Internal Comparison Group Research Articles

Case-Based Training of Evidence-Based Clinical Practice in Primary Care and Decreased Mortality in Patients With Coronary Heart Disease

We investigated the 10-year mortality rates in a trial that tested a case-based intervention in primary care aimed at reducing the gap between evidence-based goals and clinical practice in patients with coronary heart disease (CHD). A prospective randomized controlled pragmatic trial was undertaken in a primary care setting. New evidence-based guidelines, with intensified lipid-lowering recommendations in CHD, were mailed to all general practitioners in the region and presented at a lecture in 1995. General practitioners (n = 54) and patients with CHD (n = 88) were assigned according to their primary health care center to 2 balanced groups and randomly allocated to usual care as a control or to an active intervention. General practitioners in the intervention group participated in repeated case-based training during a 2-year period. Patients whose CHD was treated by specialists (n = 167) served as an internal specialist comparison group. Altogether, 255 consecutive patients were included. Cox regression analysis was used to detect any survival benefit of the intervention. At 10 years, 22% of the patients in the intervention group had died as compared with 44% in the control group (P = .02), with a hazard ratio of 0.45 (95% confidence interval, 0.20-0.95). This difference was mainly due to reduced cardiovascular mortality in the intervention group (P = .01). In addition, the mortality rate of 22% in the intervention group was comparable to the rate of 23% seen in patients treated by a specialist. Use of case-based training to implement evidence-based practice in primary care was associated with decreased mortality at 10 years in patients with CHD.

The risk of preterm birth following treatment for precancerous changes in the cervix: a systematic review and meta-analysis

Studies investigating the association between treatment for precancerous changes in the cervix and risk of preterm birth have used a variety of comparison groups. To investigate whether treatment for precancerous changes in the cervix is associated with preterm birth (<37 weeks) and to examine the impact of the type of comparison group on estimates of risk. PubMed, Embase and CENTRAL were searched for studies pubished between 1950 and 2009. Eligible studies were those that reported preterm birth outcomes for excisional and ablative treatments separately and included a comparison group. Pooled relative risks (RR) and 95% confidence intervals were computed using a random effects model. Thirty eligible studies were located. Excisional treatment was associated with an increased odds of preterm birth, when compared with an external (RR 2.19, 95% CI 1.93-2.49) or internal (RR 1.96, 95% CI 1.46-2.64) comparison group. In comparison with women who were assessed but not treated, the risk estimate was smaller (RR 1.25, 95% CI 0.98-1.58). Ablative treatment was associated with an increased risk of preterm birth when an external comparison group (RR 1.47, 95% CI 1.24-1.74) but not an internal comparison group (RR 1.24, 95% CI 0.73-2.10) or untreated comparison group (RR 1.03, 95% CI 0.90-1.18) was used. Excisional treatment was associated with a significantly increased risk of preterm birth. It provides new evidence that some types of ablative treatment may also be associated with a small increased risk. The type of comparison group used is an important consideration when comparing the outcomes of studies.

Does midtrimester cervical length ≥25 mm predict preterm birth in high-risk women?

We sought to assess pregnancy outcome along a continuum of cervical lengths (CLs) ≥25 mm. We conducted planned secondary analysis of a randomized cerclage trial of women with prior spontaneous preterm birth 17(0)-34(6/7) weeks. Outcomes of women who maintained CLs ≥25 mm were analyzed. Women with CLs <25 mm randomized to no cerclage comprised an internal comparison group. Of 1014 screened, 153 had CL <25 mm, and 672 had CL ≥25 mm. Birth <35 weeks occurred in 16% of the ≥25 mm cohort. The relationship between CLs ≥25 mm and birth gestational age was null (P = .15). In the <25 mm group, progressively shorter CLs predicted birth <35 weeks (P < .001); this relationship was null in the ≥25 mm group (P = .17). The continuum of CLs ≥25 mm measured between 16(0/7)-22(6/7) weeks does not predict gestational length in women with prior spontaneous preterm birth.

Lopinavir/Ritonavir in Pregnancy

The Antiretroviral Pregnancy Registry was established in 1989 to collect data on birth defects after pregnancy exposures to antiretroviral therapy. Using Registry data, this study estimates the birth defect risk after pregnancy exposures to lopinavir/ritonavir. The analysis population includes all prospective lopinavir/ritonavir-exposed pregnancies enrolled in the Registry from September 2000 through July 2007. Birth defect prevalence after pregnancy exposure is compared with rates from a population-based surveillance system, and first-trimester exposures are compared with combined second/third-trimester exposures. Among 955 live births prenatally exposed to lopinavir/ritonavir, 23 cases with birth defects were reported [2.4%, 95% confidence interval (CI) = 1.5 to 3.6). Among 267 live births with first-trimester exposures, 5 had birth defects (1.9%, 95% CI = 0.6 to 4.3). These rates are similar to the population-based comparator rate of 2.67% and the rate in infants with second/third-trimester exposures (2.6%, 95% CI = 1.6 to 4.1). No pattern of birth defects suggestive of a common etiology was seen. The prevalence of birth defects among infants prenatally exposed to lopinavir/ritonavir is not significantly different from internal or external comparison groups. These data provide reassuring information to patients and clinicians about the safety of lopinavir/ritonavir in the treatment of HIV-positive pregnant women.

Endoscopic and Quality‐of‐Life Outcomes After Revision Endoscopic Sinus Surgery

First, to examine the impact of endoscopic sinus surgery (ESS) on endoscopic and quality-of-life (QOL) outcomes after revision ESS as compared to primary ESS. Second, to evaluate whether or not other risk factors and/or co-morbidities influence the relationship between revision surgery status and outcomes of ESS. Prospective observational study with an internal comparison group. Preoperative computed tomography scores, pre and postoperative endoscopy scores, and two validated disease-specific QOL instruments, the Rhinosinusitis Disability Index (RSDI) and Chronic Sinusitis Survey (CSS), were collected on a prospective cohort of patients undergoing ESS for chronic rhinosinusitis. Data were analyzed using Pearson's chi and multiple logistic regression models. Mean preoperative Lund-Mackay computed tomography scan scores were similar in primary and revision surgery patients. In patients without polyps, revision ESS patients were 3.88 times more likely to improve on endoscopy scores than primary ESS patients (95% confidence interval 1.70, 8.83; P = .001). In nasal polyp patients, there was no difference by revision status (odds ratio 0.48; 95% confidence interval 0.15, 1.59; P = .23). The odds of improving on the RSDI (odds ratio 0.51, 95% confidence interval 0.25, 1.04, P = .065) and CSS (odds ratio 0.98, 95% confidence interval 0.51, 1.89, P = .950) were not significantly different by revision status. Both revision and primary ESS patients improved after ESS with regard to endoscopy, RSDI, and CSS scores. In non-polyp patients, revision ESS patients were more likely to improve on endoscopy scores than primary ESS patients; there was no difference in polyp patients by revision status. Revision ESS patients and primary ESS patients were equally as likely to improve on two QOL instruments.

Age at surgery for undescended testis and risk of testicular cancer: Pettersson A, Richiardi L, Nordenskjold A, Kaijser M, Akre O, Clinical Epidemiology Unit, Department of Medicine, Karolinska Institutet, Stockholm, Sweden

Undescended testis, which is a risk factor for testicular cancer, is usually treated surgically, but whether the age at treatment has any effect on the risk is unclear. We studied the relation between the age at treatment for undescended testis and the risk of testicular cancer. We identified men who underwent orchiopexy for undescended testis in Sweden between 1964 and 1999. Cohort subjects were identified in the Swedish Hospital Discharge Register and followed for the occurrence of testicular cancer through the Swedish Cancer Registry. Vital statistics and data on migration status were taken from the Register of Population and Population Changes for the years 1965 through 2000. We estimated the relative risk of testicular cancer using Poisson regression of standardized incidence ratios, comparing the risk in the cohort with that in the general population. We also analyzed the data by means of Cox regression, using internal comparison groups. The cohort consisted of 16,983 men who were surgically treated for undescended testis and followed for a total of 209,984 person-years. We identified 56 cases of testicular cancer during follow-up. The relative risk of testicular cancer among those who underwent orchiopexy before reaching 13 years of age was 2.23 (95% confidence interval [CI], 1.58 to 3.06), compared with the Swedish general population; for those treated at 13 years of age or older, the relative risk was 5.40 (95% CI, 3.20 to 8.53). The effect of age at orchiopexy on the risk of testicular cancer was similar in comparisons within the cohort. Treatment for undescended testis before puberty decreases the risk of testicular cancer.

Antiretroviral treatment for HIV infection/AIDS and the risk of developing hyperglycemia and hyperlipidemia

A cross-sectional study with internal comparison groups was conducted to describe sociodemographic characteristics, as well as verify the association between the type of antiretroviral treatment used and hyperglycemia and hyperlipidemia, with special attention to the use of HIV protease inhibitors. The data was obtained through an interview questionnaire, as well as blood and urine samples that were collected for the laboratory exams. A total of 418 patients were interviewed. 46 of these, however, met the exclusion criteria. The sample was therefore composed by 372 HIV positive patients, attended at the laboratory of the Correia Picanço State Hospital for the collection of blood, to estimate the HIV viral load and/or TCD4 cell counts from August to November 2000. The association between the variables was tested using the chi-square test and the p-value. A multiple logistic regression analysis was carried out to adjust for potential confounding factors. A greater frequency of patients with high glucose levels was observed among those making use of antiretroviral therapy without protease inhibitors, but the number of patients limited the comparisons. An association was verified between the total serum cholesterol level and the use of HIV protease inhibitors (p=0.047) even after controlling for age. An association was also observed between the triglyceride levels and the use of HIV protease inhibitors, which remained after adjustment for age, sex and creatinine levels (p<0.001). The levels of glucose and TSH, the presence of proteinuria and the practice of physical activity were not associated either with the levels of cholesterol or with the levels of tryglicerides thus they were not confounders of the associations described.

Anatomical and Behavioral Effects of in Utero Exposure to Antiepileptic Drugs

Anatomical and Behavioral Effects of in Utero Exposure to Antiepileptic Drugs

Increased rate of major malformations in offspring exposed to valproate during pregnancy

To determine the rate of occurrence of major malformations in infants whose mothers had taken the drug valproic acid (VPA) as monotherapy during the first trimester of pregnancy and had enrolled in the North American Antiepileptic Drug Pregnancy Registry. Data were collected from pregnant women throughout the United States and Canada through telephone-based interviews. Each woman was interviewed at enrollment, at 7 months' gestation, and postpartum. With her written permission, the medical records of each mother and her infant were obtained. The major malformations tabulated were those identified at or before 5 days of age. The prevalence of congenital malformations among offspring of monotherapy VPA-exposed women was compared with that among infants of women exposed to all other antiepileptic drugs (internal comparison group) and with that among newborns in the Active Malformations Surveillance Program at Brigham and Women's Hospital (external comparison group). Sixteen affected cases were identified among 149 VPA-exposed women (proportion: 10.7%; 95% CI: 6.3 to 16.9%). The prevalence in the internal comparison group was 2.9% (95% CI: 2.0 to 4.1%; odds ratio: 4.0, 95% CI: 2.1 to 7.4; p < 0.001). Assuming a 1.62% prevalence in the external comparison group, the relative risk of having an affected offspring for VPA-exposed women was 7.3 (95% CI: 4.4 to 12.2; p < 0.001). Maternal exposure to valproic acid during the first trimester of pregnancy significantly increased the risk of major malformations.

Deaths from nasopharyngeal cancer among waiters and waitresses in Chinese restaurants

Previous studies have shown that waiters have a high risk of developing cancers of the buccal cavity and pharynx, but nasopharyngeal cancer (NPC) has not been specifically studied. This study was carried out to investigate whether waiters/waitresses in Chinese restaurants have an increased risk of dying from NPC. A mortality odds ratio study was used to estimate the relative risk of dying from NPC for waiters/waitresses working in Chinese restaurants in Hong Kong during the period 1986-1995, using the general population as the external comparison group and deceased kitchen workers as an internal comparison group. Cases were deaths from NPC and the controls were deaths from the selected sets of reference causes. Seventeen deaths from NPC were identified among 415 deceased waiters and four NPC deaths occurred among 140 deceased waitresses. The adjusted mortality odds ratio (aMOR) for NPC was increased among waiters, being 3.02 (95% CI 1.82-5.00) and 2.61 (95% CI 1.02-6.69) in the external and internal comparisons, respectively. For waitresses, the aMOR was 4.58 (95% CI 1.63-12.86) in the external comparison. Analysis by duration of union membership suggested a dose-response relationship. An increased risk of dying from NPC was observed among waiters/waitresses and could not be fully explained by bias or confounding factors. Possible risk factors related to poor indoor air quality in the service areas of Chinese restaurants in Hong Kong should be further investigated.

Site-specific cancer incidence and mortality after cerebral angiography with radioactive thorotrast.

Few opportunities exist to evaluate the carcinogenic effects of long-term internal exposure to alpha-particle-emitting radionuclides. Patients injected with Thorotrast (thorium-232) during radiographic procedures, beginning in the 1930s, provide one such valuable opportunity. We evaluated site-specific cancer incidence and mortality among an international cohort of 3,042 patients injected during cerebral angiography with either Thorotrast (n = 1,650) or a nonradioactive agent (n = 1,392) and who survived 2 or more years. Standardized incidence ratios (SIR) for Thorotrast and comparison patients (Denmark and Sweden) were estimated and relative risks (RR), adjusted for population, age and sex, were generated with multivariate statistical modeling. For U.S. patients, comparable procedures were used to estimate standardized mortality ratios (SMR) and RR, representing the first evaluation of long-term, site-specific cancer mortality in this group. Compared with nonexposed patients, significantly increased risks in Thorotrast patients were observed for all incident cancers combined (RR = 3.4, 95% CI 2.9-4.1, n = 480, Denmark and Sweden) and for cancer mortality (RR = 4.0, 95% CI 2.5-6.7, n = 114, U.S.). Approximately 335 incident cancers were above expectation, with large excesses seen for cancers of the liver, bile ducts and gallbladder (55% or 185 excess cancers) and leukemias other than CLL (8% or 26 excess cancers). The RR of all incident cancers increased with time since angiography (P < 0.001) and was threefold at 40 or more years; significant excesses (SIR = 4.0) persisted for 50 years. Increasing cumulative dose of radiation was associated with an increasing risk of all incident cancers taken together and with cancers of the liver, gallbladder, and peritoneum and other digestive sites; similar findings were observed for U.S. cancer mortality. A marginally significant dose response was observed for the incidence of pancreas cancer (P = 0.05) but not for lung cancer. Our study confirms the relationship between Thorotrast and increased cancer incidence at sites of Thorotrast deposition and suggests a possible association with pancreas cancer. After injection with >20 ml Thorotrast, the cumulative excess risk of cancer incidence remained elevated for up to 50 years and approached 97%. Caution is needed in interpreting the excess risks observed for site-specific cancers, however, because of the potential bias associated with the selection of cohort participants, noncomparability with respect to the internal or external comparison groups, and confounding by indication. Nonetheless, the substantial risks associated with liver cancer and leukemia indicate that unique and prolonged exposure to alpha-particle-emitting Thorotrast increased carcinogenic risks.

Labour market training in Norway—effect on earnings

By means of a 6-year panel of annual earnings data, we estimate the impact of a labour market training programme (LMT) targeted at unemployed adults. Unlike most nonexperimental studies, we use an internal comparison group of rejected applicants to measure the counterfactual outcome of trainees. For a subsample of applicants, rejections were randomised on a course-by-course basis. Following Heckman and Hotz [J. Am. Stat. Assoc. 84 (1989) 862] we test alternative estimators. All models report positive effects of being offered training, though the effects are not always significantly different from zero. The linear control estimator reports significantly positive effects of training. Analysing courses with ordinary assignment, the pre-training test indicates a positive selection bias. Turning to applicants for courses with random assignment, the linear control model is not rejected by the pre-training test. The fixed effect, random growth and modified random growth models are rejected by the model restriction test; this is the case for courses with ordinary assignment as well as random assignment. We illustrate the value of an internal comparison group in estimating the counterfactual outcome of participants, by comparing the post-training outcomes of rejected applicants and unemployed nonapplicants. Rejected applicants have significantly higher post-training earnings than eligible nonapplicants, which indicates that self-selection into the programme is influenced by unobservables. In general, the results clearly support the view that reliable studies of training effects should be based on internal comparison groups. Moreover, the assumptions motivating nonexperimental estimators should be explicitly tested.

Evidence-based medicine in otolaryngology journals.

We set out to assess, within the context of evidence-based medicine, the levels of supporting evidence for therapeutic recommendations made in leading otolaryngology journals. We used a cross-sectional survey of clinical research articles published in 1999 in 4 high-circulation otolaryngology journals. We used study design methodology and level of evidence for clinical research articles with therapeutic recommendations. Outcomes were stratified by type of recommendation (positive vs negative) and by study focus (medical vs surgical therapy). Of the 1019 articles identified, 737 (72%) were clinical research and 268 (36%) made therapeutic recommendations. Median sample size was modest (27 subjects), with only 38% of studies reflecting planned research and 22% including an internal control or comparison group. Positive studies were 20 times more prevalent than negative ones, but were 69% less likely to have an internal control group (P =.042) and 93% less likely to include confidence intervals (P =.020). Moreover, the level of evidence for positive studies was lower than for negative studies (P =.037), with twice as many negative recommendations supported by analytic research. Similarly, the level of evidence for operation was lower than for medical therapy (P <.001), with 3 times as many medical recommendations supported by analytic research. Most therapeutic recommendations in otolaryngology journals are on the basis of descriptive case series (80%) and least often on randomized controlled trials (7%). A dual standard appears to exist for negative versus positive studies and for medical versus surgical recommendations. Greater scrutiny of the breadth and quality of evidence levels supporting therapeutic recommendations is likely to occur as the popularity of-and demand for-evidence-based medicine increases. Evaluation of levels of evidence in otolaryngology decision making.

A mortality, morbidity, and hematology study of petrochemical employees potentially exposed to 1,3-butadiene monomer

This three-part study is an update of a previous report that examined the mortality, morbidity, and hematological data of employees from a petrochemical facility in Texas who had potential exposure to 1,3-butadiene monomer. The first part describes the updated cause-specific mortality of 614 workers. Vital status for each cohort member was ascertained through 1998, a 9-year extension of the previous study. The second part is an examination of the morbidity experience of cohort members who were still working during 1992–1998, including 289 of the 614 mortality cohort members. The third part is an evaluation of the hematological results from routine health surveillance and/or medical examinations. Approximately 430 of the 614 employees who had complete blood count (CBC) data as of December 1999 were included in the hematological evaluations. The most recent examination containing CBC data was used and compared with similar data for over 2600 other employees from this facility. Overall mortality during the follow-up period, 1948–1998, was significantly lower than for the local comparison population (standardized mortality ratio (SMR) of 0.55 with a 95% confidence interval (CI) of 0.42–0.70). Mortality for all cancer was also significantly lower (SMR, 0.57; 95% CI, 0.32–0.92). Mortality for all lymphohematopoietic cancer was about the same as the comparison population (SMR, 1.06; 95% CI, 0.22–3.11). None of the cause-specific morbidity was in excess compared with an internal comparison group. There were no differences in the distribution or mean values of heamatological variables (e.g. white blood cells, red blood cells, hemoglobin, platelets, mean corpuscular volume) between the butadiene cohort and the comparison group, or between a subgroup of workers in jobs with the highest potential for butadiene exposure (i.e. shipping) and the rest of the butadiene cohort. The findings of this study suggest that the butadiene exposure at this facility in the last 20 years does not pose a health hazard to employees.

Review: silicone breast implants do not increase the risk for connective tissue diseases

(2000) N Engl J Med 342, 781. Janowsky EC, Kupper LL, Hulka BS. . Meta-analyses of the relation between silicone breast implants and the risk of connective-tissue diseases. . Mar...

An exact method for meta-analysis of case-control and follow-up studies.

In this paper, we describe an exact method for estimating a common relative risk across different epidemiologic study designs. The types of studies allowed by the method include case-control studies, follow-up studies with an internal comparison group, and follow-up studies with an external comparison group. Because the method is exact, sparseness of individual studies is not an issue. Those wishing to perform a meta-analysis of case-control studies and follow-up studies in which both the exposure and outcome are rare will find the method particularly useful. To allow one to perform the computations efficiently, we present a partial polynomial multiplication algorithm. We also describe a public-domain computer program that performs the necessary calculations.

Characteristics of the Cohort of Workers at the Mayak Nuclear Complex

At Branch No. 1 of the Russian State Research Center "Biophysics Institute", a registry has been created of workers at the "Mayak" Production Association, the first nuclear complex in Russia. This registry includes 18,830 persons hired at Mayak's nuclear reactors and radiochemical and plutonium production plant between 1948 and 1972. Twenty-five percent of these workers are women. As of December 31, 1994, the vital status is known for approximately 90% of the cohort members. A total of 5,118 persons have died. The cause for 97% of total deaths has been ascertained. The cohort members were exposed to both external gamma radiation and internal radiation from incorporated plutonium. The plutonium body burden has been measured in 30% of the cohort members with potential for plutonium exposure. External gamma-ray doses were in the range from tenths of milligrays to about 10 Gy, and plutonium body burdens were up to about 260 kBq. In view of the nature of the Mayak worker cohort, it has the potential to provide reasonably precise, quantitative estimates of the long-term health effects associated with chronic low-dose-rate exposure to external gamma radiation as well as internal radiation from plutonium. However, a number of issues must be addressed before credible risk estimates can be obtained from this cohort. These issues include the development of an appropriate internal comparison group and/or external rates and separating of the effects of internal and external exposures on risk estimates.

Evaluating the vocational rehabilitation program using longitudinal data. Evidence for a quasiexperimental research design.

The study presents benefit-cost ratios for 14 disability cohorts served by the Vocational Rehabilitation (VR) Program. The earnings impacts are estimated in a quasiexperimental framework using an internal comparison group. The earnings data are from a unique national panel constructed by linking client data of the Rehabilitative Services Administration with earnings histories from the Social Security Administration. These earnings data accommodate a series of statistical tests that allow us to identify and control for the presence of selection bias when estimating treatment impacts. The results indicate that the VR program is cost-effective in general, although not universally so across specific disabilities.

Characteristics of Pesticide Use in a Pesticide Applicator Cohort: The Agricultural Health Study

Data on recent and historic pesticide use, pesticide application methods, and farm characteristics were collected from 35,879 restricted-use pesticide applicators in the first 2 years of the Agricultural Health Study, a prospective study of a large cohort of private and commercial licensed pesticide applicators that is being conducted in Iowa and North Carolina. (In Iowa, applicators are actually “certified,” while in North Carolina they are “licensed”; for ease of reference the term license will be used for both states in this paper.) Commercial applicators (studied in Iowa only) apply pesticides more days per year than private applicators in either state. When the types of pesticides being used by different groups are compared using the Spearman coefficient of determination (r2), we find that Iowa private and Iowa commercial applicators tend to use the same type of pesticides (r2=0.88). White and nonwhite private applicators tended to use the same type of pesticides (North Carolinar2=0.89), as did male and female private applicators (Iowar2=0.85 and North Carolinar2=0.84). There was less similarity (r2=0.50) between the types of pesticides being used by Iowa and North Carolina private applicators. A greater portion of Iowa private applicators use personal protective equipment than do North Carolina private applicators, and pesticide application methods varied by state. This heterogeneity in potential exposures to pesticides between states should be useful for subsequent epidemiologic analyses using internal comparison groups.

Serum liver function profiles in coking workers.

Coking workers are regularly exposed to coke oven emissions (COE), which consist mainly of polycyclic aromatic hydrocarbons and volatile organic compounds. In a previous cross-sectional study, we found that coking and by-product workers with heavy exposure to COE in the older of two coke operation areas in Taiwan had higher serum activities of hepatic aminotransferase than the controls. In this study, we further examine the relationship of exposure to COE with liver function profiles in coking workers. Liver function profiles included serum aspartate aminotransferase (AST), alanine aminotransferase (ALT), gamma-glutamyl transpeptidase (GGT), alkaline phosphatase (ALP), and total bilirubin (BIL). The exposed group included 88 workers working 3 months or more in the older coke oven plant. Fifty-nine referents, not visiting the coke operation areas in the last 3 months, came from the administrative area in the same company. Each participant wore a personal monitor that was used to measure benzene soluble fraction (BSF) of total particulates, as a surrogate of COE, for 3 consecutive days between August 1995 and February 1996. Serum liver function profiles, hepatitis B surface antigens, and anti-hepatitis C antibodies were examined in the morning following the exposure measurements. Exposure levels were categorized by exposure situations (high, medium, low) among coking workers. The high exposure group (n = 23) worked topside of the oven. The medium exposure group (n = 44) worked at the sideoven for more than 4 hr/day, whereas the low exposure group (n = 21) worked at the sideoven for less than 4 hr/day and mostly remained in the control rooms. The low exposure group was used as an internal comparison group. The median BSF concentrations for various exposure situations were as follows, high exposure group: 372 micrograms/m3, medium exposure group: 61 micrograms/m3, low exposure group: 49 micrograms/m3, and referents: 10 micrograms/m3. The coking workers (n = 88) did not significantly differ from the referents (n = 59) in any of the liver function profiles. Excluding the referents, workers in the high exposure group had a mean AST level that was 31% higher (95% confidence interval (CI) = 9-57%) and a mean ALT level that was 46% higher (95% CI = 7-98%) than those in the low exposure group after adjusting for appropriate confounders in multivariate models. The prevalence of an abnormal hepatocellular pattern (AST > 37 IU/L or ALT > 39 IU/L) was more common in the high exposure group than in the low exposure group (adjusted odds ratio = 4.4; 95% CI = 0.9-22.6). However, these associations were not found in GGT, ALP, or BIL. After controlling for the possible effects of nonoccupational factors on serum activity of AST and ALT, we conclude that increased AST and ALT levels among topside coking workers may be caused by heavy inhalation exposure to COE. Additionally, the adverse hepatic effect seems to be caused by a mixture of hazards, rather than a unique identifiable chemical.