Potential conflict of interest: Nothing to report.
Primary biliary cirrhosis (PBC) and primary sclerosing cholangitis (PSC) are uncommon liver diseases of unknown etiology. Reported clustering of PBC cases may be due to environmental factors. Individuals with PBC have a high prevalence of thyroid disease and thyroid disease is reportedly more prevalent near Superfund toxic waste sites (SFS). The objective of this study was to examine the prevalence and potential clustering of individuals with PBC and PSC near SFS. De-identified clinical and demographic data were used to determine the observed prevalence for each New York City zip code (n = 174) and borough (n = 5) of patients with PBC (PBC-OLT) or PSC (PSC-OLT) who were listed for liver transplantation. The expected prevalence was calculated using Organ Procurement and Transfer Network (OPTN) and U.S. Census data. Both PBC-OLT patients and patients not listed for liver transplantation (PBC-MSSM) were included in the cluster analysis. Prevalence ratios of PBC-OLT and PSC-OLT cases were compared for each zip code and for each borough with regard to the proximity or density of SFS, respectively. SaTScan software was used to identify clusters of PBC-OLT cases and PBC-MSSM cases. Prevalence ratio of PBC-OLT, not PSC-OLT, was significantly higher in zip codes containing or adjacent to SFS (1.225 vs. 0.670, respectively, P = .025). The borough of Staten Island had the highest prevalence ratio of PBC-OLT cases and density of SFS. Significant clusters of both PBC-OLT and PBC-MSSM were identified surrounding SFS. In conclusion, toxin exposure may be a risk factor influencing the clustering of PBC cases. (HEPATOLOGY 2006;43:525–531.)
Primary biliary cirrhosis (PBC) is a chronic, cholestatic liver disease of unclear etiology characterized by nonsuppurative intrahepatic portal tract inflammation and bile duct destruction, which may progress to cirrhosis.1 Elevated serum IgM levels and high titer serum reactivity against mitochondrial autoantigens are the other hallmarks of the disease.2, 3 Most individuals diagnosed with PBC, typically middle-aged women, also suffer from other autoimmune diseases, such as autoimmune thyroid disease.4 The role of retroviral infection in the disease remains undefined and controversial.5, 6 Large geographic variations in the prevalence of PBC have been reported.7–10 Differences in both genetic susceptibility factors and exposure to environmental toxins are widely believed to contribute to this variability in the prevalence of PBC.
Several studies support the importance of genetic susceptibility factors in the pathogenesis of PBC.11–14 Immune-related gene polymorphisms are increased in frequency in PBC.15 Also, the concordance rate of PBC among monozygotic twins (60%) is significantly higher than among dizygotic twins (0%).14 The concordance rate in monozygotic twins is as high or higher than that observed in other autoimmune diseases believed to have a genetic component. Lastly, several studies have shown that the prevalence of PBC within families of affected individuals is significantly higher than in the general population.13, 16
Differences in the local prevalence of PBC have been reported7–10, 17, 18 and may reflect differences in exposure to environmental toxins. However, strong linkage to any specific environmental factor has not been identified except in a Japanese study18 showing an increased prevalence of PBC in Hiroshima among atomic bomb survivors. Interestingly, a number of common chemical reagents, particularly aromatic and halogenated hydrocarbons, present in both industrial and household compounds may elicit mitochondrial autoantigen modifications in vitro that enhance autoantibody reactivity.19–21 These chemical reagents are often the major toxins identified at toxic waste sites.22 In New York City, particularly hazardous toxic waste sites have been designated New York State Superfund sites (NYC SFS) by the Department of Environment and Conservation (DEC). Deleterious effects on the immune system of individuals living near SFS have been reported by Vine et al.,23 and thyroid disease was increased among women living near SFS in NYS containing polychlorinated biphenyls.24 Thus, we decided to determine if the prevalence of PBC is increased near NYC SFS employing data from several sources and multiple analytic methods.
PBC, primary biliary cirrhosis; NYC SFS, New York City Superfund sites; DEC, Department of Environment and Conservation; PSC, primary sclerosing cholangitis; PBC-OLT, PBC patients listed for liver transplantation; PSC-OLT, PSC patients listed for liver transplantation; OPTN, Organ Procurement and Transfer Network; MSSM, Mount Sinai School of Medicine, New York; PBC-MSSM, PBC patients followed at MSSM; spr, standardized prevalence ratio.
Study Area and Data Used.
New York City has a total area of 368 sq mi (953 km2) consisting of 174 zip codes divided among 5 boroughs (Bronx, Brooklyn, Manhattan, Queens and Staten Island). The Mount Sinai School of Medicine (MSSM) is located in Manhattan. The average zip code area is approximately 2 sq mi (5.5 km2). The residential zip codes of all PBC (n = 99) and primary sclerosing cholangitis (PSC) (n = 73) patients listed for liver transplantation in NYC (PBC-OLT and PSC-OLT patients, respectively) between 1995 and 2003 were obtained from the Organ Procurement and Transfer Network (OPTN). The use of OPTN data diminishes the effect of bias due to physician referral patterns. PSC patient data was included to control for non-specific effects of toxin exposure on cholestatic liver disease. The OPTN database contains only the residential zip code submitted at the time of listing. Street addresses are not recorded. The residential zip codes at the time of PBC diagnosis of non-listed patients (n = 172) followed at the MSSM (PBC-MSSM) were obtained from a clinical research database that includes patients cared for between 1990-2004. Informed consent in writing was obtained from each patient. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the Institutional Review Board. The New York State DEC website lists the location and major known contaminants of all state SFS.22 The website map has some acknowledged errors with regard to the zip code location of several NYC SFS. The corrected locations were used in our analyses.
A standardized expected prevalence of PBC-OLT and PSC-OLT patients was calculated for each NYC zip code based on demographic data available from OPTN for region 9 (New York and Vermont) for the years 2000-2004. Adjustments were made for gender, race, age distribution, and population differences among the 174 zip codes using U.S. Census 2000 data. Observed cases were from 1995 to 2003. The standardized prevalence ratio (Observed/stdExpected) of PBC-OLT and PSC-OLT patients listed was also calculated for each zip code. Overall, the standardized prevalence ratios for PSC-OLT were much lower than those for PBC-OLT. This was not due to fewer than expected cases of PSC in NYC compared to region 9, but was an artifact due to a surge in the number of PSC patients listed for transplantation from 2000-2004 compared to earlier years.
The mean standardized prevalence ratios in zip codes containing or adjacent to a SFS (n = 89) was compared to that in the remaining zip codes (n = 85). Additionally, the standardized prevalence ratio of PBC-OLT patients was calculated for each NYC borough along with the density of SFS within each NYC borough. The Mann-Whitney U test was used to compare values. Lastly, SaTScan, a statistical software package designed to detect geographic or temporal clustering, was used to identify clusters of PBC patients independent of SFS locations (global analysis) or with regard to SFS locations (focused analysis). The ability of this software package to identify clusters has been previously validated25 and the software was used for similar analyses.26 The software utilizes longitude and latitude data rather than zip codes. The longitude and latitude of the geographic center of NYC zip codes were used in lieu of the exact address of individual patients. In the focused analysis, the longitude and latitude of each SFS was input as well.
Mapping of Standardized Prevalence Ratios for PBC and PSC.
Primary sclerosing cholangitis (PSC) is a distinct chronic, cholestatic liver disease, also of unknown etiology, that affects predominately men (7 men: 3 women) and has a younger age distribution than PBC. The standardized prevalence ratios of PBC-OLT and PSC-OLT patients (sprPBC-OLT and sprPSC-OLT, respectively) between 1995 and 2003 were calculated and plotted for each NYC zip code. Each zip code was shaded in proportion to its sprPBC-OLT (Fig. 1, panels A, C, E, and G) and sprPSC-OLT (Fig. 1, panels B, D, F, and H). Darker shading indicates a higher prevalence ratio. The geographic patterns of distribution of the sprPBC-OLT and sprPSC-OLT were dissimilar, particularly in Queens (Fig. 1, panels C and D, respectively) and Brooklyn (Fig. 1, panels G and H, respectively). Individual zip codes with the highest sprPBC-OLT are located in Staten Island and Queens (Fig. 1, panels A and C, respectively). No obvious relationship between the locations of SFS and the prevalence patterns of PBC or PSC was immediately evident by mapping.
Comparison of Standardized Prevalence Ratios.
Zip codes containing a SFS were identified. Because the area of each zip code in NYC is small and many of the SFS were located along zip code borders, both zip codes containing a SFS and those adjacent to such zip codes were considered SFS+ zip codes (N = 89). The median radius of these SFS+ zip code clusters was less than 2 mi. The remaining zip codes were categorized as SFS− zip codes (N = 85) (Fig. 2). A standardized expected (std exp) number of cases for each zip code was calculated by indirect standardization for gender, race, age distribution, and population.27 In the former zip codes (+SFS), the number of observed PBC-OLT cases was 63 (std exp = 51 cases) and observed PSC-OLT was 39 (std exp = 98 cases). In the latter zip codes (−SFS), the number of observed PBC-OLT cases was 36 (std exp = 53 cases) and observed PSC-OLT was 34 (std exp = 99 cases). For each zip code the observed number of cases was divided by the std exp number to obtain standardized prevalence ratios.
Similar to Carpenter et al.,24 statistical analysis was done comparing the sprPBC-OLT and the sprPSC-OLT in areas with SFS and areas without SFS. The sprPBC-OLT was significantly higher in zip codes containing or bordering a SFS than in the zip codes that neither have nor border a SFS (mean 1.2 vs. 0.7, respectively P = .025) (Table 1). In contrast, the sprPSC-OLT of PSC cases was not statistically different between the two zip code groups (P = .589). The type of SFS (landfill or dump vs. building) affected the sprPBC-OLT. The mean sprPBC-OLT was 1.75 for the 15 zip codes containing SFS that were dumps or landfills and 0.33 for 10 zip codes containing SFS that were buildings (P = .045). The mean sprPSC-OLT for each type of SFS was not significantly different (P = .53). Thus, only the prevalence of PBC-OLT cases was increased near SFS, particularly near dumps and landfills designated SFS.
Table 1. Standardized Prevalence Ratio of PBC and PSC Patients Listed for Transplantation
+SFS (n = 89)
-SFS (n = 85)
NOTE. Values are shown as mean ± standard deviation.
Abbreviations: PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; +SFS, zip codes with or adjacent to Superfund sites; &miknius;SFS, remaining zip codes.
1.225 ± 1.723
0.670 ± 1.253
0.397 ± 0.818
0.341 ± 0.779
Since duration of exposure to toxins may affect disease occurrence and/or progression, patient address changes may affect the above observations. The median time reported between diagnosis of PBC and listing for transplantation varies among studies,28, 29 but probably is at least ten years. The median time between address changes for all PBC patients followed at Mount Sinai is 16 years, however, it is not known how long patients listed for transplantation lived in their current zip codes. Therefore, the prevalence ratio of PBC patients listed for transplantation in each NYC borough was compared to the density of SFS in each borough. This comparison should be less affected by address changes because U.S. Census 2000 data indicates that most NYC residents remain in the same borough when changing their address.
The density of SFS and the sprPBC-OLT were both highest in Staten Island (Table 2). Conversely, Manhattan had the lowest sprPBC-OLT and the lowest density of SFS. The overall sprPBC-OLT among Brooklyn, Queens, and the Bronx were very similar. Consequently, there was no statistically significant correlation between the density of SFS and the sprPBC-OLT among the boroughs as a group. A similar suggestive, but statistically non-significant, correlation was obtained if the percentage of each borough's area covered by SFS was correlated with the sprPBC-OLT for each borough (data not shown).
Table 2. Comparison of PBC Prevalence and Density of SFS Among New York City Boroughs
Because SFS are a heterogeneous group of toxic waste sites, the increased sprPBC-OLT near SFS may have been owing to the influence of a subset of SFS. Also sources of environmental toxins unrelated to SFS may influence the geographic distribution of PBC patients. To identify individual clusters of PBC patients, global cluster analysis was performed using a validated, statistical software package (SaTScan) that evaluates the distribution of a set of coordinates defined by their longitude and latitude. Five independent clusters of PBC-OLT were identified (Fig. 3, solid circles). None of these clusters were statistically significant (all P > .05), in part due to the relatively low number of patients in the analysis. However, SFS were present within the borders of 4 out of 5 clusters. Two clusters each were identified in Brooklyn and Queens, while the other cluster was in Staten Island. None of these clusters contained liver transplantation centers or major medical centers within their borders. The major contaminants at these sites were volatile organic compounds (e.g., benzene and toluene) and chlorinated hydrocarbons (e.g., tri- and tetra-chloroethylene).
Global cluster analysis of the residential zip codes from the time of diagnosis of the Mount Sinai cohort of non-listed PBC patients (PBC-MSSM) (n = 172) identified 3 statistically significant clusters (Fig. 3, dashed circles). As expected, one cluster was in Manhattan near Mount Sinai Hospital, which likely reflected physician referral bias or greater physician awareness and diagnosis of PBC. One PBC-MSSM cluster overlapped with the PBC-OLT cluster in Staten Island, and the other PBC-MSSM cluster was near the PBC-OLT cluster in southern Brooklyn. A review of the medical records of the patients in the latter two PBC-MSSM clusters does not suggest that these clusters were due to physician referral bias or affiliations with MSSM. The SFS within the area of overlap in Staten Island is an 11-acre dump contaminated with benzene, other volatile organic compounds, and trichloroethylene, and the SFS in southern Brooklyn is an 11-acre dump contaminated with benzene alone. PBC-MSSM and PBC-OLT patients are clustered in the same geographic areas near SFS contaminated with benzene.
Focused cluster analysis was done of the distribution of PBC-OLT and PBC-MSSM patients by including the longitude and latitude of each NYC SFS. When the locations of SFS were included, the PBC-OLT cluster previously identified in Staten Island by global analysis was statistically significant (P = .039). The same two PBC-MSSM clusters were also statistically significant by focused cluster analysis relative to the location of each NYC SFS. Subdividing the SFS by the type of site (dump and landfill vs. structure), size, class, borough location, or contaminants did not identify any additional statistically significant clusters. The New York State DEC website also provides the location of sources of toxic waste other than SFS, but these sites did not correspond to the observed clusters of PBC patients (data not shown). Thus, statistically significant overlapping clusters of both patients with early (PBC-MSSM) and late stage (PBC-OLT) PBC were identified near a SFS in Staten Island.
Environmental toxin exposure has been hypothesized to be an etiological factor in the pathogenesis of PBC based on geographic differences in the prevalence of PBC and on the antigenicity and immunogenicity of xenobiotically modified pyruvate dehydrogenase, the major PBC autoantigen.19, 21 Extensive analysis suggested modification of the autoantigenic epitope by long chain aliphatic acids, acids containing ethylene or acetylene bonds, or halogenated aromatic hydrocarbons may mimic its naturally occurring modification by lipoic acid.21 Such compounds may be used as additives in commercial products and industrial processes, and are also found in high concentrations in toxic waste sites. Epidemiological data showing increased prevalence or clustering of PBC patients near known sources of xenobiotics has been lacking. This study shows that the prevalence of PBC patients listed for transplantation was increased near NYC SFS. Additionally, a statistically significant PBC patient cluster, including both patients not listed for transplantation and those listed for transplantation, was identified in Staten Island near a SFS contaminated with volatile aromatic hydrocarbons and trichloroethylene. This epidemiological study shows a statistically significant clustering of PBC patients near known sources of environmental toxins and is one of relatively few studies linking proximity to SFS to a specific disease.
The study by Carpenter et al.24 linking proximity to SFS and thyroid disease focused on polychlorinated biphenyls-contaminated SFS. Many of the NYC SFS were contaminated with polychlorinated biphenyls, but none of these were associated with a statistically significant PBC patient cluster. The association between women with thyroid disease and proximity to polychlorinated biphenyls-contaminated SFS was obviously independent of PBC because very few patients with thyroid disease have PBC. Though many patients with PBC have autoimmune thyroid disease, the thyroid disease develops both before and after PBC. SFS may be a common source of environmental toxins influencing both diseases. Alternatively, the same environmental toxin may play a role in a number of diseases, while genetic susceptibility factors determine the outcome of toxin exposure. Polychloroethylene and benzene were the major contaminants at SFS within the area of overlap of the PBC-OLT and PBC-MSSM patient clusters on Staten Island.
Murine studies by different groups indicate that such halogenated and aromatic hydrocarbons affect hepatic and mitochondrial function as well as IgM production.30–33 Several of these compounds may form mixed disulfides with protein sulfydryl groups when glutathione levels are reduced, which may affect autoantigen immunogenicity (for review see34). Interestingly, Kanz et al. demonstrated that preexisting hypothyroidism affects metabolism of some halogenated hydrocarbons and may slightly increase their hepatotoxicity.35
The precise mechanism(s) by which proximity to SFS may increase PBC patient exposure to these toxins is not clear. A previous study by O. James' group from the United Kingdom identified significant clusters of PBC patients in the Newcastle region and speculated that differences in regional water supplies might account for clustering.10 Although the groundwater near 85% of SFS is contaminated, NYC residents do not rely on groundwater as a water supply. Consequently, inhalation of volatile organic compounds (e.g., benzene) and particle bound chlorinated hydrocarbons released into the air from the SFS represents a more likely exposure mechanism. Cigarette smoking, another source of benzene and other volatile organic compounds, and chronic bronchitis are associated with PBC.12, 36 Interestingly benzene is known to induce both hepatic oxidative stress37 and immune dysfunction.38, 39 Additionally, women have a higher pulmonary air/blood partition coefficient40 for benzene and female gender enhances hepatic37 sensitivity to benzene. The study by Amano et al.21 suggests halogenated forms of benzene may mimic the lipoylated PBC autoantigen epitope. These related findings and our current study suggest air-borne volatile organic compounds such as benzene may contribute to both loss of self-tolerance and the pathogenesis of PBC.
Certain types of work expose employees to higher levels of volatile organic compounds, but unfortunately available employment data on patients with PBC is limited. Many SFS are old chemical manufacturing buildings, where people worked before the facilities were closed and contaminants became a concern. However, our results suggest that simply living near these types of SFS does not increase the prevalence of PBC. Buildings may offer protection against the release of volatile organic compounds from within.41 The only significant clusters of PBC patient residences we identified by SatScan were near SFS that were unenclosed dump sites. The association between proximity to toxic sites and PBC-OLT prevalence may occur owing to an increased incidence of PBC or accelerated PBC progression in areas near toxic sites.
Confounding factors may have influenced the results of our analysis. For example, socioeconomics may affect both where people live and access to medical care. Data analysis according to boroughs limited the effect of patient mobility; however, patient mobility may still have affected our results. Spatiotemporal mismatches among zip codes may inadvertently affect cluster analysis as described. previously.42 Also, the increased risk of PBC among family members of those with PBC might also cause clustering of patients independent of environmental risk factors. However, at the Mount Sinai Medical Center, two members of the same family have never been listed for transplantation for PBC. MSSM patients listed for OLT were not included in the PBC-MSSM cohort to avoid having the same patient in both the PBC-OLT and PBC-MSSM group. Repeating the analysis in other regions may eliminate confounding factors that were unique to NYC.
The cluster of PBC-MSSM patients near the Mount Sinai Medical Center in Manhattan illustrates that referral pattern bias or increased physician awareness may create clusters of patients. However, the clusters in other boroughs were not near major medical centers, and the patients in these clusters were not referred by a common physician, group practice, or institution. The overlap between PBC-MSSM and PBC-OLT clusters may have occurred if patients followed at Mount Sinai were also preferentially listed for transplantation because Mount Sinai Medical Center is a liver transplantation center. Such an effect was likely limited during the years analyzed since several other area medical centers had active liver transplantation programs during the years analyzed. Additionally, there was no PBC-OLT cluster overlapping with the PBC-MSSM cluster surrounding MSSM itself.
Our observations further support the hypothesis that environmental toxin exposure is a risk factor for PBC. Researchers in countries with national patient registries and accurate toxic waste site data may be in an ideal position to investigate further the role of environmental toxin exposure in PBC. Performing similar studies for other autoimmune diseases will aid in determining the relative role of xenobiotics in the pathogenesis of autoimmune disease. The New York State DEC website also provides information on sources of toxic waste other than SFS, which may permit investigation of an association between other toxin sources and autoimmune disease.
Data from OPTN (Organ Procurement and Transplantation Network), New York State Department of Environment and Conservation (DEC), Recanati/Miller Transplant Institute were invaluable in conducting this analysis. We would like to thank S. Wallenstein for helpful advice regarding statistical analysis.