Simultaneous prediction of multiple outcomes using revised stacking algorithms.

MOTIVATION: HIV is difficult to treat because its virus mutates at a high rate and mutated viruses easily develop resistance to existing drugs. If the relationships between mutations and drug resistances can be determined from historical data, patients can be provided personalized treatment according to their own mutation information. The HIV Drug Resistance Database was built to investigate the relationships. Our goal is to build a model using data in this database, which simultaneously predicts the resistance of multiple drugs using mutation information from sequences of viruses for any new patient. RESULTS: We propose two variations of a stacking algorithm which borrow information among multiple prediction tasks to improve multivariate prediction performance. The most attractive feature of our proposed methods is the flexibility with which complex multivariate prediction models can be constructed using any univariate prediction models. Using cross-validation studies, we show that our proposed methods outperform other popular multivariate prediction methods. AVAILABILITY AND IMPLEMENTATION: An R package is being developed. In the meantime, R code can be requested by email. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

White Matter Hyperintensity Volume Influences Symptoms in Patients Presenting With Minor Neurological Deficits.

Background and Purpose- Acute minor neurological deficits are a common complaint in the emergency department and differentiation of transient ischemic attack/minor stroke from a stroke mimic is difficult. We sought to assess the ability of white matter hyperintensity (WMH) volume to aid the diagnosis in such patients. Methods- This is a post hoc analysis of the previously published SpecTRA study (Spectrometry in TIA Rapid Assessment) of adult patients that presented to the emergency department with acute minor neurological deficits between December 2013 and March 2017. WMH volumes were measured if fluid-attenuated inversion recovery imaging was available. Outcomes of interest were final diagnosis, symptoms at presentation, and 90-day stroke recurrence. Results- WMH volume was available for 1485 patients. Median age was 70 years (interquartile range, 59-80), and 46.7% were female. Mean WMH volume was higher in transient ischemic attack/minor strokes compared with stroke mimics (1.71 ln mL [95% CI, 1.63-1.79 ln mL] versus 1.15 ln mL [95% CI, 1.02-1.27 ln mL], P<0.001). In multivariable-adjusted logistic regression analysis, WMH volume was not associated with final diagnosis. However, the combination of both diffusion-weighted imaging positivity and high WMH volume led to lower odds of focal symptoms at presentation (P=0.035). Conclusions- The combination of diffusion-weighted imaging positivity and high WMH volume was associated with lower odds of focal symptoms at presentation in patients seen with minor neurological deficits in the emergency department. This suggests that WMH volume might be an important consideration and the absence of focal symptoms at presentation should not discourage clinicians from further investigating patients with suspected cerebral ischemia.

Systolic blood pressure as a predictor of transient ischemic attack/minor stroke in emergency department patients under age 80: a prospective cohort study.

BACKGROUND: Elevated blood pressure (BP) at emergency department (ED) presentation and advancing age have been associated with risk of ischemic stroke; however, the relationship between BP, age, and transient ischemic attack/minor stroke (TIA/MS) is not clear. METHODS: A multi-site, prospective, observational study of 1084 ED patients screened for suspected TIA/MS (symptom onset < 24 h, NIHSS< 4) between December 2013 and April 2016. Systolic and diastolic BP measurements (SBP, DBP) were taken at ED presentation. Final diagnosis was consensus adjudication by stroke neurologists; patients were diagnosed as either TIA/MS or stroke-mimic (non-cerebrovascular conditions). Conditional inference trees were used to define age cut-points for predicting binary diagnosis (TIA/MS or stroke-mimic). Logistic regression models were used to estimate the effect of BP, age, sex, and the age-BP interaction on predicting TIA/MS diagnosis. RESULTS: Over a 28-month period, 768 (71%) patients were diagnosed with TIA/MS: these patients were older (mean 71.6 years) and more likely to be male (58%) than stroke-mimics (61.4 years, 41%; each p < 0.001). TIA/MS patients had higher SBP than stroke-mimics (p < 0.001). DBP did not differ between the two groups (p = 0.191). SBP was predictive of TIA/MS diagnosis in younger patients, after accounting for age and sex; an increase of 10 mmHg systolic increased the odds of TIA/MS 18% (odds ratio [OR] 1.18, 95% CI 1.00-1.39) in patients < 60 years, and 23% (OR 1.23, 95% CI 11.12-1.35) in those 60-79 years, while not affecting the odds of TIA/MS in patients ≥80 years (OR 0.99, 95% CI 0.89-1.07). CONCLUSIONS: Raised SBP in patients younger than 80 with suspected TIA/MS may be a useful clinical indicator upon initial presentation to help increase clinicians' suspicion of TIA/MS. TRIAL REGISTRATION: ClinicalTrials.gov NCT03050099 (10-Feb-2017) and NCT03070067 (3-Mar-2017). Retrospectively registered.

Verification of a proteomic biomarker panel to diagnose minor stroke and transient ischaemic attack: phase 1 of SpecTRA, a large scale translational study.

OBJECTIVE: To derive a plasma biomarker protein panel from a list of 141 candidate proteins which can differentiate transient ischaemic attack (TIA)/minor stroke from non-cerebrovascular (mimic) conditions in emergency department (ED) settings. DESIGN: Prospective clinical study (#NCT03050099) with up to three timed blood draws no more than 36 h following symptom onset. Plasma samples analysed by multiple reaction monitoring-mass spectrometry (MRM-MS). PARTICIPANTS: Totally 545 participants suspected of TIA enrolled in the EDs of two urban medical centres. OUTCOMES: 90-day, neurologist-adjudicated diagnosis of TIA informed by clinical and radiological investigations. RESULTS: The final protein panel consists of 16 proteins whose patterns show differential abundance between TIA and mimic patients. Nine of the proteins were significant univariate predictors of TIA [odds ratio (95% confidence interval)]: L-selectin [0.726 (0.596-0.883)]; Insulin-like growth factor-binding protein 3 [0.727 (0.594-0.889)]; Coagulation factor X [0.740 (0.603-0.908)]; Serum paraoxonase/lactonase 3 [0.763 (0.630-0.924)]; Thrombospondin-1 [1.313 (1.081-1.595)]; Hyaluronan-binding protein 2 [0.776 (0.637-0.945)]; Heparin cofactor 2 [0.775 (0.634-0.947)]; Apolipoprotein B-100 [1.249 (1.037-1.503)]; and von Willebrand factor [1.256 (1.034-1.527)]. The scientific plausibility of the panel proteins is discussed. CONCLUSIONS: Our panel has the potential to assist ED physicians in distinguishing TIA from mimic patients.

Validation of a proteomic biomarker panel to diagnose minor-stroke and transient ischaemic attack: phase 2 of SpecTRA, a large scale translational study.

OBJECTIVE: To validate our previously developed 16 plasma-protein biomarker panel to differentiate between transient ischaemic attack (TIA) and non-cerebrovascular emergency department (ED) patients. METHOD: Two consecutive cohorts of ED patients prospectively enrolled at two urban medical centers into the second phase of SpecTRA study (training, cohort 2A, n = 575; test, cohort 2B, n = 528). Plasma samples were analyzed using liquid chromatography/multiple reaction monitoring-mass spectrometry. Logistic regression models which fit cohort 2A were validated on cohort 2B. RESULTS: Three of the panel proteins failed quality control and were removed from the panel. During validation, panel models did not outperform a simple motor/speech (M/S) deficit variable. Post-hoc analyses suggested the measured behaviour of L-selectin and coagulation factor V contributed to poor model performance. Removal of these proteins increased the external performance of a model containing the panel and the M/S variable. CONCLUSIONS: Univariate analyses suggest insulin-like growth factor-binding protein 3 and serum paraoxonase/lactonase 3 are reliable and reproducible biomarkers for TIA status. Logistic regression models indicated L-selectin, apolipoprotein B-100, coagulation factor IX, and thrombospondin-1 to be significant multivariate predictors of TIA. We discuss multivariate feature subset analyses as an exploratory technique to better understand a panel's full predictive potential.

Reducing time-to-unit among patients referred to an outpatient stroke assessment unit with a novel triage process: a prospective cohort study.

BACKGROUND: To evaluate the performance of a novel triage system for Transient Ischemic Attack (TIA) units built upon an existent clinical prediction rule (CPR) to reduce time to unit arrival, relative to the time of symptom onset, for true TIA and minor stroke patients. Differentiating between true and false TIA/minor stroke cases (mimics) is necessary for effective triage as medical intervention for true TIA/minor stroke is time-sensitive and TIA unit spots are a finite resource. METHODS: Prospective cohort study design utilizing patient referral data and TIA unit arrival times from a regional fast-track TIA unit on Vancouver Island, Canada, accepting referrals from emergency departments (ED) and general practice (GP). Historical referral cohort (N = 2942) from May 2013-Oct 2014 was triaged using the ABCD2 score; prospective referral cohort (N = 2929) from Nov 2014-Apr 2016 was triaged using the novel system. A retrospective survival curve analysis, censored at 28 days to unit arrival, was used to compare days to unit arrival from event date between cohort patients matched by low (0-3), moderate (4-5) and high (6-7) ABCD2 scores. RESULTS: Survival curve analysis indicated that using the novel triage system, prospectively referred TIA/minor stroke patients with low and moderate ABCD2 scores arrived at the unit 2 and 1 day earlier than matched historical patients, respectively. CONCLUSIONS: The novel triage process is associated with a reduction in time to unit arrival from symptom onset for referred true TIA/minor stroke patients with low and moderate ABCD2 scores.

Comparison of transcriptome responses of the liver, tail fin, and olfactory epithelium of Rana [Lithobates] catesbeiana tadpoles disrupted by thyroid hormones and estrogen.

Thyroid hormones (THs) are important developmental regulators in vertebrates, including during the metamorphosis of a tadpole into a frog. Metamorphosis is a post-embryonic developmental period initiated by TH production in the tadpole thyroid gland. The two main bioactive forms of TH are L-thyroxine (T4) and 3,5,3'-triiodothyronine (T3); these hormones have overlapping but distinct mechanisms of action. Premetamorphic tadpoles are highly responsive to TH and can be induced to metamorphose through exogenous TH exposure, making them an important model for both the study of vertebrate TH signaling and endocrine disrupting chemicals (EDCs). It is important to differentiate TH-mediated responses from estrogenic responses in premetamorphic tadpoles when assessing dysregulation by EDCs as crosstalk between the two endocrine systems is well-documented. Herein, we compare the RNA-sequencing-derived transcriptomic profiles of three TH-responsive tissues (liver, olfactory epithelium, and tail fin) in premetamorphic bullfrog (Rana [Lithobates] catesbeiana) tadpoles exposed to T3, T4, and estradiol (E2). These profiles were generated using the latest available genome assembly for the species. The data indicate that there is a clear distinction, and little overlap, between the transcriptomic responses elicited by E2 and the THs. In contrast, within the THs, the T3- and T4-induced transcriptomic profiles generally show considerable overlap; however, the degree of overlap is highly tissue-dependent, illustrating the importance of distinguishing the two THs and the affected signaling pathways within the target tissue type when evaluating hormone active agents. The data herein also show that E2 and TH treatment can uniquely induce significant changes in expression of their respective "classic" bioindicator transcripts vtg (E2) and thra, thrb, and thibz (THs). However, care must be taken in the interpretation of increased vep or esr1 transcripts as a change in transcript levels can be induced by THs rather than solely E2.

Predictors of local recurrence in a population-based cohort of women with ductal carcinoma in situ treated with breast conserving surgery alone.

BACKGROUND: To identify prognostic indicators of local recurrence (LR) in patients with ductal carcinoma in situ (DCIS) of the breast treated with breast conserving surgery (BCS) alone. METHODS: A retrospective study was conducted of all women with pure DCIS, diagnosed 1985-1999, referred for tertiary oncologic opinion in British Columbia, treated with BCS without adjuvant radiotherapy. Kaplan-Meier local control (LC) and breast cancer specific survival (BCSS) estimates for the entire group were plotted. Stratified analyses identified subgroups with high Kaplan-Meier 10-year LR. Cox multivariate modeling was used to assess predictors of LR. Kaplan-Meier BCSS rates were compared between two cohorts: those who experienced LR and those who did not have LR. RESULTS: A total of 460 women comprised the study cohort. Median follow-up was 9.4 years. The 15-year LC and BCSS rates were 82% and 97%, respectively. Stratified analyses of LR identified comedo histology, high nuclear grade, tumor size >4 cm or indeterminate size, and positive margins to be associated with significantly higher LR risk, with 10-year LR risks approximating 15-30%. The 10-year BCSS rates for the LR group were 94% compared with 99% for the NoLR group. On Cox regression modeling, high nuclear grade, the presence of comedocarcinoma, and positive margins were significant factors for higher risk of LR. CONCLUSIONS: Women with DCIS treated with BCS alone had higher LR risk, and those with a LR were more likely to die of breast cancer. Optimal local treatment is mandatory to minimize the risk of breast cancer death for women with this curable disease.

Validation of a multivariate clinical prediction model for the diagnosis of mild stroke/transient ischemic attack in physician first-contact patient settings.

We validate our previously developed (DOI: 10.1101/089227) clinical prediction rule for diagnosing transient ischemic attack on the basis of presenting clinical symptoms and compare its performance with the ABCD2 score in first-contact patient settings. Two independent and prospectively collected patient validation cohorts were used: (a) referral cohort-prospectively referred emergency department and general practitioner patients (N = 877); and (b) SpecTRA cohort-participants recruited as part of the SpecTRA biomarker project (N = 545). Outcome measure consisted of imaging-confirmed clinical diagnosis of mild stroke/transient ischemic attack. Results showed that our clinical prediction rule demonstrated significantly higher accuracy than the ABCD2 score for both the referral cohort (70.5% vs 59.0%; p < 0.001) and SpecTRA cohort (72.8% vs 68.3%; p = 0.028). We discuss the potential of our clinical prediction rule to replace the use of the ABCD2 score in the triage of transient ischemic attack clinic referrals.

Retrospective evaluation of a clinical decision support tool for effective computed tomography angiography utilization in urgent brain imaging of suspected TIA/minor stroke in the emergency department.

OBJECTIVES: The Canadian Stroke Best Practice Recommendations suggests that patients suspected of transient ischemic attack (TIA)/minor stroke receive urgent brain imaging, preferably computed tomography angiography (CTA). Yet, high requisition rates for non-cerebrovascular patients overburden limited radiological resources, putting patients at risk. We hypothesize that our clinical decision support tool (CDST) developed for risk stratification of TIA in the emergency department (ED), and which incorporates Canadian guidelines, could improve CTA utilization. METHODS: Retrospective study design with clinical information gathered from ED patient referrals to an outpatient TIA unit in Victoria, BC, from 2015-2016. Actual CTA orders by ED and TIA unit staff were compared to hypothetical CTA ordering if our CDST had been used in the ED upon patient arrival. RESULTS: For 1,679 referrals, clinicians ordered 954 CTAs. Our CDST would have ordered a total of 977 CTAs for these patients. Overall, this would have increased the number of imaged-TIA patients by 89 (10.1%) while imaging 98 (16.1%) fewer non-cerebrovascular patients over the 2-year period. Our CDST would have ordered CTA for 18 (78.3%) of the recurrent stroke patients in the sample. CONCLUSIONS: Our CDST could enhance CTA utilization in the ED for suspected TIA patients, and facilitate guideline-based stroke care. Use of our CDST would increase the number of TIA patients receiving CTA before ED discharge (rather than later at TIA units) and reduce the burden of imaging stroke mimics in radiological departments.

Sex Differences in Presentation and Outcome After an Acute Transient or Minor Neurologic Event.

IMPORTANCE: Sex differences have been described in the presentation, care, and outcomes among people with acute ischemic strokes, but these differences are less understood for minor ischemic cerebrovascular events. The present study hypothesized that, compared with men, women are more likely to report nonfocal symptoms and to receive a stroke mimic diagnosis. OBJECTIVE: To evaluate sex differences in the symptoms, diagnoses, and outcomes of patients with acute transient or minor neurologic events. DESIGN, SETTING, AND PARTICIPANTS: This prospective cohort study of patients with minor ischemic cerebrovascular events or stroke mimics enrolled at multicenter academic emergency departments in Canada between December 2013 and March 2017 and followed up for 90 days is a substudy of SpecTRA (Spectrometry for Transient Ischemic Attack Rapid Assessment). In total, 1729 consecutive consenting patients with acute transient or minor neurologic symptoms were referred for neurologic evaluation; 66 patients were excluded for protocol violation (n = 46) or diagnosis of transient global amnesia (n = 20). EXPOSURES: The main exposure was female or male sex. MAIN OUTCOMES AND MEASURES: The main outcome was the clinical diagnosis (cerebral ischemia vs stroke mimic). Secondary outcomes were 90-day stroke recurrence and 90-day composite outcome of stroke, myocardial infarction, or death. The association between presenting symptoms (focal vs nonfocal) and clinical diagnosis was also assessed. Research hypotheses were formulated after data collection. RESULTS: Of 1648 patients included, 770 (46.7%) were women, the median (interquartile range) age was 70 (59-80) years, 1509 patients (91.6%) underwent brain magnetic resonance imaging, and 1582 patients (96.0%) completed the 90-day follow-up. Women (522 of 770 [67.8%]) were less likely than men (674 of 878 [76.8%]) to receive a diagnosis of cerebral ischemia (adjusted risk ratio [aRR], 0.88; 95% CI, 0.82-0.95), but the 90-day stroke recurrence outcome (aRR, 0.90; 95% CI, 0.48-1.66) and 90-day composite outcome (aRR, 0.86; 95% CI, 0.54-1.32) were similar for men and women. No significant sex differences were found for presenting symptoms. Compared with patients with no focal neurologic symptoms, those with focal and nonfocal symptoms were more likely to receive a diagnosis of cerebral ischemia (aRR, 1.28; 95% CI, 1.15-1.39), but the risk was highest among patients with focal symptoms only (aRR, 1.45; 95% CI, 1.34-1.53). Sex did not modify these associations. CONCLUSIONS AND RELEVANCE: The results of the present study suggest that, despite similar presenting symptoms among men and women, women may be more likely to receive a diagnosis of stroke mimic, but they may not have a lower risk than men of subsequent vascular events, indicating potentially missed opportunities for prevention of vascular events among women.

In the absence of cancer registry data, is it sensible to assess incidence using hospital separation records?

BACKGROUND: Within the health literature, a major goal is to understand distribution of service utilisation by social location. Given equivalent access, differential incidence leads to an expectation of differential service utilisation. Cancer incidence is differentially distributed with respect to socioeconomic status. However, not all jurisdictions have incidence registries, and not all registries allow linkage with utilisation records. The British Columbia Linked Health Data resource allows such linkage. Consequently, we examine whether, in the absence of registry data, first hospitalisation can act as a proxy measure for incidence, and therefore as a measure of need for service. METHODS: Data are drawn from the British Columbia Linked Health Data resource, and represent 100% of Vancouver Island Health Authority cancer registry and hospital records, 1990-1999. Hospital separations (discharges) with principal diagnosis ICD-9 codes 140-208 are included, as are registry records with ICDO-2 codes C00-C97. Non-melanoma skin cancer (173/C44) is excluded. Lung, colorectal, female breast, and prostate cancers are examined separately. We compare registry and hospital annual counts and age-sex distributions, and whether the same individuals are represented in both datasets. Sensitivity, specificity and predictive values are calculated, as is the kappa statistic for agreement. The registry is designated the gold standard. RESULTS: For all cancers combined, first hospitalisation counts consistently overestimate registry incidence counts. From 1995-1999, there is no significant difference between registry and hospital counts for lung and colorectal cancer (p = 0.42 and p = 0.56, respectively). Age-sex distribution does not differ for colorectal cancer. Ten-year period sensitivity ranges from 73.0% for prostate cancer to 84.2% for colorectal cancer; ten-year positive predictive values range from 89.5% for female breast cancer to 79.35% for prostate cancer. Kappa values are consistently high. CONCLUSION: Claims and registry databases overlap with an appreciable proportion of the same individuals. First hospital separation may be considered a proxy for incidence with reference to colorectal cancer since 1995. However, to examine equity across cancer health services utilisation, it is optimal to have access to both hospital and registry files.

Effect of radiotherapy boost and hypofractionation on outcomes in ductal carcinoma in situ.

BACKGROUND: Boost radiotherapy (RT) improves outcomes for patients with invasive breast cancer, but whether this is applicable to patients with pure ductal carcinoma in situ (DCIS) is unclear. This study examined outcomes from whole breast RT, with or without a boost, and the impact of different dose-fractionation schedules in a population-based cohort of women with pure DCIS treated with breast-conserving surgery (BCS). METHODS: Data was analyzed for 957 subjects diagnosed between 1985 and 1999. RT use was analyzed over time. Ten-year Kaplan-Meier local control (LC), breast cancer specific survival (BCSS), and overall survival (OS) were compared using the log-rank test. Cox regression modeling of LC was performed. RESULTS: Median follow-up was 9.3 years. Of the patient cohort 475 (50%) had no RT (NoRT) after BCS, 338 (35%) had RT without a partial breast boost (RTNoB), and 144 (15%) had RT with boost (RT + B). Subjects with risk factors of local recurrence were more likely to receive RT. Subjects receiving adjuvant RT had a trend toward improved LC (15-year LC: NoRT 87%; RTNoB 94%; RT + B 91%; P = .065). Multivariable analysis showed that RT with or without a boost was significantly associated with improved LC (HR, 0.29 and 0.38, respectively, compared with NoRT, P = .025), with no difference associated with a boost or different dose-fractionation schedules. CONCLUSIONS: Adjuvant RT improves local control in patients with DCIS treated with BCS. Hypofractionation is as effective as standard fractionation schedules. Boost RT was not associated with improved LC compared with whole breast RT alone. Cancer 2011. © 2010 American Cancer Society.

Intervals longer than 20 weeks from breast-conserving surgery to radiation therapy are associated with inferior outcome for women with early-stage breast cancer who are not receiving chemotherapy.

PURPOSE: To determine the interval from breast-conserving surgery (BCS) to radiation therapy (RT) that affects local control or survival. PATIENTS AND METHODS: The 10-year Kaplan-Meier (KM) local recurrence-free survival (LRFS), distant recurrence-free survival (DRFS), and breast cancer-specific survival (BCSS) were computed for 6,428 women who had T1 to 2, N0 to 1, M0 breast cancer that was diagnosed in British Columbia between 1989 and 2003, and who were treated with BCS and RT without chemotherapy. Intervals from BCS to RT were grouped by weeks as follows: < or = 4 (n = 83), greater than 4 to 8 (n = 2,288; reference group); greater than 8 to 12 (n = 2,606); greater than 12 to 16 (n = 961); greater than 16 to 20 (n = 358); and greater than 20 weeks (n = 132). Cox proportional hazards models and matching were used to control for confounding variables. RESULTS: The median follow-up time was 7.5 years. The 10-year KM outcomes were as follows: LRFS, 95.4%; DRFS, 90.5%; and BCSS, 92.5%. Compared with the greater than 4 to 8 weeks group, hazard ratios (HR) were not significantly different for any outcome among patients who were treated up to 20 weeks after BCS. However, LRFS (hazard ratio [HR], 2.00; P = .15), DRFS (HR, 1.86; P = .02) and BCSS (HR, 2.15; P = .009) were inferior for women with BCS-to-RT intervals greater than 20 weeks compared with those greater than 4 to 8 weeks. The matched analysis yielded similar results. CONCLUSION: Outcomes were statistically similar for BCS-to-RT intervals up to 20 weeks, but they were inferior for intervals beyond 20 weeks. Time can be reasonably allowed for the breast to heal and for patients to consider treatment options, but RT should start within 20 weeks of BCS.