Cancer Incidence Projections in Northern Ireland to 2040

Careful monitoring of trends in cancer incidence by national cancer registries is essential if high-quality cancer diagnosis and treatment services are to be adequately maintained and resourced. However, cancer is not a disease with a single set of characteristics. There are many different types, each of which occur at different rates depending upon demographic (age, sex, socioeconomic background, and area of residence; ref. 1), lifestyle (2), and genetic factors (3). While historic trends provide evidence of changes in these risk factors and the effect of interventions such as screening or vaccination, projections provide a general guide to expected patient volumes, which is helpful in planning future health services and identifying resource requirements.

Previous analysis of cancer incidence trends in Northern Ireland (1) indicates that the age-standardized incidence rate (ASIR) of all cancers [excluding nonmelanoma skin cancer (NMSC)] is declining among men, but increasing among women. However, the strong relationship between cancer and age (1) means that the number of cases projected in future years will be largely dependent on changes in the size and age distribution of the population. Since 1993 the population of Northern Ireland has increased by 12% from 1.64 million to 1.87 million in 2017 (4). Future population projections suggest a further increase of 7% to 2.00 million by 2040 (5). However, 282,100 people were aged 60 years and over in 1993, compared with 402,800 people in 2017 (4), an increase of 43%, with the population in this age group projected to increase by a further 49% between 2017 and 2040, reaching 600,800 people (5). This large growth in the population aged 60 years and over makes it imperative that some indication of its impact on cancer incidence in future years is made available.

We thus examine historic trends in cancer incidence using data from the population-based cancer registry in Northern Ireland, which began in 1993, and project these trends forward to the year 2040 using well-established age–period–cohort (APC) models (6–9). The resulting projected incidence for all cancers combined and the 19 most common cancers will assist in planning cancer services in the future and highlight the types of cancer that may benefit from further preventative measures.

Data on all cancer cases (excluding NMSC) diagnosed in Northern Ireland between 1993 and 2017 were extracted from the Northern Ireland Cancer Registry (NICR), while historic population data were provided by the Northern Ireland Statistics and Research Agency (4) and population projections by the Office of National Statistics (5). NICR has ethical approval from the Office of Research Ethics Northern Ireland for collection of the data used in this study.

Cancer data were coded using the International Classification of Diseases 10th revision (ICD-10; ref. 10), with codes C00–C97 (excluding C44) used to identify relevant cases, and specific cancer types classified using the codes specified in Table 1. Age at diagnosis was grouped into 5-year age bands (0–4 years up to 90+ years), which along with year of diagnosis, was used to derive a 5-year birth cohort.

Five-year age group was further categorized into six broader age groups specific to each cancer site, with the boundaries for these age groups chosen so that each age group had an approximately equal number of cases. ASIRs for each cancer site, sex, and the broader age groups were generated for each calendar year using the 2013 European standard population. The annual percentage change in rates, with 95% confidence intervals (CI), for each cancer type, sex, and age group (including all ages) were determined using the Joinpoint regression program (11), which provided details of any changes in the trend direction during 1993–2017.

For each cancer site, sex, and broader age group, age-specific incidence rates (I_ay) for each 5-year age group (a) and year of diagnosis (y) were modeled using an APC model of the form:

where f_a(a) is the contribution made by 5-year age group a, f_y(y) is the contribution made by single year of diagnosis y, and f_c(c) is the contribution made by 5-year birth cohort c. This generalized linear model, which utilizes a power 5 link function, was found by Moller and colleagues (6) to provide better predictions than many common alternatives, including the frequently used Poisson model. In this model both f_a and f_c are linear functions, however if f_y were also linear, one of the three independent variables would be eliminated as a result of collinearity. As previously applied by Smittenaar and colleagues (8), we have thus used natural cubic splines to represent f_y. In this approach the linear contribution to the trend made by the year and cohort variables are collated into a single term, known as the drift, leaving the nonlinear part of the trend resulting from these components free to coexist in the regression model. The added benefit of the use of natural cubic splines is that these fit any changes in the direction of the trend much better than a simple linear approach. To ensure the best possible fit with regards to change in trend direction, three evenly spaced knots between 1993 and 2017 were used as the default when representing year using cubic splines; however, the number and placement of knots was modified for specific cancer types and/or age groups to reflect the results of the Joinpoint trend analysis. For example, if Joinpoint analysis indicated a turning point between 1993 and 2017 then an additional knot was added, with the location of this turning point included as one of the knots.

To estimate future cancer incidence, we assumed that the most recent trend will continue for the foreseeable future. However, as in previous studies (6–8), the future drift component is dampened using a geometric progression that reduces the drift by half over a 20-year period as current trends are unlikely to continue indefinitely. The degree of dampening chosen was based upon application of similar models to data from 1993 to 2012 to predict incidence in 2017 and is comparable with that used by Moller and colleagues (6).

The APC model provides estimates of the age-specific incidence rates in future years for each cancer site, sex, and 5-year age group. These projected rates are then multiplied by the population projections for that age group to give an estimate of the number of cases in future years. They are then summed to give an estimate for all ages, and are also used in combination with the European standard population to give a projection for the ASIR. An estimate of the accuracy of the prediction is provided using 95% CIs, which are provided in the Supplementary Materials and Methods.

Analysis was conducted using Stata version 15 and used the rc_spline program to derive the cubic splines (12).

During the entire 1993–2017 period under investigation, there were 190,538 cancers (excluding NMSC) diagnosed, with 50.3% of these among women. More recently, in 2013–2017, there were 4,691 male and 4,710 female cases diagnosed each year. The most common cancer types were prostate (24.2%), lung (14.5%), and colorectal (13.9%) among men and breast (29.7%), lung (13.0%), and colorectal (11.3%) among women (Table 1). The average age at diagnosis for all cancers (excluding NMSC) was 67 years (male: 68, female: 66), but ranged from 37 years for testicular cancer to 74 years for bladder cancer.

Between 1993–1997 and 2013–2017, the average number of cases of cancer (excluding NMSC) increased by 50.0% from 6,267 cases per year to 9,401 cases per year (male increase: 52.2%, female increase: 47.8%). The majority of this increase is a result of the aging of the population as the corresponding ASIRs have only increased by 11.2% (male increase: 2.2%, female increase: 15.5%) between 1993–1997 and 2013–2017.

However, the rate of change of ASIRs over time has not been constant. Among males ASIRs decreased between 1993–1999 by 1.6% per year (95% CI, 0.5–2.6), increased between 1999–2009 by 1.3% per year (95% CI, 0.8–1.8), and decreased again between 2009–2017 by 0.7% per year (95% CI, 0.2–1.3). Among women there was no significant change in ASIRs between 1993–2001, while between 2001–2017 ASIRs increased by 1.0% per year (95% CI, 0.2–1.3; Table 2).

Different cancer types also demonstrated considerably different trends over time. At the end of 2017, significant increases in male ASIRs were apparent for melanoma and head and neck, esophageal, pancreatic, and kidney cancers, while there were significant decreases for colon, prostate, stomach, and bladder cancers. Among women there were significant ASIR increases for melanoma, leukemia, and breast, lung, head and neck, liver, and kidney cancers, while significant decreases occurred for rectal, stomach, cervical, ovarian, and bladder cancers (Table 2).

Compared with rates in 2013–2017, ASIRs of cancer (excluding NMSC) are projected to decline among men with a 7% decrease by 2025 and a 9% decrease by 2040. Among women ASIRs are projected to continue to rise with a 7% increase by 2025 and a 12% increase by 2040 predicted (Fig. 1A).

By 2040, compared with the 2013–2017 average, male ASIRs are projected to decrease by more than 20% for stomach and bladder cancers and increase by more than 20% for melanoma, esophageal, liver, and kidney cancers. Also compared with the 2013–2017 average, female ASIRs are projected to decrease by more than 20% by 2040 for stomach and cervical cancers and increase by more than 20% for melanoma, head and neck, uterine, liver, kidney, pancreatic, lung, and breast cancer (Fig. 2).

Compared with the annual number of cases diagnosed in 2013–2017, the number of all cancer (excluding NMSC) cases is expected to rise by 16% for men and by 24% for women to 5,463 and 5,840 cases, respectively, by 2025. By 2040 the number of cases per year is projected to be 6,788 male and 7,450 female cases, a 45% rise among men and a 58% rise among women (Fig. 1B and Table 3).

By 2040 the number of cancers diagnosed each year among men is projected to increase for all cancer types except stomach cancer, while among women increases are expected for all cancer types except cervical and stomach cancer. In particular, the number of cases diagnosed each year is expected to more than double among males for melanoma, liver, and kidney cancers, and among females for liver, pancreatic, and lung cancers (Table 3).

More detailed information for each cancer site is available in the Supplementary Materials and Methods.

In the 25 years that cancer incidence data has been recorded in Northern Ireland, there has been an increase of 50.0% in the annual number of cases registered. This is primarily a result of increases in the population aged 60 years and over, where cancer risk is highest (1). However, additional factors are relevant as increases in ASIRs are also present for many cancer types. To aid future service delivery along with the development of strategies to deal with this increasing case volume, we have presented data on cancer incidence trends in Northern Ireland collated from a registry which has a high level of completeness and consistency in classification and collection methodology (13), and used these trends to derive estimates of cancer incidence up to the year 2040 using well-established methodology (6–9). We found that the next 25 years are likely to see a similar increase of 51.0% in cancer cases, with incidence of all types of cancer, except cervical and stomach cancer, projected to increase. Liver cancer, kidney cancer, pancreatic cancer, and melanoma are all projected to double in case volume. Increases among women are expected to be greater than those for men, as incidence rates of the most common cancers (prostate, colorectal, and lung) are expected to decline among men, while breast and lung cancer incidence rates are expected to increase among women.

In Northern Ireland, 38% of cancers (excluding NMSC) are attributable to 14 specific risk factors associated with patient lifestyle (2). Changes in these risk factors, including tobacco use, alcohol consumption, obesity, poor diet, lack of physical exercise, exposure to UV or ionizing radiation, and infections such as human papillomavirus (HPV), have the potential to alter cancer incidence rates. However, detailed trend data on lifestyle factors within the Northern Ireland population is sparse, making it difficult to assess the impact on future cancer incidence of historic, recent, or future changes in lifestyle factors.

Among the lifestyle factors associated with cancer risk, changes in prevalence of tobacco use could potentially have the greatest impact on cancer incidence trends, as tobacco use is suggested to cause 15% of all cancer cases and at least 72% of lung cancers in the United Kingdom (2). Among adults aged 16 years and over 20% of men and 18% of women were current smokers during 2017–2018, a decline on smoking prevalence in 2010–2011 when 25% of men and 23% of women smoked (14). The projected increase of 46% by 2040 in lung cancer incidence rates among females may therefore be overly pessimistic if the reported decline in smoking among women begins to impact lung cancer incidence rates in the next few decades.

A similar consideration must be given to cancer types that have a strong relationship to particular lifestyle risk factors, including melanoma and UV exposure (15), liver cancer and alcohol (16), and for obesity, physical activity, and diet-related cancers, such as uterine, esophageal, colorectal, breast, pancreatic, and liver cancers (17–19). Changes to any of these lifestyle factors could have a considerable impact on the incidence of these cancers.

In recent years (2013–2017) colorectal cancer incidence rates have declined, with this decline projected to continue for the next 5–10 years before starting to increase again. A possible contributory factor to this is the introduction of the colorectal screening program in 2010 for people aged 60–74 years old. This program uses the fecal occult blood test (20), which reduces colorectal cancer incidence by identification and later treatment of precancerous lesions (21). However, the decrease in colorectal cancer incidence rates among females has been smaller than that among men, making it difficult to fully attribute the downward trend to colorectal cancer screening. Complicating matters further is the future introduction of the fecal immunochemical test in Northern Ireland as the primary screening test for 60–74 year olds (22). This test has also been shown to reduce incidence of colorectal cancer (23) and may lead to increased uptake as a result of the test being simpler and more socially acceptable (24). The presented projections of colorectal cancer should thus be treated with caution as the impact of this planned intervention is unclear. However, colorectal cancer incidence rates were increasing prior to this intervention, thus we would expect rates to begin to increase again once colorectal cancer screening is more established, and the full reduction in colorectal cancer incidence as a result of this intervention has occurred.

Nearly all cases of cervical cancer are related to infection by HPV (25), thus girls ages 12–13 years old have been offered a vaccination against HPV genotypes 16 and 18 since 2008 (26). Modeling studies from Australia suggest that cervical cancer incidence rates will decline considerably as the cohort of vaccinated girls reaches the age when cervical cancer risk begins to increase (27). Cervical cancer incidence trends in Northern Ireland are projected to decline in a similar manner, although it is too early for the recent downward trend in cervical cancer in Northern Ireland to be explained by introduction of the HPV vaccination program. The projected decrease in Northern Ireland along the lines of the Australian study may thus be coincidental.

Prostate cancer incidence trends are highly correlated with the use of PSA testing in a population (28). In Northern Ireland, incidence rates of prostate cancer increased rapidly from 1998 to 2008, which coincided with the introduction of PSA testing in the mid-1990s. Since 2008, rates have declined slightly, with future projections based upon this assumption. Future change in PSA or other testing levels has the potential to produce prostate cancer levels that deviate considerably from current future estimates.

Compared with cancer incidence projections in the United Kingdom up to 2035, trends in Northern Ireland are broadly similar; however, some differences are apparent for particular cancer types. Rates of prostate and ovarian cancer in Northern Ireland are projected to decline, while in the United Kingdom an increase is forecast. Conversely rates of male esophageal and female lung and uterine cancer are forecast to increase in Northern Ireland, but are expected to remain steady or decline slightly in the United Kingdom. The projected increases in melanoma, liver cancer, and breast cancer in Northern Ireland are larger than the equivalent increases in the United Kingdom (8). The exact reason for any differences is unclear, but is likely to be complex and partially related to lifestyle issues. For example, prevalence of smoking varies by United Kingdom nation, and while smoking prevalence is declining across the United Kingdom, the rate of change also varies between countries (29).

The current projections, while based upon good quality data and well-established methodology, have limitations. Being based upon a smaller population than many other countries, annual cancer incidence rates can vary considerably, thus providing a degree of uncertainty in any trend analysis, projections based upon them, and the baseline against which any percentage change is measured. These projections are also susceptible to changes in population estimates, coding of cancer type, and choice of prediction methodology, as a range of viable alternatives to the approach used here exist (6, 30–34). In addition, public health initiatives targeted at cancer prevention and public awareness can cause fluctuations in the trend (35, 36), in particular, environmental and policy initiatives (e.g., smoking-free areas, screening, and vaccinations) can have a considerable impact of reducing the cancer burden. The projections presented, particularly the longer term figures, should thus be used only as a guide as to the current direction that cancer trends are taking, and should be revised periodically to take account of changes in the above mentioned factors. Comparisons with previous cancer incidence projections (37) indicates considerable agreement for some cancers (e.g., breast and melanoma), but considerable revisions for others (e.g., colorectal and cervical). These later changes are demonstrative of the impact that public health interventions can have on cancer incidence rates (e.g., screening and vaccinations, respectively), but can also indicate that an associated risk factor has, in the past, become more prevalent in the population.

Over the next two decades cancer incidence in Northern Ireland is expected to increase by 51%, with incidence doubling for many cancers. This will have a significant impact on the resources needed to diagnose and treat cancer. Cancer burden could, however, be reduced if further progress is made in improving the general health of the population through reduction of exposure to cancer risk factors especially tobacco, excessive alcohol consumption, exposure to UV radiation, and HPV.

No potential conflicts of interest were disclosed.

The funders had no role in the study design, data collection, analysis and interpretation of results, or writing of the article.

Conception and design: D.W. Donnelly, L.A. Anderson, A. Gavin

Development of methodology: D.W. Donnelly

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.W. Donnelly, A. Gavin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.W. Donnelly, A. Gavin

Writing, review, and/or revision of the manuscript: D.W. Donnelly, L.A. Anderson, A. Gavin

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.W. Donnelly

Study supervision: L.A. Anderson, A. Gavin

Northern Ireland Cancer Registry (NICR) uses data provided by patients and collected by the health service as part of their care and support. The NICR Group consists of Bernadette Anderson, Ronan Campbell, Paula Darragh, Sarah Davidson, Laura Dwyer, Deirdre Fitzpatrick, Donna Floyd, Colin Fox, Paul Frew, Tewodros Getachew, Jackie Kelly, Sinead Lardner, Ashley Levickas, Marsha Magee, Clare Marks, Brid Morris-Canter, Jacqui Napier, Eamon O'Callaghan, Jamie Roebuck, and Gerard Savage. The NICR is funded by the Public Health Agency.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.