoking and alcohol, or gallstone size and gallstone numbe r. DISCUSSION In this study, patie nts with gallbladde r carcinoma were on ave rage 20 ye ars olde r than controls with Digestive Diseases and Sciences, Vol. 44, No. 8 (August 1999)

gallstone s. Not surprisingly, many of the gallbladde r cance r case s had age associate d diseases such as hype rte nsion, diabe te s mellitus, and cardiovascular dise ase . Afte r adjusting for age with multivariate analysis, gal lb la dde r carc in om a patie nts we re pre dominantly fe male, more like ly to re port pre vious gallstone symptomology, and to have smoked. While gallstone s were not unive rsally isolate d from carcinoma patie nts at chole cystectomy, whe n prese nt, the y were le ss freque ntly classi® ed as ª chole ste rolº gallstone s base d on visual inspe ction. Gallbladde r carcinoma re mains unde te cted until it pre sents with symptoms attributable to associate d gallstone s or to proble ms re late d to the cance r itself. Fre que ntly, the carcinoma patie nt’s silent, yet far more fatal diagnosis is not re veale d until chole cyste ctomy, at which time only palliative therapy can be offere d. Atte mpts to characte rize this population have bee n dominate d by de scriptive studie s and only rarely have use d a comparison group with gallstone s. The stre ngth of this inve stigation is the combination of the case± control study de sign with multivariate analysis to de termine a risk pro® le for the gallbladde r carcinom a patie nt. V irtually all pre vious studie s describe gallbladde r carcinom a as a dise ase of e lde rly females with gallstone disease (4 ± 7). While we found fe male ge nde r to be e qually distribute d be twee n the two groups in the univariate analysis, multivariate adjustme nt for age and the pre sence of chole sterol gallstone s re veale d fe male ge nde r to be a promine nt risk factor. Seve ral pre vious studie s sugge st that diabe tes mellitus, cardiac disease, hype rtension, and porce lain gallbladde rs are more pre vale nt in cance r patie nts (2± 5), yet age adjustme nt establishe d these associations as be ing re late d to age , and not true risk factors for gallbladde r carcinoma. A past bout of gallstone symptom s was 17 time s more common in our carcinoma patie nts following age and ge nde r adjustme nt. Previous studie s inte rpre t the associations be twee n gallbladde r carcinoma and age or female gende r as re¯ e cting the incre ase d incide nce of chole lithiasis se en in these patie nts (8). Howe ve r, the se studie s matche d the comparison groups by age and ge nde r, making it impossible to assess their importance as risk factors (7, 9, 10) . We only matched our patie nts by admission date and hospital, and we spe ci® cally avoide d matching by age and gende r in orde r to assess the importance of these variable s. O ur multivariate analysis clari® es the se re lationships and shows that fe male ge nde r, advance d

1623

SCOTT ET AL

age , and pre vious gallstone s are all inde pende nt risk factors. The se associations are well acce pted risks for carcinoge ne sis. Immunologic mechanisms for ide ntifying and repairing chromosomal mutations be come le ss e f® cie nt with age and are a pote ntial cause for the rise in othe r cance rs that is see n with aging (11) . The e le vate d risk for women implie s that some associate d factor, such as hormonal e xposure may play an additional role . Epide miologic studie s have de monstrate d an association be tween this dise ase and multiparity (12, 13) . Additional support is provide d by ste roid binding studie s whe re estroge n re ceptor activity has bee n de tected in gallbladde r carcinom a specime ns (14, 15) . While our study did not de te ct any associations with parity or oral hormone the rapy, ne ithe r can we exclude them be cause of the infreque nt re porting of the se e xposure s in our subje cts’ medical records. Anothe r pote ntial mechanism for induction of gallbladde r carcinoma is e xposure to carcinoge nic che micals. Our re sults sugge st that one of the most fre que nt carcinoge nic exposure s, cigare tte smoking, is a risk factor for this dise ase . Age adjustme nt in multivariate analysis re veale d that patie nts with a past history of cigare tte e xposure were ® ve times more like ly to have gallbladde r carcinoma. Two othe r case ± control studie s sugge st a relationship for smoking e xposure and cance r of the biliary tract. O ne study e xploring the prote ctive e ffects of alcohol on cance r de te rmine d that smoking elevate d the risk for gallbladde r carcinoma in people who did not consume alcohol ( 16) . In anothe r population base d case ± control study, smoking un® ltere d cigare tte s was a signi® cant risk factor for bile duct cancer, but not for gallbladde r cance r ( 17) . An association be twe e n smoking and gallbladde r cancer might not be surprising conside ring the primary route of e xcre tion for tobacco-de rive d carcinoge ns such as be nzo[a]pyrine is via the bile (18) . One might spe culate that prolonge d storage of these carcinoge ns in the gallbladde r, as occurs betwe en meals, may allow carcinoge ne sis to initiate ( 19) . C irc um stantial e vi de nc e sugge sting benzo[ a] pyre ne as a risk factor for gallbladde r cance r can be found in e pide miologic studie s addre ssing industria l e xposure s. Those industrie s whe re gallbladde r carcinomas are most prevale nt, such as the rubbe r and metal industrie s, are the same industrie s for which be nzo[ a]pyre ne e xposure s, and in some instance s the actual benzo[ a] pyre ne± DNA adducts have be en isolate d from their worke rs (4, 20, 21) . Gallstone s are sugge sted carcinoge ns because the y

1624

irritate and in¯ ame the gallbladde r mucosa, resulting in e pithe lial dysplasia. Presumably, as gallstone s e nlarge or be come more nume rous, the probability of mucosal dam age increases. Howe ver, gallstone s are not invariably pre sent in association with gallbladde r carcinom a, sugge sting that anothe r mechanism may be at work (4). We found that 12% of our carcinoma patie nts did not have gallstone s in their gallbladde rs at surge ry. Whe n gallstone s were pre sent in the se patie nts, the y were large r but le ss nume rous than for controls. Howeve r, many of the se appare nt relationships disappe are d when age , fe male gende r, and pre se nce of chole ste rol gallstone s were include d. Conse que ntly stone size and numbe r might simply be associate d with the olde r age of this population. Rese arch into this que stion has be en con¯ icting, pe rhaps because fe w studie s adjust for age . For example , Lowe nfe ls et al (22) , using a patie nt population re stricte d to patie nts over 50 ye ars old, found an e le vate d carcinoma risk with gallstone s large r than 3 cm in diame ter. Die hl also de monstrate d a positive tre nd betwee n gallstone size and odds ratios for carcinom a, in an age and ge nde r-matche d univariate analysis (23) . In contrast, Moe rman e t al (24) , were unable to ® nd an association for ultrasonographic estimate s of stone size whe n the y e mploye d multivariate analysis to adjust for diffe rences in age . Gallbladde r cance r is linke d to e le vate d caloric intake and obesityÐ both of which may pre dispose to lithoge nic bile (7, 9, 10) . Biliary hype rse cretion of chole sterol is known to produce the nucle ation and de position of exce ss chole sterol as a chole sterol gallstone (25) . O ur obse rvation that carcinoma patie nts had fe wer chole sterol stone s than the controls se e ms at odds with spe culation that gallbladde r cancer is cause d by irritation from gallstone s. The e xplanation for this ® nding is uncle ar. Howe ver, an alte rnative possibility is that the real risk factor is e xcess se cretion of chole ste rol into bile . Continuous bathing of the gallbladde r mucosa with chole ste rol-rich bile could re sult in the direct deposition of chole ste rol into the mucosal cell membrane s, where it might conceivably in¯ ue nce cell function in a way that e nhance s cance r risk. Studie s that utilize re trospe ctive record re vie ws have se veral limitations. In this study, missing information arising from incomple te historie s in the chart made assessment of environme ntal exposure s and work history impossible , and de tails on pre vious historie s were certainly re stricted. Because of the re trospe ctive chart re vie w nature of this study, we were also force d to rely on the pathologists’ visual examiDigestive Diseases and Sciences, Vol. 44, No. 8 (August 1999)

RISK FACTORS FOR GALLBLADDER CARCINOMA

nation, and not che mical analysis, to classify gallstone composition. This approach is imprecise and some misclassi® cation may have occurre d. Howeve r, we have little re ason to belie ve that the accuracy or completene ss of the re cords varie d betwe en the two groups be cause historie s were re corde d prior to the incide ntal ® nding of the carcinoma at surge ry. In summary, elderly females are at gre atest risk for gallbladde r carcinoma. Adjustme nt for the olde r age of this population shows that age -associate d diseases, pre viously sugge sted as more freque nt in the carcinoma population, were not inde pe nde ntly associate d with gallbladde r cance r. Smoking may increase the risk of gallbladde r cance r. Furthe r cohort studie s that targe t the se populations will allow us to gain a more solid conse nsus on the risk factors for this disease. ACKNOWLEDGMENTS We are grate ful to the faculty and staff at Unive rsity Hospital, Carney Hospital, Brockton Hospital, Cape Cod Hospital, South Shore Hospital, and Quincy City Hospital for their help and patience. Special thanks go to David R. Williams, MD, Department of Surgery, Cape Cod Hospital; Anthony Dragone, MD, Department of Surgery, Quincy Hospital; Geraldine Ge ary, Medical Records Director, Carney Hospital; Mariann Doe rr, Medical Re cords Director, Brockton Hospital; Joan Bailey, Assistant Director of Me dical Re cords, Cape Cod Hospital; Cathie Nichols, Cancer Re gistrar, and Rosemarie Ames, Medical Records Director, South Shore Hospital; Kathy O wens, Tumor Re gistrar, Q uincy Hospital; and E dwin L. Prien, Laboratory for Stone Re search.

REFERENCES 1. Maram ES, Ludwig J, Kurland LT, Brian DD: Carcinoma of the gallbladder and e xtrahepatic biliary ducts in Roche ste r, Minnesota, 1931± 1971. Am J E pidemiol 109:152± 157, 1979 2. Bengmark S, Je ppsson B: Tumors of the gallbladde r. In Textbook of Gastroe nterology, 2nd e d. T Yakamada (e d). Philadelphia, JB Lippincott, 1995, pp 2739 ± 2744 3. Andrews EC, Bennett DE, Arhe lger RB: Carcinoma of the gallbladde r: Report of 45 Cases. South Me d J 62:573± 578, 1969 4. Diehl AK: Epide miology of gallbladde r cancer: A synthe sis of recent data. JNCI 65:1209 ± 1214, 1980 5. Carriaga MT, He nson DE: Live r, gallbladder, e xtrahepatic ducts, and pancre as. Cance r Suppl 75:171± 190, 1995 6. Kato K, Akai S, Tominaga S, Kato I: A case± control study of biliary tract cance r in Niigata Pre fecture, Japan. Jpn J Cance r Re s 80:932± 938, 1989 7. Zatonski WA, La V ecchia C, Prze wozniak K, Maisonneuve P, Lowenfels AB, Boyle P: Risk factors for gallbladde r carcinoma: A Polish case± control study. Int J Cancer 51:707± 711, 1992 8. Seccia M, Cavina E: Etiology and risk factors in gallbladde r cancer. Ann Ital Chir 3:163± 171, 1989

Digestive Diseases and Sciences, Vol. 44, No. 8 (August 1999)

9. Strom BL, Soloway RD, Rios-Dale nz JL, Rodriguez-Martinez HA, We st SL, Kinman JL, Polansky M, Berlin JA: Risk factors for gallbladder cance r: An international collaborative case± control study. Cance r 76:1747± 1756, 1995 10. Moe rman CJ, Bueno de Me squita HB, Sme ets FW, Runia S: Consumption of foods and micronutrients and the risk of cancer of the biliary tract. Prev Me d 24:591± 602, 1995 11. Masuda M, Nukuzuma C, Kazusaka A, Fujita S: Alterations in activation and de activation of mutage ns in aging rat liver. J Gerontol, Se r A, Biol Sci Me d Sci 50:B303± B306, 1995 12. Lambe M, Trichopoulos D, Hsie h CC, Ekbom A, Adami HO , Pavia M: Parity and cance rs of the gallbladder and the e xtrahepatic bile ducts. Int J Cance r 54:941± 944, 1993 13. The WHO Collaborative Study of Neoplasia and Steroid Contrace ptive s: Combined oral contraceptives and gallbladder cance r. Int J Epidemiol 18:309 ± 314, 1989 14. Nakamura S, Muro H, Suzuki S: Estrogen and proge ste rone receptors in gallbladder cance r. Jpn J Surg 19:189 ± 194, 1989 15. Ohnami S, Nakata H, Nagafuchi Y, Z e ze F, Eto S: Estrogen receptors in human gastric, he patoce llular, and gallbladde r carcinomas and normal live r tissue. Gan To Kagaku Ryoho 15:2923± 2928, 1988 16. Moe rman CJ, Bue no de Me squita HB, Runia S: Smoking, alcohol consumption and the risk of cancer of the biliary tract; a population-base d case ± control study in the Netherlands. E ur J Cance r Preve ntion 3:427± 436, 1994 17. Ghadirian P, Simard A, Baillargeon J: A population-based case± control study of cance r of the bile ducts and gallbladde r in Quebe c, Canada. Rev Epidemiol Sante Publ 41:107± 112, 1993 18. Bevan DR, Sadler VM: Quinol diglucuronide s are pre dominant conjugated metabolite s found in bile of rats following intratrache al instillation of benzo[ a]pyre ne . Carcinoge ne sis 13:403± 407, 1992 19. Stavric B, Klassen R: Dietary effects on the uptake of be nzo[a]pyrene. Food Chem Toxicol 32:727± 734, 1994 20. Pere ra FP, Dickey C, Sante lla R, O’Neill JP, Albertini RJ, Ottman R, Tsai WY, Mooney LA, Save la K, Hemminki R: Carcinogen-DNA adducts and gene mutation in foundry workers with low leve l e xposure to polycyclic aromatic hydrocarbons. Carcinoge ne sis 15:2905± 2910, 1994 21. Gupta P, Banerjee DK, Bhargava SK, Kaul R, Shanker V R: Abnormal patte rn of lung functions in rubber workers. J Indian Me d Assoc 92:260 ± 263, 1994 22. Lowenfels AB, Walker AM, Althaus DP, Townsend G, Dome llo L: Gallstone growth, size , and risk of gallbladde r cance r: An interracial study. Int J Epidemiol 18:50 ± 54, 1989 23. Diehl AK: Gallstone size and the risk of gallbladde r cance r. JAMA 250:2323± 2326, 1983 24. Moe rman CJ, Lagerwaard FJ, Bueno de Me squita HB, van Dale n A, van Lee uwen MS, Schrover PA: Gallstone size and the risk of gallbladde r cance r. Scand J Gastroente rol 28:482± 486, 1993 25. He aton KW: Diet and gallstone s. In He patologyÐ Rese arch and Clinical Issues, Volume 4: Gallstone s. MM Fishe r, CA Goresky, EA Shaffer, SM Strasbe rg (e ds). New York, Ple num Press, 1979, pp 371± 389

1625