Lung cancer survival and stage at diagnosis in Australia, Canada, Denmark, Norway, Sweden and the UK: a population-based study,

Transcription

1 Editor s choice Scan to access more free content Additional merial is published online only. To view please visit the journal online ( thoraxjnl ). For numbered affiliions see end of article. Correspondence to Dr Sarah Walters, Cancer Research UK Cancer Survival Group, Department of Non Communicable Disease Epidemiology, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK; sarah. walters@lshtm.ac.uk Received 18 July 2012 Revised 8 November 2012 Accepted 12 December 2012 Published Online First 12 February thoraxjnl To cite: Walters S, Maringe C, Coleman MP, et al. Thorax 2013;68: ORIGINAL ARTICLE Lung cancer survival and diagnosis in Australia, Canada, Denmark, Norway, Sweden and the UK: a populion-based study, Sarah Walters, 1 Camille Maringe, 1 Michel P Coleman, 1 Michael D Peake, 2,3 John Butler, 4 Nicholas Young, 5 Stefan Bergström, 6 Louise Hanna, 7 Erik Jakobsen, 8 Karl Kölbeck, 9 Stein Sundstrøm, 10 Gerda Engholm, 11 Anna Gavin, 12 Marianne L Gjerstorff, 13 Juanita Hcher, 14 Tom Børge Johannesen, 15 Karen M Linkler, 16 Colleen E McGahan, 17 John Steward, 18 Elizabeth Tracey, 19 Donna Turner, 20 Michael A Richards, 21 Bernard Rachet, 1 the ICBP Module 1 Working Group ABSTRACT Background The authors consider whether differences in diagnosis could explain the variion in lung cancer survival between six developed countries in Methods Routinely collected populion-based da were obtained on all adults (15 99 years) diagnosed with lung cancer in and registered in regional and nional cancer registries in Australia, Canada, Denmark, Norway, Sweden and the UK. Stage da for pients were consolided from various classificion systems. Flexible parametric hazard models on the log cumulive scale were used to estime net survival 1 year and the excess hazard up to 18 months after diagnosis. Results Age-standardised 1-year net survival from non-small cell lung cancer ranged from 30% (UK) to 46% (Sweden). Pients in the UK and Denmark had lower survival than elsewhere, partly because of a more adverse distribution. However, there were also wide internional differences in -specific survival. Net survival from TNM I non-small cell lung cancer was 16% lower in the UK than in Sweden, and for TNM IV disease survival was 10% lower. Similar pterns were found for small cell lung cancer. Conclusions There are comparability issues when using populion-based da but, even given these constraints, this study shows th, while differences in diagnosis explain some of the internional variion in overall lung cancer survival, wide disparities in -specific survival exist, suggesting th other factors are also important such as differences in trement. Stage should be included in internional cancer survival studies and the comparability of populion-based da should be improved. INTRODUCTION Lung cancer remains the leading cause of cancer deh worldwide. 1 Survival has not improved grely over the past three decades, and remains low for pients with all but the least invasive Key messages Lung cancer Wh is the key question? Do differences in diagnosis explain wide internional differences in lung cancer survival? Wh is the bottom line? Differences in diagnosis explain some of the internional variion in lung cancer survival, but wide disparities in -specific survival suggest th other factors such as trement are also important. Why read on? This is the first internional populion-based study of lung cancer survival by diagnosis and includes nearly pients. We describe how far explains internional inequalities in lung cancer survival. tumours. 2 Wide internional inequalities in lung cancer survival remain apparent, 3 4 even between European countries. 5 Governments monitor the implemention and efficacy of their cancer plans by comparing survival outcomes with those achieved in other countries with similar levels of economic development. 6 The highest survival among such countries is a benchmark against which progress can be monitored, representing wh could be tained with optimal trement and public awareness. 7 The Internional Cancer Benchmarking Partnership (ICBP), a consortium of clinicians, epidemiologists and policy makers, has described lung cancer survival differences between six developed countries in Understanding why these survival differences occur will facilite policy designed to bring survival up to the highest internional standard. Stage Walters S, et al. Thorax 2013;68: doi: /thoraxjnl

2 diagnosis is an important prognostic factor, and internional populion-based studies suggest th differences in explain some of the survival variion for several cancers, including the ICBP study for ovarian cancer Few such studies exist for lung cancer, although explains some of the survival differences between Nordic countries 13 and within countries over time. 2 In this study, using populion-based da, we investige whether lung cancer survival differences may be tributable to differences in diagnosis, reflecting delays in diagnosis or differences in staging procedures, or to differences in -specific survival which could indice differences in trement, staging or comorbidity. Clinical trials are the gold standard for testing the efficacy of new therapies, but they usually exclude elderly, frail and socially marginalised pients. Trials cannot be used for internional comparisons of the effectiveness of entire health systems to underpin public health policy: populion-based da on the survival of all pients are required. METHODS Da were originally collected for pients diagnosed during and followed up to the end of 2007 in six countries: Norway, Denmark, the UK (Northern Ireland, Wales and England), Australia (New South Wales and Victoria), Canada (Alberta, British Columbia, Manitoba and Ontario) and Sweden (clinical dabase of the Uppsala-Örebro and Stockholm-Gotland health regions). Adults (15 99 years) diagnosed with an invasive primary malignant tumour of the lung or bronchus (ICD-10 C34.0-C34.9) were included. Pients whose tumour registrion was made from a deh certifice only were excluded. Details of inclusion criteria and quality control have been published elsewhere. 8 Analyses were restricted to the period ( pients) when da were more complete, and further to registries where least 50% of pients were d, thereby excluding Ontario, Victoria, Wales and six of the eight regional registries in England. A total of pients were finally included in the analyses. We requested da in three separe fields on the extent of the tumour (T), the degree of nodal involvement (N) and the presence of metastases (M), coded to the TNM classificion. 14 We prioritised phological T and N and clinical M where both clinical and phological da were available. Norway and New South Wales provided a locally-defined classified as local, regional, distant. In order to include Norway and New South Wales in the comparisons, we mapped the TNM system used in the four other countries to the US Surveillance, Epidemiology and End Results Summary Stage 2000 classificion (SEER), 15 which is similar to both the Norwegian and the New South Wales localised, regional, distant classificions, but better documented and more widely known. 16 This mapping was straightforward where we had informion on each of T, N and M, but it was not possible for England which only provided a grouped TNM (I IV). We therefore present two sets of analyses: one in which is cegorised to the TNM classificion system, which may be more relevant for clinicians but excludes Australia and Norway, and another in which is classified as localised, regional, distant (SEER), which excludes England. We examined non-small cell (NSCLC) and small cell lung cancer (SCLC) separely because of their different aetiology and clinical behaviour. NSCLC represents over 80% of lung cancers, and we included non-microscopically verified tumours in this group. NSCLCs grow more slowly and are generally more amenable to trement, whereas SCLC is more aggressive although short-term survival can be extended with chemotherapy. Stistical analyses We estimed net survival and its mortality equivalent (the excess hazard) using a flexible parametric model implemented in the Sta command stpm2 17 (StaCorp LP, College Stion, Texas, USA, web appendix para 1). Net survival (or excess hazard) is the survival (or the mortality) of cancer pients after background mortality in the general populion has been taken into account. Background mortality was taken from life tables specific to sex, single year of age, region of residence and year of deh. 8 Analysis was first strified by diagnosis; age diagnosis and country of residence were included as covaries and their effects allowed to vary with time since diagnosis. Non-linear effects of continuous covaries and time-dependent Table 1 Sex of pients, proportion not microscopically verified and histological subtypes Australia* Canada Denmark Norway Sweden UK Number of pients % Men % Without microscopic verificion** Histological subtype as % of microscopically verified tumours (ICD-O-3 code) Adenocarcinoma ( ) Squamous cell ( ) Large cell ( , 8046) Small cell ( ) Other *Australia: New South Wales. Canada: Alberta and Manitoba. Sweden: Uppsala-Örebro and Stockholm-Gotland health regions. UK (TNM analysis): West Midlands Cancer Intelligence Unit and the Eastern Cancer Registry and Informion Centre in England and Northern Ireland; UK (SEER analysis): Northern Ireland. This table refers to pients included in the TNM analyses for Canada, Denmark, Sweden and the UK, and to the pients included in SEER analyses for Australia and Norway. **Includes pients with morphological verificion but missing morphology (ICD-O-3 code 9990 and above). Carcinomas ( ), basal cell ( ), transitional cell ( ), sarcomas ( ), other, Not Otherwise Specified ( , , ). 552 Walters S, et al. Thorax 2013;68: doi: /thoraxjnl

3 Walters S, et al. Thorax 2013;68: doi: /thoraxjnl Table 2 Number of pients with non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC) diagnosed during : country and diagnosis (TNM and SEER Summary Stage 2000), before and after impution TNM SEER Summary Stage 2000 Stage Non-small cell Small cell Non-small cell Small-cell Number diagnosis % % % % Observed impution Number diagnosis Observed impution Stage Number diagnosis Observed impution Number diagnosis Australia* All pients Missing Localised Regional Distant Canada All pients All pients Missing Missing I Localised II Regional III Distant IV Denmark All pients All pients Missing Missing I Localised II Regional III Distant IV Norway All pients All pients Missing Missing I Localised II Regional III Distant IV Sweden All pients All pients Missing Missing I Localised II Regional III Distant IV Observed impution Continued Lung cancer

4 Table 2 Continued TNM SEER Summary Stage 2000 Non-small cell Small cell Non-small cell Small-cell % % % % impution diagnosis Observed impution Number diagnosis Observed impution Stage Number diagnosis Observed impution Number diagnosis Observed Stage Number UK All pients All pients Missing Missing I Localised II Regional III Distant IV *Australia: New South Wales. Canada: Alberta and Manitoba. Sweden: Uppsala-Örebro and Stockholm-Gotland health regions. UK (TNM analysis): West Midlands Cancer Intelligence Unit Eastern Cancer Registrion and Informion Centre in England and Northern Ireland; UK (SEER analysis): Northern Ireland. Number of pients before impution. effects were modelled with cubic splines. Interactions between age and country were assessed with the likelihood rio test. Final models were selected based on the Akaike Informion Criterion and examinion of Martingale residuals (web appendix para 2). Pients were censored 3 years after diagnosis, but we provide estimes up to 18 months after diagnosis because model stability requires follow-up beyond the last survival estime. 18 We first treed pients with missing as a separe cegory. Multiple impution by chained equions was then conducted to include pients with missing in the analyses using the Sta command ice 19 (web appendix para 3). We ran the impution model 15 times and combined the results under Rubin s rules. 19 We then re-estimed -specific survival using the 15 imputed dasets by repeing the survival analysis modelling stregy described above, and compared the range of survival estimes with the estimes for pients with known. We present survival estimes by age, and country. All-ages estimes were standardised using -specific weights based on the age distribution observed across all countries (see online web appendix, tables 1 and 2). RESULTS Descriptive stistics Men represented between 52.3% (Sweden) and 62.2% (Australia) of all pients (table 1). In the UK, 26.0% of tumours were not microscopically verified compared with 5.2% in Sweden (table 1). Among microscopically-confirmed tumours, the percentage of SCLC varied from 12.9% in Australia to 18.0% in Norway. The proportion of adenocarcinomas ranged from 25.2% (UK) to 44.8% (Sweden). The age distribution diagnosis was similar in all countries, with mean age about 70 years (table 2). Pients for whom informion on diagnosis was missing tended to be older (mean age about 73 years). Pients diagnosed a more advanced tended to be younger than those diagnosed an earlier, and older pients were more likely to be missing da on (see online web appendix figure 1). Stage diagnosis The UK had the highest proportion of pients for whom da on (TNM) diagnosis were missing, both for NSCLC (30.3% vs % elsewhere) and for SCLC (38.4% vs %). The proportion was also relively high in Australia (SEER), both for NSCLC (23.4%) and SCLC (18.6%) (table 2). Lung cancer is generally diagnosed an advanced. For NSCLC, the proportion with metastic disease (TNM IV) ranged from 46.8% in Sweden to 55.0% in Denmark, or (SEER distant ) from 47.5% in Australia to 61.2% in Denmark. The proportion of early NSCLC was lowest in Denmark and the UK. For SCLC, 72.1% of pients in Denmark had metastic disease (TNM IV) compared with less than 66% elsewhere. For SEER distant, the proportion ranged from 61.3% in Australia to around 82% in Denmark and the UK (table 2). Imputing TNM or SEER where it was missing did not consistently shift distribution towards a more or less advanced. The increase in the proportion of TNM IV was % for pients with NSCLC (table 2) and 1.1% to +0.3% for those with SCLC. 554 Walters S, et al. Thorax 2013;68: doi: /thoraxjnl

5 Walters S, et al. Thorax 2013;68: doi: /thoraxjnl Table 3 One-year net survival (NS, %) overall, age-standardised and age-specific, by diagnosis and country for pients with non-small cell lung cancer diagnosed during Australia* Canada Denmark Norway Sweden UK NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI TNM All pients All ages to to to to 29.4 Age-standardised to to to to to to to to to to to to to to to to 23.3 Stage I All ages to to to to 72.9 Age-standardised to to to to to to to to to to to to to to to to 63.6 Stage II All ages to to to to 61.5 Age-standardised to to to to to to to to to to to to to to to to 49.4 Stage III All ages to to to to 35.8 Age-standardised to to to to to to to to to to to to to to to to 28.1 Stage IV All ages to to to to 16.2 Age-standardised to to to to to to to to to to to to to to to to 11.8 Missing All ages to to to to 21.5 Age-standardised to to to to to to to to to to to to to to to to 17.0 Continued Lung cancer

6 556 Walters S, et al. Thorax 2013;68: doi: /thoraxjnl Table 3 Continued Australia* Canada Denmark Norway Sweden UK NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI SEER Summary Stage 2000 All pients All ages to to to to to to 31.9 Age-standardised to to to to to to to to to to to to to to to to to to to to to to to to 26.4 Localised All ages to to to to to to 72.1 Age-standardised to to to to to to to to to to to to to to to to to to to to to to to to 61.6 Regional All ages to to to to to to 50.7 Age-standardised to to to to to to to to to to to to to to to to to to to to to to to to 37.1 Distant All ages to to to to to to 20.8 Age-standardised to to to to to to to to to to to to to to to to to to to to to to to to 17.5 Missing All ages to to to to to to 25.4 Age-standardised to to to to to to to to to to to to to to to to to to to to to to to to 21.3 *Australia: New South Wales. Canada: Alberta and Manitoba. Sweden: Uppsala-Örebro and Stockholm-Gotland health regions. UK (TNM analysis): West Midlands Cancer Intelligence Unit Eastern Cancer Registrion and Informion Centre in England and Northern Ireland; UK (SEER analysis): Northern Ireland. Lung cancer

7 Survival from non-small cell lung cancer Overall age-standardised 1-year net survival from NSCLC ranged from 30% in the UK to 46% in Sweden (table 3). Survival was also high (42%) in Australia and Canada and intermedie (34 39%) in Denmark and Norway. Overall estimes differ slightly between the TNM and SEER analyses because pients in England were only included in the TNM analyses (40% of all TNM-d pients) and Australian and Norwegian pients were only included in the SEER analyses (38% of all SEER pients). These exclusions affect the model results for the other countries. One-year survival for pients with TNM I and II disease was significantly lower in the UK ( I: 72.5%; II: 59.8%) and Denmark ( I: 73.3%; II: 59.1%) than in Canada ( I: 86.3%; II: 78.5%) and Sweden ( I: 88.4%; II: 70.7%) (table 3). Survival was also lower in the UK for III (35.3% vs % elsewhere) and IV (15.9% vs % in Denmark Figure 1 Age-standardised excess hazard ( per 1000 person-years, log scale) from non-small cell lung cancer by, country and time since diagnosis: TNM (upper panel) and SEER Summary Stage 2000 (lower panel). Notes: (1) Nional da are used for Denmark and Norway. Other countries are represented by regional registries (Australia: New South Wales; Canada: Alberta and Manitoba; Sweden: Uppsala-Örebro and Stockholm-Gotland health regions; the UK (TNM analysis): West Midlands Cancer Intelligence Unit, the Eastern Cancer Registrion and Informion Centre in England and Northern Ireland; the UK (SEER analysis): Northern Ireland). (2) Bubbles are scaled to represent distribution diagnosis. Walters S, et al. Thorax 2013;68: doi: /thoraxjnl

8 and Sweden). The apparent survival advantage in Canada for early NSCLC was not evident more advanced, and survival for IV disease was 16.8%, similar to th for pients in the UK. For TNM s I II, the excess hazard in the UK and Denmark was higher than in Canada and Sweden all time points up to 18 months, but the differences were widest soon after diagnosis, narrowing over the first year (figure 1). The excess hazard for s III IV NSCLC was consistently high in the UK. Pients with localised NSCLC in Australia had low survival, but survival for more advanced disease was average (table 3, figure 2). The excess hazard for Australian pients with regional NSCLC declined more steeply with time than in other countries (figure 1). Norwegian pients had average specific survival (table 3). The inclusion of pients with imputed made no merial difference to the survival pterns (figure 2). Survival from small cell lung cancer One-year age-standardised net survival from SCLC ranged from 24.9% in the UK to 36.6% in Sweden (TNM analyses), and from 23.1% in the UK to 38.7% in Sweden in the SEER analysis (table 4). The estimes differ because pients in England (38% of all pients) were only included in analyses by TNM, while Australian and Norwegian pients (38% of all pients) were only included in analyses by SEER. Among pients with early SCLC (TNM s I and II), survival was lower in the UK (55.9%) and Denmark (64.5%) than in Canada (71.1%) and Sweden (71.7%) (table 4). The UK had lower survival among pients with III (37.3% vs %) and IV disease (14.4% vs %). The excess hazard of deh for UK pients with III SCLC was very high immediely after diagnosis but became average by 18 months (figure 3). Among pients with IV disease, the excess hazard in the UK and in Canada was high throughout the first 18 months. Norway had average survival for pients diagnosed each SEER, and the excess hazard was generally average throughout follow-up (figure 3). Pients in Australia had quite low specific survival from SCLC (table 4). The range in the excess hazard between SEER s was narrower in Australia than in other countries throughout follow-up (figure 3). Survival estimes including pients with imputed were generally close to those for pients with known. The main exception was for the UK (Northern Ireland only) in the SEER analyses, where the number of pients was small (figure 4). DISCUSSION Lung cancer survival varied widely between these six wealthy countries in Age-standardised 1-year net survival for NSCLC ranged from 30% in the UK to 46% in Sweden. Survival was relively low in Denmark, intermedie in Norway and high in Australia and Canada. For SCLC, 1-year survival was 12 16% lower in the UK than in Sweden and Australia and intermedie elsewhere. These survival differences are partly explained by differences in diagnosis. Denmark in particular had a more adverse distribution. However, internional differences in survival were also evident within each of disease for both types of lung cancer: generally low in the UK and high in Sweden. Denmark had low survival for pients with early disease but average survival for those with more advanced disease while, in Canada, the comparively high survival for early disease was not apparent for pients with more advanced disease. Internional differences in diagnosis may arise because of differences in disease aetiology (which may affect the aggressiveness of disease), or delays in diagnosis or in the staging procedures themselves. Stage-specific survival may also vary because of differences in the quality of and access to specific trement. Populion-based survival estimes are invaluable for public health surveillance because they include all pients young and old, rich and poor, with early or le disease and with or without comorbidity not just the small percentage of younger, fitter pients typically recruited to clinical trials. This study was based on unselected pients with lung cancer from populion-based registries. However, using such da requires extensive quality control 16 to ensure comparability of da. We consider our consolidion of raw da from different staging classificions to have been largely successful; the distributions by age, survival and morphology are clinically coherent as well as internally consistent and comparable with previous estimes, but some potential for misclassificion remains. For example, small Figure 2 Age-standardised 1-year net survival from non-small cell lung cancer by diagnosis and country using known and imputed : TNM (upper panel) and SEER Summary Stage 2000 (lower panel). Notes: Australia: New South Wales; Canada: Alberta and Manitoba; Sweden: Uppsala-Örebro and Stockholm-Gotland health regions; the UK (TNM analysis): West Midlands Cancer Intelligence Unit, the Eastern Cancer Registrion and Informion Centre in England and Northern Ireland; the UK (SEER analysis): Northern Ireland. 558 Walters S, et al. Thorax 2013;68: doi: /thoraxjnl

9 Walters S, et al. Thorax 2013;68: doi: /thoraxjnl Table 4 One-year net survival (NS, %) overall, age-standardised and age-specific, by diagnosis and country for pients with small cell lung cancer diagnosed during Australia* Canada Denmark Norway Sweden UK NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI TNM All pients All ages to to to to 26.1 Age-standardised to to to to to to to to to to to to to to to to 16.2 Stage I and II All ages to to to to 63.4 Age-standardised to to to to to to to to to to to to to to to to 57.5 Stage III All ages to to to to 41.0 Age-standardised to to to to to to to to to to to to to to to to 24.5 Stage IV All ages to to to to 16.2 Age-standardised to to to to to to to to to to to to to to to to 13.6 Missing All ages to to to to 26.6 Age-standardised to to to to Not possible to estime to to to to to to to to to to to 16.4 SEER Summary Stage 2000 All pients All ages to to to to to to 27.6 Age-standardised to to to to to to to to to to to to to to to to to to to to to to to to 21.2 Continued Lung cancer

10 560 Walters S, et al. Thorax 2013;68: doi: /thoraxjnl Table 4 Continued Australia* Canada Denmark Norway Sweden UK NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI NS (%) 95% CI Localised All ages to to to to to to 63.5 Age-standardised to to to to to to to to to to to to to to to to to to to to to to to to 45.8 Regional All ages to to to to to to 67.9 Age-standardised to to to to to to to to to to to to to to to to to to to to to to to to 43.3 Distant All ages to to to to to to 18.6 Age-standardised to to to to to to to to to to to to to to to to to to to to to to to to 24.0 Missing All ages to to to to to to 32.1 Age-standardised to to to to to to to 83.5 Not possible to estime to to to to to to to to to to to to to to to to 18.1 *Australia: New South Wales. Canada: Alberta and Manitoba. Sweden: Uppsala-Örebro and Stockholm-Gotland health regions. UK (TNM analysis): West Midlands Cancer Intelligence Unit Eastern Cancer Registrion and Informion Centre in England and Northern Ireland; UK (SEER analysis): Northern Ireland. Lung cancer

11 Figure 3 Age-standardised excess hazard ( per 1000 person-years, log scale) from small cell lung cancer by, country and time since diagnosis: TNM (upper panel) and SEER Summary Stage 2000 (lower panel). Notes: (1) Nional da are used for Denmark and Norway. Other countries are represented by regional registries (Australia: New South Wales; Canada: Alberta and Manitoba; Sweden: Uppsala-Örebro and Stockholm-Gotland health regions; the UK (TNM analysis): West Midlands Cancer Intelligence Unit, the Eastern Cancer Registrion and Informion Centre in England and Northern Ireland; the UK (SEER analysis): Northern Ireland). (2) Bubbles are scaled to represent distribution diagnosis. differences between the New South Wales coding scheme and SEER could explain why New South Wales has an apparently higher proportion of localised disease than elsewhere (eg, the distance of main stem bronchial tumours from the carina is not considered in the New South Wales system whereas it is in SEER). This down-staging in New South Wales produces lower -specific survival because their localised disease cegory includes some pients who would be classified as having a more advanced within SEER. We have detailed the potential misclassificion when mapping T, N and M da to SEER and have made recommendions on the coding of diagnosis for internional surveillance of cancer control. 16 Differences in the thoroughness of staging may also contribute to internional variion in distributions and -specific survival. The proportion of histologicallyverified tumours varied widely, from 74.0% in the UK to 94.8% in Sweden. The high proportion in Sweden was consistent with published estimes, 23 and lower levels of histological verificion in the UK are also well known, with mounting evidence th elderly pients are much less likely to undergo these invasive procedures in the UK due to concerns about their frailty. 24 Low -specific survival in the UK could conceivably arise in part because of suboptimal staging, and this misclassificion of in a proportion of pients could lead to inapproprie trement and therefore overall lower survival. In order to understand the impact of different staging procedures on internional differences in survival, cancer registries will need to capture informion on the staging procedures used for each pient for example, whether sensitive investigions such as positron emission tomography (PET)-CT were used enabling pients with very low volume metastic NSCLC to be correctly identified as having advanced disease, or whether liver function tests were used to identify advanced SCLC. At the very least, registries should record whether the recorded was defined clinically (based on physical examinion, imaging or endoscopy) or following histological examinion. Stage da were missing for 2 43% of pients, despite restriction to those registries where least 50% of pients were d. Pients with missing da on tended to be older and to have poorer survival than d pients. To reduce bias, we imputed where it was missing, using the available covariables. 19 Walters S, et al. Thorax 2013;68: doi: /thoraxjnl

12 Figure 4 Age-standardised 1-year net survival from small cell lung cancer by diagnosis and country using known and imputed : TNM (upper panel) and SEER Summary Stage 2000 (lower panel). Notes: Australia: New South Wales; Canada: Alberta and Manitoba; Sweden: Uppsala-Örebro and Stockholm-Gotland health regions; the UK (TNM analysis): West Midlands Cancer Intelligence Unit, the Eastern Cancer Registrion and Informion Centre in England and Northern Ireland; the UK (SEER analysis): Northern Ireland. We used standard approaches to deal with other potential biases th face all internional comparisons of cancer survival using populion-based da, such as age standardision to control for differences in the age distribution. 25 Nonetheless, some compositional differences remained: men made up over 59% of the pients in Australia, Norway and the UK compared with less than 54% elsewhere, and men generally have lower lung cancer survival than women. 26 The histological subtypes of lung cancer also varied. Non-microscopically verified cancers were grouped with NSCLC, but those could have included some pients with SCLC or misdiagnosed cancer. Any effect would have been small, but probably more important in the UK which had the lowest proportion of pients with histological confirmion (74.0%). The UK also had the lowest proportion of adenocarcinomas, which have a better prognosis, and this was true even after imputing morphology where it was missing (results not shown). In general, the completeness of all these registries is high, but small differences in informion capture could also contribute to differences in survival. In Sweden, cancer registrions are not initied from deh certifices as they are elsewhere, which may mean some pients with very poor survival are missed. The effect on overall survival will be minimal given the very high completeness of the Swedish da. 27 Despite these issues of da quality and comparability, there remains evidence of large internional differences in lung cancer survival. For NSCLC, the low survival in Denmark is driven partly by more advanced diagnosis. In the UK, the proportion of pients diagnosed an early is also lower than elsewhere. A more advanced distribution may arise because of faster disease progression ( possibly reled to a higher incidence of smoking 28 ) or because of delays in diagnosis. Research has pointed to diagnostic delay in Denmark, and policies are now in place in Denmark and the UK to shorten the time to diagnosis. However, survival from NSCLC in the UK was also comparively low for each of disease. Compared with Sweden, 1-year survival in the UK was stistically significantly 10 16% lower within each TNM I IV and 9 19% lower than in Canada for s I III (survival for IV was low in Canada). Danish pients had low -specific survival for TNM s I II. Pients in the UK also had lower overall survival from SCLC despite the relively favourable distribution, because -specific survival each was generally lower than elsewhere. Low -specific survival may indice poorer levels of specific trement. Previous studies have shown th surgery for lung cancer is less frequently performed in the UK than elsewhere in Europe or Canada Similarly, lower provision of chemotherapy and radiotherapy has been reported in the UK 32 than in Australia, 36 Canada 20 and Sweden. 34 These studies are not directly comparable because they use different methods, da and calendar periods. The overall ptern is nevertheless suggestive of the need for wider access to optimal trement. We are currently examining how far differences in trement between the six ICBP countries may explain the variion in survival. Other explanions th may be considered include factors such as obesity, smoking, performance stus and comorbidity. It would be desirable to have da on these factors, which may be more common in the countries with lower overall survival, 33 but such factors would have to be both highly prevalent and powerfully prognostic to explain the survival differences between countries reported here. Clinical choices may also be more conservive in some countries than others in the face of comorbidities, older age and lower performance stus. This study demonstres wide differences between six wealthy countries in short-term survival from lung cancer, the most common cause of cancer deh worldwide. These differences seem unlikely to be explained by artefacts of method or unrecorded confounders, and differences in diagnosis and -specific trement are likely to be important explanions. The study also shows the importance of including diagnosis in internional survival comparisons. It is essential th the quality, completeness and comparability of da are improved, both in clinical settings and in populion-based cancer registries, in order to facilite surveillance of internional trends in survival. Such comparisons incentivise policy development and act as benchmarks for setting nional cancer plans. Author affiliions 1 Cancer Research UK Cancer Survival Group, Department of Non Communicable Disease Epidemiology, London School of Hygiene and Tropical Medicine, London, UK 2 Nional Cancer Intelligence Network, London, UK 562 Walters S, et al. Thorax 2013;68: doi: /thoraxjnl

13 3 Department of Respirory Medicine, Glenfield Hospital, University Hospitals of Leicester, Leicester, UK 4 Royal Marsden Hospital, London, UK 5 School of Social and Community Medicine, University of Bristol, Bristol, UK 6 Department of Oncology, Gävle Hospital, Gävle, Sweden 7 Velindre NHS Trust, Cardiff, UK 8 Department of Thoracic Surgery, Odense University Hospital, Odense, Denmark 9 Department of Respirory Medicine and Allergology, Karolinska University Hospital Stockholm Sweden 10 Department of Oncology, St Olavs Hospital, University Hospital of Trondheim, Trondheim, Norway 11 Department of Cancer Prevention and Documention, Danish Cancer Society, Copenhagen, Denmark 12 Northern Ireland Cancer Registry, Belfast, UK 13 Danish Cancer Registry, Stens Serum Institut Nional Institute for Health Da and Disease Control, Copenhagen, Denmark 14 Alberta Health Services, Edmonton, Alberta, Canada 15 Norwegian Cancer Registry, Oslo, Norway 16 Thames Cancer Registry, London, UK 17 Cancer Surveillance and Outcomes, British Columbia Cancer Agency, Vancouver, British Columbia, Canada 18 Cancer Intelligence and Surveillance Unit, Cardiff, UK 19 Cancer Institute New South Wales, Sydney, New South Wales, Australia 20 Cancer Care Manitoba, Winnipeg, Manitoba, Canada 21 Department of Health, Nional Cancer Action Team, London, UK Acknowledgements We thank the cancer registry staff in all jurisdictions, whose sustained efforts in da collection and quality control over many years have enabled lung cancer survival to be compared by diagnosis. The authors would like to thank Martine Bomb, Cherine Foot and Donia Sadik Cancer Research UK for their logistical support. Collaborors ICBP Module 1 Working Group: Programme Board Søren Brostrøm (Danish Nional Board of Health, Hospital Services and Emergency Management, Copenhagen, Denmark); Heher Bryant (Canadian Partnership Against Cancer, Toronto, Ontario, Canada); David Currow (Cancer Institute New South Wales, Sydney, New South Wales, Australia); Anna Gavin (Northern Ireland Cancer Registry, Belfast, UK); Gunilla Gunnarsson (Swedish Associion of Local Authorities and Regions, Stockholm, Sweden); Jane Hanson (Welsh Cancer Nional Specialist Advisory Group, Cardiff, UK); Todd Harper (Cancer Council Victoria, Carlton, Victoria, Australia); Stein Kaasa (University Hospital of Trondheim, Trondheim, Norway); Michael A Richards (Nional Cancer Action Team, Department of Health, London, UK); Michael Sherar (Cancer Care Ontario, Toronto, Ontario, Canada); Bob Thomas (Department of Health Victoria, Melbourne, Victoria, Australia). Module 1 Collaborors and Cancer Registries Jan Adolfsson (Regional Cancer Centre, Stockholm County Council and the CLINTEC Department, Karolinska Institutet, Stockholm, Sweden); Ole Andersen (Nional Board of Health, Health Planning Division, Copenhagen, Denmark); Heher Bryant (Canadian Partnership Against Cancer, Toronto, Ontario, Canada); Andy Coldman (Cancer Surveillance and Outcomes, British Columbia Cancer Agency, Vancouver, British Columbia, Canada); Dhali Dhaliwal (CancerCare Manitoba, Winnipeg, Manitoba, Canada); Gerda Engholm (Department of Cancer Prevention and Documention, Danish Cancer Society, Copenhagen, Denmark); Anna Gavin (Northern Ireland Cancer Registry, Belfast, UK); David Forman (Section of Cancer Informion, Internional Agency for Research on Cancer, Lyon, France); Marianne L Gjerstorff (Danish Cancer Registry, Stens Serum Institut Nional Institute for Health Da and Disease Control, Copenhagen, Denmark); Juanita Hcher (Alberta Health Services, Edmonton, Alberta, Canada); Charlotte Hosbond (Nional Board of Health, Copenhagen, Denmark); Tom B Johannesen (Norwegian Cancer Registry, Oslo, Norway); Ms Lambe (Regional Oncological Centre, Uppsala University Hospital, Uppsala, Sweden and the Karolinska Institutet, Stockholm, Sweden); Karen M Linkler (Thames Cancer Registry, London, UK); Loraine Marrett (Cancer Care Ontario, Toronto, Ontario, Canada); Colleen E McGahan (Cancer Surveillance and Outcomes, British Columbia Cancer Agency, Vancouver, British Columbia, Canada); John McLaughlin (Cancer Care Ontario, Toronto, Ontario, Canada); David Meechan (Trent Cancer Registry, Sheffield, UK); Richard Middleton (Northern Ireland Cancer Registry, Belfast, UK); Kamini Milnes (Cancer Care Ontario, Toronto, Ontario, Canada); Diane Nishri (Cancer Care Ontario, Toronto, Ontario, Canada); Nicola Quin (Cancer Council Victoria, Carlton, Victoria, Australia); Linda Rabenek (Cancer Care Ontario, Toronto, Ontario, Canada); Carol Russell (Alberta Health Services, Edmonton, Alberta, Canada); Janey Shin (Canadian Partnership Against Cancer, Toronto, Ontario, Canada); John Steward (Welsh Cancer Intelligence and Surveillance Unit, Cardiff, UK); Elizabeth Tracey (Cancer Institute New South Wales, Sydney, New South Wales, Australia); Donna Turner (CancerCare Manitoba, Winnipeg, Manitoba, Canada). Clinical Committee Stefan Bergström (Department of Oncology, Central Hospital, Gävle, Sweden); Michael Boyer (Sydney Cancer Care Centre, Sydney, New South Wales, Australia); John Butler (Royal Marsden Hospital, London, UK); William Evans ( Juravinski Cancer Centre Hamilton Health Sciences, Cancer Care Ontario, Toronto ON, Canada); Louise Hanna (Velindre NHS Trust, Cardiff, UK); Erik Jakobsen (Odense University Hospital, Odense, Denmark); Karl Kölbeck (Karolinska University Hospital, Stockholm, Sweden); Jason Lester (Velindre NHS Trust, Cardiff, UK); Jonhan McAleese (Belfast City Hospital, Belfast, UK); Michael D Peake (NHS Cancer Improvement, Nional Cancer Intelligence Network, Department of Respirory Medicine, Leicester, UK); Stein Sundstrøm (Department of Oncology, St Olavs Hospital, University Hospital of Trondheim, Trondheim Norway); Nicholas Young (School of Social and Community Medicine, University of Bristol, Bristol, UK); Ian Williamson (Royal Gwent Hospital, Newport, UK). Central Analytic Team Sarah Walters, Camille Maringe, Michel P Coleman, Bernard Rachet (Cancer Research UK Cancer Survival Group, London School of Hygiene and Tropical Medicine, London, UK). Contributors Programme Board (SB, HB, DC, AG, GG, JH, TH, SK, MAR, MS and BT) identified the questions and the countries to be studied, supervised the study and contributed to drafting the report. Module 1 collaborors ( JA, OA, HB, DF, AG, CH, TBJ, ML, MLG, DM, RM, NQ, LR, JS, JSt and ET) contributed to study design and drafting the report. Cancer registry contributors ( JA, HB, AC, DD, GE, AG,MLG,JH,TBJ,CL,ML,KML,LM,CEM,JMcL,DM,RM,KM,DN,CR,JSt,JS, ET and DT) collected, prepared and transmitted the raw da, advised on the protocol, corrected da after quality control and checked the results of the analyses. Clinical Committee (SB, MB, JB, WE, LH, EJ, KK, JL, JMcA, MDP, SS, NY and IW) advised on methodology, helped interpret the findings and contributed to drafting the report. Central Analytic Team (SW, CM, MPC and BR) drafted the protocol, obtained stutory and ethics approvals, had access to all the raw da, did the da preparion, quality control and analyses and drafted the report. All authors had access to the results of all steps of da preparion, quality control and analyses and contributed to the final report. Funding This work was supported by the Department of Health, England. Cancer Research UK supports the Cancer Survival Group (C1336/A11700). The Northern Ireland Cancer Registry is funded by the Northern Ireland Public Health Agency. Competing interests MAR is the Nional Cancer Director (England, funded by the Department of Health). Other authors declare th they have no conflicts of interest. Provenance and peer review Not commissioned; externally peer reviewed. REFERENCES 1 World Cancer Report. Lyon, France: Internional Agency for Research on Cancer, Janssen-Heijnen ML, Coebergh JW. The changing epidemiology of lung cancer in Europe. Lung Cancer 2003;41: Sankaranarayanan R, Swaminhan R. Cancer survival in Africa, Asia, the Caribbean and Central America. Lyon: Internional Agency for Research on Cancer, Coleman MP, Quaresma M, Berrino F, et al. Cancer survival in five continents: a worldwide populion-based study (CONCORD). Lancet Oncol 2008;9: Berrino F, De Angelis R, Sant M, et al. Survival for eight major cancers and all cancers combined for European adults diagnosed in : results of the EUROCARE-4 study. Lancet Oncol 2007;8: Department of Health. Improving outcomes: a stregy for cancer. London: Department of Health, Abdel-Rahman M, Stockton D, Rachet B, et al. Wh if cancer survival in Britain were the same as in Europe: how many dehs are avoidable? Br J Cancer 2009;101(Suppl 2): Coleman MP, Forman D, Bryant H, et al. Cancer survival in Australia, Canada, Denmark, Norway, Sweden, and the UK, (the Internional Cancer Benchmarking Partnership): an analysis of populion-based cancer registry da. Lancet 2011;377: Ciccolallo L, Capocaccia R, Coleman MP, et al. Survival differences between European and US pients with colorectal cancer: role of diagnosis and surgery. Gut 2005;54: Sant M, Allemani C, Capocaccia R, et al. Stage diagnosis is a key explanion of differences in breast cancer survival across Europe. Int J Cancer 2003;106: Woods LM, Rachet B, O Connell D, et al. Large differences in pterns of breast cancer survival between Australia and England: a comparive study using cancer registry da. Int J Cancer 2009;124: Maringe C, Walters S, Butler J, et al. Stage diagnosis and ovarian cancer survival: evidence from the Internional Cancer Benchmarking Partnership. Gynecol Oncol 2012;127: Storm HH, Dickman PW, Engeland A, et al. Do morphology and explain the inferior lung cancer survival in Denmark? Eur Respir J 1999;13: Sobin LH, Gospodarowicz M, Wittekind C. TNM classificion of malignant tumours. 7th edn. Oxford: Union for Internional Cancer Control (UICC), Wiley-Blackwell, Young JL, Roffers SD, Ries LAG, et al. SEER summary staging manual 2000: Codes and Coding Instructions. NIH Pub. No Bethesda: Nional Cancer Institute, Walters S, et al. Thorax 2013;68: doi: /thoraxjnl