Absolute Risk Estimates Research Articles

Combining common genetic variants and non-genetic risk factors to predict risk of cutaneous melanoma.

Melanoma heritability is among the highest for cancer and single nucleotide polymorphisms (SNPs) contribute to it. To date, only SNPs that reached statistical significance in genome-wide association studies or few candidate SNPs have been included in melanoma risk prediction models. We compared four approaches for building polygenic risk scores (PRS) using 12874 melanoma cases and 23203 controls from Melanoma Meta-Analysis Consortium as a training set, and newly genotyped 3102 cases and 2301 controls from the MelaNostrum consortium for validation. We estimated adjusted odds ratios (ORs) for melanoma risk using traditional melanoma risk factors and the PRS with the largest area under the receiver operator characteristics curve (AUC). We estimated absolute risks combining the PRS and other risk factors, with age- and sex-specific melanoma incidence and competing mortality rates from Italy as an example. The best PRS, including 204 SNPs (AUC=64.4%; 95% confidence interval (CI)=63-65.8%), developed using winner's curse estimate corrections, had a per-quintile OR=1.35 (95% CI=1.30-1.41), corresponding to a 3.33-fold increase comparing the 5th to the 1st PRS quintile. The AUC improvement by adding the PRS was up to 7%, depending on adjusted factors and country. The 20-year absolute risk estimates based on the PRS, nevus count and pigmentation characteristics for a 60-year-old Italian man ranged from 0.5 to 11.8% (relative risk =26.34), indicating good separation.

Standard and competing risk analysis of the effect of albuminuria on cardiovascular and cancer mortality in patients with type 2 diabetes mellitus

BackgroundCompeting risks occur when populations may experience outcomes that either preclude or alter the probability of experiencing the main study outcome(s). Many standard survival analysis methods do not account for competing risks. We used mortality risk in people with diabetes with and without albuminuria as a case study to investigate the impact of competing risks on measures of absolute and relative risk.MethodsA population with type 2 diabetes was identified in Clinical Practice Research Datalink as part of a historical cohort study. Patients were followed for up to 9 years. To quantify differences in absolute risk estimates of cardiovascular and cancer, mortality standard (Kaplan-Meier) estimates were compared to competing-risks-adjusted (cumulative incidence competing risk) estimates. To quantify differences in measures of association, regression coefficients for the effect of albuminuria on the relative hazard of each outcome were compared between standard cause-specific hazard (CSH) models (Cox proportional hazards regression) and two competing risk models: the unstratified Lunn-McNeil model, which estimates CSH, and the Fine-Gray model, which estimates subdistribution hazard (SDH).ResultsIn patients with normoalbuminuria, standard and competing-risks-adjusted estimates for cardiovascular mortality were 11.1% (95% confidence interval (CI) 10.8–11.5%) and 10.2% (95% CI 9.9–10.5%), respectively. For cancer mortality, these figures were 8.0% (95% CI 7.7–8.3%) and 7.2% (95% CI 6.9–7.5%). In patients with albuminuria, standard and competing-risks-adjusted estimates for cardiovascular mortality were 21.8% (95% CI 20.9–22.7%) and 18.5% (95% CI 17.8–19.3%), respectively. For cancer mortality, these figures were 10.7% (95% CI 10.0–11.5%) and 8.6% (8.1–9.2%). For the effect of albuminuria on cardiovascular mortality, regression coefficient values from multivariable standard CSH, competing risks CSH, and competing risks SDH models were 0.557 (95% CI 0.491–0.623), 0.561 (95% CI 0.494–0.628), and 0.456 (95% CI 0.389–0.523), respectively. For the effect of albuminuria on cancer mortality, these values were 0.237 (95% CI 0.148–0.326), 0.244 (95% CI 0.154–0.333), and 0.102 (95% CI 0.012–0.192), respectively.ConclusionsStudies of absolute risk should use methods that adjust for competing risks to avoid over-stating risk, such as the CICR estimator. Studies of relative risk should consider carefully which measure of association is most appropriate for the research question.

Abstract SY32-01: Knowledge management and decision support for commonly tested susceptibility mutations

Abstract Challenges: Germline testing for inherited genetic susceptibility to cancer has become cheaper and widely used. Multigene panels now routinely test 25 to 123 genes. Health care providers, including clinicians and genetic counselors, must interpret results, assess the risk of various cancers, and advise on clinical management strategies. Aside from a small number of well-known genes, clinicians lack simple and reliable tools to personalize prevention decisions for individuals who test positive. Evidence on the types of cancer associated with susceptibility genes, and on the magnitude of increased risks associated with a mutation, is emerging at a fast rate. However, a large number of combinations of cancer type and gene variant needs to be considered. No comprehensive databases exist, information is dispersed over a vast number of published studies, the quality of the studies is uneven, and the data presented are seldom directly applicable to precision prevention decisions, which require absolute risk estimates. Thus, when assessing cancer risks for mutation carriers, clinical practice can take little advantage of the impressive scientific advancements in this field. Approaches: This presentation describes ongoing work to develop tools that could facilitate the translation into clinical practice of research results that would otherwise be difficult to access, systematize, interpret, and convert into actionable knowledge. The long-term goal should be to develop a clinical decision-support system that allows a clinician to enter important factors about a patient with a mutation and immediately receive patient-specific absolute risk estimates. The foundation for the clinical decision-support system needs to be laid by systematically collecting and curating studies on gene cancer associations and, ideally, developing a public access knowledge base for researchers and health care providers. These steps need computational tools for efficient implementations of this plan and for sustainable long-term updating of the knowledge base and decision-support system. The ask2me.org project and the data science behind it are attempting to achieve some of these steps. This project will be used as an illustration throughout. Impact: Progress in this area will have a direct clinical impact by providing patient-specific absolute risk estimates. These risk estimates will, in turn, lead to more carefully personalized prevention and management for individuals undergoing susceptibility testing, and will help us optimize screening and behavioral interventions to reduce risk. Citation Format: Giovanni Parmigiani, Danielle Braun, Kevin S. Hughes. Knowledge management and decision support for commonly tested susceptibility mutations [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr SY32-01.

Abstract 2209: Lung cancer risk prediction using DNA methylation markers

Abstract There is an urgent need to improve lung cancer risk assessment as current screening criteria miss a large proportion of cases and result in a high rate of false positives on low-dose CT screening. A fixed effect meta-analyses of 4 epigenome-wide association studies of lung cancer revealed differential DNA methylation at 16 CpG sites (FDR<0.05). The current study aimed to evaluate the extent to which such methylation markers can improve upon smoking-based risk-discrimination among ever smokers. We used data on 662 ever smoking lung cancer case-control pairs from 4 individual prospective cohorts that measured DNA methylation using the Illumina Infinium HumanMethylation450 BeadChip in peripheral blood samples before diagnosis: The Italian part of the European Prospective Investigation into Cancer (EPIC) cohort, the Melbourne Collaborative Cohort (MCCS), the Norwegian Women and Cancer cohort (NOWAC) and the Northern Sweden Health and Disease Study (NSHDS). We adopted a training-testing design where the training was performed on MCCS and NSHDS (N=511 case-control pairs), and the testing on EPIC-Italy and NOWAC (N=151 case-control pairs). Logistic regressions with lasso penalties were performed in the training set to select the best set of CpGs jointly predicting lung cancer. A methylation score was trained by fitting a logistic regression model including the selected CpGs in the training set. A baseline score based on self-reported smoking information (duration, cigarettes/day for current smokers and time since smoking cessation for former smokers) was also developed. The discriminative performances of both scores, as well as integrated model where both smoking and methylation info was incorporated, were assessed by the AUC under the ROC curves in the validation set. The methylation score based on the 9 selected CpG sites yielded an AUC of 0.78 [0.73-0.84] compared to 0.73 [0.68-0.79] for the baseline-smoking score. The model integrating both scores yielded an AUC of 0.79 [0.73-0.84], a notable 0.06-increase in AUC from using the smoking score alone (P=0.008 for difference in AUC). In conclusion, specific methylation biomarkers have a strong potential to improve lung cancer risk assessment and current USPSTF criteria for CT-screening. During the conference, we will present absolute risk estimates based on the integrated risk prediction model. Citation Format: Florence Guida, Therese H. Nøst, Caroline Relton, Paul Brennan, Torkjel M. Sandanger, Marc Chadeau-Hyam, Mattias Johansson. Lung cancer risk prediction using DNA methylation markers [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 2209.

Implantable cardioverter-defibrillator therapy and device-related complications in young patients with inherited cardiomyopathies or channelopathies: a 17-year cohort study.

To quantify appropriate and inappropriate therapy and complications related to implantable cardioverter-defibrillator (ICD) treatment in young patients receiving an ICD for a hereditary cardiomyopathy or channelopathy. This was a retrospective study including 117 consecutive patients who had received an ICD at Aarhus University Hospital, Denmark from 1 January 1999 to 31 December 2015. Patients were followed from the date of ICD implantation until migration, death, heart transplantation, or end of follow-up on 1 February 2017. Mean age at implantation was 30.5 ± 12.8 years, and the patients were followed for a mean period of 7.1 ± 4.4 years. The cumulative incidence at 1, 5, and 10 years was 17%, 29%, and 48% for appropriate ICD therapy, 6%, 13%, and 20% for inappropriate ICD therapy, and 7%, 18%, and 33% for device-related complications, respectively. Patients with an ICD implanted for secondary prevention had a higher risk of appropriate therapy compared with patients implanted for primary prevention [adjusted hazard ratio (HR) 5.18, 95% confidence interval (CI) 2.22-12.09; P < 0.01]. There was no difference in the risk of inappropriate therapy (adjusted HR 1.58, 95% CI 0.55-4.56; P = 0.40) or device-related complications (adjusted HR 1.22, 95% CI 0.56-2.68; P = 0.62) between patients with primary and secondary preventive indication. We observed high absolute risk estimates for appropriate ICD therapy in young patients with an ICD indicated by a hereditary cardiomyopathy or channelopathy. Also risks for inappropriate ICD therapy and device-related complications were significant.

A Clinical Decision Support Tool to Predict Cancer Risk for Commonly Tested Cancer-Related Germline Mutations.

The rapid drop in the cost of DNA sequencing led to the availability of multi-gene panels, which test 25 or more cancer susceptibility genes for a low cost. Clinicians and genetic counselors need a tool to interpret results, understand risk of various cancers, and advise on a management strategy. This is challenging as there are multiple studies regarding each gene, and it is not possible for clinicians and genetic counselors to be aware of all publications, nor to appreciate the relative accuracy and importance of each. Through an extensive literature review, we have identified reliable studies and derived estimates of absolute risk. We have also developed a systematic mechanism and informatics tools for (1) data curation, (2) the evaluation of quality of studies, and (3) the statistical analysis necessary to obtain risk. We produced the risk prediction clinical decision support tool ASK2ME (All Syndromes Known to Man Evaluator). It provides absolute cancer risk predictions for various hereditary cancer susceptibility genes. These predictions are specific to patients' gene carrier status, age, and history of relevant prophylactic surgery. By allowing clinicians to enter patient information and receive patient-specific cancer risks, this tool aims to have a significant impact on the quality of precision cancer prevention and disease management activities relying on panel testing. It is important to note that this tool is dynamic and constantly being updated, and currently, some of its limitations include (1) for many gene-cancer associations risk estimates are based on one study rather than meta-analysis, (2) strong assumptions on prior cancers, (3) lack of uncertainty measures, and (4) risk estimates for a growing set of gene-cancer associations which are not always variant specific. All of these concerns are being addressed on an ongoing basis, aiming to make the tool even more accurate.

External validation of risk prediction models for incident colorectal cancer using UK Biobank

Background:This study aimed to compare and externally validate risk scores developed to predict incident colorectal cancer (CRC) that include variables routinely available or easily obtainable via self-completed questionnaire.Methods:External validation of fourteen risk models from a previous systematic review in 373 112 men and women within the UK Biobank cohort with 5-year follow-up, no prior history of CRC and data for incidence of CRC through linkage to national cancer registries.Results:There were 1719 (0.46%) cases of incident CRC. The performance of the risk models varied substantially. In men, the QCancer10 model and models by Tao, Driver and Ma all had an area under the receiver operating characteristic curve (AUC) between 0.67 and 0.70. Discrimination was lower in women: the QCancer10, Wells, Tao, Guesmi and Ma models were the best performing with AUCs between 0.63 and 0.66. Assessment of calibration was possible for six models in men and women. All would require country-specific recalibration if estimates of absolute risks were to be given to individuals.Conclusions:Several risk models based on easily obtainable data have relatively good discrimination in a UK population. Modelling studies are now required to estimate the potential health benefits and cost-effectiveness of implementing stratified risk-based CRC screening.

Ovarian cancer early detection by circulating CA125 in the context of anti-CA125 autoantibody levels: Results from the EPIC cohort.

CA125 is the best ovarian cancer early detection marker to date; however, sensitivity is limited and complementary markers are required to improve discrimination between ovarian cancer cases and non-cases. Anti-CA125 autoantibodies are observed in circulation. Our objective was to evaluate whether these antibodies (1) can serve as early detection markers, providing evidence of an immune response to a developing tumor, and (2) modify the discriminatory capacity of CA125 by either masking CA125 levels (resulting in lower discrimination) or acting synergistically to improve discrimination between cases and non-cases. We investigated these objectives using a nested case-control study within the European Prospective Investigation into Cancer and Nutrition cohort (EPIC) including 250 cases diagnosed within 4 years of blood collection and up to four matched controls. Circulating CA125 antigen and antibody levels were quantified using an electrochemiluminescence assay. Adjusted areas under the curve (aAUCs) by 2-year lag-time intervals were calculated using conditional logistic regression calibrated toward the absolute risk estimates from a pre-existing epidemiological risk model as an offset-variable. Anti-CA125 levels alone did not discriminate cases from controls. For cases diagnosed <2 years after blood collection, discrimination by CA125 antigen was suggestively higher with higher anti-CA125 levels (aAUC, highest antibody tertile: 0.84 [0.76-0.92]; lowest tertile: 0.76 [0.67-0.86]; phet = 0.06). We provide the first evidence of potentially synergistic discrimination effects of CA125 and anti-CA125 antibodies in ovarian early detection. If these findings are replicated, evaluating CA125 in the context of its antibody may improve ovarian cancer early detection.

Regional evidence and international recommendations to guide lipid management in Asian patients with type 2 diabetes with special reference to renal dysfunction.

The anticipated increase in the prevalence and incidence of type 2 diabetes in Asia, and its associated cardiovascular-renal complications, will place a significant burden on patients, caregivers, and society. Despite the proven effectiveness of lipid management in reducing these complications, there are major treatment gaps, especially in Asian patients with young-onset diabetes and chronic kidney disease (CKD). Recent international guidelines recommended the adoption of absolute risk estimation of atherosclerosis and cardiovascular disease to guide treatment intensity. These recommendations replaced the previous strategy of using low-density lipoprotein cholesterol targets to guide initiation and intensification of lipid lowering, albeit still widely practiced in Asia. The latest guidelines also highlight the high risk of atherosclerosis and cardiovascular disease (ASCVD) for people with diabetes, who should be protected with statins, except for young patients without other risk factors, who will need yearly monitoring of blood lipid levels. Given the propensity of Asian patients with diabetes to develop CKD and the amplifying effect of CKD on ASCVD, the use of statins in Asian patients is particularly important. Due to interethnic differences in drug metabolism, rosuvastatin, which is largely cleared by the kidney, should be prescribed in low dosages (5-10 mg daily) in Asian populations. Conversely, epidemiological and experimental data confirm pleotropic and organ-protective effects of atorvastatin, with proven safety in Asian populations within a daily dose range of 10-40 mg. Thus, there is a need for Asian countries to review and align their lipid-lowering treatment guidelines to reduce the substantial burden of diabetes in the Asian region.

Improving patient risk communication: Translating cardiovascular risk into standardized risk percentiles

Current cholesterol guidelines recommend using 10-year risk of atherosclerotic cardiovascular disease (ASCVD) to guide informed decisions regarding statin therapy, yet patients may have difficulty conceptualizing absolute risk estimates. Peer comparisons may provide an improved tool for patient risk comprehension. Using data from the 2009-2014 National Health and Nutrition Examination Survey (NHANES), we estimated standardized risk percentiles for various age-, sex-, and race-specific subgroups based on their 10-year ASCVD risks using the Pooled Cohort Equations. We examined 9160 adults in NHANES who were free of cardiovascular disease and had complete clinical data. Among specific age, sex, and race groups, we estimated the distribution of 10-year risk, calculating the 10-year risk corresponding to each percentile in order to generate standardized cardiovascular risk percentiles. Estimated 10-year ASCVD absolute risk varied markedly by age, sex, and race subgroups. A 10-year risk of 7.0% would put a 55 year-old black male in the 20th percentile relative to his peers (ie, at lower risk than 80% of his peers), whereas a 10-year risk of 7.0% would put a 55 year-old white female in the 95th percentile (i.e., only 5% of her peers would have higher risk). Standardized cardiovascular risk percentiles by age, race, and sex are available online at populationrelativerisk.dcri.org. Cardiovascular risk varies substantially by age, sex, and race. These data allow for 10-year absolute risks of ASCVD to be translated into a standardized cardiovascular risk percentile, providing patients with information that is easy to understanding regarding how their personal risk of cardiovascular disease compares with their age-, sex-, and race-matched peers.

Application of a quantitative entry assessment model to compare the relative risk of incursion of zoonotic bat-borne viruses into European Union Member States

This paper presents a quantitative assessment model for the risk of entry of zoonotic bat-borne viruses into the European Union (EU). The model considers four routes of introduction: human travel, legal trade of products, live animal imports and illegal import of bushmeat and was applied to five virus outbreak scenarios. Two scenarios were considered for Zaire ebolavirus (wEBOV, cEBOV) and other scenarios for Hendra virus, Marburg virus (MARV) and Middle East Respiratory Syndrome Coronavirus (MERS-CoV). The use of the same framework and generic data sources for all EU Member States (MS) allows for a relative comparison of the probability of virus introduction and of the importance of the routes of introduction among MSs.According to the model wEBOV posed the highest risk of an introduction event within the EU, followed by MARV and MERS-CoV. However, the main route of introduction differed, with wEBOV and MERS-CoV most likely through human travel and MARV through legal trade of foodstuffs. The relative risks to EU MSs as entry points also varied between outbreak scenarios, highlighting the heterogeneity in global trade and travel to the EU MSs. The model has the capability to allow for a continual updating of the risk estimate using new data as, and when, it becomes available.The model provides an horizon scanning tool for use when available data are limited and, therefore, the absolute risk estimates often have high uncertainty. Sensitivity analysis suggested virus prevalence in bats has a large influence on the results; a 90% reduction in prevalence reduced the risk of introduction considerably and resulted in the relative ranking of MARV falling below that for MERS-CoV, due to this parameter disproportionately affecting the risk of introduction from the trade route over human travel.

Can stochastic consumer phase models in QMRA be simplified to a single factor?

In quantitative microbiological risk assessment (QMRA), the consumer phase covers the part of the food chain following production and retail, where the consumer transports, stores, prepares and consumes the food products considered. These consumer practices have a crucial impact on exposure, and a consumer phase model (CPM) needs to be included in a QMRA to allow an evaluation of the effectiveness of intervention measures in food production and processing in terms of human health risk. However, the development of a CPM is complex because consumer practices can be highly variable and data are scarce. So far, it is unclear to which extent CPMs need to include data on variability and detailed descriptions of the stochastic processes that may result in exposure. We therefore compared the performance of published stochastic CPMs with a simple surrogate CPM that assumes a proportional linear relation between concentration at retail and ingested dose, described by a constant factor.A comparative study was performed for different pathogens and different food products: Campylobacter in broiler meat, Salmonella in minced pork and pork cuts and Listeria in smoked salmon. Published stochastic CPMs were re-implemented and their equivalent surrogate models were derived, basing the value of the constant surrogate model factor on the absolute risk estimate from the stochastic model. The performances of the models were evaluated by comparing the effects of hypothetical intervention measures that reduce the mean or the standard deviation of the distribution of concentrations at retail. These effects were expressed in terms of relative risk estimates, as estimated in the risk assessments using the simplified and the stochastic CPMs.Results showed that after interventions that result in a reduction of the mean or standard deviation of the distribution of concentrations at retail, the relative risk estimates obtained for the simple surrogate models are always lower than those of the stochastic CPMs, which means that simplified models tend to overestimate the effects of interventions. The difference was largest in the Listeria model, where growth during storage is expected to be the dominant process. It was found that for interventions affecting the prevalence only, a simplified surrogate CPM performs similarly to a stochastic CPM.We concluded that the use of a simple surrogate CPM, which does not include the variability inherent to consumer practices, may lead to an overestimation of the effect of intervention measures in a QMRA, especially in these interventions affect the concentrations. For adequate risk assessment, it may therefore be necessary to include the variation in consumer practices (e.g. variation in storage time and temperature, cooking time and temperature and cross-contamination), as described in more realistic and more complex CPMs, definitely if this variation is expected to be large.

Defining Optimal Health Range for Thyroid Function Based on the Risk of Cardiovascular Disease.

Reference ranges of thyroid-stimulating hormone (TSH) and free thyroxine (FT4) are defined by their distribution in apparently healthy populations (2.5th and 97.5th percentiles), irrespective of disease risk, and are used as cutoffs for defining and clinically managing thyroid dysfunction. To provide proof of concept in defining optimal health ranges of thyroid function based on cardiovascular disease (CVD) mortality risk. In all, 9233 participants from the Rotterdam Study (mean age, 65.0 years) were followed up (median, 8.8 years) from baseline to date of death or end of follow-up period (2012), whichever came first (689 cases of CVD mortality). We calculated 10-year absolute risks of CVD mortality (defined according to the SCORE project) using a Fine and Gray competing risk model per percentiles of TSH and FT4, modeled nonlinearly and with sex and age adjustments. Overall, FT4 level >90th percentile was associated with a predicted 10-year CVD mortality risk >7.5% (P = 0.005). In men, FT4 level >97th percentile was associated with a risk of 10.8% (P < 0.001). In participants aged ≥65 years, absolute risk estimates were <10.0% below the 30th percentile (∼14.5 pmol/L or 1.10 ng/dL) and ≥15.0% above the 97th percentile of FT4 (∼22 pmol/L or 1.70 ng/dL). We describe absolute 10-year CVD mortality risks according to thyroid function (TSH and FT4) and suggest that optimal health ranges for thyroid function can be defined according to disease risk and are possibly sex and age dependent. These results need to be replicated with sufficient samples and representative populations.

Abstract PR02: Validation of breast cancer risk prediction model using Nurses Health and Nurse Health II Studies

Abstract Background: Adding genetic and other biomarkers to breast cancer risk prediction models could markedly improve model discrimination; however, these expanded models have not been validated in a range of populations. In particular, the calibration of these new models how well the predicted absolute risks match observed risks has not been established. Good calibration is essential to confirm the utility of these risk models in precision prevention and treatment programs. Large cohort studies provide an ideal setting to validate risk models, as they can be used to validate both relative and absolute risks. However, in practice, genetic and biomarker data are often not available in the full cohort, but only on a sub sample of cases and controls. When the rules for sampling cases and controls into the sub sample are known, inverse-probability-of-sampling (IPW) weights can be used to estimate empirical absolute risks. When the sampling rules are unknown or complicated, the IPW weights can be estimated by regressing selection into the sub sample on matching and other inclusion criteria. Methods: We evaluated the performance of recently published breast cancer risk prediction models [Maas et al. JAMA Oncol 2016] in the Nurses Health Study (NHS) and Nurses Health Study II (NHSII). We first assess a prediction model that only includes questionnaire data (BMI, hormone replacement therapy (HRT), alcohol consumption, smoking status, height, parity, age at menarche and menopause, age at first birth, and family history of breast cancer). These data are available on all subjects in the NHS and NHSII blood subcohorts: 32,826 women in NHS (with disease follow-up from 1990-2012) and 29,611 women in NHS II (1999-2013). We will then validate a model that includes both questionnaire data and a polygenic risk score based on 92 established risk SNPs. Genetic data are available on case-control samples nested within the blood subcohorts: 2308 breast cancer cases and 3344 controls from NHS and 612 breast cancer cases and 933 controls from NHSII. We estimated IPW weights among controls using logistic regression in the blood subcohorts, with sampling as control being the outcome and the following predictors: age at baseline, menopausal status, HRT, length of HRT use for premenopausal women at baseline, and length of follow up time. We used the iCARE software package (Maas P, Chatterjee N, Wheeler W et al. 2015) to calculate predicted 5 and 10-year absolute risks of breast cancer based on the published models, empirical 5 and 10-year incidence across deciles of predicted risk, and Hosmer-Lemeshow goodness of fit and AUC statistics. Results: For the risk model without genetic information, predicted risks in the blood subcohorts ranged from 6.5/1,000 (1st decile) to 20.1/1,000 (10th decile) for NHS. Although empirical risks increased across deciles at approximately the same rate as predicted rates, empirical risks were higher than predicted (Hosmer-Lemeshow p&amp;lt;10-10). Validation analyses in the NHS II blood subcohort are ongoing. Due to matching and selection on control status, the baseline distribution of questionnaire risk factors differed between the blood subcohorts and the controls from the nested case-control samples. The IPW-weighted distribution in controls closely matched the distribution in the full subcohorts, suggesting a well-calculated weight. We will present IPW-based validation of the risk model in the nested case-control samples (work in progress). Conclusions: These results confirm that breast cancer risk prediction models can discriminate between high-risk and low-risk women, but they also highlight that the accuracy of absolute risk estimates can vary across populations. Findings from this study can add insights into model improvement and model application. Moreover, the method of using IPW weights to approximate a full cohort analysis provides a potential solution for utilizing nested case-control studies in future validation analyses. This abstract is also being presented as Poster A05. Citation Format: Chi Gao, Parichoy Pal Choudhury, Paige Maas, Rulla Tamimi, Heather Eliassen, Nilanjan Chatterjee, Montserrat Garcia-Closas, Peter Kraft. Validation of breast cancer risk prediction model using Nurses Health and Nurse Health II Studies. [abstract]. In: Proceedings of the AACR Special Conference: Improving Cancer Risk Prediction for Prevention and Early Detection; Nov 16-19, 2016; Orlando, FL. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2017;26(5 Suppl):Abstract nr PR02.

Measuring the Association Between Body Mass Index and All-Cause Mortality in the Presence of Missing Data: Analyses From the Scottish National Diabetes Register.

Incorrectly handling missing data can lead to imprecise and biased estimates. We describe the effect of applying different approaches to handling missing data in an analysis of the association between body mass index and all-cause mortality among people with type 2 diabetes. We used data from the Scottish diabetes register that were linked to hospital admissions data and death registrations. The analysis was based on people diagnosed with type 2 diabetes between 2004 and 2011, with follow-up until May 31, 2014. The association between body mass index and mortality was investigated using Cox proportional hazards models. Findings were compared using 4 different missing-data methods: complete-case analysis, 2 multiple-imputation models, and nearest-neighbor imputation. There were 124,451 cases of type 2 diabetes, among which there were 17,085 deaths during 787,275 person-years of follow-up. Patients with missing data (24.8%) had higher mortality than those without missing data (adjusted hazard ratio=1.36, 95% confidence interval: 1.31, 1.41). A U-shaped relationship between body mass index and mortality was observed, with the lowest hazard ratios occurring among moderately obese people, regardless of the chosen approach for handling missing data. Missing data may affect absolute and relative risk estimates differently and should be considered in analyses of routinely collected data.

Abstract P2-05-06: Analytical and clinical validation of a fully automated tissue-based quantitative assay (MetaSite Breast™) to detect the likelihood of distant metastasis in hormone receptor (HR)-positive, HER2-negative early stage breast cancer (ESBC)

Abstract Background: MetaSite Breast™ is a validated assay to predict risk of distant breast cancer metastasis in patients with HR+/HER2- ESBC. The assay measures the number of MetaSites defined as tumor microanatomic structures composed of MENA protein expressing tumor cells in contact with CD31+ endothelial cells and CD68+ macrophages. Previous studies have demonstrated that an increased number of these microanatomic structures is associated with distant metastasis (DM) in HR+/HER2- ESBC independent of clinicopathologic features. Analytical validation of MetaSite Breast™ demonstrated precision of 97-99% (repeat image analysis of the same slide) and performance of 91-96% (staining and image analysis of serial tumor sections). We sought to further understand the importance of the MetaSite in predicting distant breast cancer metastasis utilizing a fully automated prognostic assay in an independent large patient cohort. Methods: We conducted a nested case-control study within a cohort of 3,760 patients diagnosed between 1980 and 2000 with invasive breast cancer from the Kaiser Permanente Northwest health care system. Cases (n=259) were women who developed a subsequent distant metastasis; controls, selected using incidence density sampling, were matched closely to cases (1:1) on age at and calendar year of primary diagnosis. Of the 481 patient tumor samples evaluated in this study, 57% were HR+/HER2-, 19% were triple negative (TN), and 15% were HER2+ disease. Multivariate models were adjusted for clinical factors including: lymph node status, tumor size, tumor grade, and HRT; as well as matching variables: age and year of diagnosis. Odds ratios (ORs) and 95% confidence intervals (CIs) were estimated using logistic regression. Results: In the HR+/HER2- group, MetaSite Score (MS) ranged from 0-357 and the mean was 44.6. MS was a significant predictor of DM (P=0.039) in patients with HR+/HER2- disease. Cut-points based on tertiles of MS in all 259 controls defined intermediate (13-41) and high (&amp;gt;41) risk groups that were significantly associated with risk of DM versus the low risk group (OR=2.24; 95%CI=1.23-4.13, P=0.009) and (OR=2.94; 95%CI=1.62-5.41, P=0.0005), respectively. Univariate estimates of absolute risk of DM with cutoffs based on 90% sensitivity and specificity were 9.4% for the low risk group (MS&amp;lt;7), 14.1% for the intermediate (MS=7-91), and 23.4% for the high (MS&amp;gt;91). When adjusted for clinical factors, estimates of absolute risk of DM were 6.6%, 14.1%, and 33.0% for the low, intermediate, and high risk groups, respectively. A binary cut-point for the high risk group was determined (MS&amp;gt;14) and was significant with a 2-fold higher risk of DM versus the low risk group and adjusted for clinical covariates (P=0.036). MS was not positively associated with DM in TN or HER2+ disease. Conclusions: MetaSite Breast™ significantly predicted the risk of distant breast cancer metastasis in ESBC patients with HR+/HER2-disease, independent of classical clinicopathologic features. Citation Format: Donovan MJ, Jones JG, Entenberg DR, Condeelis JS, D'alfonso TM, Gustavson M, Molinaro A, Oktay MH, Xue X, Sparano JA, Peterson MA, Podznyakova O, Rohan TE, Shuber AP, Gertler FB, Ly A, Divelbiss ME, Hamilton DA. Analytical and clinical validation of a fully automated tissue-based quantitative assay (MetaSite Breast™) to detect the likelihood of distant metastasis in hormone receptor (HR)-positive, HER2-negative early stage breast cancer (ESBC) [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P2-05-06.

Multiple imputation was a valid approach to estimate absolute risk from a prediction model based on case–cohort data

ObjectiveTo compare weighting methods for Cox regression and multiple imputation (MI) in a case–cohort study in the context of risk prediction modeling. Study Design and SettingBased on the European Prospective Investigation into Cancer and Nutrition Potsdam study, we estimated risk scores to predict incident type-2 diabetes using full cohort data and case–cohort data assuming missing information on waist circumference outside the case–cohort (∼90%). Varying weighting approaches and MI were compared with regard to the calculation of relative risks, absolute risks, and predictive abilities including C-index, the net reclassification improvement, and calibration. ResultsThe full cohort comprised 21,845 participants, and the case–cohort comprised 2,703 participants. Relative risks were similar across all methods and compatible with full cohort estimates. Absolute risk estimates showed stronger disagreement mainly for Prentice and Self & Prentice weighting. Barlow and Langholz & Jiao weighting methods and MI were in good agreement with full cohort analysis. Predictive abilities were closest to full cohort estimates for MI or for Barlow and Langholz & Jiao weighting. ConclusionsMI seems to be a valid method for deriving or extending a risk prediction model from case–cohort data and might be superior for absolute risk calculation when compared to weighted approaches.

Endometrial cancer risk prediction including serum-based biomarkers: results from the EPIC cohort.

Endometrial cancer risk prediction models including lifestyle, anthropometric and reproductive factors have limited discrimination. Adding biomarker data to these models may improve predictive capacity; to our knowledge, this has not been investigated for endometrial cancer. Using a nested case-control study within the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort, we investigated the improvement in discrimination gained by adding serum biomarker concentrations to risk estimates derived from an existing risk prediction model based on epidemiologic factors. Serum concentrations of sex steroid hormones, metabolic markers, growth factors, adipokines and cytokines were evaluated in a step-wise backward selection process; biomarkers were retained at p<0.157 indicating improvement in the Akaike information criterion (AIC). Improvement in discrimination was assessed using the C-statistic for all biomarkers alone, and change in C-statistic from addition of biomarkers to preexisting absolute risk estimates. We used internal validation with bootstrapping (1000-fold) to adjust for over-fitting. Adiponectin, estrone, interleukin-1 receptor antagonist, tumor necrosis factor-alpha and triglycerides were selected into the model. After accounting for over-fitting, discrimination was improved by 2.0 percentage points when all evaluated biomarkers were included and 1.7 percentage points in the model including the selected biomarkers. Models including etiologic markers on independent pathways and genetic markers may further improve discrimination.

Breast Cancer Screening in the Precision Medicine Era: Risk-Based Screening in a Population-Based Trial.

Ongoing controversy over the optimal approach to breast cancer screening has led to discordant professional society recommendations, particularly in women age 40 to 49 years. One potential solution is risk-based screening, where decisions around the starting age, stopping age, frequency, and modality of screening are based on individual risk to maximize the early detection of aggressive cancers and minimize the harms of screening through optimal resource utilization. We present a novel approach to risk-based screening that integrates clinical risk factors, breast density, a polygenic risk score representing the cumulative effects of genetic variants, and sequencing for moderate- and high-penetrance germline mutations. We demonstrate how thresholds of absolute risk estimates generated by our prediction tools can be used to stratify women into different screening strategies (biennial mammography, annual mammography, annual mammography with adjunctive magnetic resonance imaging, defer screening at this time) while informing the starting age of screening for women age 40 to 49 years. Our risk thresholds and corresponding screening strategies are based on current evidence but need to be tested in clinical trials. The Women Informed to Screen Depending On Measures of risk (WISDOM) Study, a pragmatic, preference-tolerant randomized controlled trial of annual vs personalized screening, will study our proposed approach. WISDOM will evaluate the efficacy, safety, and acceptability of risk-based screening beginning in the fall of 2016. The adaptive design of this trial allows continued refinement of our risk thresholds as the trial progresses, and we discuss areas where we anticipate emerging evidence will impact our approach.

The Impact of Excluding Trials from Network Meta-Analyses - An Empirical Study.

Network meta-analysis (NMA) expands the scope of a conventional pairwise meta-analysis to simultaneously compare multiple treatments, which has an inherent appeal for clinicians, patients, and policy decision makers. Two recent reports have shown that the impact of excluding a treatment on NMAs can be substantial. However, no one has assessed the impact of excluding a trial from NMAs, which is important because many NMAs selectively include trials in the analysis. This article empirically examines the impact of trial exclusion using both the arm-based (AB) and contrast-based (CB) approaches, by reanalyzing 20 published NMAs involving 725 randomized controlled trials and 449,325 patients. For the population-averaged absolute risk estimates using the AB approach, the average fold changes across all networks ranged from 1.004 (with standard deviation 0.004) to 1.072 (with standard deviation 0.184); while the maximal fold changes ranged from 1.032 to 2.349. In 12 out of 20 NMAs, a 1.20-fold or larger change is observed in at least one of the population-averaged absolute risk estimates. In addition, while excluding a trial can substantially change the estimated relative effects (e.g., log odds ratios), there is no systematic difference in terms of changes between the two approaches. Changes in treatment rankings are observed in 7 networks and changes in inconsistency are observed in 3 networks. We do not observe correlations between changes in treatment effects, treatment rankings and inconsistency. Finally, we recommend rigorous inclusion and exclusion criteria, logical study selection process, and reasonable network geometry to ensure robustness and generalizability of the results of NMAs.