A nomogram for predicting endometrial cancer risk in asymptomatic women: Addressing disparities in patient populations with limited access and delayed diagnosis (425)

Objectives: Endometrial cancer (EC) may be the most common malignancy and the leading cause of death ahead of breast and lung cancer in certain communities (Timoteo-Liaina IJGO ‘21). Early curable stage presentation benefits many patients; however, disparities exist from the advanced stage, aggressive histology, and access (RK Lee SGO ‘20). We sought to address this with a low-resource EC Risk Nomogram (ECRN) to identify risk warranting triage of asymptomatic women in areas with disproportionate disease burden to affect favorable stage migration and eventually reduce mortality disparities. Methods: Using target community retrospective data, odds ratios were determined for the beta coefficients of interactive binary outcomes of Age/BMI, plus Nulliparity, Hormone use, and Race (ABNHR). Risk probability was calculated based on each permutation of ABNHR, and a rank order was created. A baseline risk of 1 (indicating no increased risk and therefore no screening triage) was assigned to women who were of any age with BMI <25, any BMI and <25 years old (yo), <30 yo and BMI <50, <35 yo and BMI <45, 40 yo and BMI <40, 45 yo and BMI <35, or <50 yo and BMI <30 with additional adjustments for nulliparity, hormone use (>1yr) and race other than White. Results: ECRN can be constructed from a target community’s historical data “learning set” to determine ABNHR based risk. At-risk communities can use their consensus values to determine the risk justifying the need and type of additional triage. Each ABNHR variable coefficient can be adjusted based on actuary data from the represented community or local health plan after determining PPV and NPV from a community’s historical data “testing set.” Additional triage for EC screening should be adjusted for treatment outcomes so that good prognosis communities need not participate and may continue the standard of care, e.g., biopsy upon bleeding symptoms. Additional variables, e.g., IDDM and HgA1c, can be added to enhance accuracy depending on available resources. Conclusions: EC screening based on ABNHR as univariate single parameters has few advocates since the disappointing results of ultrasonography in the previous century resulted in a thinner, more parous, lower risk US. Multivariate nomograms are more precise and tailored to outcomes in specific communities. EC disproportionate disease burden may be reduced by earlier detection based on an individualized risk threshold. We built ECRN based on factors requiring no invasive, costly lab data (ABNHR) to determine how and when to triage asymptomatic women to additional questioning, sonogram, or biopsy. Future trials will use ECRN inclusion criteria to enroll patients for prospective EC screening. Each variable can be continuously monitored and updated based on real-world data from any particular cohort so that desired positive and negative predictive value of EC screening triage can be maintained and aligned with communities’ resources, outcomes, disparities, and values. Objectives: Endometrial cancer (EC) may be the most common malignancy and the leading cause of death ahead of breast and lung cancer in certain communities (Timoteo-Liaina IJGO ‘21). Early curable stage presentation benefits many patients; however, disparities exist from the advanced stage, aggressive histology, and access (RK Lee SGO ‘20). We sought to address this with a low-resource EC Risk Nomogram (ECRN) to identify risk warranting triage of asymptomatic women in areas with disproportionate disease burden to affect favorable stage migration and eventually reduce mortality disparities. Methods: Using target community retrospective data, odds ratios were determined for the beta coefficients of interactive binary outcomes of Age/BMI, plus Nulliparity, Hormone use, and Race (ABNHR). Risk probability was calculated based on each permutation of ABNHR, and a rank order was created. A baseline risk of 1 (indicating no increased risk and therefore no screening triage) was assigned to women who were of any age with BMI <25, any BMI and <25 years old (yo), <30 yo and BMI <50, <35 yo and BMI <45, 40 yo and BMI <40, 45 yo and BMI <35, or <50 yo and BMI <30 with additional adjustments for nulliparity, hormone use (>1yr) and race other than White. Results: ECRN can be constructed from a target community’s historical data “learning set” to determine ABNHR based risk. At-risk communities can use their consensus values to determine the risk justifying the need and type of additional triage. Each ABNHR variable coefficient can be adjusted based on actuary data from the represented community or local health plan after determining PPV and NPV from a community’s historical data “testing set.” Additional triage for EC screening should be adjusted for treatment outcomes so that good prognosis communities need not participate and may continue the standard of care, e.g., biopsy upon bleeding symptoms. Additional variables, e.g., IDDM and HgA1c, can be added to enhance accuracy depending on available resources. Conclusions: EC screening based on ABNHR as univariate single parameters has few advocates since the disappointing results of ultrasonography in the previous century resulted in a thinner, more parous, lower risk US. Multivariate nomograms are more precise and tailored to outcomes in specific communities. EC disproportionate disease burden may be reduced by earlier detection based on an individualized risk threshold. We built ECRN based on factors requiring no invasive, costly lab data (ABNHR) to determine how and when to triage asymptomatic women to additional questioning, sonogram, or biopsy. Future trials will use ECRN inclusion criteria to enroll patients for prospective EC screening. Each variable can be continuously monitored and updated based on real-world data from any particular cohort so that desired positive and negative predictive value of EC screening triage can be maintained and aligned with communities’ resources, outcomes, disparities, and values.