stephane ragusa - Academia.edu (original) (raw)
Papers by stephane ragusa
American Journal of Epidemiology, Jun 1, 2006
Physical activity is associated with a reduced risk of colon cancer, but the effect of activity o... more Physical activity is associated with a reduced risk of colon cancer, but the effect of activity on colorectal adenomas, which are precursors to colon cancer, is uncertain. The influence of physical activity on colorectal adenomas among African-American women is of particular interest because African-American women have an increased risk of colon cancer relative to other U.S. women. We prospectively assessed the relation of physical activity to the incidence of colorectal polyps, which we used as a proxy for colorectal adenomas, among African-American women. We followed 45,400 women in the Black Women's Health Study from 1997 through 2003. Data were obtained by biennial mailed questionnaires. During 287,029 person-years of follow-up, 1,390 women reported having been diagnosed with colorectal polyps. We converted hours per week of vigorous exercise and hours per week of walking to MET-hours. We estimated incidence rate ratios (IRRs) with Cox proportional hazard models, controlling for age, body mass index, smoking, family history of colorectal cancer, and education. For total MET-hours per week spent in walking and vigorous exercise, the IRR decreased from 0.94 for <5 MET-hours per week to 0.72 for ! 40 MET-hours per week (P trend ¼ 0.01). The inverse association was apparent among most subgroups examined, including women who may be at higher risk of colorectal adenomas because of being obese. These findings indicate that increased physical activity is associated with a reduced incidence of colorectal polyps among African-American women.
Background: Women's awareness about their personal breast cancer (BC) risk in the general pop... more Background: Women's awareness about their personal breast cancer (BC) risk in the general population is generally low. Mass screening and mass prevention interventions have as yet been moderately efficient in breast oncology. "Personalized prevention" including risk communication, personalized screening and primary prevention recommendations is a promising. A personal history of BBL slightly increases subsequent BC risk. Objectives: the main objective was to evaluate the acceptability of a mathematical tool- based breast cancer risk assessment and subsequent proposal of a personalized BC prevention program in a BBL population. Secondary objectives were to evaluate information receipt, awareness, satisfaction, and anxiety. Methods: Women were eligible for the study if aged 40-74, were recently diagnosed with a benign breast lesion at the one stop breast Unit of the center, had no personal history of cancer or atypical lesions and were not BRCA carriers. Women were proposed a personalized risk assessment using a mathematical tool (BCSC score adapted to the French population-Ragusa et al) together with personalized information on risk, BC screening and prevention, release of a personalized program and evaluation of their receipt. The main end point was the proportion of women willing to have a risk assessment and personalized counseling. A cut-off point of 70% was considered critical to consider acceptability. Secondary end points were perceived BC risk, satisfaction, anxiety and distress levels at day 2 using standardized questionnaires, as well as adherence with the proposed programs. Results: Of 150 women proposed BC risk assessment and personalized prevention information between 02/2014 and 03/2015, 129 (86%) accepted. Median age: 53.6 years. 33% had a low BC risk (< 1.1% at 5 yrs [mean risk of 50 yrs-old women in France]), 53% a moderate risk (1.1-1.66% at 5 yrs), while 14% were high risk (> 1.66% at 5 yrs). 87% had never had any previous information on BC risk. 3 pts required a genetic assessment. Participants were globally very satisfied with physicians' and nurses' interpersonal skills, availability and provision of information (mean score > 4; range 2-5). The mean scores of clarity of the BC risk information (4.14±1;range 2-5) and screening program information (4.21±0.93; range 2-5) were high. The mean score of perceived risk level was estimated to 33.5% (SD=21.9). Mean scores of state anxiety (36.7±12.2; range 20-71), trait anxiety (39.5±8.9; range 23-59), depressive symptoms (3.4±3.3; range 0-12) and psychological distress indicated low levels of all. Higher level of state-anxiety was associated with lower scores of satisfaction with doctors and nurses human qualities (r = 0.26, p<.05) and with lower scores of clarity of information about screening program (r = 0.25, p<.05). Conclusion: The receipt of breast cancer risk assessment and personalized prevention information among women diagnosed with BBL was high (86%). Information need is high given the low level of real risk awareness. Such population may benefit from personalized prevention. Anxiety and distress scores were low and satisfaction rates high. Citation Format: Tlemsani C, Boinon D, Yung MF, Ragusa S, Mazouni C, Balleyguier C, Saghatchian M, Ghouadni A, Rivera S, Michiels S, Delaloge S. Receipt of breast cancer risk assessment and personalized prevention information among women diagnosed with a benign breast lesion (BBL) in a one stop breast unit: A prospective assessment. [abstract]. In: Proceedings of the Thirty-Eighth Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2015 Dec 8-12; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2016;76(4 Suppl):Abstract nr P3-10-03.
Cancer Research, 2017
Background: Stratified breast cancer (BC) prevention is a major option for the future but require... more Background: Stratified breast cancer (BC) prevention is a major option for the future but requires clinically meaningful internationally validated risk models. Non parametric models may be alternate methods for modeling in very large cohorts. We have previously shown that a non-parametric similarity-based k-nearest neighbors' (kNN) model performs better than the BCRAT/Gail model to on 65 000 women of the E3N French national cohort (Dartois et al 2015). We used this method to develop and validate a mammographic density-based model in larger general screening populations (pops). Methods: A modified version of a data-mining based algorithm, the kNN method, was implemented and adapted as previously described [ref Dartois]. Core concept of kNN algorithm is to gather similar profiles using a distance computation. We developed a BC risk prediction model on 629 229 women (wn) from the US Breast Cancer Research Consortium (BCSC), with 5 times random selection of learning and validation s...
American Journal of Epidemiology, 2006
New psychiatric screening instruments are often validated against expert assessment or establishe... more New psychiatric screening instruments are often validated against expert assessment or established reference standards (RS) such as the Structured Clinical Interview for DSM-IV (SCID). Measurement error in the RS can bias estimates of the new instrument's sensitivity (SE) and specificity (SP). We used two methods to estimate SE and SP of a psychiatric screening instrument, accounting for measurement error in the RS. (1) We used expert judgment to specify likely values for SE/SP of the RS. We then produced corrected SE/SP for the screening instrument, using simple algebraic manipulation and assuming conditional independence. (2) With !3 independent measures of disease (e.g., screen, RS, chart review), latent class analysis (LCA) provides simultaneous estimates of SE/SP for all measures assuming conditional independence between the tests. Results: Compared to the SCID, the Substance Abuse (SA)/Mental Illness Symptoms Screener (SAMISS) demonstrated 86% SE and 75% SP for classifying persons with any SA disorder (assuming perfect SCID measurement) (N ¼ 148). Assuming the SCID has 80% SE and 96% SP in detecting true SA, corrected estimates of SAMISS performance were 99% SE and 79% SP (method 1). Using LCA (method 2) with chart review as the 3rd test, the characteristics for the 3 measures were: SAMISS-91% SE, 81% SP; SCID-73% SE, 98% SP; chart review-85% SE, 99% SP. Implications: An imperfect RS introduces error into the estimation of test characteristics, usually biasing SE and SP downward. These 2 straightforward methods relax the often questionable assumption of perfect RS performance and generate corrected estimates of SE and SP.
Abstract—According to the World Health Organization, starting from 2010, cancer will become the l... more Abstract—According to the World Health Organization, starting from 2010, cancer will become the leading cause of death worldwide. Prevention of major cancer localizations through a quantified assessment of risk factors is a major concern in order to decrease their impact in our society. Our objective is to test the performances of a modeling method easily readable by a physician. In this article, we follow a data mining process to build a reliable assessment tool for primary breast cancer risk. A k-nearest-neighbor algorithm is used to ...
Cancer Research, 2022
Background Currently, mammographic-based breast cancer screening (BCS) using age as the single cr... more Background Currently, mammographic-based breast cancer screening (BCS) using age as the single criterion for population selection, apart from rare high-risk indications, is being questioned for its imperfect sensitivity (interval cancers) and specificity (false positive recalls), as well as the risk of over-diagnoses. BC risk scores incorporating personal and family history, breast mammographic density and genetic information based on a polygenic score (PRS) give a promisingly accurate likelihood of a woman developing invasive BC in the next 5 years. MyPeBS, a European Commission H2020-funded randomized clinical trial (NCT03672331) conducted in 6 countries (Belgium, France, Israel, Italy, Spain and UK) aims to demonstrate the usefulness of a risk-based screening approach to improve BCS in the general population. Methods MyPeBS’s primary objective is to show non-inferiority of the risk-stratified BCS arm in terms of incidence rate of breast cancer of stage 2 and higher, compared to t...
Chez les procaryotes, les proteines synthetisees comprennent un groupement formyl a leur extremit... more Chez les procaryotes, les proteines synthetisees comprennent un groupement formyl a leur extremite n-terminale. Ce groupement doit ensuite etre hydrolyse par la deformylase, avant d'aboutir a des proteines fonctionnelles. La deformylase est ainsi necessaire a la croissance bacterienne. Un inhibiteur de cette activite pourrait donc agir comme antibiotique efficace, et non toxique puisque cette enzyme est absente chez les eucaryotes. Pour determiner la nature d'un tel inhibiteur, il est utile de connaitre le mecanisme catalytique de la cible enzymatique et de pouvoir tester facilement l'affinite qu'elle presente pour differents substrats ou inhibiteurs. Pour realiser ces etudes, il est necessaire de disposer in vitro d'une forme de l'enzyme representative de sa forme physiologique. La purification de deformylase par les methodes biochimiques classiques conduit a une enzyme environ 1000 fois moins active que la forme existant in vivo. Nous avons montre qu'en...
Journal of Molecular Biology, Jun 25, 1999
Breast cancer research and treatment, 2015
Breast cancer remains a global health concern with a lack of high discriminating prediction model... more Breast cancer remains a global health concern with a lack of high discriminating prediction models. The k-nearest-neighbor algorithm (kNN) estimates individual risks using an intuitive tool. This study compares the performances of this approach with the Cox and the Gail models for the 5-year breast cancer risk prediction. The study included 64,995 women from the French E3N prospective cohort. The sample was divided into a learning (N = 51,821) series to learn the models using fivefold cross-validation and a validation (N = 13,174) series to evaluate them. The area under the receiver operating characteristic curve (AUC) and the expected over observed number of cases (E/O) ratio were estimated. In the two series, 393 and 78 premenopausal and 537 and 98 postmenopausal breast cancers were diagnosed. The discrimination values of the best combinations of predictors obtained from cross-validation ranged from 0.59 to 0.60. In the validation series, the AUC values in premenopausal and postme...
Cancer has recently become the leading cause of death worldwide according to the World Health Org... more Cancer has recently become the leading cause of death worldwide according to the World Health Organization. As a consequence, health authorities acknowledge the need to implement prevention and screening programs to decrease its incidence. The efficiency of these programs can be increased by targeting higher risk subsets of the population. Efficient tools capable of monitoring the population risk are therefore needed. Constraints to building cancer risk scores and impacts on the tools platform are presented. Major constraints beyond performance of a risk score concern the role of domain experts and their acceptability by end users. Readability is therefore an important criterion. It is shown that a simple k-nearest-neighbor algorithm can achieve good performance with the help of the domain expert. To illustrate this, a risk score made of only four attributes is presented for the French population.
According to the World Health Organization, starting from 2010, cancer has become the leading cau... more According to the World Health Organization, starting from 2010, cancer has become the leading cause of death worldwide. Prevention of major cancer localizations through a quantified assessment of risk factors is a major concern in order to decrease their impact in our society. Our objective is to test the performances of a modeling method that answers to needs and constraints of end users. In this article, we follow a data mining process to build a reliable assessment tool for primary breast cancer risk. A k-nearest-neighbor algorithm is used to compute a risk score for different profiles from a public database. We empirically show that it is possible to achieve the same performances as logistic regressions with less attributes and a more easily readable model. The process includes the intervention of a domain expert, during an offline step of the process, who helps to select one of the numerous model variations by combining at best, physician expectations and performances. A risk s...
Breast cancer research and treatment, 2015
Breast cancer remains a global health concern with a lack of high discriminating prediction model... more Breast cancer remains a global health concern with a lack of high discriminating prediction models. The k-nearest-neighbor algorithm (kNN) estimates individual risks using an intuitive tool. This study compares the performances of this approach with the Cox and the Gail models for the 5-year breast cancer risk prediction. The study included 64,995 women from the French E3N prospective cohort. The sample was divided into a learning (N = 51,821) series to learn the models using fivefold cross-validation and a validation (N = 13,174) series to evaluate them. The area under the receiver operating characteristic curve (AUC) and the expected over observed number of cases (E/O) ratio were estimated. In the two series, 393 and 78 premenopausal and 537 and 98 postmenopausal breast cancers were diagnosed. The discrimination values of the best combinations of predictors obtained from cross-validation ranged from 0.59 to 0.60. In the validation series, the AUC values in premenopausal and postme...
American Journal of Epidemiology, Jun 1, 2006
Physical activity is associated with a reduced risk of colon cancer, but the effect of activity o... more Physical activity is associated with a reduced risk of colon cancer, but the effect of activity on colorectal adenomas, which are precursors to colon cancer, is uncertain. The influence of physical activity on colorectal adenomas among African-American women is of particular interest because African-American women have an increased risk of colon cancer relative to other U.S. women. We prospectively assessed the relation of physical activity to the incidence of colorectal polyps, which we used as a proxy for colorectal adenomas, among African-American women. We followed 45,400 women in the Black Women's Health Study from 1997 through 2003. Data were obtained by biennial mailed questionnaires. During 287,029 person-years of follow-up, 1,390 women reported having been diagnosed with colorectal polyps. We converted hours per week of vigorous exercise and hours per week of walking to MET-hours. We estimated incidence rate ratios (IRRs) with Cox proportional hazard models, controlling for age, body mass index, smoking, family history of colorectal cancer, and education. For total MET-hours per week spent in walking and vigorous exercise, the IRR decreased from 0.94 for <5 MET-hours per week to 0.72 for ! 40 MET-hours per week (P trend ¼ 0.01). The inverse association was apparent among most subgroups examined, including women who may be at higher risk of colorectal adenomas because of being obese. These findings indicate that increased physical activity is associated with a reduced incidence of colorectal polyps among African-American women.
Background: Women's awareness about their personal breast cancer (BC) risk in the general pop... more Background: Women's awareness about their personal breast cancer (BC) risk in the general population is generally low. Mass screening and mass prevention interventions have as yet been moderately efficient in breast oncology. "Personalized prevention" including risk communication, personalized screening and primary prevention recommendations is a promising. A personal history of BBL slightly increases subsequent BC risk. Objectives: the main objective was to evaluate the acceptability of a mathematical tool- based breast cancer risk assessment and subsequent proposal of a personalized BC prevention program in a BBL population. Secondary objectives were to evaluate information receipt, awareness, satisfaction, and anxiety. Methods: Women were eligible for the study if aged 40-74, were recently diagnosed with a benign breast lesion at the one stop breast Unit of the center, had no personal history of cancer or atypical lesions and were not BRCA carriers. Women were proposed a personalized risk assessment using a mathematical tool (BCSC score adapted to the French population-Ragusa et al) together with personalized information on risk, BC screening and prevention, release of a personalized program and evaluation of their receipt. The main end point was the proportion of women willing to have a risk assessment and personalized counseling. A cut-off point of 70% was considered critical to consider acceptability. Secondary end points were perceived BC risk, satisfaction, anxiety and distress levels at day 2 using standardized questionnaires, as well as adherence with the proposed programs. Results: Of 150 women proposed BC risk assessment and personalized prevention information between 02/2014 and 03/2015, 129 (86%) accepted. Median age: 53.6 years. 33% had a low BC risk (< 1.1% at 5 yrs [mean risk of 50 yrs-old women in France]), 53% a moderate risk (1.1-1.66% at 5 yrs), while 14% were high risk (> 1.66% at 5 yrs). 87% had never had any previous information on BC risk. 3 pts required a genetic assessment. Participants were globally very satisfied with physicians' and nurses' interpersonal skills, availability and provision of information (mean score > 4; range 2-5). The mean scores of clarity of the BC risk information (4.14±1;range 2-5) and screening program information (4.21±0.93; range 2-5) were high. The mean score of perceived risk level was estimated to 33.5% (SD=21.9). Mean scores of state anxiety (36.7±12.2; range 20-71), trait anxiety (39.5±8.9; range 23-59), depressive symptoms (3.4±3.3; range 0-12) and psychological distress indicated low levels of all. Higher level of state-anxiety was associated with lower scores of satisfaction with doctors and nurses human qualities (r = 0.26, p<.05) and with lower scores of clarity of information about screening program (r = 0.25, p<.05). Conclusion: The receipt of breast cancer risk assessment and personalized prevention information among women diagnosed with BBL was high (86%). Information need is high given the low level of real risk awareness. Such population may benefit from personalized prevention. Anxiety and distress scores were low and satisfaction rates high. Citation Format: Tlemsani C, Boinon D, Yung MF, Ragusa S, Mazouni C, Balleyguier C, Saghatchian M, Ghouadni A, Rivera S, Michiels S, Delaloge S. Receipt of breast cancer risk assessment and personalized prevention information among women diagnosed with a benign breast lesion (BBL) in a one stop breast unit: A prospective assessment. [abstract]. In: Proceedings of the Thirty-Eighth Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2015 Dec 8-12; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2016;76(4 Suppl):Abstract nr P3-10-03.
Cancer Research, 2017
Background: Stratified breast cancer (BC) prevention is a major option for the future but require... more Background: Stratified breast cancer (BC) prevention is a major option for the future but requires clinically meaningful internationally validated risk models. Non parametric models may be alternate methods for modeling in very large cohorts. We have previously shown that a non-parametric similarity-based k-nearest neighbors' (kNN) model performs better than the BCRAT/Gail model to on 65 000 women of the E3N French national cohort (Dartois et al 2015). We used this method to develop and validate a mammographic density-based model in larger general screening populations (pops). Methods: A modified version of a data-mining based algorithm, the kNN method, was implemented and adapted as previously described [ref Dartois]. Core concept of kNN algorithm is to gather similar profiles using a distance computation. We developed a BC risk prediction model on 629 229 women (wn) from the US Breast Cancer Research Consortium (BCSC), with 5 times random selection of learning and validation s...
American Journal of Epidemiology, 2006
New psychiatric screening instruments are often validated against expert assessment or establishe... more New psychiatric screening instruments are often validated against expert assessment or established reference standards (RS) such as the Structured Clinical Interview for DSM-IV (SCID). Measurement error in the RS can bias estimates of the new instrument's sensitivity (SE) and specificity (SP). We used two methods to estimate SE and SP of a psychiatric screening instrument, accounting for measurement error in the RS. (1) We used expert judgment to specify likely values for SE/SP of the RS. We then produced corrected SE/SP for the screening instrument, using simple algebraic manipulation and assuming conditional independence. (2) With !3 independent measures of disease (e.g., screen, RS, chart review), latent class analysis (LCA) provides simultaneous estimates of SE/SP for all measures assuming conditional independence between the tests. Results: Compared to the SCID, the Substance Abuse (SA)/Mental Illness Symptoms Screener (SAMISS) demonstrated 86% SE and 75% SP for classifying persons with any SA disorder (assuming perfect SCID measurement) (N ¼ 148). Assuming the SCID has 80% SE and 96% SP in detecting true SA, corrected estimates of SAMISS performance were 99% SE and 79% SP (method 1). Using LCA (method 2) with chart review as the 3rd test, the characteristics for the 3 measures were: SAMISS-91% SE, 81% SP; SCID-73% SE, 98% SP; chart review-85% SE, 99% SP. Implications: An imperfect RS introduces error into the estimation of test characteristics, usually biasing SE and SP downward. These 2 straightforward methods relax the often questionable assumption of perfect RS performance and generate corrected estimates of SE and SP.
Abstract—According to the World Health Organization, starting from 2010, cancer will become the l... more Abstract—According to the World Health Organization, starting from 2010, cancer will become the leading cause of death worldwide. Prevention of major cancer localizations through a quantified assessment of risk factors is a major concern in order to decrease their impact in our society. Our objective is to test the performances of a modeling method easily readable by a physician. In this article, we follow a data mining process to build a reliable assessment tool for primary breast cancer risk. A k-nearest-neighbor algorithm is used to ...
Cancer Research, 2022
Background Currently, mammographic-based breast cancer screening (BCS) using age as the single cr... more Background Currently, mammographic-based breast cancer screening (BCS) using age as the single criterion for population selection, apart from rare high-risk indications, is being questioned for its imperfect sensitivity (interval cancers) and specificity (false positive recalls), as well as the risk of over-diagnoses. BC risk scores incorporating personal and family history, breast mammographic density and genetic information based on a polygenic score (PRS) give a promisingly accurate likelihood of a woman developing invasive BC in the next 5 years. MyPeBS, a European Commission H2020-funded randomized clinical trial (NCT03672331) conducted in 6 countries (Belgium, France, Israel, Italy, Spain and UK) aims to demonstrate the usefulness of a risk-based screening approach to improve BCS in the general population. Methods MyPeBS’s primary objective is to show non-inferiority of the risk-stratified BCS arm in terms of incidence rate of breast cancer of stage 2 and higher, compared to t...
Chez les procaryotes, les proteines synthetisees comprennent un groupement formyl a leur extremit... more Chez les procaryotes, les proteines synthetisees comprennent un groupement formyl a leur extremite n-terminale. Ce groupement doit ensuite etre hydrolyse par la deformylase, avant d'aboutir a des proteines fonctionnelles. La deformylase est ainsi necessaire a la croissance bacterienne. Un inhibiteur de cette activite pourrait donc agir comme antibiotique efficace, et non toxique puisque cette enzyme est absente chez les eucaryotes. Pour determiner la nature d'un tel inhibiteur, il est utile de connaitre le mecanisme catalytique de la cible enzymatique et de pouvoir tester facilement l'affinite qu'elle presente pour differents substrats ou inhibiteurs. Pour realiser ces etudes, il est necessaire de disposer in vitro d'une forme de l'enzyme representative de sa forme physiologique. La purification de deformylase par les methodes biochimiques classiques conduit a une enzyme environ 1000 fois moins active que la forme existant in vivo. Nous avons montre qu'en...
Journal of Molecular Biology, Jun 25, 1999
Breast cancer research and treatment, 2015
Breast cancer remains a global health concern with a lack of high discriminating prediction model... more Breast cancer remains a global health concern with a lack of high discriminating prediction models. The k-nearest-neighbor algorithm (kNN) estimates individual risks using an intuitive tool. This study compares the performances of this approach with the Cox and the Gail models for the 5-year breast cancer risk prediction. The study included 64,995 women from the French E3N prospective cohort. The sample was divided into a learning (N = 51,821) series to learn the models using fivefold cross-validation and a validation (N = 13,174) series to evaluate them. The area under the receiver operating characteristic curve (AUC) and the expected over observed number of cases (E/O) ratio were estimated. In the two series, 393 and 78 premenopausal and 537 and 98 postmenopausal breast cancers were diagnosed. The discrimination values of the best combinations of predictors obtained from cross-validation ranged from 0.59 to 0.60. In the validation series, the AUC values in premenopausal and postme...
Cancer has recently become the leading cause of death worldwide according to the World Health Org... more Cancer has recently become the leading cause of death worldwide according to the World Health Organization. As a consequence, health authorities acknowledge the need to implement prevention and screening programs to decrease its incidence. The efficiency of these programs can be increased by targeting higher risk subsets of the population. Efficient tools capable of monitoring the population risk are therefore needed. Constraints to building cancer risk scores and impacts on the tools platform are presented. Major constraints beyond performance of a risk score concern the role of domain experts and their acceptability by end users. Readability is therefore an important criterion. It is shown that a simple k-nearest-neighbor algorithm can achieve good performance with the help of the domain expert. To illustrate this, a risk score made of only four attributes is presented for the French population.
According to the World Health Organization, starting from 2010, cancer has become the leading cau... more According to the World Health Organization, starting from 2010, cancer has become the leading cause of death worldwide. Prevention of major cancer localizations through a quantified assessment of risk factors is a major concern in order to decrease their impact in our society. Our objective is to test the performances of a modeling method that answers to needs and constraints of end users. In this article, we follow a data mining process to build a reliable assessment tool for primary breast cancer risk. A k-nearest-neighbor algorithm is used to compute a risk score for different profiles from a public database. We empirically show that it is possible to achieve the same performances as logistic regressions with less attributes and a more easily readable model. The process includes the intervention of a domain expert, during an offline step of the process, who helps to select one of the numerous model variations by combining at best, physician expectations and performances. A risk s...
Breast cancer research and treatment, 2015
Breast cancer remains a global health concern with a lack of high discriminating prediction model... more Breast cancer remains a global health concern with a lack of high discriminating prediction models. The k-nearest-neighbor algorithm (kNN) estimates individual risks using an intuitive tool. This study compares the performances of this approach with the Cox and the Gail models for the 5-year breast cancer risk prediction. The study included 64,995 women from the French E3N prospective cohort. The sample was divided into a learning (N = 51,821) series to learn the models using fivefold cross-validation and a validation (N = 13,174) series to evaluate them. The area under the receiver operating characteristic curve (AUC) and the expected over observed number of cases (E/O) ratio were estimated. In the two series, 393 and 78 premenopausal and 537 and 98 postmenopausal breast cancers were diagnosed. The discrimination values of the best combinations of predictors obtained from cross-validation ranged from 0.59 to 0.60. In the validation series, the AUC values in premenopausal and postme...