Project Title: Is Surgery of the Primary Tumor Associated with Improved Survival in Patients with HER2+ De Novo Metastatic Breast Cancer Treated with Pertuzumab?
Faculty Mentor: Karen Daily
Student: Luderve Rosier
Research Project Description:
Metastatic breast cancer (MBC) can present as a recurrence of an earlier stage breast cancer or with an intact primary breast tumor at the time of diagnosis called “de novo” MBC. Approximately 10% of MBC cases are de novo MBC, which have been shown to have better prognosis which may be due to treatment naivety. Current standard of care treatment does not distinguish de novo MBC which is treated with systemic palliative therapy like all MBC.
The advent of the anti-HER2 agents Trastuzumab and more recently Pertuzumab has drastically changed survival of patients with HER2+ MBC. Prior to Trastuzumab, median survival was less than two years, and now with both Trastuzumab and Pertuzumab is nearly five years. (Lambertini, 2020). The CLEOPATRA trial of a largely treatment naïve population has shown Pertuzumab improves overall survival by approximately 1.5 years. Remarkably, more than one third of patients are alive at eight years of follow up and 16% remain free of disease (Swain, 2020).
The value of surgery to the primary tumor in de novo MBC has been a longstanding and as yet unresolved question (Coa et al, 2019; Khan et al, 2002). Nonrandomized retrospective data sets have shown superior survival for de novo MBC patients who undergo surgery, but the patients selected to undergo surgery had other confounding characteristics of both the tumor and the host associated with longer expected survival. Patients who receive surgery are more likely to be earlier in disease progression, younger, have fewer comorbidities, and be treated in an academic institution or comprehensive cancer center (Lane, et al., 2019) Despite a lack of evidence clearly establishing benefit,surgery remains frequently performed for de novo MBC. Lane et al demonstrated in a National Cancer Database (NCDB) analysis of 24,015 women with de novo MBC that 41.9% of patients received surgical resection after systemic therapy. The Eastern Cooperative Oncology Group prospective randomized trial (ECOG 2108) results are expected to resolve whether surgery should be offered to de novo MBC patients following systemic therapy. ECOG 2108 randomly assigned de novo MBC patients whose disease did not progress on initial systemic therapy to surgical resection of the primary versus standard treatment with no surgery. However, it is unlikely that ECOG 2108 will answer this question for HER2+ patients treated with Pertuzumab due to the trial’s timing of enrollment relative to Pertuzumab’s FDA approval and sample size (n=391).
We are specifically interested in local treatment of the primary tumor following Trastuzumab and Pertuzumab containing treatment regimens, which have been associated with high rates of complete and durable remissions of HER2+ MBC. The study population will be limited to those diagnosed in 2013 or later to correlate with both FDA approval of Pertuzumab in the metastatic setting and availability of HER2 tumor status in the National Cancer Database (NCDB). We will correlate the survival data available in NCDB to provide information on outcomes with and without local treatment of primary tumors
Project Title: Gender Bias in Semi-Annual Evaluations of GME Accredited Housestaff: A Qualitative Study and Comparison of Non-Procedural versus Procedural Specialties
Faculty Mentor:Julia Close
Student: Kristi Bears and Ashley Frye
Email: email@example.com firstname.lastname@example.org
Research Project Description:
The Accreditation Council for Graduate Medical Education (ACGME) requires that all residency and fellowship programs evaluate their residents on at least a semi-annual basis. Recommendations are made by the Clinical Competency Committee (CCC) as a group and provided to the program director for the final decision. These semi-annual evaluations are designed to track the progress of the resident and to ensure that residents are meeting their ACGME milestones.1 Previous studies have identified bias in the evaluations of residents across multiple specialties including emergency medicine (EM) and surgery.2-4 One study in EM looked qualitatively at the ACGME narrative feedback given to residents across two years and discovered that the “ideal” EM resident possesses stereotypically masculine traits such as decisiveness, confidence, and independence. Only 50% of male residents received negative comments regarding possession of “ideal” traits in contrast to 77% of females. Additionally, male residents received consistent feedback from attendings when they struggled while females received dissonant feedback.2 A second study in EM, reported that male and female residents received similar feedback at the beginning of residency. However, by graduation, the rate men progressed through the ACGME milestones was significantly higher amounting to 3-4 months of training.3 The surgical study looked at the language used in resident feedback and found there was gendered use of the job domain, disposition and humanism, reference to future, professional competency, and overall performance themes. The overall tone of the comments about men were more positive than the comments made to women.4 These studies suggest a bias in the language used to describe male and female residents that may or may not put these women at a disadvantage.
In addition to looking at gendered differences in language, a number of studies have attempted to assess whether the gender of the resident evaluator plays a role in producing gender biased evaluations. Two out of three studies looking at internal medicine residency programs did not find a significant difference in evaluations based on the gender of the evaluator. Each of the studies performed a quantitative analysis of a nine point scale utilized to rate residents.5-7 The only study with a statistical difference showed that male residents received significantly higher scores from male attendings than from female attendings in 6/9 dimensions on the evaluation form.5 One of the EM studies discussed earlier also did not find a significant difference in evaluations based on the gender of the evaluator.3 As a result, future studies may require a closer look at evaluations or the use of narrative evaluations rather than numbers based rating-scales to parse out the true effect the gender of the evaluator has on resident evaluations. Despite looking at language, the gender of resident evaluators, and trends of gender bias over the course of a resident’s training, there are no known studies that have compared gender bias in procedural versus non-procedural residency programs. With women taking up far less spots than men at approximately 35% across 10 surgical specialties which are procedurally based, this study looks to assess if the language used in the semi-annual ACGME resident evaluations recommended by the CCC and sent to the program director for finalization for the semi-annual evaluation differs across procedural and non-procedural based programs.8
Based on previous studies looking to assess the influence of gender bias on resident evaluations, we expect to see a significant difference in the themes and word choice attributed to male and female residents.2-4 With more men than women traditionally holding spaces in procedural based programs, we also hypothesize that gender bias will have a greater impact on resident evaluations within the procedural specialties than within the non-procedural or mixed specialties.8 Additionally, while recognizing that the CCC evaluating the resident/fellow and the program director may have different recommendations, we do not expect to see a significant difference in evaluations based on the gender of the program director due to few other studies citing a significant difference in evaluations across male and female evaluators.3,5-7 Finally, contrary to the findings of one of the EM studies previous discussed, we expect gender bias to dissipate as residents progress through their training due to individuation potentially reducing bias as evaluators better get to know their residents. A study conducted by the #HeForShe Taskforce describes how individuation exercises may reduce implicit bias.10
The purpose of this study is to determine if there is a gender bias present in the semi-annual evaluations of UF Health Shands residents and fellows. For the purposes of this study, this institution’s programs have been categorized as procedural, non-procedural, or mixed. If gender bias is observed, the goal is to determine if there is a difference in gender bias between procedural, non-procedural, and mixed training programs by analyzing the language used to characterize residents/fellows. The prevalence of women in procedural, non-procedural, and mixed training programs will be determined, and the role the gender of the training program director, race of the residents/fellows, and post-graduate year plays in gender bias will also be assessed. This study may increase bias awareness for resident and fellow evaluators to generate more accurate and constructive evaluations regardless of gender for current and future residents/fellows.
- Common Program Requirements. ACGME Main Page. https://www.acgme.org/What-We-Do/Accreditation/Common-Program-Requirements. Accessed April 17, 2020.
- Mueller AS, Jenkins TM, Osborne M, Dayal A, Oconnor DM, Arora VM. Gender Differences in Attending Physicians Feedback to Residents: A Qualitative Analysis. Journal of Graduate Medical Education. 2017;9(5):577-585. doi:10.4300/jgme-d-17-00126.1.
- Dayal A, O’Connor DM, Qadri U, Arora VM. Comparison of Male vs Female Resident Milestone Evaluations by Faculty During Emergency Medicine Residency Training. JAMA Internal Medicine. 2017;177(5):651. doi:10.1001/jamainternmed.2016.9616.
- Gerull KM, Loe M, Seiler K, McAllister J, Salles A. Assessing gender bias in qualitative evaluations of surgical residents. The American Journal of Surgery. 2019;217(2):306-313. doi:10.1016/j.amjsurg.2018.09.029.
- Rand VE, Hudes ES, Browner WS, Wachter RM, Avins AL. Effect of evaluator and resident gender on the American board of internal medicine evaluation scores. Journal of General Internal Medicine> 1998;13(10): 670-674. doi:10.1046/j.1525-1497.1998.00202.x
- Brienza RS, Huot S, Holmboe ES. Influence of Gender on the Evaluation of Internal Medicine Residents. Journal of Womens Health. 2004;13(1):77-83. doi:10.1089/154099904322836483.
- Holmboe ES, Huot SJ, Brienza RS, Hawkins RE. The Association of Faculty and Residents’ Gender on Faculty Evaluations of Internal Medicine Residents in 16 Residencies. Academic Medicine. 2009;84(3):381-384. doi:10.1097/acm.0b013e3181971c6d.
- ACGME Residents and Fellows by Sex and Specialty, 2017. AAMC. https://www.aamc.org/data-reports/workforce/interactive-data/acgme-residents-and-fellows-sex-and-specialty-2017. Accessed April 21, 2020.
- Corbin JM, Strauss A. Grounded theory research: Procedures, canons, and evaluative criteria. Qualitative Sociology. 1990;13(1):3-21. doi:10.1007/bf00988593.
- Dibrito SR, Lopez CM, Jones C, Mathur A. Reducing Implicit Bias: Association of Women Surgeons #HeForShe Task Force Best Practice Recommendations. Journal of the American College of Surgeons. 2019;228(3):303-309. doi:10.1016/j.jamcollsurg.2018.12.011.
Project Title: Towards prediction of CRC in population under 50 years old using EHR-based Machine Learning
Faculty Mentor: Thomas George
Student: Michael Quillen
Research Project Description:
Colorectal cancer (CRC) is the 3rd most common cancer worldwide, and approximately 4.2% of both men and women will be diagnosed during their lifetime. Of all cancer deaths, CRC represents 8% of deaths worldwide. Despite the declining overall death rates attributed to screening programs over the past three decades, the overall 5 year survival rate for patients with CRC was only 64% from 2009-2015. There were an estimated 145,600 new cases and a total of 51,020 deaths in the US in 2019 from CRC. More concerning, there has been an increase in incidence of CRC in patients under 50 and these malignancies in the younger population have been shown to be more aggressive in comparison to ages over 50.
With the introduction of CRC screening programs, national guidelines currently recommend screening assessment (via endoscopy or fecal-based assays) beginning at the age of 50 for those without CRC risk factors. This allows early detection of cancer and pre-cancerous (i.e. hyperplasia, tubular adenoma) pathology leading to improved chances for survival. For those with a known CRC risk factor (i.e., a genetic predisposition/family history or inflammatory bowel disease [IBD]), the age to begin CRC screening should be significantly lower to also allow for early cancer detection and optimal survival. Given the increasing incidence rates of young patients with CRC without concomitant increases in IBD or genetic predisposition, it is critical to accurately identify previously unrecognized risk factors for CRC for public health, patient care and policy recommendations.
The ability to access and analyze data from large registries offers an opportunity to identify previously unrecognized risk factors or combinations of factors in young CRC patients that may be associated with this unfortunate trend in oncology. Currently, there is no model that is sufficient to cover the full range of risk factors across variable demographics. Most models use a non-genetic approach with typical multivariable linear regression techniques to correlate relationships between risk factors and colorectal cancers; however, these models have limitations in focusing on details of specific risk factors.
Machine learning and other methods of mining large data sets have grown increasingly popular in the biomedical research community for the ability to swiftly glean new insights and detect new trends otherwise insurmisable to the user. Despite the presence and availability of large data sets of patients with CRC < 50 yo, no attempt has been made in literature to mine new insights using the proposed methods from large datasets. We propose the use of machine learning (i.e. supervised learning algorithms) may be useful in identifying previously unknown or unexamined risk factors for the early onset (< 50 y.o.) of CRC in patients, taking into consideration the already well established risk factors of genetic predisposition (i.e. Lynch Syndrome, Familial Adenomatous Polyposis) and Inflammatory Bowel Disease.
The general aim of this project is to use a machine learning (i.e. Neural Network, Decision Trees, Support Vector Machines) approach to identify, analyze and correlate risk factors for CRC in patients under 50.
Specific aims that will enable this goal are as follows:
- From the OneFlorida data, identify all patients under the age of 50, who do not have IBD or known genetic predisposition for CRC.
- Separate the patient population into two groups: with CRC, and without CRC.
- Identify a list of risk factors from literature and by filtering all available variables from patients’ medical records, and analyze for statistical relevance using logistic regression.
- Build machine learning models for CRC prediction.
- Moore KJ, Sussman DA, Koru- Sengul T. Age-Specific Risk Factors for Advanced Stage Colorectal Cancer, 1981–2013. Prev Chronic Dis 2018;15:170274. DOI: https://doi.org/10.5888/pcd15.170274.
- Risk Prediction Models for Colorectal Cancer: A Review Aung Ko Win, Robert J. MacInnis, John L. Hopper and Mark A. Jenkins Cancer Epidemiol Biomarkers Prev March 1 2012 (21) (3) 398-410; DOI:10.1158/1055-9965.EPI-11-0771
- Clarke WT, Feuerstein JD. Colorectal cancer surveillance in inflammatory bowel disease: practice guidelines and recent developments. World J Gastroenterol. (2019) 25:4148–57. doi: 10.3748/wjg.v25.i30.4148
- Stidham RW, Higgins PDR. Colorectal Cancer in Inflammatory Bowel Disease. Clin Colon Rectal Surg. 2018;31(3):168–178. doi:10.1055/s-0037-1602237