Data collection and quality implications

We are grateful to the many organisations across the UK which collect, analyse, and share the data which we use, and to the patients and public who consent for their data to be used.

• Haematological Malignancy Research Network (HMRN)
• Information Services Division Scotland (ISD Scotland)
• International Agency for Research on Cancer (IARC)
• NHS Digital
• Northern Ireland Cancer Registry (NICR)
• Northern Ireland Statistics and Research Agency (NISRA)
• Office for National Statistics (ONS)
• Public Health England (PHE)
• Scottish Government
• United Kingdom and Ireland Association of Cancer Registries (UKIACR)
• Welsh Cancer Intelligence and Surveillance Unit (WCISU)
• Welsh Government
• World Cancer Research Fund (WCRF)

*List is not exhaustive, please email us if you wish to be added.

Cancer registries collect and record information on cancer registrations and deaths in the UK population.  Data from cancer registries are crucial for calculating cancer incidence and survival statistics. This information is essential for undertaking research into cancer, planning high quality cancer services and informing where further progress is needed.

Comparisons of cancer waiting times data cannot currently be made between countries in the UK due to different definitions and methods of measuring each wait, and separate Ministerial targets for each government. Further investigation would be needed to establish whether differences have significant impact on the comparability of the data.

Records from discontinued organisations/commissioning hubs count towards the national performance figures. We have excluded these in our own analysis so this may affect the national figure we display on our website compared to the figure NHS England produce.

ICD codes included vary per country but cancer waiting times data groups broadly cover the same tumours.

Cancer registration has a long history in many countries of the world, particularly in the more affluent regions such as the UK, but much of the world’s populations live in regions that are not covered by such systems. Data for deaths is better, but still incomplete.

The International Agency for Research on Cancer (IARC) routinely uses the available data to estimate worldwide cancer statistics. However, data quality varies greatly by region, and this should be considered when interpreting these estimated statistics.

Cancer registrations are usually completed within two years of diagnosis but, in a small proportion of cases, the process can take longer. Consequently the cancer registration databases held in the UK, and by the Office for National Statistics (ONS), are continually being updated.

Cancer Research UK uses an extract of data from ONS as the source of the data for England to collate the UK figures for cancer incidence. Other extracts from ONS, or from alternative sources, will have been taken at different times and, thus, will have minor differences between them although the revisions are good for the quality of data such differences are nearly always trivial in terms of numbers.

We use mortality data from death registrations, which means that all deaths from cancer registered in a particular year are included. Occasionally a death may occur at the end of one year but be registered in another year, for example, a death occurring on December 31st 2013 will probably be registered in January 2014.

Mortality statistics published by the Office for National Statistics are occurrences of deaths rather than registrations of death. Comparison of our data with data from a death-occurrence dataset will have small discrepancies.

The European standard population[1] gives more weight to older people and is reflective of the age profile of people in Europe. These standard populations are artificial population profiles that allow comparisons between different proportions of people of different ages. This allows comparison of incidence and/or mortality rates across geographical areas or over time.

The European Standard Population (ESP) used for many cancer statistics was created using data for 1976 (‘ESP1976’), but it is no longer representative of European Union Member States. The population is aging and projections show a progressive shift to the older ages, particularly the proportion of the population aged 65 and over.[2] A new ESP (‘ESP2013’) was calculated by Eurostat to address the changes in the age profile in Europe to be a better reflection of the population.[2]

What are the differences between the 1976 and 2013 European Standard Populations?

The ESP2013 will better represent the underlying burden of the disease in the population.

The ESP2013 will be closer to the crude rate.

The ESP2013 is weighted more heavily towards older ages.

The ESP2013 has more age groups; 85-89, 90-94 and 95+ (where previously the limit was 85+). However, due to the availability of population estimates of assured quality for those aged 95+ the Office of National Statistics (ONS) has recommended an upper age limit of 90+ should be used.[3]

Comparison of ESP1976 and ESP2013 Standard Populations, Weight by Age Group

What does this mean for cancer rate figures?

For almost all cancers, there is an increase in the age standardised cancer incidence rate using the ESP2013 compared to ESP1976, and for some cancers this change is large.

All Cancers Excluding Non-Melanoma Skin Cancer (C00-C97 Excl. C44), Number of New Cases, Crude, 1976 and 2013 European Age-Standardised (AS) Incidence Rates, UK, 2012

For all cancers combined (excluding non-melanoma skin cancer) the ESP1976 age standardised incidence rate for the UK in 2012 is 426.5 per 100,000 for males and 376.2 per 100,000 for females. The impact of the ESP2013 is that all rates appear to increase, although this effect is larger for males than females. The incidence rates using ESP2013 are 670.1 per 100,000 (56% increase) for males and 534.8 per 100,000 for females (42% increase).

All Cancers Excluding Non-Melanoma Skin Cancer (C00-C97 Excl. C44), 1976 and 2013 European Age-Standardised (AS) Incidence Rates comparison, UK, 2012

The change in the rate for each cancer will be dependent on the age distribution of diagnoses. Although there is an apparent shift in the incidence rate when calculated with ESP2013 this difference is very similar across time, so there is very little impact on the measurement of the trend. The percentage increase for incidence rates between 1979-1981 to 2010-2012 for ESP1976 is 27.3% compared with 27.8% for ESP2013.

All Cancers Excluding Non-Melanoma Skin Cancer (C00-C97 Excl. C44), 1976 and 2013 European Age-Standardised (AS) Incidence Rates, GB, 1979 to 2012

The impact on mortality rates will be similar.

Figures using the ESP1976 and ESP2013 are not comparable because the weights used to represent the population structure are different.

It is possible to compare different points in time or different cancer figures which use the same ESP, but not between ESPs; for example:

We compile incidence data produced by the regional cancer registries in England, and the three national registries in Wales, Scotland and Northern Ireland for our UK statistics but it means we have to wait until all of the data has been published by each country before we can compile and publish it

The process of registering a cancer is complex and there are a number of processes in place to ensure the data is of a high-quality so there is usually a delay of around 18 months before the data is complete enough for them to be published. Our work to compile and analyse the data can take around 6 months before it is published on our web content.

Alcohol data are based on survey respondents’ self-reported alcohol drinking in the week before providing data.[1-3] This ‘past week’ data may not be representative of an ‘average week’ for the individual, because there is a large difference between self-reported ‘usual’ alcohol drinking frequency and self-reported alcohol drinking frequency in the week before providing data. 62% of adult males and 46% of adult females in England say they ‘usually’ drink alcohol at least once or twice a week, compared with 67% and 53% of adult males and females respectively in England who say they drank alcohol in the week before providing survey data (2012).[4] 5% of children aged 11-15 in England say they ‘usually’ drink alcohol at least once a week, compared with 9% who say they drank alcohol in the week before providing survey data (2013).[3]

Body mass index BMI is used as the measure of overweight and obesity. BMI is generally a reliable measure, and is easy to calculate from height and weight. However it does not distinguish between fat and muscle, which renders standard BMI thresholds misleading in some population groups[1-4] and may obscure relationships between proportion of adipose (fatty) tissue and health outcomes.

Other measures of obesity are less commonly used in the UK, these include central adiposity (waist circumference and waist-to-hip ratio), skin fold thickness, and bio-impedance.[5]

These are derived from the statutory death registrations (registrations of deaths certified by a doctor or coroner) in the countries of the UK. As it is a legal requirement to register deaths quickly, the mortality data for the UK can be compiled more quickly, but there is still a delay of around 6-12 months before it is available to us, and then takes around 6 months before we are able to publish it.

Data on the stage at diagnosis is not yet routinely available for the UK due to inconsistencies in the collecting and recording of staging data across the UK.

Incidence by stage at diagnosis for England data are “Experimental Statistics” provided by the National Cancer Intelligence Network.

One year survival by stage at diagnosis data for England are provided by the Office for National Statistics for some cancer types. Data concerning other cancers and other survival periods are from the former Anglia Cancer Network, which only covers around 5% of the population of England and may not be representative of the country as a whole due to differences in underlying demographic factors (such as age, deprivation or ethnicity), as well as variation in local healthcare provision standards and policies.

We mainly use survival data that are calculated using the official data from the Office for National Statistics for England and/or England and Wales, where individual patients are followed up using national death registrations, and are produced approximately on an annual basis. We work with partners on this data so timescales can vary.

Non-melanoma skin cancers (NMSCs) constitute a substantial burden to the health services across the UK because of the large number of cases diagnosed each year, however NMSC incidence figures are under-estimates because the recording of NMSC is known to be incomplete.[1] Many cancer registries record only the first NMSC of each histological type (basal cell carcinoma  (BCC) and squamous cell carcinoma (SCC)) per person, and information on small NMSCs treated in primary care or the private sector may never reach the registries.[2] An estimated 30-50% of BCC and around 30% of SCC goes unrecorded, though this may vary by registry.[3-6]

From time to time, Cancer Research UK will use personal data about individuals for the purpose of generating statistics about cancer in the UK.

We acquire personal data through various legal gateways, and with the express approval of the appropriate authorities, in cases where we are unable to obtain explicit consent from individuals. Where it is possible and practical (within reason) for us to obtain consent from individuals to use their data, we will do so.

We apply appropriate safeguards to the data we acquire, to ensure they are kept secure, used only for approved purposes and by named staff; and we will never share these data with any person or organisation outside of Cancer Research UK.

For further information about how personal data are used by Cancer Research UK for generating statistics, please contact the Data Liaison Manager. You can also find out more about how we handle personal data more generally, by checking out our Privacy Policy.

Mortality may be higher than case numbers because of the way death certification and cancer registration works, and what data is available and when.

For example, if a patient has died from cancer but the official documenting the death on the certificate can’t confirm what type of cancer caused the death, it may be recorded as a non-specific cancer type (for example Cancer of Unknown Primary (CUP)). However, on receipt of the death certificate the cancer registries may then be able to identify other information about that patients’ history and determine what type of cancer it was, and update the record regarding their case diagnosis. The data would therefore show a patient recorded as having a known type of cancer for their case data, e.g. being a stomach cancer case in the incidence data, but as a different cancer type in the mortality data, e.g. being a CUP death in the mortality data. This inconsistency will remain because the death certificate cannot be changed and the effect is potentially inflated mortality statistics for non-specific cancers like CUP at the detriment of the mortality statistics for specific cancers.

These are calculated every 4-6 years by the International Agency for Research on Cancer. Collecting cancer data from around the world takes some time, and like UK statistics, processes are implemented to ensure the data are as complete as possible. This means additional delays before the worldwide and European data are published.

The World standard population[1] is reflective of an average world age profile and so rates tend to be generally lower than European-age-standardised for the same population.