Lung cancer rate predictions using generalized additive models

Comparison of Three Prediction Models for Predicting Chronic Obstructive Pulmonary Disease in China

Purpose

To predict the future number of patients with chronic obstructive pulmonary disease (COPD) in China and compare the three prediction models.

Methods

A generalized additive model (GAM), autoregressive integrated moving average (ARIMA) model, and curve-fitting method were used to fit and predict the number of patients with COPD in China. Data on the number of patients with COPD in China from 1990 to 2019 were obtained from the Global Burden of Disease (GBD) database. The coefficient of determination (R2), mean absolute error (MAE), mean absolute percentage error (MAPE), root mean squared error (RMSE), relative error of prediction, Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC) were used to evaluate and compare the fitting effect, prediction effect, and reliability of the three models.

Results

The GAM, ARIMA, and curve-fitting methods could predict future trends in COPD in China. The performance of the GAM is the best among the three models, whereas the curve fitting method is the worst, and the ARIMA (0,1,2) model is in between. The prediction results of the three models showed that the number of patients with COPD in China is expected to increase from 2020 to 2025.

Conclusion

GAM and AIRMA models are recommended for predicting the future prevalence of COPD in China. The number of patients with COPD in China is expected to increase in the next few years. The prevention and control of COPD in China still needs to be strengthened. Using appropriate models to predict future trends in COPD will provide support for health policymakers.

Comparison of statistical models to predict age-standardized cancer incidence in Switzerland

This study compares the performance of statistical methods for predicting age-standardized cancer incidence, including Poisson generalized linear models, age-period-cohort (APC) and Bayesian age-period-cohort (BAPC) models, autoregressive integrated moving average (ARIMA) time series, and simple linear models. The methods are evaluated via leave-future-out cross-validation, and performance is assessed using the normalized root mean square error, interval score, and coverage of prediction intervals. Methods were applied to cancer incidence from the three Swiss cancer registries of Geneva, Neuchatel, and Vaud combined, considering the five most frequent cancer sites: breast, colorectal, lung, prostate, and skin melanoma and bringing all other sites together in a final group. Best overall performance was achieved by ARIMA models, followed by linear regression models. Prediction methods based on model selection using the Akaike information criterion resulted in overfitting. The widely used APC and BAPC models were found to be suboptimal for prediction, particularly in the case of a trend reversal in incidence, as it was observed for prostate cancer. In general, we do not recommend predicting cancer incidence for periods far into the future but rather updating predictions regularly.

Predicting Patient-Reported Outcomes Following Surgery Using Machine Learning

Patient-reported outcomes (PROs) enable providers to identify differences in treatment effectiveness, postoperative recovery, quality of life, and patient satisfaction. By allowing a shift from disease-specific factors to the patient perspective, PROs provide a tailored patient-centric approach to shared decision-making. Artificial intelligence (AI) and machine learning (ML) techniques can facilitate such shared decision-making and improve patient outcomes by accurate prediction of PROs. This article aims to provide a comprehensive review of the use of AI and ML models in predicting PROs following surgery through an overview of common predictive algorithms and modeling techniques, as well as current applications and limitations in the surgical field.

Estimating the costs of non-medical consumption in life-years gained for economic evaluations

Including the costs of non-medical consumption in life years gained in economic evaluations of medical interventions has been controversial. This paper focuses on the estimation of these costs using Dutch data coming from cross-sectional household surveys consisting of 56,569 observations covering the years 1978-2004. We decomposed the costs of consumption into age, period and cohort effects and modelled the non-linear age and cohort patterns of consumption using P-splines. As consumption patterns depend on household composition, we also estimated household size using the same regression modeling strategy. Estimates of non-medical consumption and household size were combined with life tables to estimate the impact of including non-medical survivor costs on an incremental cost-effectiveness ratio (ICER). Results revealed that including non-medical survivor costs substantially increases the ICER, but the effect varies strongly with age. The impact of cohort effects is limited but ignoring household economies of scale results in a significant overestimation of non-medical costs. We conclude that a) ignoring the costs of non-medical consumption results in an underestimation of the costs of life prolonging interventions b) economies of scale within households with respect to consumption should be accounted for when estimating future costs.

Multidimensional penalized splines for incidence and mortality-trend analyses and validation of national cancer-incidence estimates

BACKGROUND

Cancer-incidence and mortality-trend analyses require appropriate statistical modelling. In countries without a nationwide cancer registry, an additional issue is estimating national incidence from local-registry data. The objectives of this study were to (i) promote the use of multidimensional penalized splines (MPS) for trend analyses; (ii) estimate the national cancer-incidence trends, using MPS, from only local-registry data; and (iii) propose a validation process of these estimates.

METHODS

We used an MPS model of age and year for trend analyses in France over 1990-2015 with a projection up to 2018. Validation was performed for 22 cancer sites and relied essentially on comparison with reference estimates that used the incidence/health-care ratio over the period 2011-2015. Alternative estimates that used the incidence/mortality ratio were also used to validate the trends.

RESULTS

In the validation assessment, the relative differences of the incidence estimates (2011-2015) with the reference estimates were <5% except for testis cancer in men and < 7% except for larynx cancer in women. Trends could be correctly derived since 1990 despite incomplete histories in some registries. The proposed method was applied to estimate the incidence and mortality trends of female lung cancer and prostate cancer in France.

CONCLUSIONS

The validation process confirmed the validity of the national French estimates; it may be applied in other countries to help in choosing the most appropriate national estimation method according to country-specific contexts. MPS form a powerful statistical tool for trend analyses; they allow trends to vary smoothly with age and are suitable for modelling simple as well as complex trends thanks to penalization. Detailed trend analyses of lung and prostate cancers illustrated the suitability of MPS and the epidemiological interest of such analyses.

An R package for an integrated evaluation of statistical approaches to cancer incidence projection

Background

Projection of future cancer incidence is an important task in cancer epidemiology. The results are of interest also for biomedical research and public health policy. Age-Period-Cohort (APC) models, usually based on long-term cancer registry data (> 20 yrs), are established for such projections. In many countries (including Germany), however, nationwide long-term data are not yet available. General guidance on statistical approaches for projections using rather short-term data is challenging and software to enable researchers to easily compare approaches is lacking.

Methods

To enable a comparative analysis of the performance of statistical approaches to cancer incidence projection, we developed an R package (incAnalysis), supporting in particular Bayesian models fitted by Integrated Nested Laplace Approximations (INLA). Its use is demonstrated by an extensive empirical evaluation of operating characteristics (bias, coverage and precision) of potentially applicable models differing by complexity. Observed long-term data from three cancer registries (SEER-9, NORDCAN, Saarland) was used for benchmarking.

Results

Overall, coverage was high (mostly > 90%) for Bayesian APC models (BAPC), whereas less complex models showed differences in coverage dependent on projection-period. Intercept-only models yielded values below 20% for coverage. Bias increased and precision decreased for longer projection periods (> 15 years) for all except intercept-only models. Precision was lowest for complex models such as BAPC models, generalized additive models with multivariate smoothers and generalized linear models with age x period interaction effects.

Conclusion

The incAnalysis R package allows a straightforward comparison of cancer incidence rate projection approaches. Further detailed and targeted investigations into model performance in addition to the presented empirical results are recommended to derive guidance on appropriate statistical projection methods in a given setting.

Smoking epidemic in Europe in the 21st century

OBJECTIVE

To estimate smoking-attributable mortality in the long-term future in 29 European countries using a novel data-driven forecasting approach that integrates the wave pattern of the smoking epidemic and the cohort dimension.

METHODS

We estimated and forecasted age-specific and age-standardised smoking-attributable mortality fractions (SAMF) and 95% projection intervals for 29 European countries by sex, 1950-2100, using age-period-cohort modelling with a generalised logit link function. We projected the (decelerating) period increases (women) by a quadratic curve to obtain future declines, and extrapolated the past period decline (men). In addition, we extrapolated the recent cohort trend.

RESULTS

SAMF among men are projected to decline from, on average, 25% in 2014 (11% (Sweden)-41% (Hungary)) to 11% in 2040 (range: 6.3%-15.4%), 7% in 2065 (range: 5.9%-9.4%) and 6% in 2100. SAMF among women in 21 non-Eastern European countries, currently at an average of 16%, are projected to reach peak levels in 2013 (Northern Europe), 2019 (Western Europe), 2027 (Greece, Italy) and 2022 (Central Europe), with maximum levels of, on average, 17% (8% (Greece)-28% (Denmark)), and to decline to 10% in 2040 (range: 4%-20%), 5% in 2065 (range: 3.5%-7.6%) and 4% in 2100. For women, a short-term shift in the peak of the inverse U-shaped age pattern to higher ages is projected, and crossovers between the age-specific trends.

CONCLUSION

Our novel forecasting method enabled realistic estimates of the mortality imprint of the smoking epidemic in Europe up to 2100. The high peak values in smoking-attributable mortality projected for women warrant attention.

Exploring the usefulness of Lexis diagrams for quality improvement

Background

Visualization is important to aid practitioners in understanding local care processes and drive quality improvement (QI). Important aspects include timely feedback and ability to plot data over time. Moreover, the complexity of care also needs to be understood, as it affects the variation of care processes. However, there is a lack of QI methods visualizing multiple, related factors such as diagnosis date, death date, and cause of death to unravel their complexity, which is necessary to understand processes related to survival data. Lexis diagrams visualize individual patient processes as lines and mark additional factors such as key events. This study explores the potential of Lexis diagrams to support QI through survival data analysis, focusing on feedback, timeliness, and complexity, in a gynecological cancer setting in Sweden.

Methods

Lexis diagrams were produced based on data from a gynecological cancer quality registry (4481 patients). The usefulness of Lexis diagrams was explored through iterative data identification and analysis through semi-structured dialogues between the researcher and domain experts (clinically active care process owners) during five meetings. Visualizations were produced and adapted by the researcher between meetings, based on the dialogues, to ensure clinical relevance, resulting in three relevant types of visualizations.

Results

Domain experts identified different uses depending on diagnosis group and data visualization. Key results include timely feedback through close-to-real-time visualizations, supporting discussion and understanding of trends and hypothesis-building. Visualization of care process complexity facilitated evaluation of given care. Combined visualization of individual and population levels increased patient focus and may possibly also function to motivate practitioners and management.

Conclusion

Lexis diagrams can aid understanding of survival data, triggering important dialogues between care givers and supporting care quality improvement and new perspectives, and can therefore complement survival curves in quality improvement.

Projecting Cancer Incidence for 2025 in the 2 Largest Populated Cities in Vietnam

The population size and projected demographics of Vietnam's 2 largest cities, Ho Chi Minh City (HCMC) and Hanoi, will change dramatically over the next decade. Demographic changes in an aging population coupled with income growth and changes in lifestyle will result in a very different distribution of common cancers in the future. The study aimed to project the number of cancer incidence in the 2 largest populated cities in Vietnam for the year 2025. Cancer incidence data from 2004 to 2013 collected from population-based cancer registries in these 2 cities were provided by Vietnam National Cancer Institute. Incidence cases in 2013 and the previous decades average annual percent changes of age-standardized cancer incidence rates combined with expected population growth were modeled to project cancer incidence for each cancer site by gender to 2025. A substantial double in cancer incidence from 2013 to 2025 resulted from a growing and aging population in HCMC and Hanoi. Lung, colorectum, breast, thyroid, and liver cancers, which represent 67% of the overall cancer burden, are projected to become the leading cancer diagnoses by 2025 regardless of genders. For men, the leading cancer sites in 2025 are predicted to be lung, colorectum, esophagus, liver, and pharynx cancer, and among women, they are expected to be breast, thyroid, colorectum, lung, and cervical cancer. We projected an epidemiological transition from infectious-associated cancers to a high burden of cancers that have mainly been attributed to lifestyle in both cities. We predicted that with 16.9% growth in the overall population and dramatic aging with these 2 urban centers, the burdens of cancer incidence will increase sharply in both cities over the next decades. Data on projections of cancer incidence in both cities provide useful insights for directing appropriate policies and cancer control programs in Vietnam.

Trends in disability-adjusted life years of lung cancer among women from 2004 to 2030 in Guangzhou, China: A population-based study

BACKGROUND

Forecast of disease burden in lung cancer is an important health agenda. One of the main challenges is to predict the evolution of trends in disability-adjusted life year (DALY) of lung cancer so as to anticipate the future burden and to coordinate the supply of sufficient health services and care.

METHODS

Using 2004-2013 cancer registry data in Guangzhou, we fitted Bayesian age-period-cohort models with age, period, and cohort effects to analyze trends of lung cancer among women, and then made forecast for DALY of lung cancer until 2030.

RESULTS

During 2004-2013, there was an annual average of 10,582 DALYs for lung cancer (15.84% of total DALY). In 2014-2030, DALY is expected to reach 234,752 person-years for lung cancer (12.25% of total DALY), with an annual mean of 13,809 DALYs. Lung cancer crude DALY rate is projected to rise steadily from 257.56 (95% uncertainty interval: 165.97-361.22) in 2014 to 316.99 (219.96-419.41) per 100,000 women in 2030, and the rise is mainly seen in 45-64 years age group. Lung cancer DALY rate remains the highest in the 65-89 years age group.

CONCLUSIONS

Women at 65-89 years carry the highest lung cancer burden among other age groups in Guangzhou. The DALY rate of lung cancer is projected to increase most precipitously for the 45-64 years age group. This indicates that concerted efforts are needed to develop adequate cancer services, and to reassess health resources for control and care of lung cancer in these populations.

Statistical projection methods for lung cancer incidence and mortality: a systematic review

OBJECTIVES

To identify and summarise all studies using statistical methods to project lung cancer incidence or mortality rates more than 5 years into the future.

STUDY TYPE

Systematic review.

METHODS

We performed a systematic literature search in multiple electronic databases to identify studies published from 1 January 1988 to 14 August 2018, which used statistical methods to project lung cancer incidence and/or mortality rates. Reference lists of relevant articles were checked for additional potentially relevant articles. We developed an organisational framework to classify methods into groups according to the type of data and the statistical models used. Included studies were critically appraised using prespecified criteria.

RESULTS

One hundred and one studies met the inclusion criteria; six studies used more than one statistical method. The number of studies reporting statistical projections for lung cancer increased substantially over time. Eighty-eight studies used projection methods, which did not incorporate data on smoking in the population, and 16 studies used a method which did incorporate data on smoking. Age-period-cohort models (44 studies) were the most commonly used methods, followed by other generalised linear models (35 studies). The majority of models were developed using observed rates for more than 10 years and used data that were considered to be good quality. A quarter of studies provided comparisons of fitted and observed rates. While validation by withholding the most recent observed data from the model and then comparing the projected and observed rates for the most recent period provides important information on the model's performance, only 12 studies reported doing this.

CONCLUSION

This systematic review provides an up-to-date summary of the statistical methods used in published lung cancer incidence or mortality projections. The assessment of the strengths of existing methods will help researchers to better apply and develop statistical methods for projecting lung cancer rates. Some of the common methods described in this review can be applied to the projection of rates for other cancer types or other non-infectious diseases.

Predicting the Epidemiological Dynamics of Lung Cancer in Japan

While the prevalence of smoking has steadily declined over time, the absolute numbers of lung cancer cases and deaths have continued to increase in Japan. We employed a simple mathematical model that describes the relationship between demographic dynamics and smoking prevalence to predict future epidemiological trends of lung cancer by age and sex. Never-smokers, smokers, and ex-smokers were assumed to experience different hazard of lung cancer, and the model was parameterized using data from 2014 and before, as learning data, and a future forecast was obtained from 2015 onwards. The maximum numbers of lung cancer cases and deaths in men will be 76,978 (95% confidence interval (CI): 76,630⁻77,253) and 63,284 (95% CI: 62,991⁻63507) in 2024, while those in women will be 42,838 (95% CI: 42,601⁻43,095) and 32,267 (95% CI: 32,063⁻32,460) in 2035 and 2036, respectively. Afterwards, the absolute numbers of cases and deaths are predicted to decrease monotonically. Our compartmental modeling approach is well suited to predicting lung cancer in Japan with dynamic ageing and drastic decline in smoking prevalence. The predicted burden is useful for anticipating demands for diagnosis, treatment, and care in the healthcare sector.

The burden associated with ambient PM2.5 and meteorological factors in Guangzhou, China, 2012-2016: A generalized additive modeling of temporal years of life lost

BACKGROUND

Daily exposure to ambient particulate matter with aerodynamic diameter <2.5 μm (PM_2.5) increases deaths and is an important contributor to burden of disease in population. To better understand the disease burden associated with PM_2.5, we examined the effects of PM_2.5 on daily years of life lost (YLL) in Guangzhou, China.

METHODS

Using Guangzhou death registry, air pollution and meteorological database, we applied generalized additive models (GAM) to the relationships between YLL and PM_2.5. We then adjusted the models for age, gender, seasonality and meteorological variables. We also conducted within-data prediction of YLL while setting 2012-2014 as baseline.

RESULTS

Over 2 million YLLs (800,137 YLLs for females and 1,212,040 YLLs for males) were observed during 2012-2016. YLL was higher for the elderly people. Mean daily average PM_2.5 concentration was 47.3 μg/m³. In model comparisons, the GAM with six meteorological variables (sunshine hours, relative humidity, precipitation, atmospheric pressure, wind speed, evaporation) outperformed the others. The R² and total deviance were 0.542 and 53.0%, respectively. Non-linear trends were observed for PM_2.5 and meteorological variables. Fitted daily YLL increased to the highest level, when PM_2.5 concentration reached 134.3 μg/m³ and atmospheric pressure reached 99.4 kPa. Within-data prediction supported the fitted GAM, where low mean absolute percentage errors were observed.

CONCLUSIONS

Daily PM_2.5 exposure has a nonlinear effect on YLL and increased levels of PM_2.5 may lead to increased YLL. This study highlights the urge to reduce ambient PM_2.5 pollution in Guangzhou, in order to promote environmental health.

Lung cancer mortality in Australia: Projected outcomes to 2040

OBJECTIVES

The aim was to develop and validate a statistical model which uses past trends for lung cancer mortality and historical and current data on tobacco consumption to project lung cancer mortality rates into the future for Australia.

METHODS

We used generalized linear models (GLMs) with Poisson distribution including either age, birth cohort or period, and/or various measures of population tobacco exposure (considering cross-sectional smoking prevalence, cigarettes smoked and tar exposure per capita). Sex-specific models were fitted to data for 1956-2015 and age-standardized lung cancer mortality rates were projected forward to 2040. Possible lags of 20-30 years between tobacco exposure and lung cancer mortality were examined. The best model was selected using analysis of deviance. To validate the selected model, we temporarily re-fitted it to data for 1956-1990 and compared the projected rates to 2015 with the observed rates for 1991-2015.

RESULTS

The best fitting model used information on age, birth cohort and tar exposure per capita; close concordance with the observed data was achieved in the validation. The forward projections for lung cancer mortality using this model indicate that male and female age-standardized rates will decline over the period 2011-2015 to 2036-2040 from 27.2 to 15.1 per 100,000, and 15.8 to 11.8 per 100,000, respectively. However, due to population growth and ageing the number of deaths will increase by 7.9% for males and 57.9% for females; from 41,040 (24,831 males, 16,209 females) in 2011-2015 to 52,403 (26,805 males, 25,598 females) in 2036-2040.

CONCLUSION

In the context of the mature tobacco epidemic with past peaks in tobacco consumption for both males and females, lung cancer mortality rates are expected to continually decline over the next 25 years. However, the number of lung cancer deaths will continue to be substantial, and to increase, in Australia's ageing population.

Environmental risk factors for Toxoplasma gondii infections and the impact of latent infections on allostatic load in residents of Central North Carolina

BACKGROUND

Toxoplasma gondii infection can be acquired through ingestion of infectious tissue cysts in undercooked meat or environmental oocysts excreted by cats. This cross-sectional study assessed environmental risk factors for T. gondii infections and an association between latent infections and a measure of physiologic dysregulation known as allostatic load.

METHODS

Serum samples from 206 adults in the Durham-Chapel Hill, North Carolina area were tested for immunoglobulin (IgG) responses to T. gondii using commercial ELISA kits. Allostatic load was estimated as a sum of 15 serum biomarkers of metabolic, neuroendocrine and immune functions dichotomized at distribution-based cutoffs. Vegetated land cover within 500 m of residences was estimated using 1 m resolution data from US EPA's EnviroAtlas.

RESULTS

Handling soil with bare hands at least weekly and currently owning a cat were associated with 5.3 (95% confidence limits 1.4; 20.7) and 10.0 (2.0; 50.6) adjusted odds ratios (aOR) of T. gondii seropositivity, respectively. There was also a significant positive interaction effect of handling soil and owning cats on seropositivity. An interquartile range increase in weighted mean vegetated land cover within 500 m of residence was associated with 3.7 (1.5; 9.1) aOR of T. gondii seropositivity. Greater age and consumption of undercooked pork were other significant predictors of seropositivity. In turn, T. gondii seropositivity was associated with 61% (13%; 130%) greater adjusted mean allostatic load compared to seronegative individuals. In contrast, greater vegetated land cover around residence was associated with significantly reduced allostatic load in both seronegative (p < 0.0001) and seropositive (p = 0.004) individuals.

CONCLUSIONS

Residents of greener areas may be at a higher risk of acquiring T. gondii infections through inadvertent ingestion of soil contaminated with cat feces. T. gondii infections may partially offset health benefits of exposure to the natural living environment.

Vegetated land cover near residence is associated with reduced allostatic load and improved biomarkers of neuroendocrine, metabolic and immune functions

BACKGROUND

Greater exposure to urban green spaces has been linked to reduced risks of depression, cardiovascular disease, diabetes and premature death. Alleviation of chronic stress is a hypothesized pathway to improved health. Previous studies linked chronic stress with a biomarker-based composite measure of physiological dysregulation known as allostatic load.

OBJECTIVE

This study's objective was to assess the relationship between vegetated land cover near residences and allostatic load.

METHODS

This cross-sectional population-based study involved 206 adult residents of the Durham-Chapel Hill, North Carolina metropolitan area. Exposure was quantified using high-resolution metrics of trees and herbaceous vegetation within 500m of each residence derived from the U.S. Environmental Protection Agency's EnviroAtlas land cover dataset. Eighteen biomarkers of immune, neuroendocrine, and metabolic functions were measured in serum or saliva samples. Allostatic load was defined as a sum of potentially unhealthy biomarker values dichotomized at 10th or 90th percentile of sample distribution. Regression analysis was conducted using generalized additive models with two-dimensional spline smoothing function of geographic coordinates, weighted measures of vegetated land cover allowing decay of effects with distance, and geographic and demographic covariates.

RESULTS

An inter-quartile range increase in distance-weighted vegetated land cover was associated with 37% (95% Confidence Limits 46%; 27%) reduced allostatic load; significantly reduced adjusted odds of having low level of norepinephrine, dopamine, and dehydroepiandrosterone, and high level of epinephrine, fibrinogen, vascular cell adhesion molecule-1, and interleukin-8 in serum, and α-amylase in saliva; and reduced odds of previously diagnosed depression.

CONCLUSIONS

The observed effects of vegetated land cover on allostatic load and individual biomarkers are consistent with prevention of depression, cardiovascular disease and premature mortality.

Cancer incidence and mortality in France over the 1980-2012 period: solid tumors. Rev Epidemiol Sante Publique. 2014;62:95-108. [PMID: 24613140 DOI: 10.1016/j.respe.2013.11.073]

BACKGROUND

Cancer incidence and mortality estimates for 19 cancers (among solid tumors) are presented for France between 1980 and 2012.

METHODS

Incidence data were collected from 21 local registries and correspond to invasive cancers diagnosed between 1975 and 2009. Mortality data for the same period were provided by the Institut national de la santé et de la recherche médicale. The national incidence estimates were based on the use of mortality as a correlate of incidence. The observed incidence and mortality data were modeled using an age-period-cohort model. The numbers of incident cases and deaths for 2010-2012 are the result of short-term projections.

RESULTS

In 2012, the study estimated that 355,000 new cases of cancer (excluding non-melanoma skin cancer) and 148,000 deaths from cancer occurred in France. The incidence trend was not linear over the study period. After a constant increase from 1980 onwards, the incidence of cancer in men declined between 2005 and 2012. This recent decrease is largely related to the reduction in the incidence of prostate cancer. In women, the rates stabilized, mainly due to a change in breast cancer incidence. Mortality from most cancer types declined over the study period. A combined analysis of incidence and mortality by cancer site distinguished cancers with declining incidence and mortality (e.g., stomach) and cancers with increasing incidence and mortality (e.g., lung cancer in women). Some other cancers had rising incidence but declining mortality (e.g., thyroid).

CONCLUSION

This study reveals recent changes in cancer incidence trends, particularly regarding breast and prostate cancers.

Age-period-cohort effects in the incidence of hip fractures: political and economic events are coincident with changes in risk

An age-period cohort model was fitted to analyse time effects on hip fracture incidence rates by sex (Portugal, 2000-2008). Rates increased exponentially with age (age effect). Incidence rates decreased after 2004 for women and were random for men (period effect). New but comprehensive fluctuations in risk were coincident with major political/economic changes (cohort effect).

INTRODUCTION

Healthcare improvements have allowed prevention but have also increased life expectancy, resulting in more people being at risk. Our aim was to analyse the separate effects of age, period and cohort on incidence rates by sex in Portugal, 2000-2008.

METHODS

From the National Hospital Discharge Register, we selected admissions (aged ≥ 49 years) with hip fractures (ICD9-CM, codes 820.x) caused by low/moderate trauma (falls from standing height or less), readmissions and bone cancer cases. We calculated person-years at risk using population data from Statistics Portugal. To identify period and cohort effects for all ages, we used an age-period-cohort model (1-year intervals) followed by generalised additive models with a negative binomial distribution of the observed incidence rates of hip fractures.

RESULTS

There were 77,083 hospital admissions (77.4 % women). Incidence rates increased exponentially with age for both sexes (age effect). Incidence rates fell after 2004 for women and were random for men (period effect). There was a general cohort effect similar in both sexes; risk of hip fracture altered from an increasing trend for those born before 1930 to a decreasing trend following that year. Risk alterations (not statistically significant) coincident with major political and economic change in the history of Portugal were observed around birth cohorts 1920 (stable-increasing), 1940 (decreasing-increasing) and 1950 (increasing-decreasing only among women).

CONCLUSIONS

Hip fracture risk was higher for those born during major economically/politically unstable periods. Although bone quality reflects lifetime exposure, conditions at birth may determine future risk for hip fractures.

Short-term projection of cancer incidence in Japan using an age-period interaction model with spline smoothing

OBJECTIVE

In Japan, population-based cancer incidence data are reported several years behind the latest year of cancer mortality data. To bridge this gap, we aimed to determine a short-term projection method for cancer incidence.

METHODS

Data between 1985 and 2007 were obtained from the population-based cancer registries in four prefectures (Miyagi, Yamagata, Fukui and Nagasaki). Three projection models were examined: generalized linear model with age and period (A + P linear); generalized linear model with age, period and their interactions (A*P linear); and generalized additive model with age, period and their interactions smoothed by spline (A*P spline). We performed a 5-year projection for the years 2000 and 2005, based on the data of 1985-95 and 1985-2000, respectively. Seven cancer sites (stomach, liver, colorectal, lung, female breast, cervix uteri and prostate) and all cancers combined were analyzed. The accuracy of projection was evaluated by whether each observed number fell within the 95% confidence interval of the projected number.

RESULTS

The A*P spline model accurately projected 8 of 13 cancer site-sex combinations, whereas the number of site-sex combinations of accurate projection was 2 and 6 for A + P linear and A*P linear models, respectively. For liver and colorectal cancers, the A*P spline model alone performed accurate projections; the relative differences between projected and observed numbers of cancer incidence ranged between -0.4 and +10.9% for the A*P spline, and between +7.4 and +37.6% for the other two models. All three models failed to project sudden increases in prostate cancer between 2000 and 2005.

CONCLUSIONS

The A*P spline model is a candidate method for the projection of cancer incidence in Japan. However, we need a continuous validation for prostate cancer.

Predicting county-level cancer incidence rates and counts in the USA

Many countries, including the USA, publish predicted numbers of cancer incidence and death in current and future years for the whole country. These predictions provide important information on the cancer burden for cancer control planners, policymakers and the general public. Based on evidence from several empirical studies, the joinpoint (segmented-line linear regression) model (JPM) has been adopted by the American Cancer Society to estimate the number of new cancer cases in the USA and in individual states since 2007. Recently, cancer incidence in smaller geographic regions such as counties, and local policy makers are increasingly interested with Federal Information Processing Standard code regions. The natural extension is to directly apply the JPM to county-level cancer incidence data. The direct application has several drawbacks and its performance has not been evaluated. To address the concerns, we developed a spatial random-effects JPM for county-level cancer incidence data. The proposed model was used to predict both cancer incidence rates and counts at the county level. The standard JPM and the proposed method were compared through a validation study. The proposed method outperformed the standard JPM for almost all cancer sites, especially for moderate or rare cancer sites and for counties with small population sizes. As an application, we predicted county-level prostate cancer incidence rates and counts for the year 2011 in Connecticut.

Trends in cancer mortality in China: an update

BACKGROUND

Cancer deaths of China with the world population nearly a quarter will have a severe impact on global cancer trend and burden. The study aims to provide a comprehensive overview of long-term trends in cancer mortality in China.

MATERIALS AND METHODS

We used joinpoint analysis to detect changes in trends and generalized additive models to study birth cohort effect of risk factors between 1987 and 2009.

RESULTS

Mortality of all cancers declined steadily in urban areas, but not in rural areas. Decreasing mortality from cancers of the stomach, esophagus, nasopharynx, and cervix uteri was observed, while lung and female breast cancer mortality increased. Mortality from leukemia remained relatively stable, and cancer of liver, colorectal, and bladder had different trends between the rural and urban areas. Generational risks peaked in the cohorts born around 1925-1930 and tended to decline in successive cohorts for most cancers except for leukemia, whose relative risks were rising in the very recent cohorts.

CONCLUSION

The observed trends primarily reflect dramatic changes in socioeconomic development and lifestyle in China over the past two decades, and mortality from cancers of lung and female breast still represents a major public health priority for the government.

Projections of cancer mortality risks using spatio-temporal P-spline models

Cancer mortality risk estimates are essential for planning resource allocation and designing and evaluating cancer prevention and management strategies. However, mortality figures generally become available after a few years, making necessary to develop reliable procedures to provide current and near future mortality risks. In this work, a spatio-temporal P-spline model is used to provide predictions of mortality/incidence counts. The model is appropriate to capture smooth temporal trends and to predict cancer mortality/incidence counts in different regions for future years. The prediction mean squared error of the forecast values as well as an appropriate estimator are derived. Spanish prostate cancer mortality data in the period 1975-2008 will be used to illustrate results with a focus on cancer mortality forecasting in 2009-2011.

Trends in esophageal cancer mortality in China during 1987-2009: age, period and birth cohort analyzes

BACKGROUND

Esophageal cancer is one of the most commonly diagnosed malignant tumors in China. The aim of this study was to provide the representative and comprehensive informations about the long-term mortality trends of this disease in China between 1987 and 2009, using joinpoint regression and generalized additive models (GAMs).

METHODS

Age-standardized mortality rates (ASMR), overall and truncated (35-64 years), were calculated using the direct calculation method, and joinpoint regression was performed to obtain the estimated annual percentage changes (EAPC). GAMs were fitted to study the effects of age, period and birth cohort on mortality trends.

RESULTS

ASMR exhibited an overall remarked decline for rural females (EAPC=-2.3 95%CI: -3.3, -1.2), urban males (EAPC=-1.8 95%CI: -2.6, -1.0) and urban females (EAPC=-3.7 95%CI: -4.9, -2.4), but a small drop observed was not statistically significant for rural males (EAPC=-0.9 95%CI: -2.0, 0.3). The declines in ASMR were more noticeable for urban residents in recent years. Among all the residents, age effect showed an progressively increasing trend, whereas cohort effect declined steadily after the year corresponding to the maximum risk value. Period effect seemed to remain substantially unchanged throughout the years.

CONCLUSIONS

Although variations in mortality rates were observed according to sex and area, the overall decreasing trends in esophageal cancer mortality were found in most Chinese people, aside from rural males. The findings could correspond to the changes in age- and cohort-related factors in the population. Further study is required to understand these potential factors.

Bayesian approach to predicting cancer incidence for an area without cancer registration by using cancer incidence data from nearby areas

This paper compares three different methods for performing cancer incidence prediction in an area without a cancer registry under a Bayesian framework, using linear and log-linear age-period models with either age-specific slopes or a common slope across age groups. The three methods assume that a nearby area with a cancer registration has similar incidence and mortality patterns as the area of interest without a cancer registry where the cancer incidence prediction is carried out. The three methods differ in modeling strategies: (i) modeling the incidence rate directly; (ii) modeling the ratio of the number of incident cases to that of mortality cases; and (iii) modeling the difference between the incidence rate and the mortality rate. Strategy (iii) is a new approach in this type of projection. Empirical assessment is made using real data from the cancer registry of Tarragona, Spain, to predict cancer incidence in Girona, Spain, and vice versa. Predictions of short-term (3-4 years) incidence were made for 2001 in Tarragona using observed cancer incidence and mortality data for 1994-1998 from Girona. Short-term predictions were made for 2002 in Girona using Tarragona's 1994-1998 data. Additionally, long-term (10 years) incidence rate predictions were made for 2002 in Girona using data from Tarragona for the period 1985-1992. Our results suggest that extrapolating time-trends of incidence rates minus mortality rates may have the best predictive performance overall. These methods of population-level disease-incidence prediction are highly relevant to health care planning and policy decisions.

Projecting prevalence by stage of care for prostate cancer and estimating future health service needs: protocol for a modelling study

Introduction Current strategies for the management of prostate cancer are inadequate in Australia. We will, in this study, estimate current service needs and project the future needs for prostate cancer patients in Australia. Methods and analysis First, we will project the future prevalence of prostate cancer for 2010-2018 using data for 1972-2008 from the New South Wales (NSW) Central Cancer Registry. These projections, based on modelled incidence and survival estimates, will be estimated using PIAMOD (Prevalence, Incidence, Analysis MODel) software. Then the total prevalence will be decomposed into five stages of care: initial care, continued monitoring, recurrence, last year of life and long-term survivor. Finally, data from the NSW Prostate Cancer Care and Outcomes Study, including data on patterns of treatment and associated quality of life, will be used to estimate the type and amount of services that will be needed by prostate cancer patients in each stage of care. In addition, Central Cancer Registry episode data will be used to estimate transition rates from localised or locally advanced prostate cancer to metastatic disease. Medicare and Pharmaceutical Benefits data, linked with Prostate Cancer Care and Outcomes Study data, will be used to complement the Cancer Registry episode data. The methods developed will be applied Australia-wide to obtain national estimates of the future prevalence of prostate cancer for different stages of clinical care. Ethics and dissemination This study was approved by the NSW Population and Health Services Research Ethics Committee. Results of the study will be disseminated widely to different interest groups and organisations through a report, conference presentations and peer-reviewed articles.

Predictive model assessment for count data

We discuss tools for the evaluation of probabilistic forecasts and the critique of statistical models for count data. Our proposals include a nonrandomized version of the probability integral transform, marginal calibration diagrams, and proper scoring rules, such as the predictive deviance. In case studies, we critique count regression models for patent data, and assess the predictive performance of Bayesian age-period-cohort models for larynx cancer counts in Germany. The toolbox applies in Bayesian or classical and parametric or nonparametric settings and to any type of ordered discrete outcomes.

Predicting future excess events in risk assessment

Risk characterization in a study population relies on cases of disease or death that are causally related to the exposure under study. The number of such cases, so-called "excess" cases, is not just an indicator of the impact of the risk factor in the study population, but also an important determinant of statistical power for assessing aspects of risk such as age-time trends and susceptible subgroups. In determining how large a population to study and/or how long to follow a study population to accumulate sufficient excess cases, it is necessary to predict future risk. In this study, focusing on models involving excess risk with possible effect modification, we describe a method for predicting the expected magnitude of numbers of excess cases and assess the uncertainty in those predictions. We do this by extending Bayesian APC models for rate projection to include exposure-related excess risk with possible effect modification by, e.g., age at exposure and attained age. The method is illustrated using the follow-up study of Japanese Atomic-Bomb Survivors, one of the primary bases for determining long-term health effects of radiation exposure and assessment of risk for radiation protection purposes. Using models selected by a predictive-performance measure obtained on test data reserved for cross-validation, we project excess counts due to radiation exposure and lifetime risk measures (risk of exposure-induced deaths (REID) and loss of life expectancy (LLE)) associated with cancer and noncancer disease deaths in the A-Bomb survivor cohort.

Calibration of disease simulation model using an engineering approach

OBJECTIVES

Calibrating a disease simulation model's outputs to existing clinical data is vital to generate confidence in the model's predictive ability. Calibration involves two challenges: 1) defining a total goodness-of-fit (GOF) score for multiple targets if simultaneous fitting is required, and 2) searching for the optimal parameter set that minimizes the total GOF score (i.e., yields the best fit). To address these two prominent challenges, we have applied an engineering approach to calibrate a microsimulation model, the Lung Cancer Policy Model (LCPM).

METHODS

First, 11 targets derived from clinical and epidemiologic data were combined into a total GOF score by a weighted-sum approach, accounting for the user-defined relative importance of the calibration targets. Second, two automated parameter search algorithms, simulated annealing (SA) and genetic algorithm (GA), were independently applied to a simultaneous search of 28 natural history parameters to minimize the total GOF score. Algorithm performance metrics were defined for speed and model fit.

RESULTS

Both search algorithms obtained total GOF scores below 95 within 1000 search iterations. Our results show that SA outperformed GA in locating a lower GOF. After calibrating our LCPM, the predicted natural history of lung cancer was consistent with other mathematical models of lung cancer development.

CONCLUSION

An engineering-based calibration method was able to simultaneously fit LCPM output to multiple calibration targets, with the benefits of fast computational speed and reduced the need for human input and its potential bias.

The future burden of cancer in England: incidence and numbers of new patients in 2020

We estimated the future cancer incidence rates and the future numbers of cancer cases in England up to 2020 using cancer registration data for 1974-2003, and the official population projections from ONS up to 2023. Data were analysed using an age-period-cohort model as developed for the Nordic countries. We predict that for all cancers combined there will be relatively little change in age-standardised incidence rates in 2020. The number of new cancer cases per year in England is, however, predicted to increase by 33%, from 224,000 in 2001 to 299,000 cases in 2020. This increase is mainly due to the anticipated effects of population growth and ageing; cancer patients in 2020 will be older than today's cancer population.

Predicting the future burden of cancer

As observations in the past do not necessarily hold into the future, predicting future cancer occurrence is fraught with uncertainty. Nevertheless, predictions can aid health planners in allocating resources and allow scientists to explore the consequence of interventions aimed at reducing the impact of cancer. Simple statistical models have been refined over the past few decades and often provide reasonable predictions when applied to recent trends. Intrinsic to their interpretation, however, is an understanding of the forces that drive time trends. We explain how and why cancer predictions are made, with examples to illustrate the concepts in practice.