We read with interest the article in HEPATOLOGY by Bosetti et al. on trends in mortality from hepatocellular carcinoma (HCC) in Europe, 1 and are struck by the low mortality rates in Sweden. Our experience from the work with register studies on mortality and incidence of liver cancer 2,3 tells us that the results presented by Bosetti et al. are misleading. When studying mortality from HCC, the International Classification of Diseases (ICD) codes used in this study give a false picture of the situation.
Bosetti considered ICD 9th Revision (ICD-9) codes 155.0 and ICD 10th Revision (ICD-10) codes C22.0, C22.2, C22.3, C22.4, and C22.7. The Swedish mortality rates presented-1980-1984: 2.12 (men) and 1.24 (women), 1990-1994: 2.26 (men) and 1.31 (women), and 2000-2004: 0.68 (men) and 0.27 (women)-indicate that something changed during the last decade. As the authors briefly mentioned (second paragraph, page 141) the sharp drop in rates observed in Sweden in the mid-1990s may reflect changes in coding practices more than real variation in mortality. In spite of this observation, the authors point out the very low Swedish rates, both in the abstract and several times in the Results section.
The Swedish Cause of Death Register is based on data from the death certificates. In Sweden, all deaths where the death certificate says "primary liver cancer" or "liver cancer" are coded as C22.9; since 2000, this code is subdivided into "primary liver cancer" C22.90 (earlier ICD-9 155.0) and "liver cancer, unspecified" C22.99 (earlier ICD-9 155.2). 4 The Cause of Death Register 2002 reported 67 men and 37 women who died from HCC, according to the codes used by Bosetti et al. However, another 258 men and 165 women died from liver cancer coded as C22.9. Each year from 1997 (when ICD-10 was implemented) to 2004, only 16%-20% of the liver cancer deaths were coded with the ICD codes used by Bosetti et al., but 80%-84% were coded as C22.9.
From 2000-2004, the Swedish Cancer Registry (not connected to the Cause of Death Register) annually reported about 350-400 incident primary liver cancers (ICD-7: 155.0), of which 80% were histologically verified as HCC, 4 representing an incidence rate of 2.5 (men) and 1.0 (women) per 100,000 (age standardized, world population), to compare with the mortality rates presented by Bosetti et al. of 0.68 and 0.27. Because the long-term survival with HCC is poor, it is obvious that the incident HCCs reported to the Cancer Registry resulted in more than the approximately 100 deaths/year reported by Bosetti et al.
The trends in incidence and mortality from HCC are important issues. However, it is important that the analyses are relevant. The example above is from the Swedish registers but the results from other countries could be biased in the same way. Because of differences in coding tradition and the increase of noninvasive diagnostic methods (which do no allow histologic analysis), we think that a study on trends in mortality from HCC also should present the mortality rates when including C22.9 (and for earlier years ICD-9 155.2). Another, maybe better approach, is to study incident cancers from the Cancer registry.