Progress in cancer survival, mortality, and incidence in seven high-income countries 1995–2014 (ICBP SURVMARK-2): a population-based study – The Lancet

Background

Population-based cancer survival estimates provide valuable insights into the effectiveness of cancer services and can reflect the prospects of cure. As part of the second phase of the International Cancer Benchmarking Partnership (ICBP), the Cancer Survival in High-Income Countries (SURVMARK-2) project aims to provide a comprehensive overview of cancer survival across seven high-income countries and a comparative assessment of corresponding incidence and mortality trends.

Methods

In this longitudinal, population-based study, we collected patient-level data on 3·9 million patients with cancer from population-based cancer registries in 21 jurisdictions in seven countries (Australia, Canada, Denmark, Ireland, New Zealand, Norway, and the UK) for seven sites of cancer (oesophagus, stomach, colon, rectum, pancreas, lung, and ovary) diagnosed between 1995 and 2014, and followed up until Dec 31, 2015. We calculated age-standardised net survival at 1 year and 5 years after diagnosis by site, age group, and period of diagnosis. We mapped changes in incidence and mortality to changes in survival to assess progress in cancer control.

Findings

In 19 eligible jurisdictions, 3 764 543 cases of cancer were eligible for inclusion in the study. In the 19 included jurisdictions, over 1995–2014, 1-year and 5-year net survival increased in each country across almost all cancer types, with, for example, 5-year rectal cancer survival increasing more than 13 percentage points in Denmark, Ireland, and the UK. For 2010–14, survival was generally higher in Australia, Canada, and Norway than in New Zealand, Denmark, Ireland, and the UK. Over the study period, larger survival improvements were observed for patients younger than 75 years at diagnosis than those aged 75 years and older, and notably for cancers with a poor prognosis (ie, oesophagus, stomach, pancreas, and lung). Progress in cancer control (ie, increased survival, decreased mortality and incidence) over the study period was evident for stomach, colon, lung (in males), and ovarian cancer.

Interpretation

The joint evaluation of trends in incidence, mortality, and survival indicated progress in four of the seven studied cancers. Cancer survival continues to increase across high-income countries; however, international disparities persist. While truly valid comparisons require differences in registration practice, classification, and coding to be minimal, stage of disease at diagnosis, timely access to effective treatment, and the extent of comorbidity are likely the main determinants of patient outcomes. Future studies are needed to assess the impact of these factors to further our understanding of international disparities in cancer survival.

Funding

Canadian Partnership Against Cancer; Cancer Council Victoria; Cancer Institute New South Wales; Cancer Research UK; Danish Cancer Society; National Cancer Registry Ireland; The Cancer Society of New Zealand; National Health Service England; Norwegian Cancer Society; Public Health Agency Northern Ireland, on behalf of the Northern Ireland Cancer Registry; The Scottish Government; Western Australia Department of Health; and Wales Cancer Network.

The International Cancer Benchmarking Partnership (ICBP) is a global, multidisciplinary partnership of clinicians, academics, and policy makers seeking to understand how and why cancer survival differs across countries that have high-quality cancer registries and universal access to, and comparable expenditure on, health care.

The first phase of ICBP resulted in a series of studies that have sought to comprehensively understand and address international cancer survival variations, including an assessment of four cancers over the period 1995–2007 in six countries.

During the second phase of ICBP, the Cancer Survival in High-Income Countries (SURVMARK-2) project

was established to further explore the underlying reasons for observed survival differences across an expanded set of cancer types, with a focus on indicators that inform clinical practice and public health, and engagement with a multidisciplinary group of collaborators and stakeholders.

Several studies in high-income countries have shown consistent improvements in cancer survival over the past two decades, yet survival differences between countries have been a permanent feature for most cancer sites and continue to persist, albeit to varying extents.

^, 

^, 

These differences have been interpreted, and in some cases validated, as partly reflecting variations in progress in advancing cancer services, motivating policy reforms across countries.

However, evolving cancer control plans need to be assessed not only in light of improvements in survival, but also taking into account trends in incidence and mortality to better understand the underlying biological, epidemiological, and clinical processes involved.

Benchmarking international variations in cancer survival remains a highly complex exercise given the numerous potential factors that might explain observed variations. In addition to factors that directly affect patient outcomes, such as stage at diagnosis and treatment,

^, 

^, 

^, 

survival can also be influenced by differences in how cancer cases are registered and classified.

Additionally, national and subnational variability in factors linked to the underlying health systems, cancer policy, and clinical practice could all be important drivers. As such, public awareness and beliefs about cancer are considered to have less of an influence on international differences in survival.

^, 

Of the 3 890 934 cancer cases that were eligible for analysis in the 19 jurisdictions, 126 391 (3·2%) were excluded due to the cancer diagnosis only being based on a death certificate or at autopsy (77 578 [2·0%]); invalid or missing dates, age, or non-malignant behaviour of the cancer (4471 [0·1%]); the patient being younger than 15 years or older than 99 years at diagnosis (2438 [<0·1%]); and second or higher order cancers being present at the same site as the primary cancer (41 904 [1·1%]). Hence, 3 764 543 (96·8%) of eligible cancer cases were included in the analyses (table 1; appendix p 25).

In the most recent 5-year period of diagnosis (2010–14), the highest 1-year survival for most cancer sites was in Australia, followed by Canada and Norway (appendix pp 11–12). The lowest 1-year survival was observed for stomach, colon, rectal, and lung cancer in the UK; and for oesophageal cancer in Canada, pancreatic cancer in New Zealand, and ovarian cancer in Ireland. Similar patterns were observed for 5-year survival (table 2), with consistently higher survival in Australia than in the other countries, except for lung (Canada) and ovarian cancer (Norway), and lower survival in the UK, except for oesophageal (Denmark) and ovarian cancer (Ireland). Inclusion of cases that were diagnosed on the basis of death certificate or autopsy in secondary analyses did not substantially change the results (appendix pp 23–24). Using this scenario, 5-year survival estimates were reduced by a maximum of 0·6 percentage points across all countries in 2010–14.

Cancer survival increased in all seven countries over the 20-year study period, and improvements, at least in relative terms, were most substantial for the cancer sites associated with a generally poor prognosis (oesophagus, stomach, pancreas, and lung; figure 1), and for patients younger than 75 years at diagnosis (table 3; appendix pp 13–14). The most pronounced increases in 5-year survival were observed in Denmark for cancers of the colon, with a 16·6 percentage point absolute increase, and rectum, with a 21 percentage point absolute increase, between the 1995–99 and 2010–14 time periods (table 2, figure 1). The survival gap between countries narrowed over time for 1-year survival for all cancers except pancreatic cancer (because of substantial increases in survival in all countries), whereas for 5-year survival the survival gap only narrowed for oesophageal and rectal cancer (table 2). Differences in 5-year survival across countries were greater among patients aged 75 years and older than those younger than 75 years, and the survival gap widened for ovarian cancer in those older than 75 years over the 20-year period but remained fairly stable for other cancer sites (Table 3, Table 4).

Both 1-year and 5-year survival from oesophageal cancer increased substantially in all countries among patients diagnosed in the period 2010–14 compared with those diagnosed in 1995–99 (table 2; appendix pp 11–12). Comparing the periods 1995–99 and 2010–14, 5-year survival from oesophageal cancer increased by up to 11 percentage points (in Ireland and Norway), which was more pronounced in those younger than 75 years at diagnosis (up to 14 percentage points in Ireland and Norway) than in those aged 75 years and older (up to 7 percentage points in Denmark; Table 3, Table 4). Although mortality has been stable or decreasing across countries over time (except in Norway), incidence has increased (except for in Ireland and Australia; figure 2; appendix p 30).

The uniform improvements in cancer survival in this study are probably the direct consequence of major health care reforms and technological advances that have enabled earlier diagnosis, more effective and tailored treatment, and better patient management than in previous periods. For example, rectal cancer had one of the most substantial increases in 5-year net survival over time, increasing between 9 and 21 percentage points in all seven countries. Improvements in surgical techniques, including the implementation of total mesorectal excision and new guidelines that include preoperative radiotherapy are among the key changes that have improved patient outcomes.

Improvements in survival were largely seen among younger patients (aged <75 years) and might relate to the relatively wider access to adjuvant chemotherapy and ability of these patients to tolerate more aggressive treatments than older age groups. Additionally, better diagnosis and staging with new technologies such as PET-CT imaging, alongside greater precision in the selection of patients for targeted therapies on the basis of molecular markers, have probably contributed. In Denmark, the substantial increases in survival seen across all cancers, particularly colorectal cancer, might have partly resulted from national health-care reforms in 2007 when cancer became regarded as an acute life-threatening disease, after which cancer-specific pathways for diagnosis and treatment were accelerated, including reduced waiting times in hospitals and large investments in radiotherapy.

^, 

International differences in access to diagnostic services, screening practices, treatments, patient pathways, and health-care systems, and their effect on survival are being further explored in ICBP studies.

Differences in coding and classification systems and cancer registration practices in different jurisdictions and countries are potential challenges to international population-based comparisons of cancer survival. Taking the example of staging—a principle determinant of cancer survival—the implementation of common recording systems has been hindered by variations in the availability of clinical information to determine the stage of cancer, the application of different staging guidelines, and differences in the pace of change in diagnostic and treatment practice.

The validity of the definition of date of diagnosis, accuracy of ascertainment of death, and extent of trace-back of cases first notified by death certificates—eg. tracing notifications back to the source capable of confirming (or contradicting) the death certificate statement—are all possible source of artifacts that might vary by registry and differentially affect survival estimates.

^, 

^, 

However, previous studies have shown that only an implausibly high level of missed cases or delayed (later or incorrectly assigned) dates of diagnosis would lead to a substantial effect on cancer survival estimates.

^, 

Cancer registries included in ICBP SURVMARK-2 are covered by reasonably complete and efficient mortality measurement systems. Still, an earlier ICBP study

has shown that local registration practices, including the prioritisation of earlier incidence dates and the exclusion of second or higher order tumours, might inflate survival and partly account for survival gaps between jurisdictions. However, specific aspects of registration that might influence survival are uncollectable or beyond the control of the registry, such as the completeness of registration, characteristics of patients with cancer who are not included in the registries, and incomplete ascertainment of deaths. The individual and cumulative effect of these factors is extremely difficult, if not impossible, to measure. Insights into data collection processes and an in-depth assessment of coding, classification, and registration in collaboration with registry staff, biostatisticians, and clinicians are key to informing future strategies to improve international survival benchmarking.

The datasets assembled in this study are unique in several ways. ICBP SURVMARK-2 is based on a large international partnership that has facilitated the collation of 20 years of data from 21 longstanding cancer registries of internationally high quality. Stringent protocols for data processing and harmonisation were put in place and extensive exchanges with both registry staff and clinical advisors in each jurisdiction ensured optimal comparability in the final datasets. The resulting survival estimates are analogous to those from previous international studies,

^, 

although slight deviations might have occurred through differences in exclusion criteria and methodological approaches (eg, differences in the diagnostic window for period analyses and cancer definitions).

This study has several limitations. Although net survival is not subject to potential errors in cause of death certification, it remains a conceptual metric that is of little use to express the survival of an individual patient with cancer. Net survival is used at the population level to allow fair comparisons between countries and groups. Although we made every effort to ensure data comparability, differences in registration practices, and sources of information available to the individual registries might account for some of the variation in cancer survival seen across countries. Furthermore, the proportion of cases eligible for analysis varied by country and depended heavily on the proportion of cases of cancer determined by death certificate (or autopsy) only (varying between 2·3% and 4·8% for pancreatic cancer, and from 0·5% to 3·4% for the remaining cancer sites), and other quality control measures such as inconsistencies between dates. In New Zealand, originally higher proportions of cases diagnosed by death certificate only can be linked to the 1994 Cancer Registry Act that mandated all laboratories to submit new diagnoses of cancer to the national registry. Because registrations were thus obtained from pathology, hospital discharge, or mortality sources, some clinically diagnosed cases not admitted to a hospital were notified to the registry via death certificates only after the patients had died. In this study, an additional assessment of inclusion of cases determined by death certificate only—under the extreme assumption that survival equalled 1 day—reduced the 5-year survival estimates by a maximum of 0·6 percentage points in the most recent period. However, including cases diagnosed on the basis of death certificate or autopsy did not alter the overall patterns across countries and jurisdictions. Moreover, data from Canada and Australia were not national but covered most of their populations (76% of Canada’s population and 70% of Australia’s population), which is unlikely to restrict their representativeness. Different methods were applied to estimate survival in different time periods (cohort approach for patients diagnosed during 1995–2009, and period approach for diagnoses after 2010), which might affect the interpretation of trends.

Finally, this study was not designed to assess the drivers of survival differences in-depth across countries or over time, but rather generate hypotheses that will be explored in future studies.