National burden of cancer in Italy, 1990–2017: a systematic analysis for the global burden of disease study 2017

We monitored the burden of cancer in Italy and its trends over the last three decades, providing estimates of cancer incidence, mortality, years of life lost, years lived with disability, and disability-adjusted life-years (DALYs), for cancer overall and 30 cancer sites using data from the Global Burden of Disease study 2017. An overview of mortality trends between 1990 and 2017 was also provided. In 2017, there were 254,336 new cancer cases in men and 214,994 in women, corresponding to an age-standardized incidence rate (ASIR) of 438 and 330/100,000, respectively. Between 1990 and 2017, incident cancer cases, and, to a lesser extent, ASIRs significantly increased overall and for almost all cancer sites, but ASIRs significantly declined for lung and other tobacco-related neoplasms. In 2017, there were 101,659 cancer deaths in men (age-standardized death rate, ASDR, 158.5/100,000) and 78,918 in women (ASDR 93.9/100,000). Cancer deaths significantly increased between 1990 and 2017 (+ 18%), but ASDR significantly decreased (− 28%). Deaths significantly increased for many cancer sites, but decreased for stomach, esophageal, laryngeal, Hodgkin lymphoma, and testicular cancer. ASDRs significantly decreased for most neoplasms, with the main exceptions of cancer of the pancreas and uterus, and multiple myeloma. In 2017, cancer caused 3,204,000 DALYs. Between 1990 and 2017, DALYs and age-standardized DALY rates significantly declined (-3.4% and -33%, respectively). Age-standardized mortality rates in Italy showed favorable patterns over the last few decades. However, the absolute number of cancer cases and, to a lower extent, of cancer deaths increased likely due to the progressive ageing of the population, this calling for a continuous effort in cancer prevention, early diagnosis, and treatment.

While cancer deaths and death rate significantly increased between 1990 and 2017 by 18% and 11%, respectively, ASDR significantly decreased by 28% (Supplementary Table S2). Deaths and death rates in both sexes combined significantly increased for many cancer sites, but significantly decreased for stomach, esophageal, laryngeal, Hodgkin lymphoma, and testicular cancer. ASDRs significantly decreased for most neoplasms, with the exceptions of cancer of the pancreas, multiple myeloma, and uterus.
Between 1990 and 2017, a decline in DALYs (− 3.4%) and DALY rates (− 9.4%) was observed (Supplementary Table S4). DALYs and DALY rates significantly decreased for several cancer sites, including TBL, stomach, and leukemia, although they significantly increased for other cancer sites, such as pancreatic, prostate, kidney, multiple myeloma, uterine, cervical, and non-melanoma skin cancer. Age-standardized DALY rates declined overall (-33%) and for most cancer sites.
The patterns for YLLs, YLL rates, and age-standardized YLL rates were consistent with those for DALYs, DALY rates, and age-standardized DALY rates, with declines by − 6%, − 12% and − 34% for all cancers combined (Supplementary Table S5).
Overall and for all cancer sites, YLDs and YLD rates significantly increased between 1990 and 2017 (by 68% and 58%, respectively, for all cancers; Supplementary Table S6). Similarly, age-standardized YLD rates significantly increased overall (+ 20%) and for many cancer sites (including the most frequent ones), although they significantly decreased for some other cancers, such as bladder, stomach, and larynx. Trends in mortality rates and ASDRs over time. Figure 1 shows the trends in death rates and ASDRs for all cancers over the period 1990 and 2017, for men and women separately. Cancer ASDRs declined over the last decades in both men and women, while death rates levelled-off since the mid 2000s in men and since early 2010s in women (Fig. 1   ASDRs declined for most cancers, with long-term reductions for TBL, male upper aerodigestive tract, male bladder, female gallbladder, breast, stomach, prostate, ovarian, mesothelioma, non-melanoma skin, thyroid and Hodgkin lymphoma, and more recent declines for colorectum (since the late 2000s), liver (since the late 1990s), leukemia (since the early 2000s), non-Hodgkin lymphoma (since the late 1990s), and cervix cancer (since the late 1990s; Figs. 2a and b). ASDRs were stable over the last decades for female TBL, other female upper aerodigestive tract cancers, female bladder cancer, kidney, male gallbladder, brain, multiple myeloma, melanoma, female mesothelioma, and uterus (since the late 1990′s). Only for pancreatic cancer, ASDRs non significantly increased up to more recent calendar years, even if a levelling-off was observed since 2015.
Comparisons with Western European countries. In 2017, Italy had lower age-standardized incidence (375.5), death (121.2), and DALY (2676.6) rates compared to Western Europe overall (407.7, 128.8, and 2877.7, respectively; Table 2). In particular, these rates were lower than those of the other most populous Western European countries (i.e., France, Germany, and the United Kingdom), with the only exception of Spain. Moreover, between 1990 and 2017 ASIR increased by a lower extent in Italy (+ 9.4%) than in Western Europe (+ 13.7%) and other Western European countries, while ASDRs and age-standardized DALY rates decreased more strongly Table 1. Incidence, deaths, and corresponding age-standardized rates for all cancers and 30 cancer groups.

Discussion
This analysis of GBD cancer burden in Italy indicates that over the last decades cancer cases and incidence rates significantly increased overall and for almost all cancer sites, the main exception being stomach cancer. More favorable trends were, however, observed in ASIRs for TBL and as other tobacco-related cancers. Cancer deaths and death rates also increased, but death rates were more favorable-at least over more recent calendar years-for various neoplasms, including stomach, upper aerodigestive tract, Hodgkin lymphoma, and testicular cancer. Moreover, ASDRs steadily declined, particularly in men, for most cancer sites, the main exception being pancreatic cancer. A slight decline in DALYs and DALY rates was observed overall and for several cancer sites, and age-standardized DALY rates significantly declined for most neoplasms. Similar declines were found for YLLs, which explain about 95% of DALYs. Conversely, YLDs and YLD rates significantly increased overall and for many cancer sites, although a decrease in age-standardized YLD rates was observed for some neoplasms, including bladder, stomach, and larynx. A few considerations may help interpret the patterns of cancer in Italy. For all neoplasms and most cancers considered, we observed an increase in the absolute number of cancer cases and cancer deaths, reflected also in an increase of crude incidence and death rates. Conversely, ASIRs, and particularly ASDRs, showed favorable patterns for most neoplasms considered. The differences between crude and age-standardized rates indicate that the increase in cancer burden in Italy is mainly due to the changes in the population structure over the last decades 1,9 . The decrease in overall YLLs from 1990 to 2017, while the total deaths increased, also indicates that people are dying at older ages. Moreover, the increase in the overall YLDs, along with the decrease in YLLs, are likely due to better survival rates. The decreases in the age-standardized (death) rates for many cancers observed in Italy over the last few decades, can be likely attributable to improvements in modifiable risk factors, implementation of secondary prevention programs, as well as progresses in the management and treatment for various neoplasms. With reference to primary prevention, an important role in the reduction of cancer mortality and DALYs can be due to favorable modifications in behavioral risk factors, including particularly a decrease in www.nature.com/scientificreports/ (male) tobacco smoking [10][11][12] (the major determinant of lung and other tobacco-related neoplasms 11 ), a reduction in alcohol consumption (strongly associated to liver and head and neck cancer, and, at a lower extent, to colorectal and breast cancer 13 ), and a more affluent and varied diet (an important determinant of colorectal and other digestive tract cancers) 14,15 . The unfavorable trends in pancreatic cancer age-standardized rates may be due to the patterns in overweight/obesity 16 and diabetes 17 -two known risk factors for this neoplasm-though the prevalence of overweight/obesity in Italy has not steady increased in Italy as in the USA and many other European countries 18,19 . Improved diagnosis may have also had some role, too. The prevention and treatment of infections, such as Helicobacter pylori (which causes about 78% of stomach cancers), Hepatitis B and C virus (which accounts for most liver cancers), and Human Papilloma Virus (HPV; causally implicated in most cervical cancers) are also likely to have played a relevant role in the cancer mortality reductions [20][21][22][23][24] . Similarly, the improvement of working conditions and the reduction to occupational exposure to various carcinogens (such as asbestos) over the last decades may help explain the reduction in lung and bladder cancer ASDR 25,26 . Indoor and outdoor air pollution, a known risk factor for lung but also bladder cancer, with a synergistic effect with tobacco smoking 27 , has also substantially decreased in Italy over the last few decades, this further contributing to the favorable patterns in those neoplasms.
Screening has been the main responsible of the long-term decline in cervical cancer rates (through the Pap smear test and, more recently, the HPV test) in Italy as in most other Western European countries [28][29][30] . The implementation of organized screening programs (through fecal occult blood test, flexible sigmoidoscopy, and colonoscopy) has been shown to improve early diagnosis and reduce mortality from colorectal cancer 31,32 . Breast cancer screening (through breast examination and mammography) is also active in Italy since the early 2000s for women over age 50, although the coverage of such screening programs is still low in many Italian regions and its effectiveness is still debated 33,34 . The use of the prostate-specific antigen (PSA) test may have improved early diagnosis of prostate cancer, although the quantification of the role of the PSA test on mortality from this neoplasm is still under debate 35,36 .
A relevant role in the reduction of mortality rates from several cancers (including, among others, breast, prostate, testicular, leukemia, and Hodgkin lymphoma) is also due to the progress in the management and treatment for those cancers over the last decades, with the adoption of modern diagnostic techniques, improvements in surgery, introduction of innovative treatment approaches, recent developments in personalized medicine, as well as a better case management [36][37][38][39][40] .
We observed a relatively lower burden of cancer in Italy, as compared to other most Western European countries, this suggesting a possible improved control of modifiable risk factors 41 , as well as a better management and treatment for cancer in our country. The access to the best treatment options for cancer care in Italy is guaranteed by the presence of an efficient National Health System (NHS) with universal coverage, which assures since 1978 free of charge access to health care for all citizens 42 . Indeed, the Italian NHS ranked second globally by the World Health Organization in 2010, first by the Bloomberg Global Health Index in 2013, and ninth by the GBD Healthcare Access and Quality Index 2016 9,43,44 .

Limitations.
As in prior GBD studies, estimates presented in this study depend on the quality and quantity of the data sources available to inform the estimates 45 . Because of the lag time for data reporting, estimates for 2017 were mainly based on data and trends from recent years. Cancer mortality estimates are predominantly based on vital registration data, cancer registry data, and to a much lesser extent other data sources. In Italy, vital registration data are available since long time, has a 99% coverage, and cancer certification is reasonably reliable and valid, particularly for most common neoplasms 7 . Italian cancer registries cover approximatively 70% of the Italian population 46 . Incidence estimates used in this paper were predicted values from models that used observed mortality data as inputs and the use of MIRs. Although MIRs can change (for example as a consequence of COVID-19), incidence from mortality is estimated after careful estimation of MIRs using selected data sources 1 . Moreover, MIRs allow for a uniform method to estimate incidence and have been used in other cancer estimation frameworks which have detailed its benefits, including greater representativeness, especially in the absence of quality or complete population-based cancer registry systems 47 . Although the proportion of miscoded deaths in Italy is low, the redistribution of unspecified codes (the so-called "garbage" codes) may have somehow affected mortality, as well as incidence, estimates. Misclassification of metastatic sites as primary cancer sites (e.g., liver, TBL, and brain) is another source of potential bias, but again it should be limited in Italy, where sufficient diagnostic resources exist. Further, changes in coding practices or coding systems may also have an effect, even though mapping to the GBD causes list includes adjustments to account for different coding systems. GBD incidence and mortality estimates were somewhat higher than those provided by the Global Cancer Observatory, but differences were below 10% for many cancers, except for selected cancer sites (such as corpus uteri, brain, and stomach for incidence, and nasopharynx, cervix, and prostate for mortality) 6 . Such differences may be due to different estimation methods used and, in particular, to the redistribution of "garbage codes" made by GBD.

Conclusions
Age-standardized cancer rates in Italy showed favorable patterns over the last few decades, particularly for mortality, thanks to improvements in lifestyles, early diagnosis and treatment. However, the absolute number of cancer cases and, to a lower extent, of cancer deaths increased, likely due to the progressive ageing of the population 1 . Such increase may threaten the already progressively reducing country health resources 48 . Therefore, health care professionals, researchers, and policy makers should made a continuous effort in health promotion and prevention, to maximize the control of behavioral and environmental risk factors (namely tobacco, alcohol, www.nature.com/scientificreports/ low physical activity, high body mass index, and air pollution) for cancer, but also of other non-communicable diseases, as well as to promote early diagnosis, and further improve cancer treatment and management.

Methods
Data were extracted from the GBD 2017 results tool (http://ghdx.healt hdata .org/gbd-resul ts-tool). Methods and data sources are described in detail in previous GBD publications 1,2,49,50  Estimation framework. Details of the GBD estimation framework are provided in the eAppendix of the GBD 2017 Cancer Collaboration paper 1 . Briefly, the GBD cancer estimation process starts with mortality. Mortality estimates are based on vital registration data using an ensemble model approach 2,51 . Single-cause mortality estimates are scaled into the separately estimated all-cause estimate 2 . Cancer incidence estimates are derived from mortality estimates, dividing them by a mortality to incidence ratio (MIR). MIR is separately estimated using a spatio-temporal Gaussian process regression for each cancer type, sex, 5-year age group, location, and year 52 . The correlation between survival data and the MIR is used to estimate 10-year cancer prevalence; total prevalence is then partitioned into four sequelae (i.e., diagnosis/treatment, remission, metastatic/disseminated, and terminal phase), and each sequela prevalence is multiplied by a disability weight to estimate YLDs. Lifetime prevalence of procedure-related disability is estimated for larynx, breast, colorectal, bladder, and prostate cancer. YLLs are calculated as the difference between the corresponding standard life expectancy for a person's age, sex, and year of actual age at death 2 . DALYs are the sum of YLDs and YLLs and represent the loss in years due to premature death or morbidity (one DALY can be regarded as one lost year of "fully healthy life"). www.nature.com/scientificreports/