Evolving Trends in Female to Male Incidence and Male Mortality of Primary Biliary Cholangitis

Primary biliary cholangitis (PBC) has been regarded as female-predominant without evidence of gender difference in survival. We aimed to compare the overall survival, incidence and prevalence of PBC in two well defined population-based studies over a recent decade, considering also sex ratios and mortality. We have taken advantage of population-wide records, during 2000–2009, in Lombardia, Northern Italy, and Denmark. We focused on the incident cases of PBC, including gender and outcome, among 9.7 million inhabitants of Lombardia and 5.5 million of Denmark. In Lombardia there were 2,970 PBC cases with a female:male ratio of 2.3:1. The age/sex-adjusted annual incidence of PBC was 16.7 per million. Point prevalence was 160 per million on January 1st 2009. In Denmark there were 722 cases of incident PBC, female:male ratio was 4.2:1, and the annual incidence was 11.4 per million, a point prevalence of 115 per million in 2009. Cox regression multivariate analysis identified male sex as an independent predictor of all-cause mortality in both Italian (HR 2.36) and Danish population (HR 3.04). Our data indicate for PBC a sex ratio significantly lower than previously cited, a reversal of the usual latitudinal difference in prevalence and a surprisingly higher overall mortality for male patients.

Statistical analysis. The incidence rate was defined and computed as the number of patients with a first-time diagnosis of PBC in a particular year divided by the total number of inhabitants at the beginning of that year. For Lombardia, we used a 4 year washout period (2000)(2001)(2002)(2003)(2004) to avoid including prevalent cases due to the fact that data recording started in 2000, and determined a 6 year span time (from Jan 1 st 2004 to Dec 31 st 2009) with no significant variation of the Standardized Incidence Ratio (SIR) (Poisson regression rate P = 0.15). The incidence rate for a range of years was obtained by summing the numerator and denominator for each year. All incidence rates were age-standardized to the WHO Standard Population 41 with the direct standardization method and 95% confidence intervals (CI) based on the Poisson distribution. The prevalence was computed as the proportion of citizens with PBC divided by the total number of citizens at the beginning of each year, again standardized to the WHO Standard Population. We used the Kaplan-Meier method to estimate survival probabilities, and we used Cox regression to estimate the mortality hazard ratio for men versus women adjusting for age at diagnosis and calendar year of diagnosis, both entered as continuous variables. Relative survival analysis was performed in order to estimate PBC patients' survival with respect to that of the general population. The analysis was performed using public open access databases reporting general population all causes mortality, including separate data for sex and class of age. All statistical analyses were done using Stata statistical software 12.0 (Stata Corporation, College Station, TX, USA).

Results
Incidence, prevalence, and female:male ratio. On Jan 1 st 2010 there were 9,742,676 registered residents in Lombardia. Among these, the PBC cohort comprised 2,970 patients of whom 2,073 (69.8%) were female. The mean age at diagnosis was 61.7 (95% CI 61.5-62.7), with a slight, but statistically significant, difference between males (60.8 years) and females (62.  (Table 1A). However the trend between both populations was not homogeneous as confirmed by the regression model applied (Lombardia Poisson regression rate p < 0.0005, Denmark Poisson regression rate p = 0.48) (Fig. 1A). We hypothesize that, since no administrative data in Lombardia were available before 2000, in the first years after 2000 there would be a mix of truly incident patients and patients who were in fact diagnosed with PBC before 2000. We therefore assume a 4-year washout was needed in order to avoid SIR overestimation and determined a six-year span time (from Jan 1 st 2004 to Dec 31 st 2009) with no significant variation of SIR (p = 0.15) (Fig. 1B). The sex-specific SIR shows that the F:M ratio remained constant over the studied decade in both Lombardia and Denmark (data not shown).  The mean point prevalence of PBC in Lombardia in 2009 was 160 cases per million (234 female, 81 male) Table 1B. The prevalence in 2009 seems to be the most reliable one for Lombardia since the rate did not reach a plateau during the follow-up period (Fig. 2). Indeed, whereas point prevalence remains constant during the study time in Denmark (Poisson P = 0.12), it increases from 2000 to 2009 (P < 0.0005) ( Fig. 2A,B). Such dynamics seem to be related to the method in use for diagnosis identification. In Lombardia a diagnosis of PBC was made at admission to the hospital starting from 2000 whereas Danish registries recorded a prevalent diagnosis even if it was made before 2000. In order to avoid this limitation we put together the exemption code registry in Lombardia and the inpatient registry data, eliminate duplicates, and calculated the point prevalence in 2010 which was 295 cases per million (486 female, 159 male) Survival and prognostic factors: males fare badly/do worse. 2,970 incident PBC cases in Lombardia were observed for a mean follow-up of 4.8 years, and 14,446 person-year from PBC diagnosis. The 1, 5 and 10 year survival rate estimation was respectively 86% (CI 85-87%), 70% (CI 69-72%) and 61% (CI 59-64%) (Fig. 3A). The sex specific survival at 1, 5 and 10 years was significantly higher for females (n = 2,073), respectively 89% (CI 88-91%), 77% (CI 75-78%) and 67% (CI 65-70%) than for males (n = 897), respectively 78% (CI 75-80%), 55% (CI 52-59%) and 47% (CI 43-51%) (Fig. 3B). In Denmark a survival analysis was done by following 722 incident cases for 2,780 person-years from PBC diagnosis, with an mean follow up of 3.9 years. Survival rate estimates, at 1, 5 and 10 years, were similar to those observed in Lombardia, respectively 87% (CI 84-89%), 67% (CI 63-71%) and 54% (CI 48-60%) (Fig. 3C). The sex specific survival rate in Denmark resembled the higher mortality in men with PBC observed in Lombardia since, in Denmark, survival of females at 1, 5 and 10 years was 90% (IC 87-92%), 73% (CI 69-77%) and 60% (CI 53-67%) and in males, 72% (CI 63-79%), 42% (CI 32-51%) and 27% (CI 14-43%) (Fig. 3D). Most importantly the relative survival analysis demonstrated that the higher mortality in men with PBC was independent of the higher male mortality in the general population (Fig. 4).
Using the Cox proportional hazard model, including sex, age at diagnosis and year of diagnosis (Table 2), male sex was associated with a higher risk of all-causes death for both populations: Hazard Ratio (HR) 2.36 in Lombardia, and 3.04 in Denmark.

Discussion
In the present study we report the epidemiology and natural history of PBC for two populations, Lombardia and Denmark, reflecting two diverse latitudes of continental Europe. Our findings include a significantly lower than

Table 2. Cox regression model for predictors of mortality in PBC (2000-2009).
hitherto cited F:M ratio for both the Italian and Danish populations, a probable decrease in incidence over time (particularly for Lombardia), a lack of any latitudinal prevalence gradient, and a surprisingly higher overall mortality for male than female PBC patients for both populations. One strength of our study is the use of population-based databases, which obviates the selection bias inherent in studies restricted to just a single or a few health facilities or reference centers. The use of administrative data has been well validated for the Danish model, ensuring registration of any event within the health system of defined geographical boundaries 40 .
Previous epidemiological research on PBC, mainly European, has reported incidence rates ranging from 2 to 49 cases per million/year and prevalence of between 19 and 402 cases per million (Table 3) 32 ; our data fit within the wide range reported and are significantly different among them. Of note, our data showed that the trend between both populations was not homogeneous with an apparent decrease in incidence over time especially in Lombardia. We hypothesize that, since no administrative data in Lombardia were available before 2000, in the first years after 2000 there would be a mix of truly incident patients and patients who were in fact diagnosed with PBC before 2000. Indeed, assuming a 4-year washout was needed in order to avoid SIR overestimation, no significant variation of SIR was determined.
On the other side, the much larger number of diagnoses of PBC in Lombardia may be explicable by a heightened awareness of physicians to liver diseases since, in Italy as in other Mediterranean countries, there is a higher prevalence of viral hepatitis. Moreover, the heterogeneity of epidemiological data can be the result of demographic changes, increasing survival, earlier diagnosis, improved treatment and greater recognition of the disease by physicians and patients 32 . Furthermore, the diagnostic rate of PBC may have overall increased during these years.
The ratio of females to males with PBC in Lombardia was 2.3:1, lower than that observed in Denmark (4.2:1); however, in both cohorts the female:male ratio was far lower than any other sex ratio previously reported (Table 3). Of note, two other large epidemiological studies based on administrative data demonstrate a higher than expected presence of male PBC patients (Table 4) 44,45 . One of these is a recent paper comprising data from 33 different autoimmune disease patients from administrative databases in Sweden 44 . The authors included 2,852 PBC patients, 1,080 of which were male, showing a female: male ratio of 1.6:1, in accordance with the trend revealed by our data and other population based studies 45 . Data obtained from reference centers demonstrate that prevalence is strongly biased to females (9:1), possibly influenced by the greater use by women of medical check-ups in highly specialized centers. The present analysis of population-wide registers, however, reflects a muted difference in the frequency of PBC between the sexes more in conformity with that seen in other autoimmune diseases.
Importantly, besides the differences in incidence and prevalence in Northern Italy and Denmark, the relatively higher mortality in male PBC patients was consistent in both datasets. Indeed, male sex was confirmed to   be an independent predictor of all-cause mortality in both populations. This finding has been reported once 46 , but has not been generally replicated 47,48 ; however, the Canadian epidemiological study 45 did report male sex as an independent predictor of mortality (HR 5.06). Several hypotheses can be offered for the causes of increased male mortality, including lack of compliance, additional environmental exposures and unknown sex factors that may modulate immunity. Thus, Carbone et al. in a recent study from UK, showed that men were significantly less likely to have responded to ursodeoxycholic acid (UDCA) than women; male sex was an independent predictor of nonresponse on multivariate analysis 49 . Moreover, Carbone and colleagues reported than men were less likely to be symptomatic, which may be a cause of delay in diagnosis 49 . Unfortunately no data regarding symptoms, use of UDCA and adherence to therapy were available in our database; similarly no clinical data were available in our database that would have helped to ascertain whether liver-related deaths were more frequent. We used the need for liver transplant as a surrogate endpoint for liver related death, but only 39 PBC patients in Demark and 58 in Lombardia were transplanted during the decade of study, suggesting that there was no association. Our study has some weaknesses that could, at least in part, explain the reported differences between the two included populations. First, the diagnosis code used in Lombardia (ICD-9-CM 571.6) identifies cases of biliary cirrhosis including therefore cases of secondary biliary cirrhosis. Unfortunately, with ICD-9 secondary biliary disease that cannot be excluded, which reduces the specificity and may explain the higher number of cases of PBC reported in Lombardia. Of note, primary sclerosing cholangitis (PSC) has its own code in the Italian system and, therefore, no PSC patients were included. Second, a non-coded diagnosis of PBC at release from the hospital, as well as among outpatients with no exemption code would not be identified in the Italian database whereas the Denmark Registry does include that information. Third we obtained dates of death only from in-hospital events since it was not possible to obtain these from the exemption code. This will probably lead to underestimation of overall mortality, however it will not change the male: female ratio.
In conclusion, our population studies based on disease registers questions some long-held beliefs on determinants of PBC. Females, although more frequently affected by PBC, seem to have a better outcome. Our findings clearly define an important clinical message for the hepatologist: male PBC patients have higher mortality than their female counterparts, such that close clinical follow-up and checking adherence to therapy are strongly recommended. In addition, a steadily increasing prevalence in Lombardia points to possible unsuspected environmental effects in causation, at least in that population, although biased estimation due to factors like increasing awareness of the disease or increasing hospitalization of patients cannot be excluded. Finally, the cause(s) of increased male mortality in PBC calls for a larger and prospective study and should be addressed in the future with well-designed observational trials.