Main

Colorectal cancer is the most common cancer in Germany and the second most common cause of cancer-related deaths in both men and women (Husmann et al, 2010; GEKID, 2011). This underlines the need for effective colorectal cancer control. Analysis of population-based survival of cancer patients provides an invaluable tool for measuring progress in these activities.

Until recently, analyses of population survival in Germany were based predominantly on data from the Saarland registry covering only 1.3% of the total population (1.0 million inhabitants). This registry provided data for the international EUROCARE (Brenner et al, 2009a), EUNICE (Gondos et al, 2008) and CONCORD (Coleman et al, 2008) studies. Since the 1990s, population-based cancer registries have been built up in most of the 16 federal states of Germany. In 2009, a collaborative project involving cancer registries and the German Cancer Research Centre (DKFZ) has been initiated, which now offers the possibility to study recent data from 11 German population-based cancer registries, covering 33 million inhabitants (Hiripi et al, 2011). The aim of this study is to describe colorectal cancer survival over the period 2002–2006 stratified by sex, age, detailed subsites of left and right colon and rectum, morphology and disease extent. Moreover, our objective is to describe recent survival trends using the technique of model-based period analysis.

Materials and Methods

Sources of data

German cancer registries covering 13 of 16 federal states and 1 administrative region submitted data for the study. Only cancer registries with estimated completeness over 80% in the period 2004–2006 and reasonably low proportion of death certificate only (DCO) cases (under 20% throughout the study period or constantly decreasing to levels <20% at the end of study period) were considered for the analysis. We eventually utilised data from 11 cancer registries covering a population of 33 million inhabitants. Full details were described in detail elsewhere (Hiripi et al, 2011).

We selected patients with a primary invasive colorectal cancer (ICD-10 C18–C20) at the age of ⩾15 years in 1997–2006. We excluded cases notified by DCO. For the stage-specific analysis, stage grouping according to ENCR recommendations (localised, regional and advanced cancer) was used (Berrino et al, 2002).

Site and morphology of tumours were coded according to ICD-O-3 (Fritz, 2000). When aggregated subsites were used, right colon included caecum (C18.0), ascending colon (C18.2), hepatic flexure (C18.3) and transverse colon (C18.4). Left colon included splenic flexure (C18.5), descending (C18.6) and sigmoid (C18.7) colon. We used five morphology groups (invasive behaviour only): adenocarcinoma in polyp (8210, 8261, 8263), mucinous adenocarcinoma (8470, 8480, 8481, 8490), other adenocarcinoma (8140–47, 8190–8221, 8260–8323, 8380–8550, 8570, 8940) and others.

Statistical methods

To quantify excess mortality due to cancer, relative survival is commonly used in population-based studies. It is derived as the ratio of the observed survival of cancer patients and the expected survival of the underlying general population (Henson and Ries, 1995). In our analysis, expected survival was estimated by the Ederer II method using life tables stratified by age, sex, calendar period and federal state as obtained from participating cancer registries and the German Federal Statistical Office. To allow for comparisons between subgroups or populations with potentially different age distribution, age adjustment was done using the International Cancer Survival Standards proposed by Corazziari et al (2004).

Period analysis (Brenner et al, 2004) was employed to provide up-to-date estimates of 5-year survival in 2002–2006. Extensive empirical evaluations have shown that period analysis provides 5-year survival estimates that are very close to 5-year survival later observed for the patients diagnosed within the period of investigation (Brenner and Hakulinen, 2002; Brenner et al, 2002b). In addition, model-based period analysis (Brenner and Hakulinen, 2006) was used to investigate recent trends and to perform statistical testing. Briefly, excess numbers of deaths were modelled as a function of calendar year (numerical variable), and year of follow-up (categorical variable) in Poisson regression models, with the logarithm of the person-years at risk as an offset. Individual models were fitted for different groups of patients by age, clinical stage or disease localisation to estimate magnitude and significance of time trends. Moreover, models including year of follow-up and sex (both categorical variables) were fitted for subgroups by age to test for gender differences.

All calculations were carried out by SAS 9.2 software (SAS Institute, Cary, NC, USA), using a publicly available macro for period analysis (Brenner et al, 2002a) and its adaptation for model-based period analysis (Brenner and Hakulinen, 2006).

Results

The analysis was based on records of 164 996 colorectal cancer patients from 11 German cancer registries. Slightly more than half of the patients were males (52.5%, ranging from 46.6 to 56.2% across registries). Median age was 70 years (range 68–72). More than 94% of cases were microscopically confirmed in every registry (Table 1).

Table 1 Description of the data set used in period survival analysis, colorectal cancer patients diagnosed in 1997–2006 in Germany
Full size table

Table 2 shows sex differences and time trends in survival by age groups. Age-adjusted 5-year relative survival was higher in women by 2.6% units. Significantly higher survival in women compared with men was observed in patients under 75 years of age; however, the difference was smaller in individuals over 65. Model-based period analysis indicated an increase in age-adjusted 5-year relative survival from 60.6 to 65.0% over the period 2002–2006. Survival significantly increased in all age groups except patients aged 45–54 years.

Table 2 Age-specific 5-year relative survival (RS) of colorectal cancer for the period 2002–2006 by sex and modelled trends in age-specific survival
Full size table

Table 3 shows the distribution of patients and differences in age-adjusted survival in subgroups by subsite and morphology. Most patients (∼80%) were diagnosed with cancer of the morphological group ‘Other adenocarcinoma’. Adenocarcinomas in polyp were more often diagnosed in the left colon and rectum than in the right colon, whereas mucinous carcinomas were more prevalent in proximal sites (14.8% in right colon vs 8.5% in left colon and 7.8% in rectum).

Table 3 Age-adjusted 5-year relative survival (RS) of colorectal cancer in subsites for the period 2002–2006 by morphology
Full size table

Overall age-adjusted 5-year relative survival of 63.0% was observed for colorectal cancer patients. Survival was higher in colon cancer patients (63.8%) than in rectal cancer patients (60.7%). Localisation of the disease in left colon was associated with a slightly higher survival (65.2%) in comparison with right colon (63.8%). When looking at detailed colonic subsites irrespective of the laterality, highest survival was observed in patients with cancer in descending (67.7%) and ascending (66.5%) colon and appendix (66.4%). Cancers in caecum (61.7%), hepatic (61.2%) and splenic (60.3%) flexures were associated with somewhat lower survival. Survival was <60% in patients with cancer within unspecified or overlapping site.

Overall 5-year relative survival was substantially higher for adenocarcinomas in polyp (78.0%), which had the best prognosis in all subsites. Patients with mucinous adenocarcinomas had similar survival as the overall group when the cancer was located in the right colon, but prognosis was substantially worse when the mucinous carcinoma was located in the left colon or rectum. For adenocarcinoma in polyp, mucinous adenocarcinoma and other adenocarcinoma, survival was always worse for rectal cancer patients than for those with colon cancer. Prognosis of patients with cancer of other/unrecorded morphological types was poor in both colon (37.4%) and rectum (41.4%).

Table 4 shows distribution of clinical stages in different subsites and respective estimates of 5-year relative survival. Stage was reported for 95 422 (57.8%) cancer patients, of whom 44.2%, 27.7% and 28.1% were diagnosed in localised, regional and advanced stage, respectively. Proportion of cases with known stage was similar over all subsites (range 56–64%), exceptions are appendix (47%) and colon, NOS (35%). There was a very large gradient in prognosis by stage of disease, with overall age-adjusted 5-year relative survival of 89.5%, 65.4% and 14.9% for patients with localised, regional and advanced tumour spread, respectively. For every clinical stage, survival was higher in colon cancer patients than in rectal cancer patients. The survival was slightly higher in left colon cancer patients compared with right colon cancer patients with regional (69.0% vs 66.0%) or advanced disease (16.5% vs 13.7%). The lowest proportion of localised disease within individual colonic segments was reported in patients with cancer in caecum (41.5%) and splenic flexure (42.2%), the highest in descending colon (47.3%). The lowest proportion of advanced disease was reported in patients with cancer in ascending (26.4%) and descending colon (26.8%).

Table 4 Age-adjusted 5-year relative survival (RS) of colorectal cancer in subsites by stage for the period 2002–2006
Full size table

Table 5 provides estimates of recent trends in survival according to site and clinical stage. Significant increase in overall 5-year relative survival was noted in both colon and rectal cancer patients (by 4.6% and 4.0% units, respectively). Most pronounced increase in stage-specific survival was noted in patients with regional disease (significant increase by 4.3% and 6.0% units for colon and rectal cancer patients, respectively). A less pronounced increase was noted for patients with localised disease (not significant for rectal cancer patients). No significant changes in survival were observed for patients with metastatic colorectal cancer.

Table 5 Recent trends in age-adjusted 5-year relative survival (RS) by site and stage according to model-based period analysis, 2002–2006
Full size table

Discussion

This study provides estimates of colorectal cancer survival of German patients over the period 2002–2006, based on a population of 33 million people. Overall 5-year relative survival was 63%, and a significant increase by >4% units was seen even within this relatively short period of time. A higher 5-year relative survival was noted in younger males and females. Survival in women was significantly higher than for males <75 years of age. Prognosis of rectal cancer patients was worse than for colon cancer patients, where best prognosis was seen for tumours in ascending and descending colon. Patients with adenocarcinoma in polyp had the best survival expectations irrespective of the disease subsite. Very large differences in survival were seen between localised, regional and distant colorectal cancers. Recent increase in stage-specific survival was substantial for patients with localised cancer and especially those with regionally spread disease, whereas no increase was achieved for patients with metastatic disease.

Our overall estimate of 5-year relative survival of 63% for the 2002–2006 period is substantially higher than estimates previously reported for earlier periods from the majority of European countries including Germany (Verdecchia et al, 2007). Although previous estimates for Germany were restricted to Saarland, covering a population of only about 3% of the population included in the current study, a very similar result was obtained when we restricted our current analysis to Saarland, suggesting that the difference in survival is due to a further increase over time rather than to a regional variation. Our findings of decreasing survival with increasing age are consistent with observations from other European countries (Berrino et al, 2007). The existing gap between older and younger patients has been observed to be growing (Quaglia et al, 2009), which was also seen in our study, as survival increase in elder patients was weaker in comparison with young and middle-aged patients. Our observation of higher survival in female than in male patients, especially at younger ages, is likewise consistent with previous studies (Micheli et al, 2009). Hormonal factors have been suggested as a possible explanation for this observation (Koo et al, 2008).

We were able to provide relative survival estimates for cancer in different subsites of left and right colon. The overall estimate of 5-year relative survival was only slightly higher for left colon cancer patients; however, this difference did not apply for all subsites of left and right colon. Survival of patients with tumours located in ascending colon was considerably higher compared with the rest of the right colon. On the other hand, splenic flexure tumours were associated with lower survival compared with adjacent segments and overall survival for left colon cancers. Few reports actually provided detailed description of survival across colonic subsites. One of the most comprehensive papers dealing with the importance of colon cancer subsites and morphology was a study comparing survival between EUROCARE and SEER by Gattaet al (2003). This study was, however, concerned primarily with explanation of international differences and did not focus on differences between subsite- and morphology-specific survival estimates. Moreover, published estimates were based on patients diagnosed in 1985–1989, and were therefore considerably lower than in our study.

Right-sided colon cancer patients are older, more likely to be women and are diagnosed with more advanced disease. Survival of right colon cancer patients is still lower, even after controlling for age, sex and stage using statistical modelling (Meguid et al, 2008; Wray et al, 2009; Benedix et al, 2010). This is usually explained by differences in tumours originating in right and left colon. These may include different carcinogenic pathways, difference in exposure to risk factors or differential sensitivity to chemotherapy (Iacopetta, 2002). However, several prognostic factors change across specific subsites irrespective of colon side, notably disease extent, as measured by TNM status or disease stage. A recent German study showed caecum and splenic flexure to have least favourable distribution of clinical stages. On the contrary, diseases originating in ascending and descending colon were diagnosed early (Benedix et al, 2011). Bearing in mind very large differences in colorectal cancer stage-specific survival, these findings may likely explain the results seen in our population-based study. Indeed, these findings are in line with distributions of clinical stages across colonic segments seen in our study.

Our study also estimated up-to-date survival of patients with colorectal cancer of different morphological properties. Adenocarcinoma in polyp had the best prognosis in all subsites. Mucinous adenocarcinoma has a less favourable prognosis in left colon and rectal cancer patients only.

Stage at diagnosis has been reported as an important explanation for observable survival differences between populations (Gatta et al, 2000; Ciccolallo et al, 2005), but it is often unavailable in population-based studies (Morris et al, 2011). Our study reported stage-specific relative survival estimates in colorectal subsites and showed substantial survival advantages of patients with early-stage tumours. Our results also suggest a higher survival of left-sided colon cancer patients with regional and metastatic disease. This is similar to previous results of studies using proportional hazards regression performed in the United States (Meguid et al, 2008; Wray et al, 2009).

Progress in survival can be attributed to either earlier detection (increasing proportion of early disease) or improvements in treatment (Brenner et al, 2007). The latter is observable as an increase in stage-specific survival. In our study, both overall survival and stage-specific survival in patients with localised or regional disease were increasing over the period 2002–2006. As the stage-specific survival increases in patients with colon cancer were weaker than the overall increase, a shift of stage distribution towards earlier stages at diagnosis may partially account for the observed increase in survival. Such a shift may be explained in part by the expansion of long standing screening offers with faecal occult blood test by the offer of screening colonoscopy in Germany in 2002 (Gross et al, 2006). This suggestion is supported by a particularly strong increase in survival between 55 and 74 years of age, where most of the screening takes place (Brenner et al, 2009b). Cancer screening may add lead time or overdiagnosis bias to survival estimates (Dickman and Adami, 2006); therefore, trends in incidence and mortality have to be monitored along with trends in survival to provide a comprehensive picture of colorectal cancer control. As both incidence and mortality are decreasing since the early 2000s in Germany (Husmann et al, 2010), it is unlikely that observed trends in population survival could be fully explained by screening-associated biases. Rather, the trends in incidence and mortality point to true reduction of the burden of colorectal cancer in Germany to which both early detection and treatment may have contributed.

Contrary to localised and regional disease, no significant improvement in prognosis was observed for metastatic colorectal cancer. Even though there has been marked improvement in recent trials, median survival of these patients has gradually improved from 6 months to 24 months only (Glimelius and Cavalli-Bjorkman, 2012), and the vast majority of these patients still die within 5-years after diagnosis. The persistent lack of progress in patients with distant tumour spread in our study and also in other European populations (Brenner et al, 2011) underlines the need of enhanced efforts of primary and secondary prevention.

Worse survival in certain colonic segments may be associated with less favourable distribution of clinical stages. Colonic segments with the highest proportion of localised and the lowest proportion of advanced cancers in our study (ascending and descending colon) were also harbouring tumours with highest 5-year relative survival. The reasons for the unequal distribution of clinical stages in different colonic subsites may include worse endoscopic accessibility and visibility of certain colonic segments, notably colonic flexures and caecum (Bensen et al, 1999; Pickhardt et al, 2004; Benedix et al, 2011), which were also seen in our study to have both unfavourable stage distribution and lower survival. Efforts should therefore be made to ensure a high quality of colonoscopy, providing optimal visualisation of all parts of colon. This should be associated with an improved early detection or prevention of cancer in these problematic subsites (Rex et al, 2002).

Our study has a number of strengths and limitations. Major strengths of this study include the large number of patients and completeness of information on subsite and morphology. In particular, this collaborative study provides up-to-date survival estimates for Germany, which are based on an ∼30-fold larger database than previously available population-based colorectal cancer survival data from Germany. The size of the database enabled very detailed survival analyses by age, sex, morphological subgroups and subsites. Major limitations include the still rather substantial exclusions of patients notified by DCO and the limited information on key prognostic variables, in particular stage at diagnosis. Although high proportions of DCO cases are partly due to the fact that many of the cancer registries are still rather young, and some proportion of DCO cases may simply be cases who had not been known to the registries as their diagnosis was made before the registration period, the relatively high DCO may be indicative of less than complete registration. As completeness of registration is typically lower for patients with poor prognosis, this may have led to some overestimation of survival (Pollock and Vickers, 1994).

Our analysis employing techniques of period survival analysis using large German national dataset provided up-to-date estimates of 5-year relative survival expectations for subgroups of patients according to sex and age suffering from cancers of different localisation, morphology and extent, showing substantial survival heterogeneity between certain groups. Model-based period analysis disclosed notable increase in survival over recent years, probably attributable to progress in care for patients with localised and regional colorectal cancer and earlier detection, which was also suggested by examination of age-specific trends.