Larger men have larger prostates: Detection bias in epidemiologic studies of obesity and prostate cancer risk

Toward fairness in artificial intelligence for medical image analysis: identification and mitigation of potential biases in the roadmap from data collection to model deployment

Purpose

To recognize and address various sources of bias essential for algorithmic fairness and trustworthiness and to contribute to a just and equitable deployment of AI in medical imaging, there is an increasing interest in developing medical imaging-based machine learning methods, also known as medical imaging artificial intelligence (AI), for the detection, diagnosis, prognosis, and risk assessment of disease with the goal of clinical implementation. These tools are intended to help improve traditional human decision-making in medical imaging. However, biases introduced in the steps toward clinical deployment may impede their intended function, potentially exacerbating inequities. Specifically, medical imaging AI can propagate or amplify biases introduced in the many steps from model inception to deployment, resulting in a systematic difference in the treatment of different groups.

Approach

Our multi-institutional team included medical physicists, medical imaging artificial intelligence/machine learning (AI/ML) researchers, experts in AI/ML bias, statisticians, physicians, and scientists from regulatory bodies. We identified sources of bias in AI/ML, mitigation strategies for these biases, and developed recommendations for best practices in medical imaging AI/ML development.

Results

Five main steps along the roadmap of medical imaging AI/ML were identified: (1) data collection, (2) data preparation and annotation, (3) model development, (4) model evaluation, and (5) model deployment. Within these steps, or bias categories, we identified 29 sources of potential bias, many of which can impact multiple steps, as well as mitigation strategies.

Conclusions

Our findings provide a valuable resource to researchers, clinicians, and the public at large.

Biomarkers of glucose homeostasis and inflammation with risk of prostate cancer: A case-cohort study

BACKGROUND

Few prospective studies have examined biomarkers of glucose homeostasis or inflammation with prostate cancer risk by tumor stage or grade.

METHODS

We conducted a case-cohort study to examine associations of pre-diagnosis hemoglobin A1c (HbA1c), c-peptide, and c-reactive protein (CRP) with prostate cancer risk overall and stratified by tumor stage and grade. The study included 390 non-aggressive (T1-2, N0, M0 and Gleason score <8) and 313 aggressive cases (T3-4, or N1, or M1, or Gleason score 8-10) diagnosed after blood draw (1998-2001) and up to 2013, and a random sub-cohort of 1,303 cancer-free men at blood draw in the Cancer Prevention Study-II Nutrition Cohort. Prentice-weighted Cox proportional hazards regression models were used to estimate hazards ratios (HRs) and 95% confidence intervals (CIs).

RESULTS

In the multivariable-adjusted model without body mass index (BMI), HbA1c was inversely associated with non-aggressive prostate cancer (HR per unit increase: 0.89, 95% CI: 0.80-1.00, P=0.04). Analyses stratified by tumor stage and grade separately showed that HbA1c was inversely associated with low-grade prostate cancer (HR per unit increase: 0.89, 95% CI: 0.80-1.00) and positively associated with high-grade prostate cancer (HR per unit increase: 1.15, 95% CI: 1.01-1.30). C-peptide and CRP were not associated with prostate cancer overall or by stage or grade.

CONCLUSIONS

The present study suggests that associations of hyperglycemia with prostate cancer may differ by tumor grade and stage.

IMPACT

Future studies need to examine prostate cancer by tumor stage and grade, and to better understand the role of hyperglycemia in prostate cancer progression.

Any Correlation Between Prostate Volume and Incidence of Prostate Cancer: A Review of Reported Data for the Last Thirty Years

PURPOSE

Prostate cancer (PCa) is the most common non-skin cancer in men worldwide and more than 80% of men with PCa also have histo-anatomical findings of benign prostate hyperplasia (BPH). It is well documented that BPH develops in the transition zone (TZ), whereas 80-85% of PCa originates in the peripheral zone (PZ) of the prostate. Possible causal links between both disease entities are controversially discussed in the current literature. Some studies have reported that larger prostates have a decreased incidence of PCa compared to smaller prostates. The purpose of this systematic review is to comprehensively summarize studies analyzing any association between prostate gland volume and incidence of PCa.

METHODS

A thorough literature review was performed between 01.01.1990 through 02.28.2020 using PubMed and applying the "PRISMA" guidelines. Inclusion and exclusion criteria were defined.

RESULTS

Our systematic review found 41 articles reporting an inverse (negative) relationship between prostate gland volume and incidence of prostate cancer. Sample sizes ranged from 114 to 6692 patients in these single institutional and multi-institutional studies. Thirty-nine (95%) of the 41 articles showed a statistically significant inverse relationship. In our search, no study was found showing a positive correlation between BPH size and the incidence of PCa.

CONCLUSION

To our knowledge, this is the first systematic review on the important clinical question of interaction between prostate size and the incidence of PCa. The results are demonstrating an inverse relationship, and therefore reveal strong evidence that large prostates may be protective of PCa when compared to smaller prostates.

Increased Body Mass Index Is a Risk Factor for Poor Clinical Outcomes after Radical Prostatectomy in Men with International Society of Urological Pathology Grade Group 1 Prostate Cancer Diagnosed with Systematic Biopsies

INTRODUCTION

The association between obesity and clinically significant prostate cancer (PCa) is still a matter of debate. In this study, we evaluated the effect of body mass index (BMI) on the prediction of pathological unfavorable disease (UD), positive surgical margins (PSMs), and biochemical recurrence (BCR) in patients with clinically localized (≤cT2c) International Society of Urological Pathology (ISUP) grade group 1 PCa at biopsy.

METHODS

427 patients with ISUP grade group 1 PCa who have undergone radical prostatectomy and BMI evaluation were included. The outcome of interest was the presence of UD (defined as ISUP grade group ≥3 and pT ≥3a), PSM, and BCR.

RESULTS

Statistically significant differences resulted in comparing BMI with prostate-specific antigen (PSA) and serum testosterone levels (both p < 0.0001). Patients with UD and PSM had higher BMI values (p < 0.0001 and p = 0.006, respectively). BCR-free survival was significantly decreased in patients with higher BMI values (p < 0.0001). BMI was an independent risk factor for BCR and PSM. Receiver-operating characteristic analysis testing PSA accuracy in different BMI groups, showed that PSA had a reduced predictive value (area under the curve [AUC] = 0.535; 95% confidence interval [CI] = 0.422-0.646), in obese men compared to overweight (AUC = 0.664; 95% CI = 0.598-0.725) and normal weight patients (AUC = 0.721; 95% CI = 0.660-0.777).

CONCLUSION

Our findings show that increased BMI is a significant predictor of UD and PSM at RP in patients with preoperative low-to intermediate-risk diseases, suggesting that BMI evaluation may be useful in a clinical setting to identify patients with favorable preoperative disease characteristics harboring high-risk PCa.

Waist circumference and a body shape index and prostate cancer risk and mortality

We recently found a negative association between body mass index (BMI) and the risk of localised prostate cancer (PCa), no association with advanced PCa, and a positive association with PCa‐specific mortality. In a 15% subpopulation of that study, we here investigated the measures of abdominal adiposity including waist circumference (WC) and A Body Shape Index (ABSI) in relation to PCa risk and mortality. We used data from 58,457 men from four Swedish cohorts to assess WC and ABSI in relation to PCa risk according to cancer risk category, including localised asymptomatic and symptomatic PCa and advanced PCa, and PCa‐specific mortality. Cox regression models were used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs). During, on average, 10 years of follow‐up, 3290 men were diagnosed with PCa and 387 died of PCa. WC was negatively associated with the risk of total PCa (HR per 10 cm, 0.95; 95% CI 0.92–0.99), localised PCa (HR per 10 cm, 0.93, 95% CI 0.88–0.96) and localised asymptomatic PCa cases detected through a prostate‐specific antigen (PSA) test (HR per 10 cm, 0.87, 95% CI 0.81–0.94). WC was not associated with the risk of advanced PCa (HR per 10 cm, 1.02, 95% CI 0.93–1.14) or with PCa‐specific mortality (HR per 10 cm, 1.04, 95% CI 0.92–1.19). ABSI showed no associations with the risk of PCa or PCa‐specific mortality. While the negative association between WC and the risk of localised PCa was partially driven by PSA‐detected PCa cases, no association was found between abdominal adiposity and clinically manifest PCa in our population.

Relationship between body mass index and concentrations of prostate specific antigen: a cross-sectional study

The possible relationship between body mass index (BMI) and prostate-specific antigen (PSA) concentrations is controversial. The objective of this study was to assess the relationship between BMI and PSA concentrations in Chinese men. A total of 81,122 men who had undergone annual medical examinations at the First Affiliated Hospital of Army Medical University between 1 January 2011 and 31 December 2018 were included. Univariate and multivariate linear regression models were used to assess the relationship between BMI and PSA concentrations. The nonlinear relationship was analyzed using a generalized additive model with a spline smoothing function. Subsequently, a stratified linear regression model was used for subgroup analysis. The mean age and BMI of the participants were 45.91 ± 12.21 years and 24.79 ± 3.11 kg/m², respectively. After adjustment for age, waist circumference-hip circumference ratio, systolic blood pressure, diastolic blood pressure, fasting blood glucose, total cholesterol, triglyceride, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, alanine aminotransferase and aspartate aminotransferase, BMI was negatively related to PSA level (p < .001). A nonlinear relationship was detected, and different relationships between BMI and PSA concentrations were observed on each side of the inflection point (BMI = 23.11 kg/m²). Our study revealed an inverse, nonlinear relationship between BMI and PSA concentrations. Thus, this relationship may be a concern when establishing reference intervals or decision limits for PSA concentrations.

Obesity at Diagnosis and Prostate Cancer Prognosis and Recurrence Risk Following Primary Treatment by Radical Prostatectomy

BACKGROUND

The association of obesity at diagnosis with prostate cancer progression is uncertain. This study aimed to examine the relationship between body mass index (BMI; 18.5-<25, 25-<30, 30-<35, ≥35 kg/m²) and prognostic risk at diagnosis, compare the concordance between prognostic risk assessed at diagnostic biopsy versus pathologic risk assessed at surgery across BMI categories, and investigate the association between obesity and prostate cancer recurrence and all-cause death.

METHODS

We examined men enrolled in CaPSURE who underwent radical prostatectomy between 1995 and 2017. Multiple imputation methods were used to handle missing data and reported along with complete case findings.

RESULTS

Participants (n = 5,200) were followed for a median of 4.5 years; 685 experienced recurrence. Obesity was associated with higher prognostic risk at time of diagnosis (OR_obese = 1.5; OR_{very obese} = 1.7) and upward reclassification of disease between biopsy and surgery, driven by change in tumor stage (OR_obese = 1.3; OR_{very obese} = 1.6). We observed an association between BMI and recurrence with adjustment for disease severity using diagnostic factors (HR_{very obese} = 1.7); this association disappeared when adjusting for disease severity factors obtained at surgery.

CONCLUSIONS

Our findings suggest that residual confounding may partially explain the conflicting evidence regarding obesity's influence on prostate cancer progression. Assessing T-stage via digital rectal exam may be complicated in larger men, potentially affecting clinical treatment decisions. A strong association with all-cause mortality demonstrates healthier BMI at diagnosis may still improve overall survival.

IMPACT

Patients with greater BMI are prone to more advanced disease at diagnosis and may be more likely to have their tumor stage underestimated at diagnosis.

Preventing Lethal Prostate Cancer with Diet, Supplements, and Rx: Heart Healthy Continues to Be Prostate Healthy and "First Do No Harm" Part I

PURPOSE OF REVIEW

To discuss the overall and latest observations of the effect of diet, lifestyle, supplements, and some prescription heart healthy medications for prostate cancer prevention.

RECENT FINDINGS

The concept of maximizing heart health to prevent aggressive prostate cancer continues to be solidified with the addition of more prospective observational and randomized controlled trial data. Heart healthy is prostate healthy, but heart unhealthy is prostate unhealthy. The primary goal of reducing the risk of all-cause and cardiovascular disease (CVD) morbidity and mortality also allows for maximizing prostate cancer prevention. The obesity epidemic in children and adults along with recent diverse research has only strengthened the nexus between heart and prostate health. Greater dietary adherence toward a variety of healthy foods is associated with a graded improved probability of CVD and potentially aggressive cancer risk reduction. Preventing prostate cancer via dietary supplements should encourage a "first do no harm", or less is more approach until future evidence can reverse the concerning trend that more supplementation has resulted in either no impact or an increased risk of prostate cancer. Supplements to reduce side effects of some cancer treatments appear to have more encouraging data. Medications that improve heart health including statins, aspirin, and metformin (S.A.M.), and specific beta-blocker medications are primarily generic or low-cost and should continue to garner research interest. A watershed moment in medical education has arrived where the past perception of a diverse number of trees seemingly separated by vast distances, in reality, now appear to exist within the same forest.