Porphyria cutanea tarda increases risk of hepatocellular carcinoma and premature death: a nationwide cohort study

Data collection was derived from five nationwide Norwegian population based sources: The Norwegian Porphyria Registry [16], The Cancer Registry of Norway [17], The Cause of Death Registry of Norway [18], Statistics Norway, and the Norwegian Labour and Welfare Administration. Using a unique national 11-digit personal identification number, record linkage was performed in 2018.

Persons in Norway who are diagnosed with porphyria are invited to participate in the national Norwegian Porphyria Registry, administered by the Norwegian Porphyria Centre, established in 2000 [16]. Data is derived from patient-reported questionnaires supplemented with biochemical and genetic laboratory results. Data included in this report were updated one or two years after the first questionnaire, and thereafter every four years.

The mandatory national Norwegian Cancer Registry, records all new neoplasms since 1953. Diagnoses are based on a modified version of the 7th revision of the International Classification of Diseases (ICD-7), and completeness is greater than 98% [17]. The Cause of Death Registry of Norway has a coverage greater than 98% [18]. Diagnostic codes are based on ICD-9 and ICD-10.

Statistics Norway has maintained the Population Database, with demographic data for the entire population, since 1876, and The Norwegian Standard Classification of Education, with information about educational attainment, since 1970.

Statistics Norway and Norwegian Labour and Welfare Administration have maintained records regarding access to social security benefits: sick leave benefit, the disability pension, and medical and occupational - rehabilitation benefit since 1992. Diagnostic codes include the second revision of the International Classification of Primary Care (ICPC-2), ICD-9 and ICD-10.

All Norwegian residents, registered in the Norwegian Population Registry and 18 years or older from 2000 to 2016 were included in our cohort, and the reference population comprised of 5,451,951 adults. Of the 790 persons with a known overt diagnosis of PCT in Norway, registered at the Norwegian Porphyria Centre, alive and aged 18 years or older between January 2000 and December 2016, 612 (78%) agreed to participate (Fig. 1). Where a cancer diagnosis was the outcome, we excluded 23 persons who had a cancer diagnosis proceeding PCT symptoms (Fig. 1).

PCT diagnosis was based on biochemical testing performed at the Department of Medical Biochemistry and Pharmacology, Haukeland University Hospital [3], applying diagnostic algorithms as described by Badminton and colleagues [19]. Urinary porphyrins were expressed in nmol/mmol creatinine. Sequencing of the UROD gene was performed at the Center for Medical Genetics and Molecular Medicine at Haukeland University Hospital to determine if the patient had familial or sporadic PCT. Participants where DNA analysis had not been performed, were registered as unclassified PCT.

For our sub-group analyses, we identified 30,468 persons with a diagnosis of chronic alcohol abuse/dependence through the Norwegian Labour and Welfare Administration registrations of social security benefits.

Primary cancer diagnoses and underlying causes of death of theoretical interest, or with a count of three or greater, were investigated. This resulted in the following primary cancers: all sites (ICD-7: 140–207; ICD-10: C00–96, D45–47), colon/rectum (ICD-7: 153–154; ICD-10: C18-C21), HCC (ICD-7: 155.0; C22.0), gallbladder and biliary tract (ICD-7: 156; ICD-10:C23–24), pancreas (ICD-7: 157; ICD-10; C25), lung (ICD-7: 162; C33–34); non-melanoma skin (ICD-7: 191; ICD-10: C44), breast (ICD-7: 170; ICD-10: C50), prostate (ICD-7: 177; ICD-10: C61), and leukemia (ICD-9: 207; ICD-10: C91–95, D45–47). We included any new cases of primary cancer diagnosis from January 2000. We also assessed second primary cancers specifically. Similarly, the following underlying causes of death were investigated: malignant neoplasms (ICD-10: C00–96, D45–47), diabetes mellitus (ICD-10: E10–14), ischemic heart disease (ICD-10: I20–25), cerebrovascular diseases (ICD-10: I60-I69), chronic obstructive pulmonary disease (ICD-10: J43-J44), diseases of the liver (ICD-10: K70–77, B15–19, E83.1), renal failure (ICD-10: N17–19), and alcohol or drug overdose (ICD-10: X40–49). Chronic alcohol abuse/dependence was defined by the following codes: ICPC-2: P15; ICD-9: 303, 305.0; ICD-10: F10.

Stata/SE Version 14 for Windows was used for all statistical analyses (StataCorp Stata Statistical, Software: Release 14, College Station, TX, USA). Annual incidence was calculated by dividing the number of new cases by the total person-years of all participants. The entry time was age at baseline or 18 if younger than 18 years in 2000 (i.e., left truncated). The exit time was age at event (cancer/mortality) or censoring (death due to other factors, or end of study follow-up in 2016, whichever occurred first). To assess the relationship between PCT and cancer/mortality, a Cox regression was used to estimate the hazard ratios (HRs) with 95% confidence intervals (CIs), in which PCT diagnosis (no/yes) was the exposure and the cancer or cause of death of interest (no/yes) the outcome. Where numbers permitted, we additionally investigated PCT type (0 = reference, 1 = sporadic PCT, 2 = familial PCT, 3 = PCT type unclassified PCT) as the exposure. Age was used as the time scale, stratifying the model by birth cohorts (< 1937, 1937–1960, 1961–1983, > 1983), to adjust for calendar effects. The multivariate models were adjusted for sex, age (as the time-scale), and educational attainment. The proportionality assumption was assessed and covariates entered as time-dependent when the assumption was violated. Obtained level of education was categorized into: no schooling, compulsory education (year 1–10), upper secondary education (11–13, or 14 years if including post-secondary non-tertiary education), tertiary education (14+ years), and unspecified/missing. Interactions of exposure by educational obtainment (tertiary or upper secondary vs less education) and by sex were explored, and none were found (although limited power for interaction analyses). Differences in age at time of death were investigated by independent samples t-tests.

For each adjusted hazard ratio (aHR), we calculated the corresponding evidence value (E-value). The E-value is defined as the minimum risk ratio an unmeasured confounder requires to explain away the outcome and is calculated by the formula: E-value = hazard ratio (HR) + sqrt[HR x (HR-1)] [20].

The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the Regional Committees for Medical and Health Research Ethics, Norway (reference number: 2012/753).

Major strengths of our study are the population-based cohort design, the use of valid information from compulsory national registries, and a long period of follow-up. Persons with PCT were drawn from a nationwide registry with a good participation rate (78%). However, it is possible that participants differed from non-participants and thus selection bias may affect the validity of the results. Given the complex relationship between PCT and cancer, we chose to exclude cancer cases preceding PCT. This design reduces the risk of selection and information bias. We extend previous analyses by adjusting for education (as a proxy of socioeconomic position and lifestyle), comparing outcomes for patients with familial, sporadic and unclassified PCT, and comparing risks of PCT patients with persons with a history of alcohol abuse/dependence. A limitation of the study was the small numbers of cases. We also did not have access to liver biopsies of patients with HCC or gallbladder and biliary tract cancer. Therefore, we are uncertain regarding the proportion of these persons with concurrent fibrosis/cirrhosis.

Several common causes of PCT and cancer exist, such as alcohol use, liver disease and smoking. This information is not available in our study for the general population. Given the relationship between lifestyle factors such as excessive alcohol intake [25] and smoking habits [26] with socio-economic disadvantage and lower educational attainment, we were able to extend previous findings by controlling for educational attainment as a proxy for lifestyle.

Residual confounding by lifestyle factors is likely. We modelled residual confounding by calculating the E-value and found that an unmeasured confounder would need to have a large risk ratio of at least 39 to explain away the findings of increased HCC risk. Given the strong association between alcohol abuse/dependence and hepatitis C and both PCT and HCC, it is conceivable that unmeasured confounding may explain away any direct relationship. We were, however, able to identify persons with a history of chronic alcohol abuse/dependence to further investigate this to partially account for their generally younger age at the time of death (i.e., may have died before they could develop HCC), we conducted a competing risks survival analysis. Results, suggested persons with PCT had a 3-fold excess risk of HCC even when compared to persons with a history of alcohol abuse. The remaining high risk, albeit lower than when comparing to the general population, suggests that alcohol is an important confounder, but may not completely explain the association between PCT and HCC.

It has been proposed that the accumulation of uroporpohyrins, of which levels are high in (untreated) PCT, may cause liver injury, resulting in the formation of a hepatic tumor [27]. This may occur in parallel or by interaction with exposure to hepatotoxic factors, such as alcohol. However, if the increased risk of HCC is related in part to PCT itself or is due only to confounding lifestyle factors, is difficult to conclude. Our findings give some support to the former, given that we observed this increase in the risk ratio even after controlling for educational attainment and comparing our cohort to persons with a history of chronic alcohol abuse/dependence. Of the six persons in our cohort with PCT and HCC, two had familial PCT, which is less associated with other types of liver disease or excessive alcohol intake than sporadic PCT [3]. There was a trend towards higher concentrations of total porphyrins and uroporphyrins in PCT patients with HCC compared to those without. However, the small number of cases and variability in the findings limit our ability to draw any firm conclusions. We also observed an increased risk of gallbladder and biliary tract cancer in persons with PCT. However, given we only found three cases and tested over multiple cancers, this could reflect a chance finding.

Additionally, we found an increased risk of premature death overall, death byliver diseases and drug or alcohol overdose in persons with PCT. Liver diseases, such as chronic hepatitis C virus, are known precipitating factor for PCT [10, 28], and are also associated with excessive alcohol intake and injection drug use. We also found that persons with sporadic PCT were generally more likely to die from causes associated with lifestyle factors than persons with familial PCT. Therefore, increased mortality rates in persons with PCT, and specifically sporadic PCT, could be caused by lifestyle factors.

According to the European Association for the Study of the Liver (EASL) and European Organization for Research and Treatment of Cancer (EORTC) clinical practice guidelines for the management of HCC, an annual incidence of 1.5% would warrant surveillance of HCC in cirrhotic patients and 0.2% would warrant surveillance in non-cirrhotic patients [29]. We found an annual incidence of HCC of 0.07% in our PCT patients, with an upper bound 95% CI of 0.17%. Additionally, unlike in acute forms of hepatic porphyria, confounding by concurrent cirrhosis cannot be ruled out in patients with PCT. Therefore, surveillance cannot be currently recommended based on a PCT diagnosis alone.