In this nationwide population-based study of the 1988–2005 period we found that, in the population younger than 45 years, the incidence rate of alcoholic cirrhosis peaked in 1994 and then decreased. It appeared that men and women born in 1960 or later had progressively lower age-specific incidence rates than earlier birth cohorts. In the population aged 45–64 years, a steep increase in incidence rate from 1993 to 1994 was followed by a less markedly increasing trend. In the population aged 65 years or older, the incidence rate increased throughout the 1988–2005 period. From 1996 onwards, the trends in prevalence and hospitalization rates generally followed the incidence trends, but the number of inpatient hospitalizations per alcoholic cirrhosis patient per year increased.
The primary strength of our study is that it was based on nearly three decades of individual-level data from a nationwide population-based hospital registry with complete follow-up .
We may have underestimated the incidence rate because we failed to include alcoholic cirrhosis patients who were never hospitalized. However, we assume that this bias is small because Danish clinical practice is to refer all patients with signs or symptoms of cirrhosis to diagnostic workup and treatment in a public hospital. Another limitation is that hospital diagnoses may be wrong, so that patients with cirrhosis are not registered with the diagnosis (low completeness of registration), or patients registered with the diagnosis do not have cirrhosis (low positive predictive value of registration). A study of patients referred to Danish non-specialized medical departments in 1985–1989 found a completeness of 93 percent and a positive predictive value of 85 percent for a hospital diagnosis of cirrhosis, using a gold standard of either liver biopsy or a combination of clinical findings . The different gold standards are likely to explain the lower positive predictive value in our study . Therefore, the incidence rates may be slightly inaccurate, but the time trends in them were probably not substantially biased.
The prevalence rates depended on accurate incidence rates as well as on accurate information on the survival time of alcoholic cirrhosis patients, which was ensured by the Civil Registration System . They could be underestimated by as much as 10 percent in 1996, but by gradually smaller amounts thereafter. Thus, the prevalence rate increases seen among men and women older than 45 years could be due to bias, but the decreasing trend in prevalence rates among younger men and women could not. The hospitalization rates depended on accurate prevalence rates and emphasize that cirrhosis is a risk factor for a number of conditions that require hospitalization [19–23]. The increasing trend in the number of hospitalizations per alcoholic cirrhosis patient is probably unbiased and clearly indicates that alcoholic cirrhosis put an increasing burden on the Danish tax-funded healthcare system.
The 32 percent increase in alcoholic cirrhosis incidence from 1993 to 1994 is striking. The addition of data from outpatient and emergency room visits to the National Patient Registry increased the incidence rates by around 10 percent, accounting for a third of the increase. We also found that the increase was seen throughout Denmark, and did not depend on whether we considered only primary diagnoses. Increasing diagnostic activity and sensitivity could have contributed, but they could not explain such an abrupt increase, and they are not consistent with the increasing age at diagnosis during the study period. We find it likely that the shift from ICD8 to ICD10, which took place on 1 January 1994, may somehow have contributed although both versions contain one unambiguous code for alcoholic cirrhosis. Thus, part of the increase is an artifact caused by changes in the National Patient Registry, but our findings indicate that the incidence rate did in fact increase. The most likely reason is that the 1930–50 birth cohorts reached the typical age of alcoholic cirrhosis diagnosis, 45–64 years, around 1994. It is very unlikely that an increasing prevalence of hepatitis C infection, which is a risk factor for cirrhosis among alcohol abusers , could have contributed because hepatitis C infection has a very low prevalence in the Danish population, 0.2 percent in 1997 .
Changes in doctors' coding practice also affected our findings. In earlier years, alcoholic cirrhosis was less likely to be coded if it was not the primary diagnosis, as indicated by the attenuation of the increasing incidence trends when we considered only primary diagnoses. Thus, we may have underestimated the incidence rates in the beginning of the study period, but such a bias could not explain the decreasing trend among men and women younger than 45 years.
Our findings are consistent with data from Statistics Denmark and findings from surveys. In Denmark, the per adult (>14 years) alcohol consumption increased from 7 liters in 1965 to 12 liters in 1975 , and it is likely that those born between 1925 and, say, 1950, were responsible for this increase. The per adult consumption was stable from 1975 to 2005, and this may be explained by a continued high consumption by the 1925–1950 birth cohort combined with a low consumption by those born after 1960. Based on three surveys of Danish men and women aged 30, 40, 50, or 60 years in 1982–84, 1986–87, and 1991–92, Gerdes et al. found that alcohol consumption decreased among 30-year-olds, particularly from the 1956–57 birth cohort to the 1961–62 birth cohort . Såbye-Hansen et al. found that men and women aged 15 to 39 years consumed less alcohol in 1992 than in 1979, whereas men and women aged 40 to 79 years consumed more . Similarly, Bjørk et al. surveyed men and women in 1987, 1994, 2000, and 2003  and found an increase in alcohol consumption among men and women aged 50 years or older and a decrease among men and women aged 16–49 years. Thus, we find it plausible that Danes born around 1960 or later consume less alcohol than the generations before them. The relatively high alcohol consumption by the generations born before 1960 may be explained by social and cultural changes during their lifetime, particularly for women [7, 8].