Cox Proportional Hazards Regression Models Research Articles

Insulin resistance surrogates are associated with all-cause mortality and cardiovascular mortality in population with metabolic syndrome: a retrospective cohort study of NHANES

The study aimed to assess the association of five insulin resistance surrogates, namely HOMA-β (Homeostasis Model Assessment of Beta-cell Function), QUICKI (Quantitative Insulin Sensitivity Check Index), IGR (Insulin Glucose Ratio), e-IS (Estimated Insulin Sensitivity), and Bennett ISI(Bennett’s insulin sensitivity index) with all-cause and cardiovascular mortality in respondents with metabolic syndrome(MetS). The prospective cohort of 6662 participants aged 18 years and older with metabolic syndrome was extracted from the National Health and Nutrition Examination Survey(NHANES 1999–2016). The multivariate Cox proportional hazards regression model, Kaplan-Meier survival curve, and log-rank tests were applied to determine the association between the five indices and all-cause mortality and cardiovascular mortality in the MetS population. Restricted cubic splines, a two-piece segmented Cox proportional hazards model, and threshold effect analyses were performed to evaluate the nonlinear relationship. Sensitivity analyses were then conducted by removing individuals with CKD, CHF, CAD, stroke, or cancer, respectively. All five insulin resistance (IR) surrogates displayed a negative association with all-cause and cardiovascular mortality in participants with metabolic syndrome. Restricted cubic spline curves showed QUICKI, IGR, and e-IS had a nonlinear relationship with statistical significance. MetS population at the highest quartile of HOMA-β or IGR exhibited lower all-cause and cardiovascular event probabilities compared with those at the lowest quartile, and e-IS had a similar correlation with cardiovascular events. Threshold effect analyses showed that there were inflection points of IGR for all-cause and cardiovascular mortality. As IGR gradually approached inflection points, the two types of mortality risks descended by 15%[HR 0.85(0.80,0.91)] and 19%[HR 0.81(0.71,0.92)], respectively. Sensitivity analysis indicated most results were robust, but Bennett ISI did not exhibit significant outcomes in participants without CKD. Our study provides evidence that HOMA-β, QUICKI, IGR, e-IS, and Bennett ISI displayed a reverse correlation with all-cause mortality and cardiovascular mortality in participants with metabolic syndrome. The five IR surrogates should be given more attention during the follow-up of MetS population.

Acetylsalicylic acid and subarachnoid haemorrhage in the Nurses' Health Study.

Acetylsalicylic acid (aspirin) is known to increase the risk of bleeding throughout the body. However, there is also evidence to suggest that acetylsalicylic acid may have a protective role in the formation and rupture of intracranial aneurysms. Previous studies investigating acetylsalicylic acid and subarachnoid haemorrhage (SAH) have so far provided conflicting results. The aim of this study was to analyse the Nurse's Health Study (NHS) using serial assessments to evaluate differences in rates of SAH in those participants taking acetylsalicylic acid and those not taking acetylsalicylic acid while considering dose, frequency, and duration as well as different types of SAH. The Nurse's Health Study (NHS) is a prospective population-based cohort study of female nurses. Information on acetylsalicylic acid was first reported in 1980 until 2016 and included acetylsalicylic acid use, dose, frequency and duration. All stroke cases were classified by physicians. Cox proportional-hazards regression models were used to estimate the hazard ratio (HR) associated with acetylsalicylic acid use. A total of 117,648 NHS participants were eligible for analysis with 357 cases of SAH observed over 4,091,239 years of follow up. There was no association between acetylsalicylic acid use and SAH (HR 1.02 [0.82, 1.28], p=0.85), aneurysmal SAH (1.04 [0.78, 1.39], p=0.78), or idiopathic SAH (HR 0.94 [0.65, 1.34], p=0.72). The number of acetylsalicylic acid tablets per week was associated with SAH (HR 1.03 [1.00, 1.06], p=0.02), specifically fatal SAH (HR 1.04 [1.00, 1.08], p=0.03). There was no association between frequency and SAH (HR 1.06 [0.99, 1.13], p=0.07). There was no evidence to support a protective association between acetylsalicylic acid and either SAH or aneurysmal SAH in female participants. In fact, there was some evidence to suggest increased SAH risk with increased acetylsalicylic acid dose in some but not all analyses. Data are available by request from the Brigham and Women's Hospital, Harvard Medical School, and Harvard T.H. Chan School of Public Health.

Prognostic value of the CONUT score with immune checkpoint inhibitors as first-line therapy for metastatic malignant melanoma.

The recent availability of immune checkpoint inhibitors (ICIs) has revolutionized the treatment of advanced malignant melanoma (MM). However, many patients with MM do not benefit from ICI treatment. As immunotherapy is associated with significant toxicity and high treatment costs despite its excellent efficacy, it is pertinent to select patients who are likely to respond to ICIs. In this single-center, retrospective study we investigated whether the controlling nutritional status (CONUT) score is a useful prognostic marker in Japanese patients with advanced-stage cancer. We analyzed 123 patients with stage IV MM treated with ICIs as first-line systemic treatment at our hospital between February 2012 and July 2024. Receiver operating characteristic curve analysis was used to calculate the CONUT cut-off value and CONUT into two groups of ≥3 and ≤2. Progression-free survival (PFS) and overall survival (OS) were determined using the Kaplan-Meier method, and differences in survival were assessed using the log-rank test. The Cox proportional hazard regression model was used to evaluate independent prognostic factors. Objective response rate (ORR), PFS, and OS were significantly low in the CONUT ≥3 group, characterized by low nutritional status and high inflammation. Multivariate analysis identified the CONUT score as an independent prognostic factor for both PFS and OS. The CONUT score was not significantly associated with the development of serious immune-related adverse events. The simplicity of the CONUT score may aid in identifying patients with MM who are suitable candidates for ICI treatment.

Association of baseline and trajectory of triglyceride-glucose index with the incidence of cardiovascular autonomic neuropathy in type 2 diabetes mellitus

BackgroundCardiovascular autonomic neuropathy (CAN), characterized by disrupted autonomic regulation of the cardiovascular system, is a frequent complication associated with diabetes. The triglyceride-glucose (TyG) index represents a precise insulin resistance indicator. However, the influence of baseline and prolonged TyG index patterns on CAN risk in type 2 diabetes remains unclear.MethodsBased on the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial, multivariate logistic regression models and restricted cubic splines (RCS) were deployed for elucidating the relation between baseline TyG index and the incidence of CAN. The area under the curve (AUC) of receiver operating characteristic (ROC) curve was used to assess the diagnostic value of the TyG index in predicting the risk of CAN. The relationship between TyG trajectory and the occurrence of CAN in individuals with diabetes was examined using Kaplan-Meier curve and a multivariable Cox proportional hazards regression model. Subgroup analysis was used to assess the robustness of the results. Additionally, we explored the impact of intensive glycemia treatment on the relationship between trajectory of TyG index and CAN risk.ResultsIn this study, these in the top quartile of the TyG index had a greater likelihood of developing CAN (TyG index Q4 vs. Q1 in Model II, OR = 1.29, 95% CI 1.03–1.62, P = 0.027). RCS indicated a rising trend in the TyG index in relation to the incidence of CAN. The AUC of the TyG index for predicting the occurrence of CAN was 0.636 (95% CI 0.620–0.651; P < 0.001), with the cut-off value of 0.208. During a 7-year follow-up period, three unique TyG trajectories were recognized: class 1 (n = 431, 23.26%), class 2 (n = 798, 27.57%), and class 3 (n = 293, 31.71%). Notable discrepancies in CAN risk across various trajectories were identified in Kaplan-Meier curve (P < 0.001). Cox regression analysis indicated that individuals in class 3 experienced a greater incidence of CAN in comparison to those in class 1 after adjusting for all covariates. Subgroup analysis found no significant effect modification in this relationship. Additionally, in the intensive glycemia group, class 2 had a reduced risk of CAN, while class 3 had an increased risk when compared to standard glycemia group.ConclusionIncreased baseline levels and long-term trajectory of TyG index are associated with an increased incidence of CAN. Intensive glycemic therapy might influence the association between the trajectory of TyG index and the chance of developing CAN.

Incidence and predictors of respiratory tract infections among birth cohorts in Ethiopia, 2023

BackgroundRespiratory tract infection (RTI) has been a predominant health problem worldwide, including Ethiopia. It is one of the major causes of under-five mortality in Ethiopia, accountable for 18% of total deaths. Nationwide studies on the incidence of respiratory infections and maternal risk factors are limited. Hence, this study was aimed to determine the incidence and predictors of respiratory infections among infants aged less than six months in Ethiopia, 2023.MethodsThis was a retrospective follow up study using data from Performance Monitoring for Action Ethiopia (PMA Ethiopia). A two-stage cluster sampling technique was used to select a total 2,246 eligible study participants from 206 enumeration areas. A Cox proportional hazards regression model was used to identify independent predictors of respiratory tract infection incidence. Hazard ratios with 95% confidence intervals and p value < 0.05 were used to declare statically significant associations between variables.ResultsThe incidence rate of respiratory tract infections among birth cohorts was 22.99 per 1000 infant weeks of observation. Infants born from mothers who sustained convulsion during labor had nearly doubled [AHR = 1.65, 95%CI (1.20, 2.28)] hazard or risk of developing respiratory tract infections than their counterparts. Similarly, infants born from mothers having prolonged labor (more than 12 h) had one and half times [AHR = 1.48, 95%CI (1.13, 1.93)] increased risk of developing respiratory tract infections as compared to their counterparts.ConclusionThe incidence of respiratory tract infections is a significant public health concern in Ethiopia. Infants born to mothers with prolonged labor and convulsions need closer monitoring for early signs of respiratory tract infections.

Unwinding of Continuous Medicaid Coverage Among Pediatric Community Health Center Patients

During the COVID-19 pandemic public health emergency, states provided continuous Medicaid coverage to enrollees. In April 2023, states began to unwind this continuous coverage, prompting concern about the impact of this on pediatric patients. To examine loss of coverage during the unwinding of continuous Medicaid coverage among pediatric patients seen at community-based health care organizations. This cohort study with statistical analysis in November 2024 included pediatric patients from a multistate network of community-based health care organizations. Participants were aged up to 17 years at both their last Medicaid-insured visit during the continuous coverage period and at their first visit during the unwinding period (April 1, 2023, to March 31, 2024). Age, sex, race and ethnicity, language, and medical complexity. The main outcome was Medicaid disenrollment to uninsured status during the unwinding period. Logistic regression was used to estimate the odds of ever being uninsured during unwinding, and Cox proportional hazards regression models were used to examine the time to uninsured status from the start of unwinding in each patient's state of residence through the end of March 2024. Associations between age, sex, race and ethnicity, language, and medical complexity and the outcome were assessed. Among 450 146 pediatric patients, mean (SD) patient age was 8.11 (5.07) years and 50.1% were male. Overall, 8.7% were disenrolled from Medicaid to uninsured status. Patients aged 12 to 17 years had the highest estimated disenrollment among age groups (10.5%), but after adjustment, all other age groups had lower odds of disenrollment compared with those younger than 1 year. Females had higher odds of disenrollment (adjusted odds ratio [AOR], 1.15 [95% CI, 1.13-1.18]; adjusted hazard ratio [AHR], 1.14 [95% CI, 1.12-1.17]) than males. American Indian or Alaska Native patients had higher odds of disenrollment (AOR, 1.95 [95% CI, 1.81-2.09]; AHR, 1.81 [95% CI, 1.05-3.13]) than White patients, with estimated disenrollment of 17.1% vs 9.4%. Compared with patients with low medical complexity, those with either chronic but noncomplex (AOR, 1.83 [95% CI, 1.79-1.88]; AHR, 1.80 [95% CI, 1.44-2.27]) or complex chronic (AOR, 1.95 95% CI, 1.89-2.00]; AHR, 1.92 [95% CI, 1.67-2.21]) medical complexity had higher odds of disenrollment. In this cohort study of previously Medicaid-insured pediatric patients seen at community-based health care organizations, a meaningful proportion of patients were disenrolled to uninsured status during the unwinding of continuous Medicaid coverage, with associated differences by demographic and clinical characteristics. This could impact access to care and health outcomes for the patients.

The association between stress-induced hyperglycemia ratio and cardiovascular events as well as all-cause mortality in patients with chronic kidney disease and diabetic nephropathy

The stress hyperglycemia ratio (SHR) is an emerging biomarker used to assess blood glucose levels under acute stress conditions and has been linked to the incidence of adverse clinical outcomes. However, the precise role of SHR in patients with diabetic kidney disease (DKD) and chronic kidney disease (CKD), particularly in relation to mortality, remains poorly understood. This study seeks to investigate the clinical value of SHR as a predictive tool for all-cause and cardiovascular mortality in these patient groups. This study analyzed data from the National Health and Nutrition Examination Survey (NHANES) spanning from 1999 to 2018, encompassing 3,507 individuals diagnosed with diabetic kidney disease (DKD) or chronic kidney disease (CKD). The primary endpoints included all-cause mortality and cardiovascular mortality, with mortality data obtained from the National Death Index (NDI) through December 31, 2019. Participants were categorized into quartiles based on the stress hyperglycemia ratio (SHR), and Cox proportional hazards regression models were employed to examine the association between SHR and mortality. Model 1 did not account for any covariates, Model 2 adjusted for age, sex, and race, while Model 3 additionally incorporated adjustments for educational attainment, marital status, body mass index, smoking behavior, hypertension, hyperlipidemia, and cardiovascular disease. The study comprised 3,507 patients with a mean age of 60.7 years, of whom 56% were female. The overall incidence of all-cause mortality was 38,000 per 100,000 person-years, while cardiovascular mortality was 11,405 per 100,000 person-years. Kaplan–Meier survival analysis revealed that the second quartile of the stress hyperglycemia ratio (SHR) (Q2) exhibited the lowest all-cause mortality (log-rank P = 0.003). Cox regression analysis indicated that the hazard ratio (HR) for all-cause mortality in Q2 was 0.76 (95% CI: 0.63, 0.92), whereas the HR for Q4 was 1.26 (95% CI: 1.04, 1.52). Restricted cubic spline (RCS) analysis revealed a J-shaped association between SHR and all-cause mortality, as well as a U-shaped association with cardiovascular mortality. The minimum risk values for SHR were 0.923 for all-cause mortality and 1.026 for cardiovascular mortality. In patients with diabetic kidney disease (DKD) and chronic kidney disease (CKD), SHR demonstrated a J-shaped relationship with all-cause mortality and a U-shaped relationship with cardiovascular mortality. Subgroup analyses indicated that the effect of spontaneous hypertension on mortality was consistent across all subgroups. This study highlights a significant association between the stress hyperglycemia ratio (SHR) and both all-cause and cardiovascular mortality in patients with diabetic kidney disease (DKD) or chronic kidney disease (CKD). SHR may serve as a critical biomarker for prognostic assessment in these populations, enabling clinicians to identify high-risk patients and tailor personalized treatment strategies that enhance patient quality of life and mitigate mortality risk.Graphical abstract

Association of household carbon consumption with cardiovascular diseases: a multicohort study.

The household sector is a major contributor to CO2 emissions. Despite growing concerns about the sector's impacts on public health, an exploration of its association with cardiovascular disease (CVD) using a longitudinal dataset across multiple countries remains scarce. We conducted a multicohort study across three countries (China, Japan, and Australia) with 36,824 participants aged 18-64, free of diagnosed CVD. Cox proportional hazard regression models were employed to examine the relationship between household carbon consumption and CVD among adults with varied sociodemographic characteristics. To measure household energy-related practice, we use two metrics, one only includes energy use (electricity and fuel), and another is more comprehensive, including energy use, water, and transportation, and is named household carbon consumption (HCC). We also used the restricted cubic spline to analyze the dose-response association of household carbon consumption per month and CVD risk. Of all participants, 12.34% developed CVD. Higher energy consumption(EC) consistently correlated with an increased risk of CVD development (HR 1.18, 95% CI [1.10-1.26]), irrespective of age, gender, household income, and exercise frequency, in both China (1.17 [1.06-1.28]) and Australia (1.14 [1.02-1.26]). Additionally, high household carbon consumption was also a risk factor (1.09 [1.01-1.18]) in both China (1.10 [1.00-1.22]) and Japan (1.24 [1.09-1.42]). Notably, distinctions were discerned, with high household carbon consumption exerting a more pronounced detrimental effect on older participants (1.11 [1.02-1.21]), women (1.11 [1.00-1.23]), individuals exercise less (1.13 [1.03-1.23]), high-income households (1.15 [1.02-1.29]), and urban residents (1.14 [1.03 to 1.26]). Our findings highlight that high household carbon consumption is associated with an increased risk of developing CVD, particularly for individuals who spend prolonged periods indoors and engage in limited physical activity. Promoting a low energy consumption lifestyle and encouraging physical activity may help mitigate the risk of CVD in later life.

Scheduling, Income, and School Closures: Unlocking the Key Drivers of Home Care Personal Support Unplanned Absences Through Time-to-Event Regression Analysis.

Healthcare workers' unplanned workplace absences are a global challenge with consequences for care recipients, employers, and healthcare systems. With rising home care demand, understanding drivers of home care personal support worker (PSW) absences can inform management of labor supply through government and employer policies. In this paper, we examined predictors of PSWs' unplanned absences within a large, administrative, longitudinal dataset (2019-2021) from a home care organization in Ontario, stratified by time (pre-, early-, and mid-pandemic). After an initial spike, unplanned absence rates were generally lower during the pandemic. Cox-proportional hazard regression models for unplanned absences highlighted how increasing income and reducing travel distance between visits can be expected to decrease the hazard of unplanned absences. School closures significantly correlated with unplanned absences, highlighting disruptions within the broader care economy. As demand for home care accelerates, reducing unplanned absences will improve care consistency for those relying on PSWs to remain safe at home.

Clinical trial participation, clinical care, and patient outcomes by practice setting: a real-world database analysis of patients with Chronic Lymphocytic Leukemia or Mantle Cell Lymphoma

ABSTRACT Objective This study was designed to compare treatment patterns, clinical trial participation, and clinical outcomes among patients with small lymphocytic lymphoma/chronic lymphocytic leukemia (CLL) or Mantle cell lymphoma (MCL) by site of care. Methods A nationwide electronic health record (EHR)-derived de-identified database was utilized for this study. Eligible patients were diagnosed with either CLL or MCL from 2013–2022 who received systemic therapy for their disease. Overall survival (OS) was analyzed using Kaplan Meier method, censoring patients without events at last observation in the database. Cox proportional hazards regression model was used to adjust for baseline covariates. Results A total of 6,372 patients with CLL and 3,411 with MCL met eligibility criteria for this analysis; 13.9% and 22.2%, respectively, were treated in academic settings. Academic settings were associated with higher patient volume and were more likely to treat MCL with CAR-T, enroll patients with CLL or MCL to clinical trials, and care for patients who were younger, White, and for CLL with higher rates of del(17p) mutations (all p < 0.01). Survival was significantly longer among patients treated in academic vs community settings (median OS not reached vs 80.5 months for CLL; 95.6 vs 68.7 months for MCL from start of first-line therapy). Discussion Patients who received care in academic settings differed from those treated in the community; care in academic settings was associated with significantly longer OS and higher trial participation. Further research is warranted to better understand the factors that may contribute to the observed outcomes.

Nonlinear relationship between hepatic steatosis index and reversion to normal glucose regulation in Chinese adults with prediabetes.

The hepatic steatosis index (HSI) has been demonstrated to have a significant correlation with prediabetes and diabetes; however, its potential association with reversion to normal glucose regulation (NGR) from prediabetes remains insufficiently investigated. The primary objective of this study was to elucidate the relationship between HSI levels and the probability of reversion to NGR among Chinese adults with prediabetes. This retrospective cohort study utilized health examination data from 11,241 Chinese adults with prediabetes. Cox proportional hazard models were employed to calculate hazard ratios (HR) and 95% confidence intervals (CI) to assess the association between HSI levels and reversion to NGR. Additionally, a Cox proportional hazards regression model incorporating cubic spline functions and smooth curve fitting was utilized to determine any nonlinear relationships between HSI levels and reversion to NGR. The results of the multivariate analyses demonstrated a significant association between reduced HSI levels and an increased likelihood of reversion to NGR (HR = 0.92, 95%CI: 0.89-0.95, P < 0.0001). Furthermore, a nonlinear relationship was identified between HSI levels and the reversion to NGR, with a critical threshold at an HSI value of 43.08. Below this threshold, a strong negative association was observed, markedly enhancing the probability of returning to normoglycemic status (HR = 0.91, 95% CI: 0.88-0.94, P < 0.0001). This study reveals a negative, nonlinear correlation between HSI levels and the reversion to NGR in individuals with prediabetes. These findings highlight the essential role of effectively managing HSI as part of comprehensive prediabetes treatment strategies, which may significantly enhance the likelihood of achieving normoglycemic status.

Hospitalization for cardiovascular disease in the year after delivery of twin pregnancies.

Increased cardiovascular demand in twin pregnancies, even those without hypertensive disease of pregnancy (HDP), may pose a greater risk for cardiovascular complications compared with singletons. In this study, the risk of cardiovascular disease (CVD)-related hospitalizations and mortality within the year following delivery in relation to HDP was compared between twin and singleton pregnancies. Using the Nationwide Readmissions Database of US hospitals from 2010 to 2020, the rates of CVD readmission in four exposure groups (twin deliveries with and without HDP and singleton deliveries with and without HDP) were estimated. Cox proportional hazard regression models were used to determine associations with singletons without HDP as the reference. Of 36 million delivery hospitalizations, the rates of CVD readmission in twin and singleton pregnancies were 1105.4 and 734.1 per 100 000 delivery admissions, respectively. Compared with singletons without HDP, the adjusted hazard ratio (HR) of CVD readmission was highest for twins with HDP [HR 8.21, 95% confidence interval (CI) 7.48-9.01], followed by singletons with HDP (HR 5.89, 95% CI 5.70-6.08) and then twins without HDP (HR 1.95, 95% CI 1.75, 2.17). Compared with singletons without HDP, twin pregnancies, even in the absence of HDP, are associated with increased risks for CVD complications in the first year post-partum. These findings highlight the increased strain twin pregnancies place on the maternal cardiovascular system. These findings advocate the need for appropriate pre-conception counselling for those with cardiovascular risk factors undergoing infertility treatment, which increase the risks of multi-foetal gestation, and increased post-partum surveillance in twin pregnancies.

Analysis Method of Real-World Digital Biomarkers for Clinical Impact in Cancer Patients

Introduction: Wearable technologies can enhance measurements completed from home by participants in decentralized clinical trials. These measurements have shown promise in monitoring patient wellness outside the clinical setting. However, there are challenges in handling data and its interpretation when using consumer wearables, requiring input from statisticians and data scientists. This paper describes three methods to estimate daily steps to address gaps in data from the Apple Watch in cancer patients and uses one of these methods in an analysis of the association between daily step count estimates and clinical events for these patients. Methods: A cohort of 50 cancer patients used the DigiBioMarC app integrated with an Apple Watch for 28 days. We identified different gap types in watch data based on their length and context to estimate daily steps. Cox proportional hazards regression models were used to determine the association between step count and time to death or time to first clinical event. Decision tree modeling and participant clustering were also employed to identify digital biomarkers of physical activity that were predictive of clinical event occurrence and hazard ratio to clinical events, respectively. Results: Among the three methods explored to address missing steps, the method that identified different step data gap types according to their duration and context provided the most reasonable estimate of daily steps. Ten hours of waking time was used to differentiate between sufficient and insufficient measurement days. Daily step count on sufficient days was the most promising predictor of time to first clinical event (p = 0.068). This finding was consistent in participant cluster and decision tree analyses where the participant cluster that emerged naturally from the different levels of daily steps was the group with the highest steps on sufficient days and lowest hazard probability of mortality and clinical events. Additionally, daily steps on sufficient days can also be used as a predictor of whether a participant will have clinical events with an accuracy of 83.3%. Conclusion: We have developed an effective way to estimate daily steps from consumer wearable data containing unknown data gaps. Daily step counts on days with sufficient sampling is a strong predictor of the timing and occurrence of clinical events, with individuals exhibiting higher daily step counts having reduced hazard of death or clinical events.

Joint ensemble learning-based risk prediction of Alzheimer's disease among mild cognitive impairment patients.

Due to the recognition for the importance of early intervention in Alzheimer's disease (AD), it is important to focus on prevention and treatment strategies for mild cognitive impairment (MCI). This study aimed to establish a risk prediction model for AD among MCI patients to provide clinical guidance for primary medical institutions. Data from MCI subjects were obtained from the NACC. Importance ranking and the SHapley Additive exPlanations (SHAP) method for the Random Survival Forest (RSF) and Extreme Gradient Boosting (XGBoost) algorithms in ensemble learning were adopted to select the predictors, and hierarchical clustering analysis was used to mitigate multicollinearity. The RSF, XGBoost and Cox proportional hazard regression (Cox) models were established to predict the risk of AD among MCI patients. Additionally, the effects of the three models were evaluated. A total of 3674 subjects with MCI were included. Thirteen predictors were ultimately identified. In the validation set, the concordance indices were 0.781 (RSF), 0.781 (XGBoost), and 0.798 (Cox), and the Integrated Brier Score was 0.087 (Cox). The prediction effects of the XGBoost and RSF models were not better than those of the Cox model. The ensemble learning method can effectively select predictors of AD risk among MCI subjects. The Cox proportional hazards regression model could be used in primary medical institutions to rapidly screen for the risk of AD among MCI patients once the model is fully clinically validated. The predictors were easy to explain and obtain, and the prediction of AD was accurate.

Outcomes of surveillance in patients with intraductal papillary mucinous neoplasms (IPMNs) who tested negative for high-risk mutations (HRMs) in next-generation sequencing (NGS).

704 Background: HRMs in NGS have demonstrated high sensitivity and specificity for predicting advanced neoplasia in patients with IPMNs. However, the outcomes of surveillance in patients who tested negative for these HRMs have not been investigated in prior studies. This study aimed to identify the predictors of progression in patients with negative NGS results for HRMs. Methods: We conducted a retrospective review of prospectively collected data from patients who underwent NGS testing between 2016 and 2023 in our pancreatic cancer prevention program. Patients without IPMNs (i.e., negative for KRAS/GNAS), with HRMs, and those without at least one follow-up imaging post-NGS were excluded from the study. Progression was defined as the development of new or additional worrisome features or high-risk stigmata (per Kyoto guidelines), or advanced neoplasia (HGD/PDAC) following NGS testing. A Cox proportional hazards regression model was used to identify predictors of progression. Results: A total of 543 patients underwent NGS testing during the study period, of which 210 patients met the inclusion criteria. The median follow-up duration was 22 months (IQR = 12-36). Of these, 72 patients developed progression following NGS testing. A total of 11 patients underwent surgical resection, and 5 of them had HGD. Factors predictive of progression included a history of smoking (HR = 1.74; 95% CI = 1.06 - 2.86), larger cyst size (HR = 1.09; 95% CI = 1.03 - 1.16), and the presence of a dilated main pancreatic duct (MPD) (HR = 2.56; 95% CI = 1.5 - 4.4) prior to NGS testing. The median time to progression for patients with worrisome features was 33 months, compared to 59 months for those without worrisome features at baseline. Similarly, the median time to progression for patients with cysts ≥ 2 cm was 38 months, compared to 57 months for those with cysts &lt; 2 cm. Conclusions: Patients who test negative for high-risk mutations in NGS are not exempt from the risk of progression. Any smoking history and presence of any worrisome features before NGS testing showed to be independent predictors of disease progression. For patients with smoking history, worrisome features or cysts ≥ 2 cm, frequent surveillance should be recommended following NGS testing. Results of Cox proportional hazards regression model for identifying the predictors of progression. Variables Hazards ratio (HR) 95% Confidence Interval P-value Age 0.99 0.96 – 1.01 0.257 Male sex 1.02 0.62 – 1.7 0.93 White race 1.07 0.63 – 0.83 0.8 Family history of pancreatic cancer 0.6 0.26 – 1.36 0.22 Personal history of other cancers 0.95 0.56 – 1.62 0.86 Diabetes 1.22 0.72 – 2.05 0.46 Any history of smoking 1.74 1.06 – 2.86 0.03* Pre-NGS cyst size (cm) 1.09 1.03 – 1.16 0.006* Pre-NGS dilated main pancreatic duct (≥5mm) 2.56 1.5 – 4.4 0.001* Pre-NGS presence of mural nodules 0.32 0.04 – 2.37 0.26 *Significant p-value.

The prognostic value of HER2 gene dosage and heterogeneity at a single-cell resolution in gastroesophageal adenocarcinoma.

473 Background: HER2 gene dosage (i.e. level of amplification) and heterogeneity are key prognostic variables for HER2+ gastroesophageal adenocarcinoma (GEA), yet no standardized approach to test these in routine clinical use exists. Here we modify the analytical pipeline for HER2 FISH to quantitatively estimate these prognostic variables and correlate them with clinical outcomes. Methods: Patients with metastatic HER2+ GEA who received trastuzumab-based therapy between 2011-2024 at Dana-Farber Cancer Institute and had available pre-treatment tissue for FISH analysis were included. HER2 FISH was conducted in a CLIA approved clinical pathology laboratory. Standard HER2/CEP17 ratio (an average of 50 cells) and individual cell HER2/CEP17 ratio were calculated. Kaplan-Meier method was used to estimate progression free (PFS) on trastuzumab-based therapy and overall survival (OS). Survival outcomes were adjusted for age, gender, tumor site, grade, prior resection, number of metastatic sites, and HER2 IHC using cox proportional hazard regression models. Results: A total of 77 patients were included. Median age was 60 years, 88% were male, and 92% had HER2 3+ tumors. Individual cells with HER2/CEP17 ratio ≥ 2.0 were defined as amplified (same cutoff as current bulk standard) and those with HER2/CEP17 ratio ≥ 4.95 (population median) were labeled highly HER2 amplified providing objective metrics to assess heterogeneity and gene dosage. Using these metrics we defined three subgroups: 1. Heterogenous (Het, n = 15): &lt;75% cells HER2+; 2. Homogenous, low gene dosage (HL, n = 30): &gt;75% cells HER2+, &lt;50% cells with high HER2 amplification; 3. Homogenous, high gene dosage (HH, n = 32): &gt;75% cells HER2+, &gt;50% cells with high HER2 amplification. Cutoffs for both metrics were determined using data distribution patterns. The subgroups were prognostically relevant and PFS and OS were significantly lower for heterogenous tumors and numerically lower for homogenous tumors with low gene dosage after multivariable adjustment (Table). Conclusions: HER2 gene dosage and heterogeneity significantly impact outcomes in HER2+ GEA. More importantly by deconvoluting clinical HER2 FISH data at single cell level, we demonstrate an objective metric with prognostic value that could be adopted for routine clinical use and for patient stratification. Further insights into differing biology of these molecular subtypes are needed. Outcomes of patients by HER2 heterogeneity and gene dosage. Hazards ratio (HR) and p values are adjusted using multivariable cox proportional hazard regression model. PFS OS Patients, No. Median, mo HR (95%CI) P Value Median, mo HR (95%CI) P Value HH 32 16.2 1 16.4 1 HL 30 12.1 1.70 (0.79-3.66) 0.173 15.7 1.71 (0.93-3.14) 0.085 Het 15 6.7 3.12 (1.35-7.2) 0.008 12.4 2.72 (1.28-5.75) 0.009

Do Existing Magnetic Resonance Imaging Definitions of Knee Osteoarthritis Identify Knees That Will Develop Clinically Significant Disease Over Up To 11 Years of Follow-Up?

In individuals without radiographic knee osteoarthritis (OA), we investigated whether magnetic resonance imaging (MRI)-defined knee OA at baseline was associated with incident radiographic and symptomatic disease during up to 11 years of follow-up. Osteoarthritis Initiative participants without tibiofemoral radiographic knee OA at baseline were assessed for MRI-based tibiofemoral cartilage damage, osteophyte presence, bone marrow lesions, and meniscal damage/extrusion. We defined MRI knee OA using alternative, reported definitions (Def A and Def B). Kellgren-Lawrence (KL) grade, joint space narrowing (JSN), and frequent knee symptoms (Sx) were assessed at baseline, 1-, 2-, 3-, 4-, 6-, 8-, and 10/11-year follow-up visits. Incident tibiofemoral radiographic knee OA (outcome) was defined as (1) KL ≥2, (2) KL ≥2 and JSN, or (3) KL ≥2 and Sx. Adjusted Cox proportional hazards regression models examined associations of baseline MRI-defined knee OA (Def A and Def B) with incident outcomes during up to 11 years of follow-up. Among 1,621 participants (mean ± SD age 58.8 ± 9.0 years, mean ± SD body mass index 27.2 ± 4.5 kg/m2, 59.5% women), 17% had MRI-defined knee OA by Def A and 24% by Def B. Baseline MRI-defined knee OA was associated with incident KL ≥2 (odds ratio 2.94 [95% confidence interval (95% CI) 2.34-3.68] for Def A and 2.44 [95% CI 1.97-3.03] for Def B). However, a substantial proportion of individuals with baseline MRI-defined knee OA did not develop incident KL ≥2 during follow-up (59% for Def A and 64% for Def B). Findings were similar for the other two outcomes. Current MRI definitions of knee OA do not adequately identify knees that will develop radiographic and symptomatic disease.

Association between adjuvant component of perioperative chemotherapy and survival outcomes in resected gastric cancer (GC).

378 Background: Perioperative chemotherapy is the standard of care for resectable GC but many patients do not proceed to or complete adjuvant chemotherapy (AC). It remains unclear whether AC following neoadjuvant chemotherapy (NAC) improves survival. Methods: Patients with GC who received NAC and underwent curative-intent gastrectomy from 2006-2022 were identified from an institutional database. Patients with mismatch repair-deficient/microsatellite instability-high tumors, GEJ tumors, M1 disease, neoadjuvant radiotherapy, total neoadjuvant intent treatment, incomplete NAC (&lt;66% of planned cycles), and incomplete data were excluded. A lymph node ratio (LNR) risk cutoff was determined using maximally selected rank statistics. Kaplan-Meier analysis with log-rank tests were used to compare overall survival (OS) and disease-free survival (DFS). Cox proportional hazards regression models were used to identify independent predictors of OS and DFS. All analyses were performed in R statistical software. Results: Of 340 patients included, 264 (78%) received AC. Most common reasons for incomplete AC were postoperative complications (20%), poor performance status (16%), minimal pathologic response to NAC (14%), patient choice (14%), and NAC toxicity (9%). Median follow-up was 6.4 years. In the overall cohort, there was no significant difference in OS (median 10 vs. 8.3 years; p=0.14) or DFS (7.0 vs. 7.5 years; p=0.12) between AC and non-AC groups. In node-negative (ypN-) patients (n=170), there was no association between receipt of AC and OS (12 vs. 10 years; p=0.6) or DFS (12 vs. 10 years; p=0.8). On univariable analysis, lack of AC receipt was not associated with shorter OS (p=0.3) or DFS (p=0.4). In the node-positive (ypN+) cohort (n=170), AC receipt was associated with significantly longer DFS (4.8 vs. 2.7 years; p=0.031) but not OS (5.5 vs. 2.8 years; p=0.071). However, on multivariate analysis, AC only showed a non-significant trend towards longer DFS (HR 0.77 [95% CI 0.49-1.19]; p=0.20). High-risk patients based on LNR (LNR&gt;0.0684; n=112) who received AC (n=87; 78%) had longer OS (p=0.033) and DFS (p=0.014). No such differences were observed in the low risk LNR subgroup (LNR&lt;0.0684; n=227). Conclusions: Following NAC, AC was not associated with improved survival in patients with ypN- GC. The association between AC and survival in ypN+ GC remains unclear but LNR may be used to identify patients with ypN+ disease who may benefit from AC. These observations require prospective validation but may have implications for clinical care and future trial design.

Impact of sirtuin genes expression on first-line treatment outcomes in patients (pts) with metastatic colorectal cancer (mCRC) enrolled in CALGB/SWOG 80405 (Alliance).

269 Background: Sirtuins (SIRT1-7) are class III histone deacetylases with different roles in CRC as tumor suppressive and oncogenic, mainly influencing proliferation, migration, invasion, DNA damage and chemoresistance. In the present study, we explored the association between tumor gene expression of sirtuins and first-line treatment outcomes in mCRC pts. Methods: We evaluated the impact of the tumor gene expression of sirtuin genes ( SIRT1 -7) on treatment outcomes (overall survival: OS; progression free survival: PFS) in 433 pts enrolled in the randomized, phase III CALGB/SWOG 80405 (NCT00265850) trial. RNA extracted from FFPE tumor samples was sequenced using the HiSeq 2500 (Illumina). Subgroup analyses were based on treatment (bevacizumab [bev] (N = 226) vs. cetuximab [cet] (N = 207)). For each gene, expression was divided into tertiles of high, medium and low. The associations between tumor gene expression and outcomes (PFS and OS) were assessed using log-rank test in univariate analysis. Multivariable Cox proportional hazards regression model was used to compute hazard ratios (HR) and 95% confidence intervals while adjusting for baseline characteristics. Treatment-gene interactions were analyzed using the likelihood ratio test. Results: Of the seven genes analyzed, SIRT3 high was associated with both longer PFS (HR: 0.59 [95% CI: 0.46-0.74], P &lt; 0.001) and OS (HR:0.64 [95% CI: 0.50-0.83], P = 0.002) overall. Treatment-gene interaction testing found significant interactions for SIRT1 and SIRT5 with PFS ( P values: 0.021 and 0.016, respectively) and OS ( P values: 0.038 and 0.014, respectively) with targeted therapies (bev vs. cet). Multivariate analysis showed SIRT1 high associated with longer PFS (HR: 0.62 [95% CI: 0.42-0.89], P = 0.014) and OS (HR: 0.59 [95% CI: 0.40-0.89], P = 0.036) in cet-treated pts; in contrast with shorter observed PFS (HR: 1.38 [95% CI: 0.96-1.99], P = 0.22) and OS (HR: 1.31 [95% CI: 0.90-1.93], P = 0.23) in bev-treated pts. SIRT5 high showed similar results and associated with longer PFS (HR: 0.64 [95% CI: 0.45-0.91], P = 0.039) and OS (HR: 0.57 [95% CI: 0.39-0.83], P = 0.013) in cet-treated pts, with no similar associations in bev-treated pts. Conclusions: Our results highlight a potential role for SIRT3 as prognostic and, SIRT1 and SIRT5 as predictive biomarkers for first-line treatment in mCRC, with differential effects based on targeted agents. Significant interaction favored cet treatment in SIRT1 high and SIRT5 high tumors. However, previous reports suggest an opposite role of SIRT5 that has been linked to chemotherapy and/or cet resistance in CRC via the inactivation of succinate dehydrogenase complex subunit A. Therefore, further studies are warranted to dissect the mechanism behind the interaction of SIRT5 with cetuximab activity and establish the potential of targeting SIRT1/5 in CRC.

Association between the triglyceride glucose-body mass index and memory-related diseases: A nationwide cohort study.

This study aimed to examine the longitudinal association between the triglyceride glucose-body mass index (TyG-BMI) index and memory-related diseases (MDs) among Chinese adults. Nationwide cohort study. Data were obtained from the China Health and Retirement Longitudinal Study (CHARLS), which is an ongoing nationally representative prospective cohort study. The exposure was the participants' TyG-BMI index at baseline from 2011 to 2012. MDs were acquired by self-reporting questions in 2018. Cox proportional hazards regression models were conducted to assess the relationship between the TyG-BMI index and MDs. During a median follow-up of 7.0 years, a total of 197 incident MDs events occurred. After multivariate adjustments for established MD-related risk factors, participants with the highest tertile of the TyG-BMI index had a higher risk of MDs (HR: 1.65; 95% CI: 1.02-2.09; P-trend <0.001) relative to individuals with the lowest tertile. Furthermore, the subgroup analyses of the association remained consistent. A high TyG-BMI index was associated with an increased risk of MDs. Our study suggests that monitoring long-term changes in the TyG-BMI index could assist with the early identification and prevention of MDs among individuals at high risk.