Despite the existence of ample population-based cancer registry data for several decades in both the United States (US) and many European countries, international cancer survival comparisons between the two sides of the Atlantic are still scarce, and they are mostly restricted to the aggregate estimates of the EUROCARE project, involving patients from rather heterogeneous European countries diagnosed in the late 1980s and early 1990s.1, 2, 3, 4 According to these comparisons, cancer survival in Europe was lower than in the US, and for many cancer sites with a large burden of disease, survival was higher in the US than in any of the European countries involved in the EUROCARE project.2, 5
At the same time, several recent evaluations have demonstrated that substantial improvements in cancer survival have been achieved in both Europe and the US in the recent years.6, 7, 8, 9, 10, 11, 12 Given the strong positive correlation between overall economic indicators such as the gross domestic product (GDP) and cancer survival outcomes,13 comparisons of cancer survival for nations with comparable economic development and health care related resources but major differences in the health care system are of particular interest. In recent years, the life expectancy of the general population was very similar in Germany and the US,14 and both countries were among the 3 countries worldwide, which spent more than 10% of their yearly GDP on health care services.15 However, notable differences exist in cancer control activities, particularly in screening, which may have substantial influence on survival estimates. In our study we compare up-to-date period survival estimates of cancer survival in Germany and the US for 23 common cancer sites.
Data sets and inclusion criteria
For Germany, data from the Saarland Cancer Registry were used, the only population-based cancer registry in Germany, which has provided internationally accepted high-quality data throughout the last 35 years.16 Saarland is a state located in the South West of Germany with a population of about 1.1 million (which is around 1.3% of the total population of Germany). The population structure and the health care system are very similar to Germany as a whole. For the US, data from 9 cancer registries included in the Surveillance, Epidemiology and End Results (SEER) Program [San Francisco-Oakland, Connecticut, Metropolitan Detroit, Hawaii, Iowa, New Mexico, Seattle (Puget Sound), Utah, Metropolitan Atlanta (SEER9)], which cover around 10% of the US population and provide data on population based cancer incidence and survival in the US since 1975, were used.17
Regarding national representativeness, age-adjusted cancer mortality in Saarland was somewhat higher for males compared to Germany for years of the study period (2000–2002), whereas rates for females were very similar.18, 19 The SEER program includes a somewhat higher percentage of urban- and foreign-born population than the national average in the US,20 and small differences in mortality trends have been reported.21 The ascertainment of patients' vital status is done by record linkage with statewide mortality data of Saarland and nationwide mortality data of the US, respectively.
At the time of this analysis, both data sets included patients diagnosed and followed with respect to vital status up to and including the year 2002. From both data sets, patients who were at least 15 years of age and were diagnosed with a first primary tumor were included in this analysis, unless they were notified by death certificate or diagnosed by an autopsy only. The analyses are restricted to the 23 most common forms of cancer, as registered by the Saarland Cancer Registry in 2000–2002. Overall, from Saarland, after excluding 458 patients because they were notified by death certificate only [DCO, n = 457 (3.1%)] or had critical missing data (n = 1), 14,192 (96.9%) patients were retained for analyses. The proportions of DCO (and autopsy) cases were lower (1.3%), and the overall number of patients included in the analyses (n = 273,654) was much higher for the US.
Survival analysis and the calculation of age-adjusted survival estimates
To provide the most up-to-date possible survival comparisons, we carried out period analyses of 5- and 10-year survival for the 2000–2002 period. These estimates pertain to patients diagnosed in 1995–2002 and 1990–2002, respectively, even though the survival experience is considered during the 2000–2002 period only, the most recent period for which data were available at the time of this analysis. Period analysis for a given period has been shown to quite closely predict long-term survival estimates of patients diagnosed in that period.7, 22 We present relative survival estimates, which were calculated as the ratio of observed (absolute) and expected survival estimates.23 The latter were calculated according to the Hakulinen method,24 using calendar-year, sex and race (US only) specific life tables. For the US patients, 1990 life tables included in the SEER database were used for the calendar years 1990–1995, and life tables for the year 2000 were used for the calendar years 1996–2002.25 These latter life tables were only stratified for white and black Americans, and we assigned the life table for whites for the category “other races” (7.4% of patients in 2000–2002) that are separately listed in the SEER database.
To account for possible age differences, the survival estimates of the German patients were adjusted to the age distribution of the US patients according to the method proposed by Brenner et al.,26 using 4 age groups (15–54, 55–64, 65–74, 75+) for all cancer sites except testicular cancer, for which 3 age groups were used (15–54, 55–64, 65+). All survival estimates were calculated with the SAS macro adperiodh.26
Differences between the relative survival of patients in the 2 countries were tested for statistical significance after adjusting for age in a Poisson regression model for relative survival as described by Dickman et al.,27 and Brenner and Hakulinen,28 using an alpha level of 0.05 as criterion for statistical significance.
Age- and stage-specific analyses
Age-specific analyses were restricted to the 4 most common cancer sites (colorectal, lung, breast and prostate), for which reasonably precise subgroup estimates could be derived for both countries. Stage-specific analyses are shown for the same cancers except prostate cancer due to the different stage classification system used in the 2 databases. In this analysis, the stage-specific estimates of patients from Germany were age adjusted to the stage-specific age distribution of US patients.
Table I describes the patient populations from Germany and the US diagnosed in 2000–2002 with one of the 23 examined cancer sites. The median age of patients was similar in the 2 populations for most cancers. Patients from Germany with cancers of the liver, cervix, uterine corpus, brain and nervous system, thyroid gland and skin melanomas, were on average 3–4 years older, whereas patients with esophageal and lung cancer from Germany were younger than those from the US. For all other sites, the difference between the median ages of patients was less than 3 years.
Table I. Number and Characteristics of Patients Diagnosed in 2000–2002 Included in the Survival Comparison Between Germany and the United States
Patient populations included in the analysis
Colon and rectum
Brain and nervous system
The proportion of male patients was, for most cancer sites, quite similar in both countries. Marked differences were only seen for patients with lung cancer, of whom 73.6% vs. 55.2% were men in Germany and the US, respectively. A difference of 7–10% units in the proportion of men was seen for cancers of the oral cavity and the larynx (surplus of men in Germany), and skin melanomas (more men in the US). For all other sites, differences in the proportions of men were equal to or below 5% units.
Table II provides age-adjusted 5- and 10-year relative survival estimates for patients from Germany, and 5- and 10-year survival estimates for patients from the US, along with the p values for the difference between the survival curves. Patients in Germany had significantly higher relative survival for 1 site (stomach cancer) 5 years, and 2 cancer sites (stomach and lung cancer) 10 years after diagnosis, whereas US patients had significantly higher relative survival for 8 (5) sites 5 (10) years after diagnosis, respectively.
Table II. Period Estimates of 5- and 10-Year Relative Survival (%) of Cancer Patients in Germany (Adjusted to the Age Distribution of US Patients) and the United States, 2000–2002
p values refer to differences in survival over the entire 5- and 10-year follow-up, respectively.
Colon and rectum
Brain and nervous system
Consistent, large and statistically significant differences in both 5- and 10-year relative survival were seen among patients with prostate, breast, cervical, stomach and oral cavity cancer. Among these, stomach cancer patients in Germany had a relative survival advantage of 10.9% (12.6%) units 5 (10) years after diagnosis, respectively. For patients with esophageal cancer from Germany, a smaller but statistically not significant advantage of 5.7% (4.8%) units was also seen. For the other sites with large and statistically significant relative survival differences, 5 (10) year relative survival was lower among patients from Germany than among patients from the US. Differences in 5 (10) year relative survival were 12.9% (20.4%) units for patients with prostate cancer, whereas they were 8.4% (13.5%), 9.4% (11.5%) and 8.3% (9.7%) units for patients with breast, cervical and oral cavity cancer, and 4.8% (2.7%) units for patients with colorectal cancer.
Figure 1 shows the age distributions and the age-specific 5-year relative survival estimates for patients with the 4 cancers included in the age-specific analysis. For all sites except lung cancer, the proportion of patients was lower in the youngest age group in Germany, and essentially equal in the oldest age group in both countries. Among patients with lung cancer, there was much higher proportion of patients in the oldest age group in the US.
Patterns of age-specific relative survival were different for the analyzed cancer sites. For colorectal cancer, US patients had a moderately higher relative survival in all age groups. Although there was no strong age-related decline in relative survival, relative survival of patients above 75 years of age was clearly lower than in the younger age groups in both countries, and patients in Germany had a stronger age-related decline in prognosis than US patients.
For lung cancer, the survival of patients was low in both countries. Patients in the youngest age group had a 5-year relative survival of 20% only. For older age groups, relative survival declined steadily in the US to around 10% for patients above 75 years of age, whereas it was uniformly around 14% for patients above 55 years in Germany.
For breast cancer, the prognosis of US patients was better in each age group. Interestingly, relative survival of US patients did not depend on age, 5-year relative survival was around 90% in all age groups. In contrast, relative survival declined with age in Germany, and 5-year relative survival of patients in the oldest age group was around 12% units lower than in the youngest age group. In the oldest age group, patients in Germany had an almost 20% units lower relative survival than patients in the US.
Finally, prognosis of patients with prostate cancer was higher in the US, with largest differences (about 18% units) in the youngest and the oldest age group. The relative survival of US patients with prostate cancer was close to 100% in all age groups, indicating that patients had hardly any excess mortality due to the disease within 5 years after diagnosis.
Table III shows the stage distribution of patients in Germany and in the US with cancers of the colon and rectum, lung and breast. Stage was available for at least 95% of registered patients in the SEER data set, compared to 74–83% in Germany. For each of the cancer sites, relative survival of patients with unknown stage was very close to the overall relative survival estimates for Germany but consistently lower in the US.
Table III. Number of Patients in the Two Registry Databases, Proportions with Known Stage, Stage Distributions (of Those with Known Stage Only) and Age-Adjusted (for Germany Only) Stage-Specific 5-Year Period Relative Survival Estimates for Selected Cancer Sites, Germany and the United States, 2000–2002
L, localized; R, regional; M, metastasis; X, unknown.
Estimates for Germany were adjusted to the stage-specific age distribution of US patients.
Colon and rectum
For colorectal cancers, only minor differences between the 2 countries were seen in the stage distribution of patients. However, the stage-specific survival estimates were considerably higher in the US for patients with localized, and particularly for regionally spread colorectal cancer.
Patients with lung cancer had the least favorable stage distribution among the examined cancer sites, with over 80% of patients diagnosed with regionally spread or metastasized disease. Stage-specific survival was also poorest for patients with lung cancer, with only around 50% of patients surviving after 5 years with localized disease. There were no major differences in the stage-specific outcomes between both countries.
Patients with breast cancer had a more favorable stage distribution in the US, with 63% of patients diagnosed with localized disease compared to 50% in Germany. Stage-specific outcomes were slightly better in the US for all 3 stages, with survival advantages of 3.8, 5.6 and 1.6% units for localized, regional and metastasized disease, respectively.
To our knowledge, this is the first comparison of up-to-date long-term survival estimates of cancer patients between the US and an affluent European country by the period analysis methodology. Overall long-term survival continues to be somewhat better in the US than in Germany, but there is substantial variation between cancer sites. Five years after diagnosis, patients with stomach cancer had statistically significantly higher relative survival in Germany, whereas no significant difference was seen for 14 sites, and patients with 8 forms of cancer had a significantly higher relative survival in the US.
In theory, differences in patient survival may arise from various factors, including better clinical care (i.e. more effective treatment), or differences in the distribution of prognostic factors, such as age (which was adjusted for in our analysis) or stage, which may be influenced by more intensive screening efforts or better awareness.
There are several cancers for which very effective treatment is available, such as testicular cancer, thyroid cancer and several hematological cancers, and survival estimates of patients with these cancers are often considered as proxy measures of the effectiveness of cancer care in a country.29 The age-adjusted survival estimates of patients with these cancers were either slightly (although not statistically significantly) better in Germany (testicular cancer) or similar (thyroid cancer, leukemia's). These patterns support the idea that the overall effectiveness of cancer treatment is similar in the 2 countries.
Nevertheless, for some of the most common cancers, notably colorectal, breast and prostate cancer, survival expectations were considerably higher in the US. These differences were not explained by age, as there were no major differences in the age distribution of patients, and age was adjusted for in our analyses. For colorectal cancer, differences in survival were also not explained by stage distribution, which was similar in the 2 countries. In both countries, effective screening options such as fecal occult blood testing and colonoscopy or sigmoidoscopy are available, although utilized only by a minority of the eligible populations.30, 31 Patients with localized and regional colorectal cancers had lower survival in Germany than in the US, pointing to potential differences in the management of these cancers.
By contrast, higher survival of patients with breast cancer in the US is explained to a large extent by a more favorable stage distribution of patients (63.0% localized in the US vs. 49.8% in Germany): if patients in Germany had the same stage distribution as US patients, the overall difference in 5-year relative survival would be reduced to 4.0% units (instead of 8.4; results of additional stage-adjusted analyses, data not shown). In 2000, 80% of US women above 40 years reported that they had a screening mammography within the last 2 years.30 In Germany, a national screening mammography program for women aged 50–69 years was launched only in 2004.31 The high utilization of mammography screening in the US may account for much of the difference between the stage distributions of the 2 patients populations.
Patients with prostate cancer had the largest difference in their survival expectations in our study. Although a direct comparison of stage distribution was not possible in our analysis, it is very likely that differences in screening activities and stage distribution account for much of the survival difference. By 2001, 57% of US men above 50 years of age reported a PSA test within the last 12 months.32 In Germany, PSA testing is not covered by health insurance, and only around 1.5 million tests were privately purchased in 200231 (which corresponds to about 12% of the male population of Germany above 50 years of age), indicating a substantially lower utilization than in the US. The strong rise in relative survival of patients with prostate cancer along with a major shift toward earlier stages at diagnosis in the US was previously demonstrated.33 Whether more widespread use of PSA screening in the US also reduces mortality and increases survival not just by lead time cannot be answered by this analysis, but requires a more comprehensive evaluation including trends in incidence and mortality.
In the comparison carried out within the framework of the EUROCARE study, patients from Germany diagnosed in 1985–1989 had lower 5-year relative survival than US patients for all 12 assessed sites, except for stomach cancer.2 Of the 11 sites involved in both the EUROCARE and the current comparison (stomach, colon, rectum, lung, breast, skin melanoma, cervix, corpus, ovary, prostate, non-Hodgkin lymphoma), 5-year relative survival in Germany was now higher than that seen in the US for patients with stomach cancer and non-Hodgkin lymphoma, although the difference was not significant for the latter, and very similar in both countries for patients with lung cancer. For the other 8 cancer sites, prognosis continued to be better in the US. Among the 13 cancer sites not involved in the previous comparison, 5-year relative survival was significantly higher in the US for 2 sites (oral cavity and bladder cancers), and not significantly different for the remaining 11 cancer sites.
Our study has several strengths and limitations. Among the former, this study is the first to provide up-to-date comparative survival data from Europe and the US using the period analysis approach, thereby taking recent and ongoing increases in cancer survival into account.8, 11 Furthermore, survival from 2 large nations, which both are highly developed and invest a very large proportion of their national resources into health care, were compared, whereas previous comparisons had used aggregate data of rather heterogeneous European countries. Apart from staging of prostate cancer, coding practices were highly consistent between registries, but less detailed and complete clinical and histopathological data and much smaller sample sizes in the Saarland Cancer Registry prohibited more detailed comparison according to patient subgroups (including gender-specific analyses). In particular, stage-specific analysis was only possible for a few of the most common cancers and therefore no detailed analysis on potential reasons for differences in survival could be undertaken.
In conclusion, while the survival of cancer patients has increased remarkably in both countries in recent years,8, 11 the comparison of survival estimates between Germany and the US for the 2000–2002 period indicates persisting survival advantages in the US for many of the most common forms of cancer. For breast and prostate cancer, differences in screening intensity, rather than clinical care, are the most likely explanation for a large part of the substantial survival advantage seen among US patients.
This study was carried out in the framework of the IMPROVE Study.