Advancing equitable and personalized cancer care: Novel applications and priorities of artificial intelligence for fairness and inclusivity in the patient care workflow.

Patient care workflows are highly multimodal and intertwined: the intersection of data outputs provided from different disciplines and in different formats remains one of the main challenges of modern oncology. Artificial Intelligence (AI) has the potential to revolutionize the current clinical practice of oncology owing to advancements in digitalization, database expansion, computational technologies, and algorithmic innovations that facilitate discernment of complex relationships in multimodal data. Within oncology, radiation therapy (RT) represents an increasingly complex working procedure, involving many labor-intensive and operator-dependent tasks. In this context, AI has gained momentum as a powerful tool to standardize treatment performance and reduce inter-observer variability in a time-efficient manner. This review explores the hurdles associated with the development, implementation, and maintenance of AI platforms and highlights current measures in place to address them. In examining AI's role in oncology workflows, we underscore that a thorough and critical consideration of these challenges is the only way to ensure equitable and unbiased care delivery, ultimately serving patients' survival and quality of life.

Metformin use and mortality in Asian, diabetic patients with prostate cancer on androgen deprivation therapy: A population-based study.

BACKGROUND: This study aims to examine the associations between metformin use concurrent with androgen deprivation therapy (ADT) and mortality risks in Asian, diabetic patients with prostate cancer (PCa). METHODS: This study identified diabetic adults with PCa receiving any ADT attending public hospitals in Hong Kong between December 1999 and March 2021 retrospectively, with follow-up until September 2021. Patients with <6 months of medical castration without subsequent bilateral orchidectomy, <6 months of concurrent metformin use and ADT, or missing baseline HbA1c were excluded. Metformin users had ≥180 days of concurrent metformin use and ADT, while non-users had no concurrent metformin use and ADT or never used metformin. The primary outcome was PCa-related mortality. The secondary outcome was all-cause mortality. The study used inverse probability treatment weighting to balance covariates. RESULTS: The analyzed cohort consisted of 1971 patients (1284 metformin users and 687 non-users; mean age 76.2 ± 7.8 years). Over a mean follow-up of 4.1 ± 3.2 years, metformin users had significantly lower risks of PCa-related mortality (weighted hazard ratio [wHR]: 0.49 [95% confidence interval, CI:  0.39-0.61], p < 0.001) and all-cause mortality (wHR 0.53 [0.46-0.61], p < 0.001), independent of diabetic control or status of chronic kidney disease. Such effects appeared stronger in patients with less advanced PCa, which is reflected by the absence of androgen receptor antagonist or chemotherapy use (p value for interaction: 0.017 for PCa-related mortality; 0.048 for all-cause mortality). CONCLUSIONS: Metformin use concurrent with ADT was associated with lower risks of mortality in Asian, diabetic patients with PCa.

Artificial intelligence in cancer research and precision medicine: Applications, limitations and priorities to drive transformation in the delivery of equitable and unbiased care.

Artificial intelligence (AI) has experienced explosive growth in oncology and related specialties in recent years. The improved expertise in data capture, the increased capacity for data aggregation and analytic power, along with decreasing costs of genome sequencing and related biologic "omics", set the foundation and need for novel tools that can meaningfully process these data from multiple sources and of varying types. These advances provide value across biomedical discovery, diagnosis, prognosis, treatment, and prevention, in a multimodal fashion. However, while big data and AI tools have already revolutionized many fields, medicine has partially lagged due to its complexity and multi-dimensionality, leading to technical challenges in developing and validating solutions that generalize to diverse populations. Indeed, inner biases and miseducation of algorithms, in view of their implementation in daily clinical practice, are increasingly relevant concerns; critically, it is possible for AI to mirror the unconscious biases of the humans who generated these algorithms. Therefore, to avoid worsening existing health disparities, it is critical to employ a thoughtful, transparent, and inclusive approach that involves addressing bias in algorithm design and implementation along the cancer care continuum. In this review, a broad landscape of major applications of AI in cancer care is provided, with a focus on cancer research and precision medicine. Major challenges posed by the implementation of AI in the clinical setting will be discussed. Potentially feasible solutions for mitigating bias are provided, in the light of promoting cancer health equity.

Association between duration of gonadotrophin-releasing hormone agonist use and cardiovascular risks: A population-based competing-risk analysis.

BACKGROUND: Although androgen deprivation therapy has known cardiovascular risks, it is unclear if its duration is related to cardiovascular risks. This study thus aimed to investigate the associations between gonadotrophin-releasing hormone (GnRH) agonist use duration and cardiovascular risks. METHODS: This retrospective cohort study included adult patients with prostate cancer receiving GnRH agonists in Hong Kong during 1999-2021. Patients who switched to GnRH antagonists, underwent bilateral orchidectomy, had <6 months of GnRH agonist, prior myocardial infarction (MI), or prior stroke was excluded. All patients were followed up until September 2021 for a composite endpoint of MI and stroke. Multivariable competing-risk regression using the Fine-Gray subdistribution model was used, with mortality from any cause as the competing event. RESULTS: In total, 4038 patients were analyzed (median age 74.9 years old, interquartile range (IQR) 68.7-80.8 years old). Over a median follow-up of 4.1 years (IQR 2.1-7.5 years), longer GnRH agonists use was associated with higher risk of the endpoint (sub-hazard ratio per year 1.04 [1.01-1.06], p = 0.001), with those using GnRH agonists for ≥2 years having an estimated 23% increase in the sub-hazard of the endpoint (sub-hazard ratio 1.23 [1.04-1.46], p = 0.017). CONCLUSION: Longer GnRH agonist use may be associated with greater cardiovascular risks.

Salvage brachytherapy for multiply recurrent metastatic brain tumors: A matched case analysis.

Background: Patients with recurrent brain metastases who have exhausted external radiation options pose a treatment challenge in the setting of advances in systemic disease control which have improved quality of life and survival. Brachytherapy holds promise as salvage therapy given its ability to enforce surgical cytoreduction and minimize regional toxicity. This study investigates the role of salvage brachytherapy in maintaining local control for recurrent metastatic lesions. Methods: We retrospectively reviewed our institution's experience with brachytherapy in patients with multiply recurrent cerebral metastases who have exhausted external radiation treatment options (14 cases). The primary outcome of the study was freedom from local recurrence (FFLR). To capture the nuances of tumor biology, we compared FFLR achieved by brachytherapy to the preceding treatment for each patient. We further compared the response to brachytherapy in patients with lung cancer (8 cases) against a matched cohort of maximally radiated lung brain metastases (10 cases). Results: Brachytherapy treatment conferred significantly longer FFLR compared to prior treatments (median 7.39 vs 5.51 months, P = .011) for multiply recurrent brain metastases. Compared to an independent matched cohort, brachytherapy demonstrated superior FFLR (median 8.49 vs 1.61 months, P = .004) and longer median overall survival (11.07 vs 5.93 months, P = .055), with comparable side effects. Conclusion: Brachytherapy used as salvage treatment for select patients with a multiply recurrent oligometastatic brain metastasis in the setting of well-controlled systemic disease holds promise for improving local control in this challenging patient population.

Artificial intelligence for prediction of treatment outcomes in breast cancer: Systematic review of design, reporting standards, and bias.

BACKGROUND: Artificial intelligence (AI) has the potential to personalize treatment strategies for patients with cancer. However, current methodological weaknesses could limit clinical impact. We identified common limitations and suggested potential solutions to facilitate translation of AI to breast cancer management. METHODS: A systematic review was conducted in MEDLINE, Embase, SCOPUS, Google Scholar and PubMed Central in July 2021. Studies investigating the performance of AI to predict outcomes among patients undergoing treatment for breast cancer were included. Algorithm design and adherence to reporting standards were assessed following the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) statement. Risk of bias was assessed by using the Prediction model Risk Of Bias Assessment Tool (PROBAST), and correspondence with authors to assess data and code availability. RESULTS: Our search identified 1,124 studies, of which 64 were included: 58 had a retrospective study design, with 6 studies with a prospective design. Access to datasets and code was severely limited (unavailable in 77% and 88% of studies, respectively). On request, data and code were made available in 28% and 18% of cases, respectively. Ethnicity was often under-reported (not reported in 52 of 64, 81%), as was model calibration (63/64, 99%). The risk of bias was high in 72% (46/64) of the studies, especially because of analysis bias. CONCLUSION: Development of AI algorithms should involve external and prospective validation, with improved code and data availability to enhance reliability and translation of this promising approach. Protocol registration number: PROSPERO - CRD42022292495.

Financial Worry and Psychological Distress Among Immigrants in the United States, 2013-2018.

The coronavirus disease 2019 (COVID-19) pandemic has highlighted links among economic stability, health outcomes, and migration. The facets of financial worry and their associated psychological burden have been understudied among the immigrant population. The goal of this study was to determine the specific facets of financial worry and associated psychological burden in immigrants. This cross-sectional study, which used data from the 2013 to 2018 National Health Interview Survey (NHIS), examined patient-reported measures of worry regarding financial strain. The NHIS is a household survey of noninstitutionalized, nonmilitary adults in the United States. Multivariable ordinal logistic regressions were used to define adjusted odds ratios (AORs) for financial worry and psychological distress, adjusting for various sociodemographic variables. Among 131,669 US-born and 26,155 non-US-born participants who responded to all 6 questions on the 6-item Kessler Psychological Distress Scale (K6), the overall prevalence of participants reporting any serious psychological distress (K6 score ≥13) was 3.0% and 2.25%, respectively. Despite these overall prevalence data, there were specific areas of financial worries that were higher in non-US-born participants than in US-born participants. Compared with US-born participants, non-US-born participants had higher rates of financial worries regarding retirement [75.78% vs. 69.08%, AOR=1.37, 95% confidence interval (CI) 1.29-1.45, P<0.001], medical costs due to illness (worry about not being able to pay medical costs of a serious illness or accident) (74.94% vs. 65.27%, AOR=1.37, 95% CI: 1.29-1.45, P<0.001), standard of living (74.25% vs. 65.29%, AOR=1.42, 95% CI: 1.34-1.51, P<0.001), and medical cost of health care (worry about not having enough to pay medical costs for normal health care) (66.52% vs. 52.67%, AOR=1.51, 95% CI: 1.43-1.60, P<0.001), among other costs. Notably, serious psychological distress in non-US-born individuals was associated with increased financial worry relative to US-born individuals with a similar level of psychological distress. Further research is needed to evaluate the role physicians can play in mitigating psychological distress in patients with increased financial worry.

Risk of cardiovascular mortality with androgen deprivation therapy in prostate cancer: A secondary analysis of the Prostate, Lung, Colorectal, and Ovarian (PLCO) Randomized Controlled Trial.

BACKGROUND: For men with radiation-managed prostate cancer, there is conflicting evidence regarding the association between androgen deprivation therapy (ADT) and cardiovascular mortality (CVM), particularly among those who have with preexisting comorbidities. The objective of this study was to analyze the association between ADT and CVM across patient comorbidity status using prospectively collected data from a large clinical trial. METHODS: In total, 1463 men were identified who were diagnosed with clinically localized, intermediate-risk/high-risk prostate cancer (T2b-T4, Gleason 7-10, or prostate-specific antigen >10 ng/mL) from 1993 to 2001 and managed with either radiation therapy (RT) alone or RT plus ADT during the randomized Prostate, Lung, Colon, and Ovarian (PLCO) Cancer Screening Trial. Adjusted hazard ratios (aHRs) for cause-specific mortality (prostate cancer-specific mortality vs other-cause mortality-including the primary end point of CVM [death from ischemic heart disease, cerebrovascular accident, or other circulatory disease]) were determined using Fine and Gray competing-risk regression analysis and stratified by comorbidity history. RESULTS: There was no difference in the risk of 5-year CVM between ADT plus RT and RT alone (2.3% vs 3.3%, respectively; aHR, 0.69; 95% CI, 0.38-1.24; P = .21) overall or on subgroup analysis among men with a history of ≥1 preexisting comorbidities (3.2% vs 5.3%, respectively; aHR, 0.83; 95% CI, 0.43-1.60; P = .58), ≥2 preexisting comorbidities (6.9% vs 8.3%, respectively; aHR, 0.95; 95% CI, 0.40-2.25; P = .90), or cardiovascular disease/risk factors (3.6% vs 4.3%, respectively; aHR, 0.85; 95% CI, 0.44-1.65; P = .63). These results were all similar when each component of CVM was analyzed separately-either cardiac, stroke, or other vascular mortality (P > .05). CONCLUSIONS: This study provides prospectively collected evidence that the use of ADT plus RT, compared with RT alone, is not associated with an increased risk of CVM, even among subgroups of men who have preexisting comorbidities and cardiovascular disease.

Recent Trends and Overall Survival of Young Versus Older Adults With Stage II to III Rectal Cancer Treated With and Without Surgery in the United States, 2010-2015.

BACKGROUND: The omission of surgery via nonoperative management (NOM) for rectal cancer may be increasing, and this strategy could be particularly attractive for younger patients, whose incidence of rectal cancer has been rising. We sought to assess trends in NOM in young (younger than 55 y) versus older adult (55 y and older) rectal cancer cohorts. METHODS: The National Cancer Database was used to identify patients diagnosed with stage II to III rectal cancer between 2010 and 2015. Multivariable logistic regression defined the association between sociodemographic variables and odds of NOM, including an age (18 to 54 vs. 55+ y)×surgery (surgery vs. NOM) interaction term. Adjusted Cox regression models compared overall survival between NOM versus surgery. RESULTS: Among 22,561 patients with a median follow-up of 37.5 months, the utilization rate of NOM increased from 10.7% (2010) to 15.2% (2015). Older patients were more likely to receive NOM, although rates also increased among young (7.1% to 10.6%). Black patients were also more likely to receive NOM (P<0.001). Among the entire cohort, NOM was associated with worse overall survival (adjusted hazard ratio [AHR]=2.90, 95% confidence interval [CI]: 2.67-3.15) and there was a statistically significant age×NOM interaction (P=0.01) such that the effect of NOM on survival was worse for younger (AHR=3.37, 95% CI: 2.82-4.02) as compared with older patients (AHR=2.49, 95% CI: 2.27-2.74). CONCLUSIONS: The increasing trend for NOM in stage II to III rectal cancer may be driven by disparities in treatment. Management with NOM appears to be associated with poorer survival, particularly in younger patients and could worsen outcomes for groups already at risk for suboptimal cancer care.

Obesity and younger versus older onset colorectal cancer in the United States, 1998-2017.

The etiology behind the increasing incidence of early onset colorectal cancer (EOCRC) are incompletely elucidated, but could be attributed in part to lifestyle factors. We assessed the association between obesity and colorectal cancer (CRC) in younger versus older adults in the National Health Institute Survey. Multivariable logistic regression defined adjusted odds ratios (AORs) and associated 95% confidence intervals (CIs) for CRC including an age (< vs. ≥50 years) *BMI (< vs. ≥30.0 kg/m2) interaction term. Among 583,511 study participants with a total of 3,173 CRC cases, there was a significant age*BMI interaction term (P=0.02) such that for participants aged 18-49 years, BMI ≥30.0 kg/m2 was associated with diagnosis of CRC (34.1% vs. 27.4%, AOR 1.39, 95% CI: 1.00-1.92) but not for participants aged ≥50 (29.6% vs. 31.4%, AOR 0.93, 95% CI: 0.85-1.03). Obese BMI appears to be associated with diagnosis of EOCRC, thus weight control by early adulthood, among other healthy lifestyle behaviors, could serve as potential risk reduction strategies for CRC.

Validation of a subclassification for high-risk prostate cancer in a prospective cohort.

BACKGROUND: A subgroup of men with favorable high-risk prostate cancer (T1c with either a Gleason score of 4 + 4 = 8 and a prostate-specific antigen [PSA] level <10 ng/mL or a Gleason score of 6 and a PSA level >20 ng/mL) has been associated with improved outcomes in comparison with other standard high-risk patients. This study was designed to validate the prognostic utility of a subclassification for high-risk disease with a prospectively collected data set. METHODS: This study identified 3033 men from the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial who had been diagnosed from 1993 to 2001 with clinically localized prostate cancer-either intermediate-risk disease (clinical stage T2b-c, a Gleason score of 7, or a PSA level of 10 to 20 ng/mL) or high-risk disease (clinical stage T3-T4, a Gleason score of 8-10, or a PSA level >20 ng/mL)-that was managed with radical prostatectomy or radiation therapy. Multivariable logistic regression was used to calculate adjusted odds ratios (aORs) for pathological T3 to T4 or N1 (pT3-T4/pN1) disease. Fine and Gray competing risks regression was used to determine adjusted hazard ratios (aHRs) of prostate cancer-specific mortality (PCSM). RESULTS: The median follow-up was 5.7 years. Patients with favorable high-risk disease had lower 8-year PCSM in comparison with patients with standard high-risk disease (2.2% vs 10.8%; aHR, 0.26; 95% confidence interval [CI], 0.09-0.73; P = .01) but similar PCSM in comparison with patients with intermediate-risk disease (2.2% vs 2.2%; aHR, 0.90; 95% CI, 0.32-2.54; P = .84). Among those who underwent surgery, those with favorable high-risk disease had lower odds of pT3-T4/pN1 disease than those with standard high-risk disease (46.2% vs 63.3%; aOR, 0.50; 95% CI, 0.27-0.94; P = .03). CONCLUSIONS: This study validates the prognostic utility of a subclassification for high-risk disease in a prospectively collected patient cohort. Patients with favorable high-risk disease have PCSM similar to that of patients with intermediate-risk disease and significantly better than that of patients with standard high-risk disease. Future trials are needed to assess possible de-intensification of therapy for favorable high-risk disease.

Does Universal Insurance and Access to Care Influence Disparities in Outcomes for Pediatric Patients with Osteomyelitis?

BACKGROUND: Healthcare disparities are an issue in the surgical management of orthopaedic conditions in children. Although insurance expansion efforts may mitigate racial disparities in surgical outcomes, prior studies have not examined these effects on differences in pediatric orthopaedic care. To assess for racial disparities in pediatric orthopaedic care that may persist despite insurance expansion, we performed a case-control study of the outcomes of children treated for osteomyelitis in the TRICARE system, the healthcare program of the United States Department of Defense and a model of universal insurance and healthcare access. QUESTIONS/PURPOSES: We asked whether (1) the rates of surgical intervention and (2) 90-day outcomes (defined as emergency department visits, readmission, and complications) were different among TRICARE-insured pediatric patients with osteomyelitis when analyzed based on black versus white race and military rank-defined socioeconomic status. METHODS: We analyzed TRICARE claims from 2005 to 2016. We identified 2906 pediatric patients, of whom 62% (1810) were white and 18% (520) were black. A surgical intervention was performed in 9% of the patients (253 of 2906 patients). The primary outcome was receipt of surgical intervention for osteomyelitis. Secondary outcomes included 90-day complications, readmissions, and returns to the emergency department. The primary predictor variables were race and sponsor rank. Military rank has been used as an indicator of socioeconomic status before and during enlistment, and enlisted service members, particularly junior enlisted service members, may be at risk of having the same medical conditions that affect civilian members of lower socioeconomic strata. Patient demographic information (age, sex, race, sponsor rank, beneficiary category [whether the patient is an insurance beneficiary from an active-duty or retired service member], and geographic region) and clinical information (prior comorbidities, environment of care [whether clinical care was provided in a civilian or military facility], treatment setting, and length of stay) were used as covariates in multivariable logistic regression analyses. RESULTS: After controlling for demographic and clinical factors including age, sex, sponsor rank, beneficiary category, geographic region, Charlson comorbidity index (as a measure of baseline health), environment of care, and treatment setting (inpatient versus outpatient), we found that black children were more likely to undergo surgical interventions for osteomyelitis than white children (odds ratio 1.78; 95% confidence interval, 1.26-2.50; p = 0.001). When stratified by environment of care, this finding persisted only in the civilian healthcare setting (OR 1.85; 95% CI, 1.26-2.74; p = 0.002). Additionally, after controlling for demographic and clinical factors, lower socioeconomic status (junior enlisted personnel) was associated with a higher likelihood of 90-day emergency department use overall (OR 1.60; 95% CI, 1.02-2.51; p = 0.040). CONCLUSIONS: We found that for pediatric patients with osteomyelitis in the universally insured TRICARE system, many of the historically reported disparities in care were absent, suggesting these patients benefitted from improved access to healthcare. However, despite universal coverage, racial disparities persisted in the civilian care environment, suggesting that no single intervention such as universal insurance sufficiently addresses differences in racial disparities in care. Future studies can address the pervasiveness of these disparities in other patient populations and the various mechanisms through which they exert their effects, as well as potential interventions to mitigate these disparities. LEVEL OF EVIDENCE: Level III, prognostic study.

Predictors of Need for Liver Transplantation in Children Undergoing Hepatoportoenterostomy for Biliary Atresia.

PURPOSE: The purpose of this study was to determine perioperative risk factors for need of liver transplantation following hepatoportoenterostomy. METHODS: A retrospective review of patients undergoing hepatoportoenterostomy for biliary atresia at our institution from 1990 to 2016 was completed. RESULTS: A total of 81 patients were identified with a median age of 51 days (IQR: 33-68) at hepatoportoenterostomy and a median follow-up time of 5.7 years (IQR: 1-11.6). Ten-year overall survival was 93% (95% CI: 84-97). Thirty-six patients (44%) ultimately required transplantation at a median time from hepatoportoenterostomy of 8.9 months (IQR: 5.2-19). The 10-year transplant-free survival was 36% (95%CI: 24-49). Steroid use (N=42) was not associated with improved 10-yr transplant-free survival (33% vs. 38%, p=0.690). Age at hepatoportoenterostomy was not significantly associated with the need for transplantation. Multivariable logistic regression analysis demonstrated that total bilirubin >2mg/dL (OR: 97, p<0.001) and albumin < 3.5g/dL (OR: 24, p=0.027) at 3 months after surgery were independent predictors of the need for transplantation, while adjusting for age, sex, prematurity, and steroid use. CONCLUSION: Overall survival for children with biliary atresia is excellent, although most patients will ultimately require liver transplantation. Total bilirubin and albumin level at 3 months following hepatoportoenterostomy are predictive of the need for transplantation. Steroid use is not associated with improved outcomes.

Disparity in online health information in pediatric vs. adult surgical conditions.

BACKGROUND: Although the quality of online health information (OHI) for adult surgical conditions is well described, the availability of quality OHI for pediatric surgical conditions, and the comparison to that of adult surgical OHI, remains undefined. METHODS: Medical and lay terms for 15 pediatric and 15 adult surgical conditions were searched using Google in English. The Health on the Net Foundation, a non-governmental OHI accreditation body, designates approval for quality websites. We compared the role of patient population while controlling for disease incidence (pediatric vs. adult), term complexity (medical vs. lay), and order (earlier vs. later listing of websites) on availability of quality OHI among the first 100 websites for each term. RESULTS: Among the first 100 websites, the adjusted mean number of quality websites was 11.80 for pediatric vs. 17.92 for adult medical search terms, and 13.27 for pediatric vs. 18.20 for adult lay search terms (P < 0.05 for all). Term complexity did not affect quality, and earlier appearing results were more likely to be of high quality. CONCLUSION: Availability of quality pediatric surgical OHI lags behind that of adult surgical OHI, even when controlling for disease incidence. These findings highlight the potential need for increased quality OHI in pediatric surgery.