Deceptive learning in histopathology.

AIMS: Deep learning holds immense potential for histopathology, automating tasks that are simple for expert pathologists and revealing novel biology for tasks that were previously considered difficult or impossible to solve by eye alone. However, the extent to which the visual strategies learned by deep learning models in histopathological analysis are trustworthy or not has yet to be systematically analysed. Here, we systematically evaluate deep neural networks (DNNs) trained for histopathological analysis in order to understand if their learned strategies are trustworthy or deceptive. METHODS AND RESULTS: We trained a variety of DNNs on a novel data set of 221 whole-slide images (WSIs) from lung adenocarcinoma patients, and evaluated their effectiveness at (1) molecular profiling of KRAS versus EGFR mutations, (2) determining the primary tissue of a tumour and (3) tumour detection. While DNNs achieved above-chance performance on molecular profiling, they did so by exploiting correlations between histological subtypes and mutations, and failed to generalise to a challenging test set obtained through laser capture microdissection (LCM). In contrast, DNNs learned robust and trustworthy strategies for determining the primary tissue of a tumour as well as detecting and localising tumours in tissue. CONCLUSIONS: Our work demonstrates that DNNs hold immense promise for aiding pathologists in analysing tissue. However, they are also capable of achieving seemingly strong performance by learning deceptive strategies that leverage spurious correlations, and are ultimately unsuitable for research or clinical work. The framework we propose for model evaluation and interpretation is an important step towards developing reliable automated systems for histopathological analysis.

Usage of DNA Fingerprinting Technology to Check Sample Error and Contamination in Molecular Laboratories.

Sample identification error is a severe medical error in clinical molecular diagnostic laboratories, which can lead to reporting the wrong results for the patient involved. Sample contamination can also lead to incorrect test reports. Avoiding sample identification error and sample contamination could be life-saving. Sample switch and sample contamination could happen on laboratory bench works, especially when pipetting into multi-well plates. It is difficult to realize such errors during laboratory bench work. Laboratory staff may not be aware of such an error when it happens. DNA fingerprinting technology can be used to determine sample identity and subsequently identify sample switch and sample contamination in the laboratory. Our laboratory has explored the usage of this technology in our quality control process and successfully established that DNA fingerprinting can be used to monitor sample switch and sample contamination in next-generation sequencing and BCR/ABL1 real-time PCR bench work.

Landscape of EGFR-Dependent and -Independent Resistance Mechanisms to Osimertinib and Continuation Therapy Beyond Progression in EGFR-Mutant NSCLC.

PURPOSE: Osimertinib was initially approved for T790M-positive non-small cell lung cancer (NSCLC) and, more recently, for first-line treatment of EGFR-mutant NSCLC. However, resistance mechanisms to osimertinib have been incompletely described. EXPERIMENTAL DESIGN: Using cohorts from The University of Texas MD Anderson Lung Cancer Moonshot GEMINI and Moffitt Cancer Center lung cancer databases, we collected clinical data for patients treated with osimertinib. Molecular profiling analysis was performed at the time of progression in a subset of the patients. RESULTS: In the 118 patients treated with osimertinib, 42 had molecular profiling at progression. T790M was preserved in 21 (50%) patients and lost in 21 (50%). EGFR C797 and L792 (26%) mutations were the most common resistance mechanism and were observed exclusively in T790M-preserved cases. MET amplification was the second most common alteration (14%). Recurrent alterations were observed in 22 genes/pathways, including PIK3CA, FGFR, and RET. Preclinical studies confirmed MET, PIK3CA, and epithelial-to-mesenchymal transition as potential resistance drivers. Alterations of cell-cycle genes were associated with shorter median progression-free survival (PFS, 4.4 vs. 8.8 months, P = 0.01). In 76 patients with progression, osimertinib was continued in 47 cases with a median second PFS (PFS2) of 12.6 months; 21 patients received local consolidation radiation with a median PFS of 15.5 months. Continuation of osimertinib beyond progression was associated with a longer overall survival compared with discontinuation (11.2 vs. 6.1 months, P = 0.02). CONCLUSIONS: Osimertinib resistance is associated with diverse, predominantly EGFR-independent genomic alterations. Continuation of osimertinib after progression, alone or in conjunction with radiotherapy, may provide prolonged clinical benefit in selected patients.See related commentary by Devarakonda and Govindan, p. 6112.

The Implementation of Integrated Behavioral Health Protocols In Primary Care Settings in Project Care.

PURPOSE: The majority of adults with mental health (MH) and substance use (SU) disorders in the United States do not receive treatment. The Affordable Care Act will create incentives for primary care centers to begin providing behavioral health (MH and SU) services, thus promising to address the MH and SU treatment gaps. This paper examines the implementation of integrated care protocols by three primary care organizations. METHODS: The Behavioral Health Integration in Medical Care (BHIMC) tool was used to evaluate the integrated care capacity of primary care organizations that chose to participate in the Kern County (California) Mental Health Department's Project Care annually for 3years. For a subsample of clinics, change over time was measured. Informed by the Conceptual Model of Evidence-Based Practice Implementation in Public Service Sectors, inner and outer contextual factors impacting implementation were identified and analyzed using multiple data sources and qualitative analytic methods. RESULTS: The primary care organizations all offered partially integrated (PI) services throughout the study period. At baseline, organizations offered minimally integrated/partially integrated (MI/PI) services in the Program Milieu, Clinical Process - Treatment, and Staffing domains of the BHIMC, and scores on all domains were at the partially integrated (PI) level or higher in the first and second follow-ups. Integrated care services emphasized the identification and management of MH more than SU in 52.2% of evaluated domains, but did not emphasize SU more than MH in any of them. Many of the gaps between MH and SU emphases were associated with limited capacities related to SU medications. Several outer (socio-political context, funding, leadership) and inner (organizational characteristics, individual adopter characteristics, leadership, innovation-values fit) contextual factors impacted the development of integrated care capacity. CONCLUSIONS: This study of a small sample of primary care organizations showed that it is possible to improve their integrated care capacity as measured by the BHIMC, though it may be difficult or unfeasible for them to provide fully integrated behavioral health services. Integrated services emphasized MH more than SU, and enhancing primary care clinic capacities related to SU medications may help close this gap. Both inner and outer contextual factors may impact integrated service capacity development in primary care clinics. Study findings may be used to inform future research on integrated care and inform the implementation of efforts to enhance integrated care capacity in primary care clinics.

Mortality, causes of death, and health status among methamphetamine users.

This study examines causes of death, years of life lost, and health and drug use characteristics associated with mortality over an 8 to 10 year period in a sample of methamphetamine users who had and had not received substance use disorder treatment (N = 563). Decedents reported initiating their methamphetamine use for different reasons than surviving methamphetamine users, and some of these differences varied by treatment status. Study findings provide additional detail on long-term health and mortality outcomes in a diverse sample of methamphetamine users, which may inform public health strategies targeting the comparable and divergent needs of treated and untreated populations.

A qualitative study of methamphetamine users' perspectives on barriers and facilitators of drug abstinence.

To better understand methamphetamine (MA) use patterns and the process of recovery, qualitative interviews were conducted with adult MA users (n = 20), comparing a sample that received substance abuse treatment with those who had not received treatment. Respondents provided detailed information on why and how they changed from use to abstinence and factors they considered to be barriers to abstinence. Audio recordings and transcripts were reviewed for common themes. Participants reported a range of mild/moderate to intensely destructive problems, including loss of important relationships and profound changes to who they felt they were at their core; e.g., "I didn't realize how dark and mean I was … I was like a different person." Initial abstinence was often facilitated by multiple external forces (e.g., drug testing, child custody issues, prison, relocation), but sustained abstinence was attributed to shifts in thinking and salient realizations about using. The treatment group reported using more and different resources to maintain their abstinence than the no-treatment group. Findings indicate individualized interventions and multiple, simultaneous approaches and resources were essential in reaching stable abstinence. Understanding long-term users' experiences with MA use, addiction, and abstinence can inform strategies for engaging and sustaining MA users in treatment and recovery.

Poly-drug and marijuana use among adults who primarily used methamphetamine.

This study examines health and legal problems associated with use of commonly reported substances and combinations of substances in a sample of adults with long histories of methamphetamine (meth) use. Data are from a 2009-11 eight-year follow-up interview in an intensive natural history study (N = 373). Respondents who had not used illicit substances in the year preceding the follow-up interview (38%) were compared to users of marijuana-only (16%), meth-only (7%), and poly-drug users who used meth + marijuana but not heroin or cocaine (19%), and poly-drug users who used heroin and/or cocaine (20%). Multinomial regression results indicate that compared to drug-abstinent individuals, greater depressive symptomatology was reported for poly-drug users of meth + marijuana (p = .001), and arrest rates were higher for poly-drug users who used heroin/cocaine (p = .006); no differences in health, mental health, or criminal involvement were observed for meth-only users compared to abstinent individuals. Users of marijuana-only and poly-drug users of heroin/cocaine experienced poorer physical health status than those who were abstinent. To further explore this finding, use of marijuana for medical reasons was examined by drug use group. Overall, health and criminal outcomes varied based on type and combination of substances used, and these differences should be considered when planning treatment strategies.

Patterns of treatment utilization and methamphetamine use during first 10 years after methamphetamine initiation.

The study examined joint trajectories of methamphetamine (MA) use and substance abuse treatment utilization and identified differences among pattern groups for a sample of 348 treated for MA use. Results from group-based trajectory modeling showed that treatment utilization during the first 10 years after initiation of MA use could be categorized into three distinctive patterns: about half the MA users have a pattern of low treatment utilization; one-fourth follow a quicker-to-treatment trajectory with higher probability of treatment during the first 5 years of MA use and less treatment in the next 5 years; and one-fourth have a slower-to-treatment trajectory with more treatment during the second half of the 10-year period. Four MA use patterns were identified: consistently low use, moderate, and high use, as well as a decreasing use pattern. Periods of greater likelihood of treatment participation were associated with periods of decreasing or lower frequency of MA use.

Managing port-site bleeding during less invasive coronary artery bypass grafting.

INTRODUCTION: Robotic-assisted coronary artery bypass grafting (r-CABG) requires the placement of ports bluntly through the chest wall. When removed, these ports create bleeding sites that can be difficult to detect and treat. This study evaluated whether a topical hemostatic agent placed locally within these sites helps to reduce bleeding and blood product requirements. METHODS: We retrospectively analyzed outcomes for r-CABG cases where 5 mL of a flowable hemostatic agent was injected locally within all port sites (hemostat group, n = 62) compared with patients whose port sites were untreated (controls, n = 131). Outcomes included chest tube output, red blood cell (RBC) transfusions, length of hospital stay, and the risk of reoperation for bleeding. Analyses were adjusted for risk factors known to influence bleeding and Society of Thoracic Surgeons (STS) risk score as a weighted composite of variables, which controls for patient and clinical variables. RESULTS: The 2 study groups had similar baseline characteristics and underwent the same r-CABG procedure. The hemostat group had significant reductions in RBC transfusion (24.2% versus 40.8% receiving blood; P = .026; 0.44 versus 1.39 U transfused postoperatively, P = .024). After adjustment for bleeding risks (using STS risk score), differences in transfusions remained significant. Reoperation rates for bleeding, length of stay, chest tube drainage, and intraoperative transfusions were not significantly different in the 2 groups. CONCLUSIONS: There was significantly reduced postoperative bleeding and less exposure to blood products in the hemostat group. These findings suggest that undetected bleeding from sites used for port access serves as an underappreciated source of morbidity after r-CABG.

Using facebook to maximize follow-up response rates in a longitudinal study of adults who use methamphetamine.

This study examines the process and effects of using facebook (FB) to locate and re-contact study participants targeted for follow up in a longitudinal study of adult methamphetamine users (N = 649). A follow-up interview was conducted in 2009-11 approximately 8 years after previous study participation. Our paper describes re-contact efforts involving FB, including IRB regulatory issues and the effectiveness of using FB compared to mailings and phone calls. A total of 48 of the 551 surviving non-incarcerated participants who agreed to be contacted for follow up studies were contacted via FB, of whom 11 completed the follow-up interview. Those contacted through FB were more likely to be younger, female, relocated out-of-state, and reported somewhat higher rates of anxiety and cognitive problems compared to those not located on FB. Although participants contacted through FB are likely to differ demographically from those contacted by phone or mail, FB provides a potentially effective means to expand conventional methods of correspondence for contacting hard to reach participants.

Gender differences in physical and mental health outcomes among an aging cohort of individuals with a history of heroin dependence.

BACKGROUND: This paper examines the health status and functioning of an aging cohort of individuals with a history of heroin dependence with a focus on gender differences. METHOD: Study subjects were originally sampled from methadone maintenance clinics in California in the 1970s and completed follow-up interviews in 2005-09. Out of the original study sample (N=914), 343 participants (44.3% female) were interviewed (70.6% of those not deceased). Bivariate analyses examined gender differences in participants' overall health status and physical and mental health problems. Scores on SF-36 scales were compared with general population norms by gender and age, as well as between participants in the study sample who did and did not report past-year drug use. RESULTS: Average age of the study sample was 58.3 (SD=4.9) years for males and 55.0 (SD=4.1) years for females. There were no significant gender differences in past-year drug use (38% of sample) or injection drug use (19%). Women reported significantly more chronic health problems and psychological distress compared with men, and overall poorer health and functioning compared with general population norms. Men under 65 had poorer physical health and social functioning compared with population norms. Men in the study sample reporting past-year substance use had poorer physical functioning, but less bodily pain, than non-users, whereas women with past-year substance use had poorer mental health than other women. CONCLUSION: Individuals with a history of heroin dependence have poorer health and functioning than their counterparts in the general population. At a younger age, women reported poorer overall health status and more chronic health and mental health problems than men. Study findings may inform interventions for this population, particularly related to gender-specific treatment needs.

30-year trajectories of heroin and other drug use among men and women sampled from methadone treatment in California.

BACKGROUND: This study examines 30-year trajectories of heroin and other drug use among men and women who were in methadone maintenance treatment in California in the late 1970s and interviewed in 1978-1981. METHODS: Nearly half (N=428; 46.8%) of the original study sample (N=914) was deceased. Of the remaining 486 subjects, 343 (44.3% female) completed a follow-up interview in 2005-2009 (70.6% of those not deceased). Average age at follow-up was 58.3 (SD=4.9) years for males and 55.0 (SD=4.1) years for females. Longitudinal data was obtained on their drug use, treatment participation, and criminal justice status over the follow-up period. Trajectory group modeling was used to identify distinctive trajectory groups based on monthly averages of heroin and other drug use per year; group differences were examined. RESULTS: Four heroin and five alcohol and other drug (AOD) trajectory groups were identified. A greater proportion of women (60%) were in the "rapid decrease" heroin group (odds of use less than 10% by 10 years following initiation of use) as compared with the other groups. More rapid decrease of heroin use was associated with increases in AOD use, whereas a gradual decrease in heroin use was associated with a gradual decrease in AOD use. More school problems and earlier age at onset of heroin use and first arrest were associated with more persistent heroin use. CONCLUSION: Heroin-use trajectories were linked with changes in AOD use. Childhood antecedents of heroin-use trajectories were identified as well as gender differences.