Serum Aminotransferase Activity and Mortality Risk ina United States Community

Tae Hoon Lee,1 W. Ray Kim,1,2 Joanne T. Benson,2 Terry M. Therneau,2 and L. Joseph Melton III2

Serum aminotransferase [such as aspartate aminotransferase (AST) and alanine aminotrans-ferase (ALT)] is commonly used as an indicator of liver disease. The aim of the study was todetermine the degree to which aminotransferase results are associated with increased mor-tality at the population level. All adult residents of Olmsted County, Minnesota, who had ahealth care encounter at Mayo Clinic, Rochester, in 1995 were identified and their AST orALT results extracted from a laboratory database. These subjects were followed forward fromJanuary 1995 to April 2006 and their survival determined. To exclude patients with abnor-mal results because of a terminal illness, deaths within the first 2 years were excluded. Themain outcome measure was survival. Standardized mortality ratios (SMRs) were calculated,based on Minnesota White death rates. During 1995, AST was measured at least once in18,401 community residents, of whom 2,350 (13%) had results greater than the upper limitof normal (ULN). Of 6,823 subjects who had their ALT measured, 911 (13%) had resultshigher than ULN. Abnormal AST was associated with a significantly increased SMR (1.32 for1-2� ULN and 1.78 for >2� ULN). SMR was also higher for abnormal ALT (SMR � 1.21for 1-2� ULN and 1.51 for >2� ULN). In contrast, normal AST or ALT was associated witha risk of death lower than expected (SMR 0.95 for AST, 0.61 for ALT). Conclusion: Serumlevels of AST and ALT obtained in a routine medical care setting are associated with futuremortality in community residents. (HEPATOLOGY 2008;47:880-887.)

Serum aminotransferase activities, including aspar-tate aminotransferase (AST) and alanine amino-transferase (ALT), are commonly referred to as

“liver enzymes,” because they are abundantly presentwithin hepatocytes, catalyzing transfer of amino groups togenerate products in gluconeogenesis and amino acid me-tabolism. Because these enzymes are released from dam-aged hepatocytes into the blood, their activities measuredin the serum have been widely recognized as a tool todetect liver disease.1-3 Although used routinely in clinicalpractice for decades, their role as a predictor of mortalityhas not been examined until recently. A German study of

male construction workers found that elevated AST wasassociated with increased mortality from all causes.4 Like-wise, in a large population-based study from South Korea,serum ALT activity measured at baseline was correlatedwith subsequent mortality, and it was shown that elevatedserum ALT was associated with a marked increase in mor-tality from liver disease.5 However, that particular studywas conducted in an area endemic for hepatitis B virusand yet lacked information about hepatitis B virus in thestudy participants. Whether these data are applicable tothe United States may be questionable.

This is important because data from the NationalHealth and Nutrition Examination Survey (1999-2002)showed that the prevalence of elevated serum of AST orALT in the United States was 9.8%, most of which couldnot be explained on the basis of easily identifiable liverdisease such as hepatitis C or excessive alcohol consump-tion.6 Presumably, many Americans with abnormal ami-notransferases have nonalcoholic fatty liver disease(NAFLD).7 Whereas NAFLD is an important source ofmorbidity and mortality from liver disease, it also may bean important indicator of increased cardiovascular riskmediated through metabolic consequences of insulin re-sistance.8,9 Thus, aminotransferases may be an importantindicator of health, yet the magnitude of their impact on

Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransfer-ase; NAFLD, nonalcoholic fatty liver disease; REP, Rochester Epidemiology Project;SMR, standardized mortality ratios; ULN, upper limit of normal.

From the 1Division of Gastroenterology and Hepatology; and the 2Department ofHealth Sciences Research, Mayo Clinic College of Medicine, Rochester, MN.

Received June 1, 2007; accepted October 12, 2007.Supported by grants from the National Institutes of Health (DK-34238, DK-

61617, and AR-30582).Address reprint requests to: W. Ray Kim, M.D., Division of Gastroenterology and

Hepatology (PL 6), Mayo Clinic College of Medicine, 200 1st Street SW, Rochester,MN 55905. E-mail: kim.ray@mayo.edu; fax: 507-538-3974.

Copyright © 2007 by the American Association for the Study of Liver Diseases.Published online in Wiley InterScience (www.interscience.wiley.com).DOI 10.1002/hep.22090Potential conflict of interest: Nothing to report.

880

mortality in the general population of the United Stateshas not been established.

In this work, we use unique data resources available inOlmsted County, Minnesota, to evaluate the relation be-tween serum aminotransferase activities and subsequentmortality. The aims of the study were (1) to describe theutilization of serum aminotransferase assessments in thecommunity; (2) to determine whether abnormal amino-transferase results are associated with increased mortality;and (3) to explore correlation between aminotransferaseresults and causes of death.

Patients and Methods

Data Sources and Elements. Population-based epi-demiologic research can be conducted in OlmstedCounty, Minnesota, because medical care is virtually self-contained within the community, and there are only a fewproviders. In fact, just 2 major health care providers servethe population, namely, the Mayo Medical Center andOlmsted Medical Center. Moreover, each of these pro-viders uses a unit medical record whereby all (inpatientand outpatient) medical information for each individualresident of Olmsted County is accumulated in a singlerecord. These distinct features of the health care environ-ment in Olmsted County led to a creation of a unifiedmedical index system (the Rochester EpidemiologyProject, REP) by which the details of the medical careprovided to Olmsted County residents can be studied.10

Thus, REP makes it possible to identify a group of pa-tients with certain characteristics and follow them for-ward, assessing long-term outcomes such as mortality andcauses of death.

Based on the REP database, all adult (age �18) Olm-stead County residents were identified who had at leastone healthcare encounter at any of the Mayo Clinic facil-ities in Olmsted County in calendar year 1995. AlthoughMayo has a well-established reputation as a quaternaryreferral center for a variety specialty and subspecialty con-ditions, it also delivers the majority of primary care to thelocal population: each year, approximately half the popu-lation of Olmsted Country is seen at one of the MayoClinic facilities for reasons such as pre-employment exam-inations, minor illnesses, sports physicals, and routinemedical care.

The year 1995 was chosen to ensure at least 10 years offollow-up. These subjects’ demographic information(age, sex, and race) was extracted from the database.Mayo’s laboratory database (“Laboratory InformationSystem”) was then queried to extract all AST and ALTdata on the same individuals. In 1995, AST and ALTactivities in the serum were assayed by colorimetric meth-

ods using commercial kits (Roche Diagnostics, Indianap-olis, IN). The upper limit of normal (ULN) of AST was31 IU/L for both men and women, whereas that for ALTwas 45 IU/L for men and 29 IU/L for women. Based onthe AST and ALT data, subjects were divided into 4groups: (1) those who did not have their aminotrans-ferases tested in 1995; (2) those whose results were lessthan the ULN; (3) those with results between 1 and 2times the ULN; and (4) those with results higher than 2times the ULN.

Survival Analysis. Community residents thus char-acterized were followed forward in time for death. Be-cause the study objective was to assess the correlationbetween aminotransferase results and long-term survival,subjects who died or were lost to follow-up within first 2years were excluded, a priori, from further analysis. Forexample, because AST is also one of the “cardiac enzymes”that increases acutely in the setting of myocardial infarc-tion, it is likely to be associated with high short-termmortality. In subjects who survived at least 2 years, the lastday of known follow-up in the REP database was used toassess survival. Thus, survival analysis was conductedbased on the date of death in those who died and the dateof last follow-up in those who survived. In those whodied, several sources of information, including institu-tional registration file, death certificates, and medicalrecords, were used to ascertain the date of death and todetermine the cause. Causes of death were classified intohepatobiliary (International Classification of Diseasesversion 9 codes: 006.3, 070.0-070.9, 155-156, 275.0-275.1, 277.3, 331.81, 452, 453.0, 456.2, 570-576,751.6, 782.4, 789.1), cardiovascular (International Clas-sification of Diseases version 9 codes: 390-459, 745-747,except 452, 453.0, 456.2, 572.1), and all others. Fol-low-up was ended as of April 2006.

The Kaplan-Meier method was used to estimate sur-vival of community residents in the 4 groups according tothe aminotransferase data. Standardized mortality ratios(SMRs) were calculated to determine the association be-tween aminotransferase data and subsequent mortality,independent of age and sex. Mortality data of the refer-ence population (that is, Minnesota whites) were ob-tained from the Minnesota State Department ofHealth.11 Based on these data, expected numbers ofdeaths were derived for age-specific and sex-specific strata.SMRs were calculated by the ratio between observed andexpected numbers of deaths. Ninety-five percent confi-dence intervals were calculated assuming a Poisson errorstructure. To assess the overall relation between amino-transferase values and subsequent risk of mortality, gen-eralized additive models were used to model a smooth butpotentially nonlinear conformation.12,13 Again, relative

HEPATOLOGY, Vol. 47, No. 3, 2008 LEE ET AL. 881

risk was calculated using the expected mortality based onthe age-specific and sex-specific mortality rates for theMinnesota white population.

ResultsDuring 1995, 54,586 residents (49% of community

residents) had at least one healthcare encounter at anyMayo Clinic facility in Olmsted County. Of those, 7,404residents who died or who moved from the communitywithin the following 2 years were excluded according tothe a priori inclusion criteria. Of the remaining 47,182community residents, 18,401 (39.0%) had their ASTmeasured and 6,823 (14.5%) had an ALT assessment. In22,448 residents, either of the aminotransferases wasmeasured, and 2,776 (5.9%) had both AST and ALT(Fig. 1).

Table 1 compares the characteristics of OlmstedCounty residents with respect to AST and ALT data.Individuals who did not have their AST measured wereyounger than those who did (P � 0.001), whereas thereverse was true with ALT (P � 0.001). Overall, there wasfemale preponderance, except in those with abnormalAST results. Most of our study subjects were of a whiterace, although race data were not always available.

Of the 18,401 residents with AST data, 16,051(87.2%) had results within normal limits, 1897 (10.3%)between ULN and 2 times ULN, and 453 (2.5%) higherthan 2 times ULN. The distribution of AST was skewedto the right, with a mean of 25.7 (�37.2) and median of21.0 (range, 4-1860; interquartile range, 17-26). Individ-uals with the highest AST results were younger than theother AST groups. The proportion of nonwhites washigher among those with abnormal AST than amongthose with normal AST.

With regard to ALT, the result was within normallimits in 5912 (86.6%), whereas 911 (13.4%) had abnor-mal ALT values. Of those, 241 (3.5%) had ALT that wasgreater than twice normal. The distribution of ALT wasalso skewed, with a mean of 29.4 (�80.9) U/L and me-dian of 21.0 (range, 4-4860; interquartile range, 15-29).In contrast to the AST data, those with ALT greater than2 times ULN were older and more likely to be female thanthose with lower ALT results, although these differenceswere not statistically significant. Similar to AST results,the proportion of whites was smaller among those withabnormal ALT than among those with normal ALT.

When the study population was followed forward intime, 4639 deaths were observed. The median duration of

Fig. 1. Study population.

882 LEE ET AL. HEPATOLOGY, March 2008

follow-up among the survivors was 10.9 years (interquar-tile range: 10.1-11.2 years). Figure 2 shows the survival ofOlmsted County residents by their AST and ALT resultswithout adjustment for age and sex. For both AST andALT, abnormal results were associated with progressivedecreases in survival. Survival of those who were nottested differed between AST and ALT. The group whoseAST was not tested had better survival than those withAST results. In contrast, survival in those without ALTresult was better than those with ALT greater than 2 timesULN but poorer than those with ALT between 1 and 2times the ULN. This observation could be accounted forby the differences in the age distribution between groups.For example, those without AST were much younger thanthose with AST data, whereas those without ALT werethe oldest of the group (Table 1).

Standardized mortality ratios were calculated to ex-clude confounding effects of age and sex (Table 2). Com-munity residents who did not have their aminotransferasedetermined had better survival than expected (SMR �0.79 for AST and 0.93 for ALT). Those whose resultswere normal also had better survival than expected(SMR � 0.95 for AST and 0.61 for ALT). For both ASTand ALT, abnormal results were associated with an in-creased risk of death. For AST, abnormal AST up to 2times ULN was associated with a 32% increase in the riskof death, and more than 2 times ULN with a 78% in-crease, compared with the reference population. Simi-larly, abnormal ALT up to 2 times ULN was associatedwith a 21% increase in the risk of death, and more than 2times ULN, with a 59% increase, compared with thereference population.

Figure 3 illustrates the relation between aminotransfer-ase results and the likelihood of subsequent death. In bothsexes, there was a progressive rise in mortality as AST

increased. This tendency was also observed within thenormal range and the AST value that corresponded to arelative risk of 1 (in other words, that of the referencepopulation) was 22 IU/L in women and 32 IU/L in men.

Fig. 2. Kaplan Meier survival of Olmsted County residents by AST andALT results. * Indicates P � 0.01 in comparison with the �ULN groupin each panel.

Table 1. Characteristics of Residents of Olmsted County, MN, According to Aspartate Aminotransferase (AST) and AlanineAminotransferase (ALT) Results

AST ALT

<ULN 1–2� ULN >2� ULN No Test <ULN 1-2� ULN >2� ULN No Test

No. 16,051 (34.0%) 1,897 (4.0%) 453 (1.0%) 28,781 (61.0%) 5,912 (12.5%) 670 (1.4%) 241 (0.5%) 40,359 (85.5%)Age (mean, SD) 54.0 (18.6) 56.2 (18.5)* 50.7 (18.6)* 41.2 (15.8)* 43.5 (12.5) 43.6 (13.1) 46.0 (17.1) 46.7 (18.7)*Sex (% male) 40.2 53.3† 52.0† 42.2† 45.7 47.0 42.4 41.9†Race (%‡)

Caucasian 96.5 94.3 90.6 95.7 98.2 95.7 92.5 95.6Asian/Pacific Is 1.5 2.6 3.8 1.8 0.5 1.3 2.5 1.9African American 0.8 1.0 2.1 0.9 0.3 1.1 0.5 1.0Native Am/Alaskan 0.3 0.6 0.8 0.3 0.2 0.4 1.5 0.3Hispanic 0.2 0.3 1.3 0.3 0.3 0.4 0.0 0.3Other 0.7 1.2 1.3 1.0 0.5 1.1 3.0 0.9

Unknown/ missing (n) 2733 343 80 6510 1446 133 40 8047

*P � 0.05 compared with the respective �ULN group (Wilcoxon rank sum test).†P � 0.05 compared with the respective �ULN group (�2 test).‡% excluding missing race.

HEPATOLOGY, Vol. 47, No. 3, 2008 LEE ET AL. 883

A similar trend was seen with ALT; however, because ofthe smaller sample size, the estimates are not as precise.For example, the downward trend above an ALT of 100IU/L in women is probably an artifact because of therelatively few observations in that range of ALT (n � 52).

Seventy-three subjects (1.6% of the 4639 decedents,out of the 47,182 study subjects) died of liver disease(including hepatobiliary malignancies);1559 (33.6%), of

cardiovascular causes, 1,015 (21.9%), of malignancy (ex-clusive of hepatobiliary neoplasms); leaving 1,992(42.3%) who died of other or undetermined causes. Theage at death was much younger in decedents from hepa-tobiliary causes (69.6 � 5.8) than others (82.3 � 12.0 forcardiovascular, 74.0 �13.5 for nonliver malignancies and80.4 � 15.2 for other/unknown causes). There was astrong male preponderance among decedents from livercauses (58.9%), whereas those from other causes weremore likely to be women (45.5% male for cardiovascular,45.7% for non-liver malignancies and 40.8% for other/unknown causes). Of decedents with available data, ASTwas significantly higher among those who died of hepato-biliary causes (Fig. 4). The proportion of decedents whohad their ALT measured was higher among those withhepatobiliary causes of death, although the proportionswith ALT data were much smaller in general than thosewith AST data. Of those with available data, decedentsfrom hepatobiliary causes had higher ALT than thosewith other causes of death.

DiscussionIn this work, we demonstrate that serum AST and ALT

activities are associated with increased risk of mortality inthe ensuing decade. The relation between the aminotrans-ferase results at the outset and the subsequent risk of death(excluding the first 2 years) was almost linear. It is clearlyshown that these simple, inexpensive blood tests may rep-resent valuable indicators of long-term outcome. One ofthe main reasons why AST was measured much morecommonly than ALT (40% versus 15%) is that AST wasa part of an automated multi-channel chemistry panelcommonly ordered by clinicians before a change in theMedicare reimbursement policy in 1998. Because thepanel became unbundled since, it is likely that AST isordered less commonly now compared with 1995. Our

Table 2. Standardized Mortality Ratios (SMR) Among Residents of Olmsted County, MN, According to AspartateAminotransferase (AST) and Alanine Aminotransferase (ALT) Results

Total Numberof Subjects

Number ofObserved Deaths

Number ofExpected Deaths SMR (95% CI) P

ASTNot done 28,758 1,461 1,850 0.79 (0.75–0.83) �0.001� ULN 15,991 2,671 2,825 0.95 (0.91–0.98) 0.0041–2� ULN 1,885 374 283 1.32 (1.19–1.46) �0.001� 2� ULN 452 102 57 1.78 (1.45–2.16) �0.001

ALTNot done 40,235 4,292 4,617 0.93 (0.90–0.96) �0.001�ULN 5,885 193 315 0.61 (0.53–0.71) �0.0011–2� ULN 667 39 32 1.21 (0.86–1.66) 0.228� 2� ULN 240 32 20 1.51 (1.04–2.14) 0.018

NOTE. SMR based on Minnesota whites as the reference population. 141 (AST) and 202 (ALT) results were excluded from this analysis because of missing sex orbirth date, or because their laboratory tests were performed on different dates: therefore, although they had �2 years of follow-up for 1 test, they did not for the other.

Fig. 3. Relative risk of death according to the amnotransferaselevel.

884 LEE ET AL. HEPATOLOGY, March 2008

data raises a question whether aminotransferase must beconsidered as a screening tool to be applied more widely.

The SMRs we found in association with abnormalaminotransferases range between 1.21 and 1.78. SMRs ofa similar magnitude have been reported in associationwith conditions well established to have a substantial pub-lic health impact. Thus, Bender et al.14 estimated thatbody mass indices higher than 25kg/m2 were associatedwith SMRs of 1.73 in male and 1.46 in female. Similarly,Ringbaek et al.15 showed an SMR of 1.54 in men and 1.91in women with asthma after 15 years of follow-up. Inaddition to these comparisons, the high proportion(13%) of abnormal aminotransferase results among com-munity residents seen in our practice makes our resultsimportant. As expected, we found that normal amino-transferase values are “protective”; once subjects with ab-normal results and thus higher risk of death are removedfrom the population, the remainder will have survivalbetter than average.

There may be several mechanisms by which serumAST and ALT may be associated with an increased risk ofdeath. First, as Fig. 4 suggests, AST and ALT, beingmarkers of serious liver diseases, increase the risk of mor-

tality from liver disease.3,5,16,17 Although that associationmay be self-evident, it is less certain what level of amino-transferase best distinguishes the presence and absence ofliver disease. Prati et al.18 found that when asymptomaticindividuals with hepatitis C and those at high risk ofNAFLD were removed, the ULN may defined as ALT of30 IU/L in men and 19 IU/L in women, much lower thanconventional ULN (40 IU/L and 30 IU/L, respectively).Piton et al.19 proposed that, in addition to sex, body massindex be taken into account in the definition of the ULNfor ALT. Although these data are helpful in identifyingpatients who may have clinically relevant liver disease,they do not necessarily indicate that these normal valuesare helpful in estimating the risk of mortality. In our data,Fig. 3 suggests that the risk of death (from all causes)increased as AST/ALT levels increase even within the nor-mal range. However, the level of risk did increase beyond1 (average risk of death in this study population). Forexample, in women, in whom the ULN for ALT was 30,the risk of death is higher for ALT of 30 compared withALT of 20, but the risk of death at 30 is about that of anaverage woman included in this study.

Second, aminotransferases, particularly ALT, mightalso be a marker of cardiovascular diseases, and this mayadd to the risk of death. According to a study based on theThird National Health and Nutrition Examination Sur-vey, ALT was associated with the risk of coronary heartdisease among nonobese participants without viral hepa-titis or excessive alcohol consumption.9 The link betweenALT and coronary artery disease risk is thought to be viathe metabolic syndrome, which creates a biochemical mi-lieu to facilitate atherosclerosis, as well as NAFLD.8,20

AST also may become elevated in patients with acutemyocardial injury or congestive heart failure.21 By theintent of our study design, deaths within the first 2 yearsof observation were eliminated to exclude many of thesepatients with advanced end organ damage.

Third, aminotransferase elevations may reflect anotherserious comorbid condition that increases the risk ofdeath. For example, chronic alcohol consumption orabuse is one of the common causes of AST/ALT eleva-tion. The increased risk of death in that population notonly may stem from alcohol-induced end organ damage(such as alcoholic liver disease) but also may be related tomorbidity associated with concurrent depression, poly-substance use including tobacco products, and acci-dents.22-26 Similarly, although hepatitis C virus infectionis an important source of morbidity and mortality fromliver disease, it also may be a marker of injection drug useand its associated physical, psychological, and social con-sequences.27 Finally, aminotransferases can be elevated in

Fig. 4. Comparison of aminotransferase levels by the cause ofdeath.

HEPATOLOGY, Vol. 47, No. 3, 2008 LEE ET AL. 885

various kinds of chronic nonhepatic disease such as mus-cle and thyroid disease.28

A unique feature of our data is that they are derivedfrom a self-contained healthcare setting in OlmstedCounty, Minnesota.10,29 The well-established infrastruc-ture and proven track record of the REP help us be con-fident of the quality and completeness of the follow-up inour large number of community residents. In addition,we would like to point out that the subjects included inthis study were seen at Mayo Clinic, which principallyfunctioned as a primary care health provider to the localresidents, rather than as the specialty referral center it isreputed to be.

It is also important to underline the limitations to thisstudy. First, we point out that our data were obtained in aroutine medical care setting, as opposed to screening ofrandom individuals in the community. Obviously, thereis a selection process embedded in our data that insofar asto be captured in this study, the community resident musthave been seen for healthcare and the provider must haveordered AST/ALT. When the prevalence of aminotrans-ferase elevation in this study (13%) is compared againstthat in the US population in general (10%, NationalHealth and Nutrition Examination Survey 1999-2002),the degree of oversampling of those with abnormal ami-notransferase in this study appear rather minor.6 Our datadid show that physicians’ decisions to order AST/ALThave a demonstrable correlation with the subsequent riskof death in that lack of AST/ALT data was associated withsmall yet significant decrease in mortality. Thus, our re-sults apply to individuals seen in healthcare setting ratherthan population at large. Second, along the same vein,only a minority of community residents had their AST/ALT measured. This reduced the sample size in specificsubgroups. For example, the proportion with AST/ALTdata was quite small among decedents, which preventedmore meaningful analysis of cause-specific mortality.Third, this study was based on a single aminotransferaseresult. According to a study based on blood donors, ami-notransferase elevation was found to be persistent in 28%,intermittent in 36%, and singular in 33%.30 Incorpora-tion of follow-up laboratory results may further refine ourunderstanding of the relation between aminotransferaseand subsequent morbidity and mortality. Furthermore,given the long lag time between asymptomatic amino-transferase elevation and death of liver disease, a studywith longer follow-up may find a stronger association be-tween the two. Lastly, the large proportion of whites inour study may reduce the generalizability of the data.Although there may be differences in the prevalence ofelevated ALT between races, they may be more indicativeof disparities in liver disease prevalence. Further investi-

gation is needed as to whether there are inherent biolog-ical differences in the distribution of ALT or thesignificance of abnormal values among difference races.

Serum aminotransferases levels are not recommendedas a routine preventive service by the US Preventive Ser-vices Task Force.31 The US Preventive Services TaskForce emphasizes instead the role of individual physiciansin ordering screening tests according to their judgment ofbenefits and risks. Our data lend some support to the USPreventive Services Task Force recommendation, becausecommunity residents in whom AST/ALT were not or-dered had a lower risk of death than expected. Conversely,liver disease is generally asymptomatic until complica-tions of advanced disease occur. Measurement of amin-otransferases may allow early detection and treatment ofconditions that could lead to significant morbidity andmortality in the future. Although our data are not able toanswer these questions, we believe they contribute to un-derstanding the role of these simple blood tests in improv-ing the health of the population in general. In conclusion,these data, based on a large number of residents in a UScommunity, suggest that serum levels of AST and ALTobtained in a routine medical care setting are associatedwith future mortality.

References1. Wroblewski F, Ladue JS. Serum glutamic pyruvic transaminase in cardiac

with hepatic disease. Proc Soc Exp Biol Med 1956;91:569-571.2. Wroblewski F, Ladue JS. Serum glutamic oxalacetic aminopherase

(transaminase) in hepatitis. J Am Med Assoc 1956;160:1130-1134.3. Wroblewski F. The clinical significance of transaminase activities of serum.

Am J Med 1959;27:911-923.4. Arndt V, Brenner H, Rothenbacher D, Zschenderlein B, Fraisse E, Flied-

ner TM. Elevated liver enzyme activity in construction workers: prevalenceand impact on early retirement and all-cause mortality. Int Arch OccupEnviron Health 1998;71:405-412.

5. Kim HC, Nam CM, Jee SH, Han KH, Oh DK, Suh I. Normal serumaminotransferase concentration and risk of mortality from liver diseases:prospective cohort study. BMJ 2004;328:983.

6. Ioannou GN, Boyko EJ, Lee SP. The prevalence and predictors of elevatedserum aminotransferase activity in the United States in 1999-2002. Am JGastroenterol 2006;101:76-82.

7. Daniel S, Ben-Menachem T, Vasudevan G, Ma CK, Blumenkehl M. Pro-spective evaluation of unexplained chronic liver transaminase abnormali-ties in asymptomatic and symptomatic patients. Am J Gastroenterol 1999;94:3010-3014.

8. Marchesini G, Brizi M, Morselli-Labate AM, Bianchi G, Bugianesi E,McCullough AJ, et al. Association of nonalcoholic fatty liver disease withinsulin resistance. Am J Med 1999;107:450-455.

9. Ioannou GN, Weiss NS, Boyko EJ, Mozaffarian D, Lee SP. Elevatedserum alanine aminotransferase activity and calculated risk of coronaryheart disease in the United States. HEPATOLOGY 2006;43:1145-1151.

10. Melton L. History of the Rochester Epidemiology Project. Mayo Clin Proc1996;71:266-274.

11. Therneau TM, Offord J. Technical Report Series No. 63, Expected Sur-vival Based on Hazard Rates (update). Department of Health ScienceResearch, Rochester, MN: Mayo Clinic; 1999.

886 LEE ET AL. HEPATOLOGY, March 2008

12. Hastie TJ, Tibshirani RJ. Generalized additive models. London: Chapman& Hall, 1990.

13. Venables WN, Ripley BD. Modern applied statistics with S-PLUS. NewYork: Springer, 1997.

14. Bender R, Zeeb H, Schwarz M, Jockel KH, Berger M. Causes of death inobesity: relevant increase in cardiovascular but not in all-cancer mortality.J Clin Epidemiol 2006;59:1064-1071.

15. Ringbaek T, Seersholm N, Viskum K. Standardised mortality rates in femalesand males with COPD and asthma. Eur Respir J 2005;25:891-895.

16. Zimmerman HJ, West M. Serum enzyme levels in the diagnosis of hepaticdisease. Am J Gastroenterol 1963;40:387-404.

17. Ellis G, Goldberg DM, Spooner RJ, Ward AM. Serum enzyme tests indiseases of the liver and biliary tree. Am J Clin Pathol 1978;70:248-258.

18. Prati D, Taioli E, Zanella A, Della Torre E, Butelli S, Del Vecchio E, et al.Updated definitions of healthy ranges for serum alanine aminotransferaselevels. Ann Intern Med 2002;137:1-10.

19. Piton A, Poynard T, Imbert-Bismut F, Khalil L, Delattre J, Pelissier E, etal. Factors associated with serum alanine transaminase activity in healthysubjects: consequences for the definition of normal values, for selection ofblood donors, and for patients with chronic hepatitis C. MULTIVIRCGroup. HEPATOLOGY 1998;27:1213-1219.

20. Clark JM, Brancati FL, Diehl AM. The prevalence and etiology of elevatedaminotransferase levels in the United States. Am J Gastroenterol. 2003;98:960-967.

21. Giallourakis CC, Rosenberg PM, Friedman LS. The liver in heart failure.Clin Liver Dis 2002;6:947-967, viii-ix.

22. Reed MB, Wang R, Shillington AM, Clapp JD, Lange JE. The relationshipbetween alcohol use and cigarette smoking in a sample of undergraduatecollege students. Addict Behav 2007;32:449-464.

23. Sher KJ, Gotham HJ, Erickson DJ, Wood PK. A prospective, high-riskstudy of the relationship between tobacco dependence and alcohol usedisorders. Alcohol Clin Exp Res 1996;20:485-492.

24. Jackson KM, Sher KJ, Wood PK, Bucholz KK. Alcohol and tobacco usedisorders in a general population: short-term and long-term associationsfrom the St. Louis epidemiological catchment area study. Drug AlcoholDepend 2003;71:239-253.

25. Williams AF. Alcohol-impaired driving and its consequences in the UnitedStates: the past 25 years. J Safety Res 2006;37:123-138.

26. Regier DA, Farmer ME, Rae DS, Locke BZ, Keith SJ, Judd LL, et al.Comorbidity of mental disorders with alcohol and other drug abuse: resultsfrom the Epidemiologic Catchment Area (ECA) Study. JAMA 1990;264:2511-2518.

27. Williams I. Epidemiology of hepatitis C in the United States. Am J Med1999;107:2S-9S.

28. Green RM, Flamm S. AGA technical review on the evaluation of liverchemistry tests. Gastroenterology 2002;123:1367-1384.

29. Kurland LT, Molgaard CA. The patient record in epidemiology. Sci Am1981;245:54-63.

30. Friedman LS, Dienstag JL, Watkins E, Hinkle CA, Spiers JA, Rieder SV, etal. Evaluation of blood donors with elevated serum alanine aminotransfer-ase levels. Ann Intern Med 1987;107:137-144.

31. USPSTF. The guide to clinical preventive services. Available at: http://www.ahrq.gov/clinic/pocketgd.htm. Accessed Jan 10, 2007.

HEPATOLOGY, Vol. 47, No. 3, 2008 LEE ET AL. 887