Comparative cancer survival in European countries

Abstract

Background

Survival for patients with cancer varies widely across Europe. This review is an attempt to explore some of the factors that influence this variation.

Sources of data

The data on cancer survival come from EUROCARE-5 and a recent OECD report. These figures have been analysed together with data from a variety of other sources: other OECD data sets; EUROSTAT; The World Bank; Gallup and the World Health Organisation.

Areas of agreement

This study confirms the importance of national socio-economic factors in influencing the outcomes for patients with cancer.

Areas of controversy

The usual suspects (limited access to expensive new cancer drugs; delayed diagnosis and late presentation) may have less influence on cancer survival than is usually assumed.

Growing points

Disparities in outcomes challenge systems of health care to re-evaluate their strategies. The key point is that these new strategies need to be informed by facts, rather than suppositions.

Areas timely for developing research

The role and scope of national cancer registries should be enhanced. We need to record more detailed information on each patient with cancer and, in an era of linked data, cancer registries are ideally placed to collect and curate such information.

cancer, survival, cancer registry, neoplasms/mortality, Europe

Introduction and background

Observed disparities in survival for patients with cancer

Two recent publications^1,2 have drawn attention to international disparities in the survival of European patients with cancer. Both reports were covered widely in the British press.³ The headlines suggested that the NHS was failing its patients with cancer and that, based on a previous analysis, if the low survival in the UK could be brought up to the level of the best in Europe, then this would save 10 000 lives each year.⁴ The aim of this review is to look behind the figures and, systematically, to analyse those factors, acting across Europe, that might produce the inequalities observed in these studies. A further aim is to suggest how we might improve and enrich the quality of information upon which international comparisons are based.

The EUROCARE-5¹ study built on the foundations established by the previous four studies. The report, with its appendices, gives a comprehensive account of the methods used and further detailed information can be obtained from the EUROCARE website (http://www.eurocare.it/Eurocare5/ProtocolsEU5/tabid/89/Default.aspx). The data were obtained from 107 population-based cancer registries within 29 European countries. For each country, age-standardized 5-year survival was reported for 10 cancers (stomach, colon, rectum, lung, melanoma of skin, breast, ovary, prostate, kidney, non-Hodgkin lymphoma). For several countries there was only partial coverage of the population: Belgium (58%); France (23%); Germany (23%); Switzerland (30%); Italy (35%); Portugal (76%); Spain (17%); Poland (13%).

The provenance of the data used in the OECD report² is not so clearly established. The data for breast cancer, cervical cancer and colorectal cancer are compiled from national governmental sources, the results for lung cancer are taken from the EUROCARE-4 study.⁵ The OECD report appears to rely heavily on data provided by secondary sources within central governments rather than figures obtained directly from cancer centres or registries and this may affect the accuracy of some of the information included within the report. For this reason, the OECD data on survival have not been used in this analysis.

What are the origins of the observed disparities?

There are several possible explanations for the observed variation in cancer survival across Europe. The first question to ask is: are the differences genuine or are they artefactual? Does the process of data capture and analysis introduce biases that might inflate estimates of survival in some countries and underestimate survival in other countries?

Data capture and analysis

If population coverage is incomplete then it cannot be assumed that the patients available for analysis are representative of the population as a whole. For example, only 23% of the French population was included in the EUROCARE-5 study—the large urban centres, Paris, Marseilles, were excluded.

If follow-up is incomplete for a significant number of patients then the proportion of censored observations will increase—survival figures may be artificially inflated by failure to record deaths in patients whose vital status is classed as unknown but who, if follow-up had been complete, would have been identified as having died from cancer. A similar problem can arise with migration: patients who are diagnosed in one country but die in a different country may skew the results from both. The EUROCARE-5 data showed that, overall, 1.1% of patients were lost to follow-up. The figures for the UK were lower than this (England 0.5%; Scotland 0.1%; Wales 0%; Northern Ireland 0%); the figures for France and Switzerland were much higher, 4.6% and 8.2%, respectively. According to OECD figures, only four countries in the EUROCARE-5 study have over 8% of their population classed as ‘foreign’⁶—Belgium, Switzerland, Germany and Austria. These also happen to be the only countries in which the 5-year survival after a diagnosis of lung cancer exceeds 15%. It is possible that selective losses from outward migration, which is turn related to ethnicity, may have inflated the survival estimates from these countries.

Diagnostic criteria may differ between nations. If a condition were defined as pre-malignant in one country but classified as an invasive cancer in another then, ceteris paribus, survival would be higher in the country that registered such patients as having invasive cancers. This phenomenon has been invoked to explain international differences in survival, for example higher survival for Japanese patients with stomach cancer compared with those diagnosed in the West.⁷

The extent to which these considerations might influence relative estimates of survival is, unfortunately, uncertain. The EUROCARE team have tried very hard to employ procedures that are as uniform as possible but, until all countries have cancer registries that cover 100% of the population and attain equivalent rates for completeness of follow-up, there will be some doubt about the true comparability of data. The data from France and Switzerland are, on the basis of population coverage and completeness of follow-up, likely to be somewhat less reliable than the data from countries with full coverage of the population and losses to follow-up of 1% or less. The figures from Iceland are based on a total of just over 10 000 patients and, given this relatively small number, the confidence intervals on the estimates of survival are comparatively wide: for stomach cancer the estimate spans a range of 13.5% (from 27.8% to 41.3%); the corresponding figure for England is 0.8% (from 16.6 to 17.4%).

These potential biases are very unlikely to explain more than a small proportion of the variation observed in both the EUROCARE-5 study and the OECD report. We need to consider other explanations. These can be classified as follows: national socio-economic factors; factors related to the general health of nations; the national resources provided for health care in general; the specific resources available for the care of patients with cancer. Fortunately, much of the information required to address the potential effects of these various factors is publicly available and what follows is a systematic attempt to explore their relationships with some of the outcomes reported by EUROCARE-5 and the OECD.

Methods

The data used in this analysis are from publicly accessible sources. I retrieved information covering the following domains: general socio-economic measures; general measures of population health and lifestyle; availability of resources for health care; specific resources available for the management of cancer. Wherever possible, a corroborative approach has been applied to the data—for each potentially influential variable a variety of different measures have been used.

General socio-economic measures

The most widely used measure for comparing economic prosperity is gross domestic product (GDP) per capita. The data used here come from the OECD⁸ using average figures for the years 2000–2007—values are in current US$ adjusted for purchasing power parity (PPP). The Cantril score is a measure of general well-being. The Gallup organisation has modified the Cantril Self-Anchoring Striving Scale⁹ to form a ‘Life Evaluation Well-Being Index’—the values used¹⁰ are the raw scores for each country obtained from surveys carried out between 2005 and 2009.

The Human Development Index (HDI)^11,12 is based on the work of the economist Amartya Sen. It encompasses three important dimensions of human experience: the opportunity to live a long and healthy life; the availability of education; being able to have a decent standard of living. Life expectancy at birth, adult literacy rate, combined enrolment ratio (proportion of children enrolled in education as a proportion of the theoretical total number who would be eligible for enrolment) and GDP per capita are combined into a single index which can vary between 0 (minimum achievement) and 1.0 (maximum achievement).

The proportion of citizens classified as ‘foreign’ can be used as an estimate of migration. The data used here come from the OECD and are for the year 2000.⁶ Income inequality is a measure of the size of the gap between those who have and those who have not. I have used two sources of data: the EUROSTAT data for 2005¹³ which measures inequality as the ratio of total income received by the 20% of the population with the highest income to that received by the 20% of the population with the lowest income. The Gini coefficient¹⁴ is a measure of ‘the extent to which the distribution of income or consumption expenditure among individuals or households within an economy deviates from a perfectly equal distribution’. Values range from 0 (perfect equality) to 100 (total inequality). The values used in this analysis are from the year 2000.

Civic engagement is a significant component of general well-being.¹⁵ The OECD have measured civic engagement using a composite figure, based on voter turnout and the extent to which citizens feel involved in the making of rules and laws, the values range from 0 (no engagement) to 10 (maximal engagement).

General measures of population health and lifestyle

The most widely accepted measure of the general health of a nation is life expectancy at birth. The figures used here are for the year 2000 and were obtained from EUROSTAT.¹⁶ The World Health Organisation provides national data on disability-adjusted life years (DALYs) for all ages and for those aged over 60¹⁷—these figures have been used to provide an indication of the general health of the whole population and that of older citizens. For this analysis the DALYs lost due to cancer have been subtracted from the estimate of total DALYs lost for each country. The ‘Adjusted DALY’ figure indicates that the raw estimate has been modified to correct for differences between countries in the age structure of their populations. The European Commission publishes national data on self-reported illness¹³ the figure used in this study is the percentage of the population reporting ‘a long-standing illness or health problem’; data are for the year 2005. Similarly, data are available for EUROSTAT¹³ on the proportion of the population who feel that they have needs for health care that, for reasons of money, geography or excessive waiting times, are not being met; data used are from 2005. Smoking, obesity and alcohol use can all affect, not just the incidence of cancer, but also the outcomes of treatment in patients with cancer. The data on smoking prevalence are from a recent systematic review;¹⁸ the figures on obesity are from EUROSTAT;¹³ alcohol use is measured as litres of pure alcohol consumed per adult (>15 years) per annum.¹³

Availability of resources for general health care

The percentage of GDP spent on health care uses OECD data⁸ to calculate an average for each country for the years 2000–2007. The figure is the total of individual and collective spending on health care expressed as a percentage of national GDP. The health spending per capita is for the year 2005 and is obtained from the OECD—the units are US$ adjusted for PPP. OECD^19,20 data were also used to calculate the average spending on prescription drugs per capita from 2000 to 2007. The figures on numbers of doctors per 100 000 population and the number of acute beds per 100 000 were obtained from EUROSTAT¹³ and used to calculate the ratio of doctors to acute hospital beds. Data on megavoltage radiotherapy treatment machines per million population came from the ESTRO QARTS project²¹ Information on the number of CT scanners and MRI scanners per million population was obtained from the OECD^19,20 and the average numbers over the period 2000–2008 were used in the analysis.

Annual spending on cancer research

The national figures on cancer research spending come from a European Cancer Research managers report published in 2007.²² The figures are for spending in 2004, mid-way through the EUROCARE-5 study period.

Specific resources available for the management of cancer

The data on spending on cancer have been extracted from the recent review by Luengo-Fernandez et al.²³

Cancer survival

The figures on cancer survival across Europe were abstracted from the EUROCARE-5 publication (and its appendices). Data from the European Cancer Observatory²⁴ were used to calculate the ratio of mortality to incidence as a proxy measure of current survival. The Office of National statistics (ONS) website²⁵ provided the data on site-specific survival for each of the English cancer networks. The more detailed analysis has concentrated on colorectal cancer. This cancer was chosen because it can act as a bellwether—it is common (third most frequent cancer in Europe, second only to lung cancer as a cause of death from cancer); as the population ages, its incidence is rising; it is potentially curable; screening programmes are being introduced; the sex incidence is approximately equal; its successful treatment involves close multidisciplinary collaboration; surgery, chemotherapy and radiotherapy all have important roles to play in its management. The EUROCARE-5 study reports data separately for colon cancer and for rectal cancer. The 5-year survival estimates for colorectal cancer can be estimated by using incidence data²⁶ for the two tumours and calculating overall colorectal cancer survival on a pro rata basis. Survival figures are for cause-specific survival, with relative survival used as a proxy for this measure.

Given the heterogeneity of the data used in this study, detailed statistical analysis would be inappropriate. The only statistical test used has been the Spearman rank correlation coefficient. The following convention was adopted for interpreting the correlation coefficient: highly significant association r ≥ 0.7; significant association r > 0.5 r < 0.7; modest association r > 0.32 r < 0.5; no association r 0–0.32. The scatterplots are presented to show that, even when significant correlations exist, there are still potentially instructive exceptions to the general trends.

Results

There were considerable international variations in the factors used as the independent variables in this analysis (Fig. 1). For most of these variables there was a smooth gradient from lowest/worst to highest/best, the exception was the HDI which showed a steep gradient over 11 countries between 0.795 (Bulgaria) and 0.896 (Portugal) and a shallower gradient thereafter, with 15 countries (Italy, Spain, Germany, France, Austria, Denmark, Finland, Ireland, UK, Switzerland, Belgium, Netherlands, Sweden, Iceland and Norway) all with HDI values in the range 0.916 to 0.944.

Fig. 1

The variability of input variables in the nations studied in the EUROCARE-5 report. GDP, gross domestic product; DALY, disability-adjusted life year; PPP, purchasing power parity; MRI, magnetic resonance imaging. The sources of data are given in the text and references.

For most of these measures, the UK ranking is close to the middle. The exceptions are general spending on pharmaceuticals and spending on drugs for cancer, where the UK ranking is in the lower third.

The correlation matrix (Table 1) illustrates the relationship between the variables that were considered and population-based estimates of the survival of patients with cancers of the large bowel, lung and breast. Survival after a diagnosis of cancer is clearly associated with socio-economic factors, with measures reflecting general health and lifestyle, and with estimates of the resources that are available for health care. These factors have less correlation with the incidence of cancer. The strongest and most consistent associations are for those measures—life-expectancy, DALY's lost—that serve as proxies for the general health of the population. Of the lifestyle factors analysed (alcohol consumption, obesity and smoking), it is alcohol intake that is most closely associated with survival.

Table 1

Correlation matrix exploring relationships between input variables and cancer survival, mortality and incidence

The cells are shaded according to the convention: dark grey, coefficient ≥0.7; medium grey, coefficient >0.5 to <0.7; light grey, coefficient >0.32 to <0.5; white, coefficient 0 to 0.32. The values are Spearman's rank correlation coefficients for each pair of variables. GDP, gross domestic product; DALY, disability-adjusted life year; CI5 IX, cancer incidence on five continents volume 9; EUCAN, European Cancer Observatory.

The scatter plots shown in Figure 2 illustrate the relationships for colorectal cancer that were summarized in Table 1. It is clear that some factors have a linear relationship with survival (life-expectancy, DALY's lost, percentage of GDP spent on health, HDI, MRI scanners per million population, megavoltage treatment units per million population) whilst for other factors there is no clear linear relationship (proportion of population classed as foreign, medication as a proportion of per capita spending on cancer). The relationship between survival for patients with colorectal cancer and healthcare costs per capita seems particularly steep. Another feature of this graph is that it shows quite clearly that similar levels of investment in cancer services can have very different outcomes. The UK, Ireland, Spain, France, Italy and Sweden all spend similar amounts per capita on cancer care but the difference in 5-year survival from the worst (the UK 53%) to best (Sweden 62%) is almost 10%.

Fig. 2

The relationships, for each nation in EUROCARE-5, between input variables and 5-year survival after a diagnosis of colorectal cancer. CSS- Cause-specific (relative) Survival. GDP, gross domestic product; DALY, disability-adjusted life year; PPP, purchasing power parity; HDI, human development index; MV, megavoltage radiotherapy machines; MRI, magnetic resonance imaging; In this, and subsequent figures, the country codes are as follows: country code, Austria AUS, Belgium BEL, Bulgaria BGR, Croatia HRV, Czech Rep. CZE, Denmark DNK, Estonia EST, Finland FIN, France FRA, Germany DEU, Iceland ISL, Ireland IRL, Italy ITA, Latvia LVA, Lithuania LTU, Malta MLT, Netherlands NLD, Norway NOR, Poland POL, Portugal PRT, Slovakia SVK, Slovenia SVN, Spain ESP, Sweden SWE, Switzerland CHE, UK GBR, The sources of data are given in the text and references.

The pairwise comparisons shown in Figure 3 show that there is not necessarily a clear correlation in survival between types of cancer. The relationships between colon cancer survival and rectal cancer survival, and between breast cancer survival and ovarian cancer survival are close to linear. There is no such relationship between oesophageal cancer survival and pancreatic cancer survival, or between renal cancer survival and colorectal cancer survival. This suggests that it is not some overall inadequacy in cancer care that causes lower survival in some countries. Examples of this include the comparatively high survival for oesophageal cancer in Ireland, and for pancreatic cancer in Croatia. Conversely, the survival for ovarian cancer in Ireland and the UK seems particularly poor relative to these countries' results for other types of cancer.

Fig. 3

Relationship between 5-year survival for different tumour types and different countries. CSS, cause-specific (relative) survival.

The results shown in Figure 4 suggest that results for some types of tumour (colorectal, breast, ovary) are more affected by the general health of the older population than is the case for other tumour types (lung, pancreas, kidney). Thus, although the factors summarized in Table 1 may be significantly associated with outcome for certain types of cancer, these relationships are not universal, and for each type of cancer there are likely to be different patterns of factors influencing outcome.

Fig. 4

Relationship, for different types of tumour and for different nations, between DALY lost and 5-year cause-specific (relative) survival (CSS).

The relationships between per capita expenditure on cancer research and survival for patients with breast cancer and colorectal cancer (Fig. 5) show a steep initial rise followed by a plateau. The UK is a clear outlier. The plot for lung cancer shows no obvious relationship between spending and outcome.

Fig. 5

Relationship between 5-year cause-specific (relative) survival (CSS) and spending on cancer research for different nations and three different tumour types: lung, breast and colorectal.

Figure 6 shows the extent to which survival estimates for the English cancer networks interdigitate with those for northern European countries. Although the majority of networks have survival figures significantly lower than those in northern European countries the results from certain networks, Dorset for example, are above the overall European average.

Fig. 6

Forest plots (estimates of 5-year relative survival with 95% confidence intervals) for English cancer networks and northern European nations for breast cancer, stomach cancer, lung cancer and colon cancer. The vertical line indicates the figure for overall European average survival. The sources of data are given in the text and references.

Discussion

Limitations of this study

One major problem with the methods used in this study is the assumption that overall population measures of health and disability can be applied to that sub-population of citizens who are unfortunate enough to develop cancer. There is no adjustment for socio-economic deprivation, ethnicity or the myriad other factors that might selectively influence not just the incidence of cancer but also patients' subsequent survival. Unfortunately, there is really no way round this problem—large-scale prospective longitudinal studies would be required to address the issue, and studies that are sufficiently large simply do not exist.

The analysis has relied on data from published sources. This means that it is impossible to apply methods of quality control: the provenance of the data can be described, but its integrity simply has to be taken on trust. It has only been possible to ask questions for which there are data available. The main problem here is the lack of national information on the timeliness and quality of surgery, on pathways of care, on access to specialized care and on adherence to guidelines.

There is abundant opportunity for confounding amongst the variables used. Richer countries will spend more on health, more upon cancer care and more upon cancer research. These variables cannot be regarded as separate but reflect different aspects of the same underlying property—in this case, affluence. Much of the variation between Western and Eastern Europe originates from the comparative affluence of the countries in the West. There is a flip side to this problem—the presence of confounding means, effectively, that variables are being explored from multiple perspectives and the overall consistency of the results implies that genuine effects are present.

Spending on drugs is not necessarily a comparative indicator of availability. Some countries may be able to negotiate discounts for the more expensive new drugs²⁷ and this would alter the relationship between expenditure and prescribing.

There is only limited data available on national differences in rates of recruitment to clinical trials. The proportion of patients enrolled in clinical trials has been used as a measure of the quality of care. This analysis has been unable to assess whether or not differences in rates of trail recruitment have any effects on cancer survival figures.

Advantages of this study

The study takes a broad perspective—there is no ideologically based agenda. It complements a recent analysis of the last iteration of the EUROCARE data²⁸ but provides additional specific information on a number of other influential variables specifically related to spending on cancer. It is predominantly descriptive and makes few assumptions, deliberately eschewing any elaborate analyses or models.

False expectation or under-achievement?

Despite numerous caveats,^29–31 there are, amongst the nations of Europe, genuine and significant differences in survival after a diagnosis of cancer. This is scarcely surprising: we are dealing with radically different systems of health care, with wide variations in available resources and disparate populations. The real interest in the data provided by the EUROCARE-5 analysis¹ and by the OECD² lies in the fact that the results from some countries, specifically the UK and Denmark, are far worse than might be expected. This raises two main questions: if worse, why? were our (disappointed) expectations reasonable ones?

The systematic approach adopted here can go some way to answering these questions. Survival after a diagnosis of cancer depends upon the general wealth and health of the population as a whole. It is also significantly related to the resources, technological, human and financial, available for the care of patients with cancer. The UK, despite its post-imperial dreams, is well down the international rankings for most of these influential factors. Given that our expenditure and resources are half-way down the league, it is not surprising that our results are also mid-table.

The surprise from the European data on cancer survival is not, therefore, the results themselves but that our expectations were, in the first place, so unrealistically high. So why do we assume that we will do better than the evidence suggests that we have any right to expect? One explanation may lie in the disconnection, particularly evident in the UK, between spending on cancer research and spending on cancer treatment. We spend over twice as much per capita on cancer research as our nearest European rival, Sweden: UK spending per capita in Euros (2004) = 13.3; Sweden 6.8.^22,32 Our spending on cancer treatment ranks 11th out of the 25 countries in the EUROCARE-5 analysis: on a per capita basis, we spend only ∼55% of what Germany spends.²³ There is a climate of expectation that is driven by spending on cancer research, with its talk of breakthroughs and attendant publicity, that is not, at least in proportional terms, matched by spending on the management of patients with cancer. The clear message is that increasing spending on cancer research is unlikely to compensate for spending relatively less on cancer treatment. We get what we pay for—we should not hanker after salmon if we only have a herring income.³³

The economy, stupid³⁴

It is difficult to look at the data presented here without concluding that the solution to the international disparities in outcome would be to increase spending on health in general, and cancer services in particular, in those countries whose survival figures are significantly below the average. This is undoubtedly a naive view, and one that can be criticized from several perspectives.³⁵ However, it has the virtue of simplicity and provides a departure point for a more considered analysis: the root of the problem is that it is not simply about spending, it is about spending wisely. It is not just about treatment, it is also about prevention, screening and earlier diagnosis. It also raises a troubling question: if, as a result of the economic collapse of 2008, spending on health declines in real terms, will this be associated with a decline in survival for patients with cancer, particularly in those countries hardest hit by the recession?

Diagnostic delay

It is unlikely that delay in diagnosis provides the sole explanation for the observed differences in cancer survival figures. There is not necessarily a simple linear relationship between the duration of symptoms before the start of treatment and the probability of long-term survival.^36,37 Patients with short symptomatic intervals and systemic symptoms may fare worse than patients with longer times to presentation and symptoms exclusively related to the primary tumour.³⁸ This does not mean that strategies aimed at decreasing the interval between first symptom and first treatment^39,40 will be unsuccessful. It simply implies that the relationship between the duration of symptoms and outcome may be more complex than it initially appears. There is, as yet, no firm evidence that variation in outcome is related to delays in primary care,⁴¹ despite the assumption in the media (based on advice from un-named ‘experts’) that ‘many of the problems in the UK were caused by delays in diagnosis, resulting in cancers not being treated until they had reached an advanced stage and leading to especially poor survival for older patients’.³ The title of this article is typically inaccurate—‘Cancer Survival in Britain the Worst in Europe’. The text refers only to cancer survival figures being below the European average, the worst survival figures in Europe are, in fact, in the countries of the former Eastern bloc.

The drugs don't work?

There is a belief, originating in economic research funded by the pharmaceutical industry,⁴² and enthusiastically promulgated by free marketeers,⁴³ that increased spending on new cancer drugs will be the solution to the problem of international disparities in cancer survival. The evidence from the countries of the former Eastern Bloc suggests that this is not the case. These countries spend a higher proportion of their cancer budgets on drugs, the opportunity cost being less spending on surgery and radiotherapy, but their results are still, compared with the European average, poor. Absolute spending per patient is lower than in Western Europe and there may be problems with access to, and administration of, even relatively cheap cytotoxic drugs. Nevertheless, it is easier to spend money on drugs than it is to reconfigure or develop surgical or radiotherapeutic services. The lesson here is that it is vital to have a suitable infrastructure, based on adequate resources for diagnosis, surgery and radiotherapy. Without this solid foundation, increased spending on drugs is futile.

Epidemiology is about people

This study draws attention to the need for a more personal approach to the epidemiology of cancer care. We should be asking more of our cancer registries and, importantly, supporting them financially in an enhanced role. Unfortunately, at least in the UK, the trend appears to be in the opposite direction. The coalition government has withdrawn funding for some of the programmes run by the ONS.⁴⁴ It is ironic that GCHQ and the NSA can know so much about our personal lives but a cancer registry can only provide limited information. In order to make sense of the differences in outcomes, both between and within nations, we need more detailed information about each individual patient diagnosed with cancer. It would also be helpful if those countries whose cancer registries cover only a proportion of their population felt able to extend their coverage so as to include all their citizens.

A minimum set of data, whose origins are based on the results summarized in this study is shown in Table 2. Some of this information can be obtained automatically through electronic linkage of existing sets of data and so its acquisition may not add significantly to the costs of registering patients with cancer. Other items, such as quantitative details of consumption of tobacco and alcohol, are not uniformly available: the items in Table 2 represent a set of aspirations rather than information that is currently available.

Table 2

A summary of the variables that are useful for a deeper understanding of differences in cancer survival between nations

Expanded data set for comparative studies of survival

Cancer-related

Age at diagnosis

Gender

Diagnosis

Site

Morphology

Stage at diagnosis

Staging investigations

Dates of milestones on route from symptoms to diagnosis to treatment to follow-up

Primary treatment (with dates)

Surgery

Chemotherapy

Radiotherapy

Relapse/recurrence

Date

Site

Treatment

Current status and date

Cause of death

Specify any long-term complications of treatment

Personal medical

Previous or concurrent illnesses with measure of their severity

Regular drug therapy

Family history

Height, weight, BMI

Blood pressure

Personal social

Marital status

Living alone

Married/cohabiting

Tobacco use

Alcohol use

Exercise

Diet

Occupation

Income

Level of education

Demographic

Place of residence

Urban/rural

Access to services

Links to census data on neighbourhood characteristics

The literature does not help much

Attempts to use the published literature on international variations in process and outcomes for patients with colorectal cancer have proved somewhat unfruitful. The inchoate nature of the available information means that it is extremely difficult to identify any meaningful patterns or to draw any useful conclusions.⁴⁵ We will really only be able to make proper sense of international survival figures once we have, at the level of the individual patient, information that is sufficiently rich in detail to provide information on the context, processes and outcomes of cancer care. It is unlikely that this level of international co-operation will be achieved in the immediate future, although important efforts are being made with high-resolution studies, with detailed sets of data, looking at comparisons of survival between various countries.^46–49

Medicine is a social science

The main lesson from the EUROCARE-5 and OECD studies is this: there are no simple, unifactorial, explanations for international differences in survival for patients with cancer. The roots of the observed inequalities are to be found deep within societies. If disparities are to be eliminated then a systems-based approach will be required. It is insufficient to concentrate attention on one small area within a complex whole.

Rudolf Virchow, the founder of cellular pathology, pointed out a long time ago: ‘Medicine is a social science, and politics is nothing else but medicine on a large scale. Medicine, as a social science, as the science of human beings, has the obligation to point out problems and to attempt their theoretical solution: the politician, the practical anthropologist, must find the means for their actual solution’.⁵⁰ His words ring as true now as they did then.