Abstract
Background. Incidence rates of cancers of the pancreas and biliary tract, typically derived from cancer registers, have been reported to be decreasing. This study tested whether pancreatic and biliary tract cancers are underreported in the Swedish Cancer Register (CR). Methods. The concordance of pancreatic and biliary tract cancer diagnoses in 1990–2009 between CR and the Swedish Patient Register (PR) were evaluated through record linkage. To further assess the completeness of these cancer diagnoses in both CR and PR, record linkage was also made to the Swedish Causes of Death Register (DR). Results. A total of 31 067 cases of pancreatic cancer and 14 273 cases of biliary tract cancer were identified in CR or PR. Altogether, 44% of the pancreatic cancers and 44% of the biliary tract cancers were registered in PR only, and not in CR. The concordance between CR and PR declined from 63% in the years 1990–1994 to 44% in 2005–2009 for pancreatic cancer. The corresponding figures for biliary tract cancer were 60% and 37%. This decline in concordance was also observed with increasing age, e.g. the concordance between CR and PR for pancreatic cancer declined from 62% in patients < 60 years to 36% among patients ≥ 80 years. The corresponding figures for biliary tract cancer were 52% and 38%. Conclusion. There is an overwhelming underreporting of pancreatic and biliary tract cancers within the Swedish Cancer Register, which has increased during recent years. The reported decreasing incidence rates for pancreatic and biliary tract cancers might therefore be incorrect.
Cancers of the pancreas and biliary tract are characterized by rapid spread and a worse prognosis than that of virtually any other cancer [Citation1]. Recent studies have reported declining incidence trends of these tumors [Citation2–4]. Estimations of cancer incidence are typically derived from cancer registers, emphasizing the importance of completeness of such registers [Citation5]. The Swedish Cancer Register (CR) has excellent overall quality and completeness [Citation6]. However, cancers without histologic confirmation are less likely to be reported [Citation7]. The rapid developments in medical imaging techniques during recent years, making cancer diagnosis possible even in patients that are probably unfit for treatment, might have resulted in less frequent use of histologic confirmation of these tumors. Furthermore, an increased awareness of the risk of complications following trans-abdominal biopsies of pancreatic and biliary tract cancers among patients not considered for surgical treatment [Citation8], may influence the reporting of pancreatic and biliary tract cancers to cancer registers.
To our knowledge, no previous study has evaluated the potential underreporting of biliary tract cancer in CR. Two studies have found an underreporting of pancreatic cancer in CR compared to Swedish Patient Register (PR) [Citation9], and the Swedish Cause of Death Register (DR) [Citation2]. However, no study has used both PR and DR to more thoroughly assess the completeness of pancreatic cancer in CR.
The aim of this study was to assess the completeness of CR regarding pancreatic and biliary tract cancers, by comparing the concordances of these diagnoses in CR, PR, and DR.
Methods
Design
The records of pancreatic and biliary tract cancers in CR or PR between 1987 and 2009 were retrieved. Three sub-populations for each cancer were identified: 1) patients with records of pancreatic or biliary tract cancer diagnosis in both CR and PR (CR-and-PR); 2) patients with records of pancreatic or biliary tract cancer in PR, but not in CR (PR-not-CR); and 3) patients with records of pancreatic or biliary tract cancer in CR, but not in PR (CR-not-PR). Due to the extraordinarily high mortality rates for pancreatic and biliary tract cancers, the accuracy of cancer diagnoses in the three sub-groups was further evaluated by linkage to DR. The personal identity number, uniquely identifying each person in Sweden, was used for all record linkages between the registers.
Data sources
The Swedish Cancer Register. Since its establishment in 1958, CR has collected information on all newly diagnosed cancers in Sweden. Both clinicians and pathologists are required to report new cases of cancer to the register, resulting in an estimated overall completeness of a minimum of 96% [Citation7]. All diagnoses are coded according to the 7th version of the International Classification of Disease (ICD-7).
The Swedish Patient Register. PR has recorded information on individual hospital discharge diagnoses in Sweden since 1964. The diagnoses are based on the final hospital record notes for each hospitalization. Presumably, all patients diagnosed with pancreatic or biliary tract cancers are at some point admitted to a hospital. Consequently, hospital discharge records provide a good reference against which to assess the completeness of CR. The percentage of the Swedish population covered by this register is 100% from 1987 and onwards and the accuracy of the diagnoses codes in PR has been reported to be between 85% and 95% [Citation10]. The discharge diagnosis codes were registered according to the 9th version of ICD between 1987 and 1996 and the 10th version of ICD from 1997 onwards.
The Swedish Causes of Death Register. Since 1961, DR has collected data on causes of deaths nationwide in Sweden. The data is derived from death certificates, which provide information about underlying and secondary causes of death coded according to the ICD-codes [Citation11]. Although the accuracy of cancer records registered in DR might be of limited specificity [Citation12], the validity of these diagnoses is high if they are concordant with those recorded in CR or PR [Citation13,Citation14]. Similar to PR, the ICD versions 9 and 10 were used to define pancreatic and biliary tract cancer.
Identification of cases
All records representing pancreatic (ICD-7 code 157) and biliary tract cancers (ICD-7 codes 155.1-3, 155.8, and 155.9) were identified in CR. The corresponding 9th and 10th version of ICD-codes for pancreatic (ICD-9 code 157 and ICD-10 code C25) and biliary tract cancers (ICD-9 code 156 and ICD-10 codes C23-C24) were used for identification of cancer records in PR and DR.
Exclusions
To ensure that the date of cancer diagnosis for the identified cases was the first report of cancer, the study period was restricted to 1990 and onwards. All cases of pancreatic or biliary tract cancer reported before the study period were excluded. Further exclusions were made for cancer in situ, a diagnosis first made at autopsy, reported deaths before cancer diagnosis, or if the date of diagnosis was missing.
Statistical analysis
The concordance of diagnoses between CR and PR was estimated as relative proportions in the three sub-populations. The date of cancer diagnosis was the date recorded in CR. The hospital admission date was used as the date of diagnosis for cases reported to PR only. For the three sub-populations, the overall concordance between CR and PR for pancreatic and biliary tract cancers were stratified for sex, age group (< 60, 60–69, 70–79, or ≥ 80 years), and calendar period of diagnosis (1990–1994, 1995–1999, 2000–2004, or 2005–2009). Furthermore, it is expected that a person with pancreatic or biliary tract cancer would be hospitalized more than once during the course of the disease. Thus, the analysis was also stratified for the number of hospital records in PR including a diagnosis of pancreatic or biliary tract cancer. In patients who died during the study period, the concordance between the diagnosis of pancreatic or biliary tract cancer and the cause of death was calculated. The cause of death was categorized into seven groups: pancreatic cancer, biliary tract cancer, esophageal or gastric cancer, liver cancer, metastatic or advanced cancer, non-gastrointestinal or hematologic cancer, and non-malignant. Finally, the proportion of deaths in relation to time since cancer diagnosis was calculated.
Results
Pancreatic cancer in the Cancer Register and the Patient Register
Altogether, 31 067 unique cases of pancreatic cancer were identified in CR or PR. The distribution of these cases is presented in and . Overall, 53% (n = 16 344) of the total number of cases was identified in both registers (CR-and-PR), 44% (13 605) were identified in PR only (PR-not-CR), and 3% (1118) were found in CR only (CR-not-PR). There were no apparent gender differences. The proportion of patients with diagnosis of pancreatic cancer in CR declined with increasing age; among individuals aged 80 years or older, 40% were registered in CR compared to 66% of those younger than 60 years. A similar trend was noted for cases of a more recent calendar period; 46% were registered in 2005–2009, while 68% were registered in 1990–1994.
Figure 1. Distribution of cases with pancreatic cancer in the Swedish Cancer Register (CR) and the Swedish Patient Register (PR).
Table I. Number (%) of patients with diagnosis of pancreatic cancer in the Swedish Cancer Register (CR) and the Swedish Patient Register (PR) between 1990 and 2009.
Biliary tract cancer in the Cancer Register and the Patient Register
A total of 14 273 cases of biliary tract cancer were reported in CR or PR. The distribution of these cases is presented in and . Altogether, 48% (6799) were identified as CR-and-PR, 44% (6303) were identified as PR-not-CR, and 8% (1171) were identified as CR-not-PR. Registration of cases of biliary tract cancer in CR was slightly more frequent in women (58%) than in men (52%). Similarly to what was found for pancreatic cancer, the proportion of individuals 80 years or older was lower in CR (46%) compared to those aged less than 60 years (59%). Additionally, only 43% of biliary tract cancer cases were registered in CR in 2005–2009, while the corresponding figure was 73% in 1990–1994.
Figure 2. Distribution of cases with biliary tract cancer in the Swedish Cander Register (CR) and the Swedish Patient Register (PR).
Table II. Number (%) of patients with diagnosis of biliary tract cancer in the Swedish Cancer Register (CR) and the Swedish Patient Register (PR) between 1990 and 2009.
Pancreatic cancer diagnosis and cause of death
Among the cases of pancreatic cancer in PR-not-CR, 93% (12 581) died during the study period and pancreatic cancer was recorded as a cause of death in 80% of these cases (). The concordance between pancreatic cancer diagnosis and cause of death increased with increasing number of hospital discharges including pancreatic cancer diagnosis, from 70% in cases with one discharge record of pancreatic cancer, to 92% for those with five or more such records. Additionally, metastatic or advanced cancer was noted to be a common cause of death (35%).
Table III. The concordance between the diagnosis of pancreatic cancer in the Swedish Cancer Register (CR) and the Swedish Patient Register (PR) and cause of death in the Causes of Death Register stratified by the number of hospital discharges.
For cases in the sub-group CR-not-PR, 96% (1068) were reported to have died and pancreatic cancer was recorded as a cause of death in 71% of these cases. Metastatic or advanced cancer was reported as a cause of death in 51% of the cases.
Concerning cases in CR-and-PR, 97% (15 891) died during the study period and pancreatic cancer was reported as a cause of death in 96% of these cases. The trend towards increased concordance with number of hospital discharges was weaker than for cases in PR-not-CR. Again, metastatic or advanced cancer was the second most frequently reported cause of death (39%).
Overall, 88% of the pancreatic cancer cases that died during the study period did so within one year of diagnosis (Supplementary Table I, available online at http://informahealthcare.com/doi/abs/10.3109/0284186X.2013.857429, and ). The CR-and-PR sub-population had a tendency towards longer survival. The corresponding figures were 91% among cases in PR-not-CR, 96% in CR-not-PR, and 86% in CR-and-PR. The concordance between diagnosis of pancreatic cancer and such cancer as cause of death declined with survival time, from 85% within six months of follow-up among cases in PR-not-CR to 16% after > 5 years after the date of the diagnosis. The corresponding figures for cases in CR-and-PR were 96% and 73%, respectively.
Figure 3. Survival probability of cases with pancreatic cancer registered in the Swedish Cancer Register (CR) and the Swedish Patient Register (PR).
Biliary tract cancer diagnosis and cause of death
In total, 5833 (93%) of the cases with a record of biliary tract cancer in PR-not-CR died during the study period (). The overall concordance between a diagnosis of biliary tract cancer and such cause of death was 59%. The concordance increased with increasing number of hospital discharges, from 50% for cases with one discharge record including biliary tract cancer in PR, to 71% in cases with four such records. Metastatic or advanced cancer was a common cause of death (32%).
Table IV. The concordance between the diagnosis of biliary tract cancer in the Swedish Cancer Register (CR) and the Swedish Patient Register (PR) and cause of death in the Causes of Death Register stratified by the number of hospital discharges.
Concerning cases in CR-not-PR, 92% (1082) died during the study period, and biliary tract cancer was reported as a cause of death in 51% of these cases.
The concordance between a diagnosis of biliary tract cancer and such cause of death was higher (84%) among cases in CR-and-PR. There were no substantial differences when stratifying the results for the number of hospital discharges.
In total, 85% of biliary tract cancer cases and reported death during the study period, died within the first year of cancer diagnosis (Supplementary Table II, available online at http://informahealthcare.com/doi/abs/10.3109/0284186X.2013.857429, and ). The CR-and-PR sub-population had a tendency towards longer survival. The concordance between a diagnosis of biliary tract cancer and such cancer as a cause of death was highest in CR-and-PR and lowest in CR-not-PR. The concordance between a diagnosis of biliary tract cancer and such cancer as a cause of death decreased with survival time.
Figure 4. Survival probability of cases with biliary tract cancer registered in the Swedish Cancer Register (CR) and the Swedish Patient Register (PR).
Discussion
This study indicates a substantial underreporting of pancreatic and biliary tract cancers in the Swedish Cancer Register, particularly during later calendar periods and in older patients.
Among methodological strengths of this study is the availability of three nationwide registers, each with separate and independent registration of pancreatic and biliary tract cancer. Moreover, the large study size and the population-based design counteract chance errors and selection bias, respectively. There are, however, some potential limitations of our study that mainly relate to the accuracy of the data of the comparison registers. The concordance between cancer diagnosis in PR and the cause of death in DR increased with the number of hospital discharges including cancer diagnosis. This may indicate erroneous discharge diagnosis codes in PR that were later corrected when the definitive diagnosis was determined, possibly before the case was reported to CR, making an element of misclassification possible. However, the high cumulative mortality during the first year after pancreatic and biliary tract cancer diagnosis is an indirect measure of the accuracy of the data. Furthermore, previous validation studies of PR regarding diagnoses such as acute pancreatitis [Citation15], ischemic heart disease [Citation16], and Guillain-Barre Syndrome [Citation17] have reported a generally high reliability. Furthermore, evaluations of the accuracy of diagnoses have been undertaken by the Swedish National Board of Health and Welfare and showed a generally high level of accuracy [Citation18]. After review of almost 900 hospital charts in 1986 and an additional 900 in 1990, the diagnosis codes were correct in 87–88% of cases on a three-digit level. A recent meta-analysis of the reliability of PR reported an overall positive predictive value of up to 95% [Citation10], with higher figures for more severe diseases, such as malignancies. Thus, although misclassification may to some extent have influenced our results, it is unlikely that it explains the large population not reported to CR.
Conversely, any over-reporting of pancreatic or biliary tract cancers to PR may affect the reported results, however, if we compare with other tumors, the discrepancies between the Cancer Registry and the Patient Registry is limited [Citation7]. In parallel, the frequency of cystic lesions of the pancreas in the adult population has been described to be significant [Citation19], which might introduce some problems with over-diagnosis. However, since only malignant tumors were included in this study, any such problem should be limited, and not play any major role in the overall results of this study. Therefore, it is unlikely that any over-reporting to the Patient Registry would explain more than a limited part of the discrepancy between this registry and the Cancer Registry.
Moreover, the results suggest that the diagnoses of cases identified only in PR are less accurate than diagnoses of cases reported also to CR, particularly for cancers of the biliary tract. Additionally, the results concerning concordance between cancer diagnosis and cause of death indicate that biliary tract cancer may not be as clearly characterized as pancreatic cancer because other causes of death than cancer were more common. While pancreatic cancer is a more locally defined cancer, biliary tract cancer is often found with amorphous and locally advanced tumors that might reduce the accuracy of diagnosis even if tissue samples are present [Citation20].
In line with previous studies [Citation7,Citation9], our results indicate incomplete coverage of CR with regards to the diagnosis of pancreatic cancer. However, no previous study has assessed the completeness of biliary tract cancer in CR. Between 1960 and 2003, 14% of all cases of pancreatic cancer found in CR were not registered in DR [Citation2], and in a recent report, the records of pancreatic cancer in CR in 1987–1999 were compared to PR, showing an overall concordance of 73% [Citation9]. However, these studies did not include the most recent time period and did not include all available registers for the assessment of the completeness of pancreatic cancer in CR. The extent of underreporting revealed in the present study has not been described previously.
Our results may in part reflect the lack of tissue samples for histologic confirmation of these cancers. A previous study found a considerable underreporting to CR of cancers without histological confirmation [Citation12]. Moreover, the rapid development in medical imaging techniques, particularly for pancreatic and biliary tract cancers [Citation21,Citation22], may substantially contribute to the underreporting to CR, since imaging is increasingly a sufficient ground for diagnosis, circumventing the need for biopsies. In this study, only 40% of the elderly patients (≥ 80 years) with pancreatic cancer and 46% with biliary tract cancer were registered in CR, compared to 66% and 59% of young patients (< 60 years) with pancreatic cancer and biliary tract cancer, respectively. It is reasonable to assume that for patients with a poor prognosis, particularly elderly patients who are more often burdened with multiple comorbidities, biopsies might not be obtained, as the results of such histological evaluation would not be likely to change the clinical decision-making. However, it is mandatory to report new primary cancers to CR regardless of the histological confirmation [Citation23,Citation24]. Based on clinical experience, the lack of awareness among clinicians about this fact may be an important contributing factor for this underreporting, particularly in the elderly.
For patients considered for palliative chemotherapy, histological verification is usually a prerequisite. Thus, for patients treated with palliative chemotherapy, the degree of histological verification should be high, but for patients where palliative treatment is not considered, the rate of tissue sampling may be lower. Additionally, oncology centers treating cancer patients may have more functioning routines for reporting patients to the CR, compared to the units where the patients are diagnosed. This in turn would influence the reporting rates for different patients depending on treatment status.
Furthermore, the tendency towards longer survival in the CR-and-PR sub-population in both pancreatic and biliary tract cancer cases indicates that these patients have a better prognosis. This group probably consists of younger and healthier patients, where active treatment may be more frequent. Thus, the rate of reporting may be higher.
The main implication of our results is that the reported population-based incidence of pancreatic and biliary tract cancers is underestimated, and that this underestimation increases with calendar time. This error might in fact explain the reported decrease in incidence of these tumors during recent years.
In conclusion, this study of the completeness of the registration of pancreatic and biliary tract cancers in CR indicates considerable underreporting of these cancers, which has increased during more recent years. Since the reported incidence of these cancers is typically based on cancer registers of similar design in other countries, there might be a global underestimation of the true incidence of pancreatic and biliary tract cancers, and an erroneously reported decreasing trend of the incidence of these tumors.
Supplementary Tables I–II
Download PDF (24.4 KB)Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.
This work was supported by the Swedish Research Council. The Swedish Research Council had no involvement in the study design, in the collection, analysis and interpretation of data; in the writing of the manuscript; and in the decision to submit the manuscript for publication.
References
- WHO World Cancer Report 2008. Geneva: WHO; 2008.
- Luo J, Adami HO, Reilly M, Ekbom A, Nordenvall C, Ye W. Interpreting trends of pancreatic cancer incidence and mortality: A nation-wide study in Sweden (1960–2003). Cancer Cause Control 2008;19:89–96.
- Fitzsimmons D, Osmond C, George S, Johnson CD. Trends in stomach and pancreatic cancer incidence and mortality in England and Wales, 1951–2000. Br J Surg 2007; 94:1162–71.
- von Hahn T, Ciesek S, Wegener G, Plentz RR, Weismuller TJ, Wedemeyer H, et al. Epidemiological trends in incidence and mortality of hepatobiliary cancers in Germany. Scand J Gastroenterol 2011;46:1092–8.
- Khan SA, Emadossadaty S, Ladep NG, Thomas HC, Elliott P, Taylor-Robinson SD, et al. Rising trends in cholangiocarcinoma: Is the ICD classification system misleading us?. J Hepatol 2012;56:848–54.
- National Board of Health and Welfare. Ett halvt sekel med svenska cancerregistret. Stockholm, Sweden. 2008. [cited 2013 Oct 15]. Available from: http://www.socialstyrelsen.se/publikationer2008/2008-126-48.
- Barlow L, Westergren K, Holmberg L, Talback M. The completeness of the Swedish Cancer Register: A sample survey for year 1998. Acta Oncol 2009;48:27–33.
- Heimbach JK, Sanchez W, Rosen CB, Gores GJ. Trans- peritoneal fine needle aspiration biopsy of hilar cholangiocarcinoma is associated with disease dissemination. HPB 2011; 13:356–60.
- Lambe M, Eloranta S, Wigertz A, Blomqvist P. Pancreatic cancer; reporting and long-term survival in Sweden. Acta Oncol 2011;50:1220–7.
- Ludvigsson JF, Andersson E, Ekbom A, Feychting M, Kim JL, Reuterwall C, et al. External review and validation of the Swedish national inpatient register. BMC Public Health 2011;11:450.
- National Board of Health and Welfare. Dödsorsaksstatistik. Historik, produktionsmetoder och tillförlitlighet. Stockholm, Sweden. 2010. [cited 2013 Oct 15]. Available from: http://www.socialstyrelsen.se/Lists/Artikelkatalog/Attachments/18019/2010-4-33.pdf.
- Mattsson B, Wallgren A. Completeness of the Swedish Cancer Register. Non-notified cancer cases recorded on death certificates in 1978. Acta Radiol Oncol 1984;23:305–13.
- Johansson LA, Westerling R. Comparing Swedish hospital discharge records with death certificates: Implications for mortality statistics. Int J Epidemiol 2000;29:495–502.
- Godtman R, Holmberg E, Stranne J, Hugosson J. High accuracy of Swedish death certificates in men participating in screening for prostate cancer: A comparative study of official death certificates with a cause of death committee using a standardized algorithm. Scand J Urol Nephrol 2011;45:226–32.
- Razavi D, Ljung R, Lu Y, Andren-Sandberg A, Lindblad M. Reliability of acute pancreatitis diagnosis coding in a National Patient Register: A validation study in Sweden. Pancreatology 2011;11:525–32.
- Rosen M, Alfredsson L, Hammar N, Kahan T, Spetz CL, Ysberg AS. Attack rate, mortality and case fatality for acute myocardial infarction in Sweden during 1987–95. Results from the national AMI register in Sweden. J Intern Med 2000;248:159–64.
- Jiang GX, de Pedro-Cuesta J, Fredrikson S. Guillain-Barre syndrome in south-west Stockholm, 1973–1991, 1. Quality of registered hospital diagnoses and incidence. Acta Neurolog Scand 1995;91:109–17.
- National Board of Health and Welfare. Kvalitet och innehåll i patientregistret – Utskrivningar från slutenvården 1964–2007 och besök i specialiserad öppenvård (exklusive primärvårdsbesök) 1997–2007. Stockholm, Sweden. 2009. [cited 2013 Oct 15]. Available from: http://www.socialstyrelsen.se/publikationer2009/2009-125-15.
- Brugge WR, Lauwers GY, Sahani D, Fernandez-del Castillo C, Warshaw AL. Cystic neoplasms of the pancreas. N Engl J Med 2004;351:1218–26.
- Blechacz B, Komuta M, Roskams T, Gores GJ. Clinical diagnosis and staging of cholangiocarcinoma. Nat Rev Gastroenterol Hepatol 2011;8:512–22.
- Valls C, Guma A, Puig I, Sanchez A, Andia E, Serrano T, et al. Intrahepatic peripheral cholangiocarcinoma: CT evaluation. Abdom Imaging 2000;25:490–6.
- Ramsay D, Marshall M, Song S, Zimmerman M, Edmunds S, Yusoff I, et al. Identification and staging of pancreatic tumours using computed tomography, endoscopic ultrasound and mangafodipir trisodium-enhanced magnetic resonance imaging. Australas Radiol 2004;48: 154–61.
- National Board of Health and Welfare. Ordinance for cancer register at the National Board of Health and Welfare. Stockholm, Sweden: National Board of Health and Welfare; 2001.
- National Board of Health and Welfare. Ändring i föreskrifterna och allmänna råden (SOSFS 2003:13) om uppgiftsskyldighet till cancerregistret vid Socialstyrelsen. SOSFS 2006:15. Stockholm, Sweden. 2006. [cited 2013 Oct 15]. Available from: http://www.socialstyrelsen.se/sosfs/2003-13/Documents/2006_15.pdf