Cancer Projections. Incidence to

Transcription

1 Cancer Projections Incidence to

2 Ministry of Health Cancer Projections: Incidence to Wellington: Ministry of Health. Published in January 2010 by the Ministry of Health PO Box 5013, Wellington, New Zealand ISBN (online) HP 5029 This document is available on the Ministry of Health s website:

3 Contents Authorship and Acknowledgements Executive Summary v vi Introduction 1 Background 1 Method 1 Reporting 4 Updated Projections: Summary of Key Findings 6 Projections of total cancer registrations 9 How similar are the updated and the earlier projections? 9 Updated Projections by Site 10 References 36 Appendix: Prostate Cancer Incidence 37 List of Tables Table 1: percentage change in incidence, by site and sex, to Table 2: Cancer sites for which major change in incidence ASR is projected from to Table 3: Key incidence projection results for selected sites, to Table 4: Updated projected total cancer count,* annualised average, List of Figures Figure 1: All adult cancer 10 Figure 2: Childhood cancer 11 Figure 2a: Childhood leukaemia 12 Figure 3: Bladder cancer 13 Figure 4: Bone and connective tissue cancer 14 Figure 5: Brain cancer 15 Figure 6: Colorectal cancer 16 Figure 7: Gallbladder cancer 17 Figure 8: Hodgkin s disease 18 Figure 9: Kidney cancer 19 Figure 10: Laryngeal cancer 20 Figure 11: Leukaemia 21 Cancer Projections: Incidence to iii

4 Figure 12: Lip, mouth and pharynx cancer 22 Figure 13: Liver cancer 23 Figure 14: Lung cancer 24 Figure 15: Melanoma 25 Figure 16: Myeloma 26 Figure 17: Non-Hodgkin s lymphoma 27 Figure 18: Oesophageal cancer 28 Figure 19: Pancreatic cancer 29 Figure 20: Stomach cancer 30 Figure 21: Thyroid cancer 31 Figure 22: Male cancers 32 Figure 23: Female cancers I 33 Figure 24: Female cancers II 34 Figure 25: Cancer of all other sites 35 Figure 26: Prostate cancer registrations: estimates and projections by age 38 Figure 26a: Prostate cancer registrations (estimates and projections), by individual five-year age group 39 iv Cancer Projections: Incidence to

5 Authorship and Acknowledgements The modelling was done and the report written by Robert Templeton, Martin Tobias and Anna Davies (Health and Disability Intelligence, Health and Disability System Strategy Directorate, Ministry of Health). The authors would like to acknowledge constructive input from the peer reviewers: John Childs, Tony Blakely, Simon Bidwell, Deborah Woodley, Susan Hanna, Simon Ross and Helen Jones. Cancer Projections: Incidence to v

6 Executive Summary This report updates the cancer incidence projections produced by the Ministry in 2001 (Ministry of Health 2002) and first updated in (Ministry of Health 2007), using newly available data for the 2004 period. The main projection method has not been changed. For a more complete view of the projected burden of cancer, both previous reports should be consulted, together with reports on the projected mortality from cancer (Ministry of Health 2002, 2008, 2009). Overall, the risk of cancer is projected to stabilise over the coming decade ( 2016) for males and actually decline, by about 11%, for females. Nevertheless, the burden of incident cancers will still increase, by about 29% for males and 12% for females, as a result of demographic trends (increasing size and older age structure of the New Zealand population). This projected stabilisation or decline in the risk of cancer represents a notable public health success story, coming as it does after half a century (or more) of steadily increasing incidence rates. Substantial risk reductions are projected for tobacco-related cancers in males with stabilisation or lesser reduction in females. Among non-tobacco related cancers, substantial risk reductions are projected for colorectal cancer in both sexes (even without an organised screening programme), stomach cancer (both sexes) and cervical and ovarian cancers in females. By contrast, continuing increases in risk are projected for thyroid cancer (both sexes), primary liver cancer (both sexes), and lymphomas and prostate cancer in males (the latter even excluding the impact of PSA testing). Importantly, breast cancer risk is projected to stabilise, while the risk of melanoma is projected to stabilise in males and may actually decline slightly in females. vi Cancer Projections: Incidence to

7 Introduction Background Projections of cancer incidence (and mortality and survival) are useful for making investment decisions in cancer treatment facilities and in planning for the oncology workforce, and can also contribute to the formulation and evaluation of cancer control policy (Black and Stockton 2001; McDermid 2005; Moller et al 2002). In 2001 the Ministry of Health prepared incidence and mortality projections for 26 types of cancer, forecasting from or (as the most recent five-year period for which cancer registration or mortality data were then available) out to or respectively (Ministry of Health 2002). This work was undertaken on behalf of the Cancer Control Taskforce to inform the development of the Cancer Control Strategy (Ministry of Health 2003). A further five years registration data became available in 2007 (covering ), allowing the incidence projections to be updated to (Ministry of Health 2007). Rather than waiting another five years before carrying out the next update, we have introduced a dual projection method. This allows us to carry out a short-term projection (two years) given an additional three years of data, so deriving estimates for the next five-year period. We then use this new data to derive a long-term projection for the next decade, exactly as before. As we now have cancer incidence data for , we used the short-term projection methodology to give us new data for the five-year period , and then applied our long-term projection methodology to produce projections out to Using this dual approach, we will be able to update the cancer incidence projections three- rather than five-yearly in future. A similar approach has been adopted to enable the cancer mortality projections to be updated (reported separately). Both the updated incidence and mortality projections should be used in tandem, along with estimates and projections of cancer survival (also reported separately) to obtain the full picture of the projected cancer burden. For example, the updated projections can be used for modelling future capacity requirements for oncology services, along with work currently being done by the Ministry in relation to oncology workforce projections. The updated cancer incidence projections will also feed in to a cancer burden of disease study being undertaken jointly by the Ministry and the University of Otago. This study will provide the epidemiological input into a tool (currently under development) for evaluating the cost effectiveness of cancer-related interventions. Method In updating the projections we have been careful to avoid method drift so as to ensure comparability with our earlier projections. We have, however, incorporated revisions made by the New Zealand Cancer Registry to the incidence data for some cancers (in particular bladder cancer), and have used updated population projections (based on the rather than the 2001 Census) provided by Statistics New Zealand. Bladder cancer rates have been revised to reflect a change in cancer registration practice from 2005, whereby superficial (non-invasive) bladder cancers are no longer Cancer Projections: Incidence to

8 included. That is, rates prior to 2005 have been adjusted (downwards) to reflect the ratio of non-invasive to invasive cancers reported to the New Zealand Cancer Registry in the period. This adjustment assumes that this ratio has remained stable over time. In addition to projecting all childhood cancers, projections of childhood leukaemia (acute lymphoblastic leukaemia, ALL) have been added. ALL comprises approximately one-third of all childhood cancers; even so, counts for boys and girls had to be pooled to generate stable rates. Projections of prostate cancer incidence could not be updated (although they have been extended), as these are based only on data up to This constraint has been imposed in order to avoid the discontinuity resulting from the widespread use of prostate specific antigen (PSA) testing since the late 1980s and early 1990s. All adult cancer for males therefore also excludes the PSA effect. Once prostate cancer incidence returns to its pre PSA level or thereabout (see Appendix), updating will become possible once again. Note that all adult cancer, childhood cancer and adult cancer of all other sites are very heterogeneous categories. Trends in the incidence of these sites should therefore be interpreted cautiously as they may reflect divergent trends in the mix of specific cancers included in the respective category. We have not extended the new projections out beyond two five-year periods, that is, (henceforward 2016 ) as our earlier experience revealed that a projection horizon of more than 10 years yields estimates that are too uncertain to be useful for planning. For technical information relating to the cancer incidence and population data, the selection of cancer sites, and the age/period/cohort modelling methodology used for the long-term projections, please see our earlier report (Ministry of Health 2002). A brief summary of the regression modelling approach is provided here (Box 1), and the shortterm projection methodology is outlined in Box 2. 2 Cancer Projections: Incidence to

9 Box 1: Long-term projection methodology Cancer incidence rates can be thought of as realisations of three time dimensions: age (at diagnosis), period (calendar year of diagnosis) and cohort (year of birth). Given a sufficiently long historical time series of cancer incidence data, statistical models can be constructed to project incidence based on the historical pattern of age, period and cohort effects. Model identification The approach adopted was to fit the data by five year period from to for each cancer by five-year age group (the method requires data aggregated by five year period and age group, so generating ten year overlapping cohorts) to four different models: generalised linear models (the classical age/period/cohort or APC models); nonparametric generalised additive models (GAM models); non-linear models developed by Dyba and Hakulinen (so-called DH models); and Bayesian versions of APC models using second differences as autoregressive priors for the parameters, fitted using markov chain monte carlo (MCMC) methods (Bayesian Age-period-cohort Modelling and Prediction or BAMP models). Model selection For each cancer, all four models were fitted and then tested using measures of goodness of fit and ex-post tests. Those models that failed either or both statistical tests were eliminated. Note that the number and type of models finally used thus differed depending on the particular cancer. Model averaging Final fitted values (for each cancer) were calculated as the mean of the modelled number of cases, for each age-period combination, of the models finally selected and fitted for each cancer. Projection Once the final set of models for each cancer had been determined, the models were then fitted again using all available data together with the short-term projected data for 2007 and 2008 (see Box 2 below). For the BAMP, GAM and DH models, projections are automatically obtained by extending the modelled risk surface to future time periods using the fitted parameters. To project the classical APC models, however, an assumption had to be made that the age effects would remain stable (which is usually accepted as a reasonable assumption), so that the fitted age parameters could be used for projection. Future period and cohort effect parameters were estimated by fitting a linear regression model to the three most recent period or cohort parameters and then extrapolating the next two. This assumes that the recent observed period and cohort trends will continue unchanged into the future, which is a strong assumption and somewhat limits the usefulness of APC models for projection. Cancer Projections: Incidence to

10 Uncertainty estimation A BAMP 90% credible interval was obtained during the model fitting process. This includes both uncertainty associated with the Bayesian model and random fluctuations in the data. Such credible intervals are generally acknowledged to be wide, thereby providing a conservative estimate of uncertainty for the projection obtained through the model averaging process. Note that the credible interval is not centred about the average projection (for each cancer) but about the BAMP projection. Hence we show both the 90% credible interval (purple coloured area in figures 1-25) and the highest and lowest rates obtained from the individual models selected for each cancer (broken red lines in figures 1-25). Box 2: Short-term projection methodology Our projection method is based on data aggregated into five-year periods (see Box 1). For cancer incidence, the most recent five-year period is , but the two most recent years of data (2007, 2008) are not available yet. In order to make sure all the latest registration data (that is, all years up to and including ) is used to help make the best projections, we estimate the number of cancer registrations for 2007 and This then gives us a new five-year period, (henceforth ), to use in our longterm projections (so enabling the long-term projection horizon to extend out to (henceforth 2016 ). To carry out the short-term projection (that is, in this case, to derive estimates for cancer incidence in 2007 and 2008) we use the forecasts for the missing years 2007 and 2008 from the previous round of long-term projections, for which data up to 2003 was available, as the starting point. We then adjust these earlier projections to account for the observed forecast errors from the years 2004, 2005 and. Suppose X f t i the original forecast number of cancer registrations for year = t+i and X t j the observed number of cancer registrations for year = t-j, then the estimated number of cancer registrations for year t+i is 2 X t j ˆ f j 0 f X t i R X t i X 2 t i f X t j j 0 So for the most recent five-year period the number of cancer registrations is 2008 X i Xˆ i 2004 i 2007 i Reporting The updated projections from to 2016 are presented in summarised form in the body of the report (in alphabetical order, except for some sex-specific cancers and all other sites ). Full results are available from the authors. Box 3 provides a guide to interpreting the tables and charts. 4 Cancer Projections: Incidence to

11 We also present a brief summary of the key findings (page 6). The appendix (page 37) provides a more detailed analysis of prostate cancer incidence estimates and projections. Box 3: Interpreting Figures 1 25 and associated tables Except for sex-specific sites, projections for males are shown on the left and females on the right. The upper charts show empirical data points and fitted and projected models for incidence rates by age group. (All rates are per 100,000.) The lower charts show empirical data points and the fitted and projected models for age standardised incidence rates (standardised by the direct method to the World Health Organization [WHO] World Standard Population). The solid red line is the updated projection. The broken lines show the highest and lowest rates obtained for the individual models that were averaged to obtain the final updated projection (the red line) for each cancer site. The purple coloured space shows the 90% Bayesian credible interval, used to represent the prediction interval or uncertainty around the updated projection (ie, uncertainty around the red line). The tables below the charts show rates (upper tables) and counts (lower tables) by age group, for (, most recent data available) and for ( 2016, the projection horizon), and the percentage change between these two dates (that is, over the next decade). Note that the estimates shown are not the empirical (observed) data for these years, but rather the estimate for this period obtained from the fitted model. This approach is taken to reduce the stochastic noise inherent in the observed data. As a consequence, the percentage change shown in the table will be slightly different from that obtained by simply comparing the observed data for with the projection for Also note that the data for 2007 and 2008 are not in fact observed (empirical) data, but are short-term forecasts derived as explained in Box 2 above. Finally, note that the age standardised rates for adult cancers reported here are for the age group 15+ years. These rates differ from the age standardised cancer registration rates published in Cancer: New registrations and deaths (Ministry of Health 2010), which include all ages (0+ years). Cancer Projections: Incidence to

12 Updated Projections: Summary of Key Findings Updated projections by cancer site (largely in alphabetical order) are provided in Figures 1 25 and their associated tables. Key results are summarised in four tables below. Table 1: percentage change in incidence, by site and sex, to Site Male (%) Female (%) ASR count ASR count All adult Childhood Bladder Bone and CT Brain Breast Cervix Colorectal Gallbladder Hodgkins Kidney Laryngeal Leukemia Lip, mouth Liver Lung Melanoma Myeloma NHL Oesophageal Ovary Endometrium 8 38 Pancreatic Prostate Testis 2 10 Stomach Thyroid All Other Percentage change estimated from fitted (smoothed) estimates, not empirical estimates, for ASR = age standardised rate per 100,000 (directly standardised to the WHO World Population). All adult cancer for males excludes prostate cancers diagnosed solely by PSA testing (the PSA effect ). 6 Cancer Projections: Incidence to

13 Table 1 shows that the overall risk of being diagnosed with cancer is projected to reduce slowly in females, reducing by approximately 11% over the decade. Among males, the overall risk of cancer excluding the PSA effect is projected to remain stable (or decline very slightly). At the same time the burden (count) of new cancer patients is projected to increase by about 12% in females and a more substantial 29% in males, reflecting the offsetting effect of demographic trends (the expected increase in size and structural ageing of the New Zealand population). Cancer sites projected to show major changes in incidence risk (ASR) or burden (total count) over the next decade ( 2016) are summarised in Table 2 below. Table 2: Cancer sites for which major change in incidence ASR is projected from to 2016 Male >10% increase in risk* Hodgkins Liver NHL Prostate Thyroid >40% increase in burden** Hodgkins Kidney Liver NHL Prostate Thyroid >10% decrease in risk* Colorectal Gallbladder Larynx Lung Pancreas Stomach Female Hodgkins Liver Thyroid Liver Thyroid Bone Cervix Colorectal Myeloma Ovary Stomach >10% decrease in burden** Cervix Childhood cancer, all adult cancer and adult cancer in all other sites are excluded from this table as these are heterogenous categories. * Risk = age standardised rate. ** Burden = total count. Several cancers are projected to reduce by >10% in risk over the decade, including colorectal cancer in both sexes (even in the absence of an organised screening programme), cervical cancer in females (as a result of screening, not immunisation against HPV), and lung cancer in males (reflecting the differential phasing of the tobacco epidemics between the sexes). Because of demographic trends, however, no cancer actually decreases substantively in burden (count of new cases per year) except for cervical cancer. Cancer Projections: Incidence to

14 By contrast, several cancers are projected to buck the overall trend and continue to increase in incidence. These include primary liver cancer and thyroid cancer in both sexes and lymphomas and prostate cancer in males (the latter even after correction for the PSA effect ). Although relatively rare, primary liver cancer and thyroid cancer have been steadily increasing in both sexes for at least half a century. While the former may be related mainly to persistent infection with hepatitis B virus, the cause of the latter is unknown. The sharp rise in NHL noted in previous reports is projected to continue in males, while risk of NHL is projected to increase more slowly in future in females. Again the reason for this trend, or the projected sex difference in the trend, is unknown. Table 3 below summarises the key incidence projection results for the major cancer sites and those of special policy interest. Table 3: Key incidence projection results for selected sites, to 2016 Selected site Colorectal Lung Melanoma Breast Cervix Comment Rates are projected to decline in all age groups except for both sexes, falling overall by approximately one-quarter in the age group. This occurs without an organised screening programme. Burden still increases, however (by about 15%), because of offsetting demographic effects. Rates continue their steady long-term decline in males (all ages), falling by onequarter over the decade. As a result, overall burden remains stable. Trends in incidence rates vary by age group among females, most notably increasing substantively (~15%) in young adults, but the overall outcome is stability. However, female and male rates most probably will not cross over by the projection horizon. Given stable rates, burden must increase for females an increase of one-quarter is projected over the decade. Trends in rates are projected to vary by age among males, decreasing in younger and increasing in older age groups (most probably reflecting cohort effects), such that the overall rate remains stable and the total burden increases by one-third. Note, however, that the credible interval around these projections is exceptionally wide (reflecting the very divergent trends by age group). Female rates are projected to decline in all age groups except the oldest, so the overall rate falls slightly. The burden increases by about one-sixth overall. Rates are projected to fall among women younger than 45 years of age, remain stable in those aged year olds and increase slightly in those aged >75 years, with the result that the overall rate remains stable. This reflects the complex interaction of underlying epidemiological trends with the impact of screening. burden nevertheless increases by about one-fifth, reflecting the impact of demographic trends. Both rates and counts are projected to continue to fall sharply, although exact estimates are imprecise because of relatively small numbers. This is entirely due to the ongoing effect of the screening programme (insufficient time has elapsed for HPV immunisation to have had a measurable impact on incidence). The steady decline in cervical cancer incidence (since the introduction of the National Cervical Screening Programme) is a notable public health success story. 8 Cancer Projections: Incidence to

15 Selected site Prostate Comment Rates are projected to increase slowly, even after correcting for the impact of opportunistic PSA screening, with the result that the burden of new cases is expected to increase by approximately 70% (and would more than double if cases detected solely by PSA screening were included). The steep increase in burden reflects the impact of population ageing in particular, as this cancer has a particularly right-shifted age distribution. See the appendix to this report for more details. Projections of total cancer registrations Table 4 summarises the updated projections for the total cancer count (all sites combined) in 2016, derived independently by two methods: 1. using the all adult cancers model (that is, treating all cancers as a single site ) 2. summing the total counts projected for each separate site (that is, sum of all sites other than childhood cancers ). Table 4: Updated projected total cancer count,* annualised average, 2016 Male* Female Derived from all adult cancer model Derived by summing across all individual cancer sites Difference (%) 11,893 11, (2%) 10,049 10, (-3%) 21,942 22, (0%) * Excludes PSA effect. As Table 4 shows, the two counts agree closely (within 3% by sex and within 1% overall), which provides a valuable test as to the internal consistency of the models. How similar are the updated and the earlier projections? The updated (2009) projections can be compared with those produced in with respect to their respective forecasts for 2011, the projection horizon for the earlier set. In brief, both sets of projections agree closely, except for six sites where the updated projection (age standardised rate) lies below the 90% credible interval of the corresponding earlier projection. These sites are ovary, adult leukaemia, kidney and non-hodgkins lymphoma in females, and adult leukaemia and childhood cancers in males. Changes in ICD classification or coding may explain these findings, at least for ovarian cancer and adult leukaemia. Cancer Projections: Incidence to

16 Updated Projections by Site Figure 1: All adult cancer (excluding PSA effect ) All adult sites Age Male Female Rates (12,36) (10,26) (67,94) (100,158) (428,604) (425,664) -11 1,919 1,798 (1430,2097) -6 1,351 1,210 (871,1399) -10 3,669 3,726 (2840,4016) 2 1,928 1,837 (1312,2065) (360,489) (268,408) -11 Counts (37,117) (30,75) (384,542) (604,958) -14 2,508 2,955 (2436,3437) 18 3,288 3,521 (2586,4036) 7 2,554 3,440 (2737,4014) 35 1,930 2,476 (1783,2863) 28 3,510 4,902 (3737,5284) 40 2,696 3,162 (2258,3554) 17 9,195 11,893 (9550,12997) 29 8,953 10,049 (7378,11246) Cancer Projections: Incidence to

17 Figure 2: Childhood cancer Childhood Age Male Female Rates (12,28) (13,28) (10,21) (10,21) (8,17) (8,17) (11,21) (11,21) 6 Counts (19,44) (18,40) (15,34) (15,32) (12,26) (12,25) (50,96) (50,90) 9 Cancer Projections: Incidence to

18 Figure 2a: Childhood leukaemia Childhood leukaemia Males and females Age Rates (6,11) (4,7) (2,4) (4,7) 5 Counts (18,33) (12,21) (7,13) (39,63) 8 12 Cancer Projections: Incidence to

19 Figure 3: Bladder cancer Bladder Age Male Female Rates (0,1) (0,0) (8,12) (2,4) (41,63) (11,18) (110,165) (29,47) (10,14) (3,4) -12 Counts (2,4) (1,2) (43,68) (15,26) (79,120) (22,36) (145,217) (51,82) (276,401) (91,141) 15 Note: Counts for have been adjusted to reflect the ratio of non-invasive to invasive cases in This adjustment introduces an additional source of uncertainty into the projections (not reflected in the credible interval). Cancer Projections: Incidence to

20 Figure 4: Bone and connective tissue cancer Bone Age Male Female Rates (1,3) (1,2) (2,3) (1,2) (4,7) (2,4) (8,16) (4,9) (17,30) (8,15) (3,5) (2,3) -11 Counts (4,8) (3,6) (9,16) (7,13) (23,40) (14,26) (16,30) (9,18) (22,39) (13,25) (77,127) (49,83) 8 14 Cancer Projections: Incidence to

21 Figure 5: Brain cancer Brain Age Male Female Rates (1,2) (1,1) (3,5) (2,3) (11,17) (5,9) (19,32) (8,15) (17,28) (8,15) (7,10) (3,5) -10 Counts (3,5) (2,4) (19,30) (11,19) (63,96) (33,55) (36,60) (17,32) (23,37) (15,26) (152,216) (81,129) 8 Cancer Projections: Incidence to

22 Figure 6: Colorectal cancer Colorectal Age Male Female Rates (4,6) (4,6) (45,67) (42,60) (205,310) (163,240) (501,744) (432,624) (48,70) (41,58) -12 Counts (21,36) (23,37) (258,383) (254,367) (393,594) (334,492) (659,979) (744,1073) (1353,1963) (1377,1944) Cancer Projections: Incidence to

23 Figure 7: Gallbladder cancer Gallbladder Age Male Female Rates (0,0) (0,0) (1,2) (2,4) (5,9) (7,12) (11,18) (15,23) (1,2) (2,3) -4 Counts (0,1) (1,2) (7,14) (12,21) (10,17) (15,25) (14,24) (25,40) (34,53) (56,85) 24 Cancer Projections: Incidence to

24 Figure 8: Hodgkin s disease Hodgkins Age Male Female Rates (1,4) (1,4) (2,6) (1,4) (2,6) (1,2) (3,8) (1,3) (3,10) (1,2) (2,6) (1,3) 21 Counts (5,14) (4,13) (14,37) (8,24) (13,34) (5,15) (6,16) (2,6) (4,13) (1,4) (43,111) (20,59) Cancer Projections: Incidence to

25 Figure 9: Kidney cancer Kidney Age Male Female Rates (2,3) (1,2) (23,32) (11,17) (57,83) (26,42) (80,112) (35,54) (14,19) (7,10) 3 Counts (12,19) (7,13) (129,182) (65,101) (110,159) (54,87) (105,148) (59,92) (370,491) (195,279) 32 Cancer Projections: Incidence to

26 Figure 10: Laryngeal cancer Laryngeal Age Male Female Rates (0,0) (0,0) (2,5) (1,1) (8,15) (2,4) (11,20) (2,3) (2,3) (0,1) -8 Counts (0,1) (0,1) (14,26) (4,9) (16,28) (3,8) (15,26) (3,6) (47,78) (11,21) Cancer Projections: Incidence to

27 Figure 11: Leukaemia (adult) Leukaemia Age Male Female Rates (1,3) (1,2) (2,5) (2,4) (9,22) (8,15) (36,86) (25,51) (84,196) (55,109) (10,21) (7,13) -1 Counts (4,9) (3,7) (12,28) (11,22) (54,125) (46,91) (68,165) (50,104) (111,258) (94,188) (254,557) (211,393) 28 Cancer Projections: Incidence to

28 Figure 12: Lip, mouth and pharynx cancer Lip, mouth and pharynx Age Male Female Rates (2,4) (1,2) (12,23) (4,8) (23,45) (9,16) (26,52) (16,29) (7,13) (3,5) -3 Counts (12,25) (6,12) (68,132) (26,46) (43,86) (18,33) (35,69) (27,50) (162,307) (80,137) Cancer Projections: Incidence to

29 Figure 13: Liver cancer Liver Age Male Female Rates (1,3) (0,1) (13,21) (3,7) (29,47) (9,21) (40,62) (14,31) (8,12) (2,5) 16 Counts (9,16) (3,7) (76,119) (19,42) (55,91) (18,42) (52,82) (24,53) (205,290) (69,140) 47 Cancer Projections: Incidence to

30 Figure 14: Lung cancer Lung Age Male Female Rates (1,2) (3,5) (37,57) (40,64) (138,211) (123,198) (244,371) (155,247) (30,44) (27,41) -3 Counts (7,14) (15,29) (213,325) (242,388) (263,404) (251,405) (321,488) (267,425) (815,1215) (789,1229) Cancer Projections: Incidence to

31 Figure 15: Melanoma Melanoma Age Male Female Rates (1,8) (3,8) (7,41) (18,34) (33,187) (46,85) (88,486) (85,162) (143,799) (123,231) (24,132) (31,56) -7 Counts (4,26) (9,23) (41,234) (111,209) (190,1063) (278,514) (169,931) (175,331) (188,1051) (212,398) (597,3299) (793,1448) 18 Note: Unusually wide credible interval for male rate is unexplained. Further work is in progress to investigate this. Cancer Projections: Incidence to

32 Figure 16: Myeloma Myeloma Age Male Female Rates (0,1) (0,1) (6,11) (4,8) (22,37) (15,26) (47,76) (27,46) (5,8) (3,5) -15 Counts (3,6) (1,3) (37,61) (26,46) (43,72) (30,52) (62,101) (46,79) (149,231) (107,174) Cancer Projections: Incidence to

33 Figure 17: Non-Hodgkin s lymphoma (NHL) Non- Hodgkins lymphoma Rates Counts Age Male Female (1,2) 7 (5,8) 33 (30,44) 86 (79,118) 127 (118,171) 23 (21,29) 6 (3,6) 40 (30,45) 191 (172,249) 165 (151,225) 166 (155,225) 568 (528,723) (1,1) 4 (3,5) 20 (17,28) 55 (47,78) 82 (71,115) 14 (12,19) 3 (2,4) 24 (18,30) 124 (105,170) 113 (97,159) 141 (122,198) 406 (351,547) Cancer Projections: Incidence to

34 Figure 18: Oesophageal cancer Oesophagus Age Male Female Rates (0,1) (0,0) (8,13) (2,4) (30,47) (10,15) (53,82) (27,41) (7,10) (2,4) -9 Counts (2,4) (1,1) (46,74) (13,22) (57,90) (20,32) (70,108) (46,71) (182,266) (83,122) Cancer Projections: Incidence to

35 Figure 19: Pancreatic cancer Pancreas Age Male Female Rates (1,1) (1,1) (8,13) (7,11) (30,46) (27,41) (52,79) (61,91) (7,10) (6,9) -2 Counts (3,6) (3,6) (48,73) (43,66) (57,87) (54,83) (68,104) (105,156) (181,265) (213,303) 27 Cancer Projections: Incidence to

36 Figure 20: Stomach cancer Stomach Age Male Female Rates (1,2) (1,2) (10,15) (6,9) (29,44) (15,24) (50,76) (31,48) (7,11) (4,6) -15 Counts (7,13) (7,13) (57,87) (34,54) (55,85) (31,49) (66,100) (53,82) (189,277) (129,192) 7 30 Cancer Projections: Incidence to

37 Figure 21: Thyroid cancer Thyroid Age Male Female Rates (0,1) (2,4) (2,4) (8,13) (5,9) (12,18) (6,13) (11,19) (6,13) (10,16) (3,5) (8,12) 31 Counts (1,4) (5,12) (11,24) (47,76) (26,50) (72,111) (11,24) (23,40) (8,17) (17,28) (65,108) (176,251) 51 Cancer Projections: Incidence to

38 Figure 22: Male cancers (excluding the PSA effect ) Age Prostate Testicular Rates (5,10) (0,1) (15,23) (41,107) (5,8) (361,840) (2,4) (1039,2253) (2,4) (79,173) (9,13) 2 Counts (17,31) (1,5) (86,130) (236,608) (29,44) (691,1608) (5,8) (1368,2964) (3,6) (2336,5100) (145,211) Cancer Projections: Incidence to

39 Figure 23: Female cancers I Note: For key see Figure 22. Age Breast Cervical Rates (41,61) (5,11) (193,271) (6,13) (268,388) (4,10) (325,461) (3,6) (100,137) (4,9) -28 Counts (251,371) (30,69) (1171,1646) (35,77) (548,794) (8,20) (559,793) (5,11) (2572,3536) (81,170) -21 Cancer Projections: Incidence to

40 Figure 24: Female cancers II Note: For key see Figure 22. Age Ovarian Endometrial Rates (1,2) (4,7) (3,6) (19,28) (30,56) (31,47) (50,95) (39,58) (49,91) (11,15) (15,27) 8 Counts (3,6) (27,42) (17,35) (115,168) (183,341) (64,97) (102,195) (67,100) (84,157) (284,400) (394,718) Cancer Projections: Incidence to

41 Figure 25: Cancer of all other sites Other sites Age Male Female Rates (1,2) (1,2) (2,5) (2,5) (14,34) (10,24) (52,129) (29,73) (130,314) (74,182) (14,32) (8,20) -29 Counts (2,6) (2,5) (13,32) (11,29) (79,191) (59,145) (99,246) (60,150) (171,413) (127,313) (370,875) (263,631) -9 Cancer Projections: Incidence to

42 References Bidwell S, Jo E Incidence and prevalence of PSA testing in New Zealand men >50 years. Unpublished report. Wellington: Ministry of Health. (see Appendix to this report). Black R, Stockton D Cancer Scenarios: An aid to planning cancer services in Scotland in the next decade. Edinburgh: Scottish Executive Health Department. McDermid I Cancer Incidence Projections for Australia 2002 to Canberra: Australian Institute for Health and Welfare. Ministry of Health Cancer in New Zealand: Trends and projections. Wellington: Ministry of Health. Ministry of Health The New Zealand Cancer Control Strategy. Wellington: Ministry of Health. Ministry of Health Cancer Incidence Projections: Update. Wellington: Ministry of Health. Moller B, Fekjaer H, Hakulinen T, et al Prediction of cancer incidence in the Nordic countries up to the year Euro J Cancer Prevention 11 (Supplement 1). 36 Cancer Projections: Incidence to

43 Appendix: Prostate Cancer Incidence Figure 26 shows observed (solid lines) and projected (broken lines) registration rates for prostate cancer by single calendar year and five-year age group (from age 60 prostate cancer incidence is very low below this age). Unlike all other projections in this report, the prostate cancer projections begin in 1987, not. This is to avoid the projections being influenced by prostate cancers detected only through PSA testing. Opportunistic screening with PSA began in the mid 1980s, and by the early 1990s had become widespread. By , over half of men aged 50+ were being screened at least once every three years, although rates were lower in younger age groups (<70 years) and in Māori and Pacific men (Bidwell et al 2009). This explains the findings shown in the middle and lower panels of Figure 26: a sudden and dramatic rise in prostate cancer registrations in all 60+ age groups (but more so in older than younger age groups, at least on an absolute scale) beginning around 1990, far in excess of that projected in the absence of PSA testing. However, within 10 years of this upturn, registration rates reverse direction, falling in older age groups from the late 1990s almost as steeply as they had risen, to levels well below those projected for the early 2000s. For the oldest age groups (80+), there is some indication that this phase of rapid decline may now also be coming to an end, with rates possibly beginning to level off. In younger age groups (60 64 years in particular), the phase of declining rates has not yet begun, but there is clear evidence that the (earlier) phase of rapid acceleration is over, with rates now essentially stable. These changing trends by age group can be more clearly seen in Figure 26a, which shows each five-year age group separately (in relation to its own projection, based on data to 1986). What explains this somewhat complex pattern? It is almost certainly the result of widespread if opportunistic PSA screening. PSA testing leads to cancers being diagnosed (and registered) earlier than was previously the case so registration rates initially rise in all age groups. But earlier diagnosis also necessarily implies younger diagnosis so registration rates subsequently fall in older age groups while remaining elevated in younger age groups. That is, screening leads to a permanent shift in the age distribution of prostate cancer incidence. 1 Ultimately we would expect age-specific prostate cancer registration rates to level off and resume the trajectory indicated by the pre PSA projections (that is, for the foreseeable future, a gradual increase). The levels at which these rates level off will determine the extent of over-diagnosis induced by PSA screening (that is, the extent to which the excess of cases detected at younger ages exceeds the deficit of cases at older ages induced by screening). We will continue to monitor and report on this issue in subsequent updates as data accumulates. 1 Phasing of screening by age (ie higher rates of PSA testing in older age groups initially) may also have contributed to this pattern. Cancer Projections: Incidence to

44 Figure 26: Prostate cancer registrations: estimates and projections by age Broken line = projected rates (per ) based on data to 1986 (ie excluding cancers detected by PSA screening). Solid line = observed rates (per ) to (ie including cancers detected by PSA screening). 38 Cancer Projections: Incidence to

45 Figure 26a: Prostate cancer registrations (estimates and projections), by five-year age group from to 85+ Broken line = projected rates (per ) based on data to 1986 (ie excluding cancers detected by PSA screening). Solid line = observed rates (per ) to (ie including cancers detected by PSA screening). Cancer Projections: Incidence to