scholarly journals A General Propensity Score for Signal Identification using Tree-Based Scan Statistics

2021 ◽  
Author(s):  
Shirley V Wang ◽  
Judith C Maro ◽  
Joshua J Gagne ◽  
Elisabetta Patorno ◽  
Sushama Kattinakere ◽  
...  
Keyword(s):  
Propensity Score ◽  
Multiple Testing ◽  
Scan Statistics ◽  
Data Driven ◽  
Signal Identification ◽  
Drug Pair ◽  
Cohort Design ◽  
Specific Outcome ◽  
Level Of Concern ◽  
Residual Confounding

Abstract Tree-based scan statistics (TreeScan) are a data-mining method that adjusts for multiple testing of correlated hypotheses when screening thousands of potential adverse events for signal identification. Simulation has demonstrated the promise of TreeScan with a propensity score (PS) matched cohort design. However, it is unclear which variables to include in a PS for applied signal identification studies to simultaneously adjust for confounding across potential outcomes. We selected 4 drug pairs with well understood safety profiles. For each pair, we evaluated 5 candidate PSs with different combinations of: predefined general covariates (comorbidity, frailty, utilization), empirically-selected (data driven) covariates, and covariates tailored to the drug pair. For each pair, statistical alerting patterns were similar with alternative PSs (≤11 alerts in 7,996 outcomes scanned). Including covariates tailored to exposure did not appreciably impact screening results. Including empirically-selected covariates can provide better proxy coverage for confounders but can also decrease power. Unlike tailored covariates, empirical and predefined general covariates can be applied “out of the box” for signal identification. The choice of PS depends on level of concern about residual confounding versus loss of power. Potential signals should be followed by pharmacoepidemiologic assessment where confounding control is tailored to the specific outcome(s) under investigation.

Inflammatory cytokines and distant recurrence in HER2-negative early breast cancer in the ECOG-ACRIN 5103 trial.

2021 ◽  
Vol 39 (15_suppl) ◽  
pp. 520-520
Author(s):  
Joseph A. Sparano ◽  
Anne M. O'Neill ◽  
Noah Graham ◽  
Donald W. Northfelt ◽  
Chau T. Dang ◽  
...  
Keyword(s):  
Breast Cancer ◽  
Adjuvant Chemotherapy ◽  
Propensity Score ◽  
Propensity Score Matching ◽  
Early Breast Cancer ◽  
Inflammatory Cytokines ◽  
Multiple Testing ◽  
Distant Recurrence ◽  
Stage Ii ◽  
P Value

520 Background: Systemic inflammation may contribute to cancer progression (PMC2803035), including recurrence of early breast cancer (PMC4828958). We hypothesized that inflammatory cytokines and/or chemokines may be associated with distant recurrence (DR). Methods: We performed a case:control study in women with stage II-III Her2-negative breast cancer, all of whom had surgery and adjuvant chemotherapy (doxorubicin/cyclophosphamide, then weekly paclitaxel) with/without bevacizumab, plus endocrine therapy if ER-positive (PMC6118403). Propensity score matching was used to identify approximately 250 case:control pairs (with/without DR). Serum samples obtained before adjuvant chemotherapy were analyzed using the MSD V-Plex Human Cytokine 36-Plex Kit for detection of human cytokines and chemokines involved in the Th1/Th2 pathway, chemotaxis, the Th17 pathway, angiogenesis, and immune system regulation. Conditional logistic regression analysis, with models fit via maximum likelihood, were used to estimate hazard ratios (HRs) and test for associations. Due to skewed nature of cytokines, HRs are reported on log base 2 scale. If adjusted for multiple testing including 36 markers, a p value of < 0.0014 would be required for statistical significance. Results: A total 249 matched pairs (498 patients) were identified. Covariates used for propensity score matching included age, menopausal status (post 54% vs. pre/peri 46%), ER/PR status (one/both pos 64% vs. both neg 36%) tumor size ( < = 2cm 17%, > 2-5cm 67%, > 5cm 16%) nodal status (neg 15%,1-3+ 32%, 4+ 53%), and grade (low 3%, int. 31%, high 66%). The only biomarker associated with a significantly increased DR risk when adjusted for multiple testing was the proinflammatory cytokine IL-6 (HR 1.37, 95% confidence intervals [CI] 1.15, 1.65, p = 0.0006). Others associated with a 2-sided p value < 0.05 included the chemokine MDC(macrophage-derived chemokine/CCL22) (1.90, 95% CI 1.17, 3.1, p = 0.0098), the T helper cell inflammatory cytokine IL-17A (HR 1.36, 95% CI 1.10, 1.67, p = 0.0052), and the cytokine VEGF-A (HR 1.13 for, 95% CI 1.01, 1.27, p = 0.037). There was no statistical interaction between VEGF-A and bevacizumab benefit. The median and mean value for IL-6 was 0.95 and 7.5 pg/ml (range 0.04-2761.24 pg/ml). Conclusions: This analysis provides level 1B evidence indicating that higher levels of the cytokine IL-6 at diagnosis are associated with a significantly higher DR risk in high-risk stage II-III breast cancer despite optimal adjuvant systemic therapy. This provides a foundation for confirmatory validation of IL-6 as a prognostic biomarker, and potentially as a predictive biomarker for testing therapeutic interventions targeting the IL-6/JAK/STAT3 pathway. Supported by NCI U10CA180820,180794,180821; UG1CA189859,232760,233290, 233196; Komen Foundation; Breast Cancer Research Foundation. Clinical trial information: NCT00433511.

Breastfeeding, Cognitive and Noncognitive Development in Early Childhood: A Population Study

2018 ◽  
pp. 111-125
Author(s):  
Lisa-Christine Girard ◽  
Orla Doyle ◽  
Richard E. Tremblay
Keyword(s):  
Propensity Score ◽  
Propensity Score Matching ◽  
Problem Behaviors ◽  
Cognitive Abilities ◽  
Multiple Testing ◽  
Standardized Assessments ◽  
Growing Up ◽  
Expressive Vocabulary ◽  
Maternal Report ◽  
Teacher Reports

BACKGROUND AND OBJECTIVES There is mixed evidence from correlational studies that breastfeeding impacts children’s development. Propensity score matching with large samples can be an effective tool to remove potential bias from observed confounders in correlational studies. The aim of this study was to investigate the impact of breastfeeding on children’s cognitive and noncognitive development at 3 and 5 years of age. METHODS Participants included ∼8000 families from the Growing Up in Ireland longitudinal infant cohort, who were identified from the Child Benefit Register and randomly selected to participate. Parent and teacher reports and standardized assessments were used to collect information on children’s problem behaviors, expressive vocabulary, and cognitive abilities at age 3 and 5 years. Breastfeeding information was collected via maternal report. Propensity score matching was used to compare the average treatment effects on those who were breastfed. RESULTS Before matching, breastfeeding was associated with better development on almost every outcome. After matching and adjustment for multiple testing, only 1 of the 13 outcomes remained statistically significant: children’s hyperactivity (difference score, –0.84; 95% confidence interval, –1.33 to –0.35) at age 3 years for children who were breastfed for at least 6 months. No statistically significant differences were observed postmatching on any outcome at age 5 years. CONCLUSIONS Although 1 positive benefit of breastfeeding was found by using propensity score matching, the effect size was modest in practical terms. No support was found for statistically significant gains at age 5 years, suggesting that the earlier observed benefit from breastfeeding may not be maintained once children enter school.

scholarly journals Data-driven hypothesis weighting increases detection power in genome-scale multiple testing

2016 ◽  
Vol 13 (7) ◽  
pp. 577-580 ◽  
Author(s):  
Nikolaos Ignatiadis ◽  
Bernd Klaus ◽  
Judith B Zaugg ◽  
Wolfgang Huber
Keyword(s):  
Multiple Testing ◽  
Data Driven ◽  
Detection Power ◽  
Genome Scale

scholarly journals Model Misspecification When Excluding Instrumental Variables from PS Models in Settings Where Instruments Modify the Effects of Covariates on Treatment

2014 ◽  
Vol 0 (0) ◽  
Author(s):  
Richard Wyss ◽  
Alan R. Ellis ◽  
Mark Lunt ◽  
M. Alan Brookhart ◽  
Robert J. Glynn ◽  
...  
Keyword(s):  
Propensity Score ◽  
Instrumental Variables ◽  
Model Misspecification ◽  
Interaction Effects ◽  
Full Model ◽  
Multiple Outcomes ◽  
Negative Effects ◽  
Individual Study ◽  
Calendar Time ◽  
Specific Outcome

AbstractTheory and simulations show that variables affecting the outcome only through exposure, known as instrumental variables (IVs), should be excluded from propensity score (PS) models. In pharmacoepidemiologic studies based on automated healthcare databases, researchers will sometimes use a single PS model to control for confounding when evaluating the effect of a treatment on multiple outcomes. Because these “full” models are not constructed with a specific outcome in mind, they will usually contain a large number of IVs for any individual study or outcome. If researchers subsequently decide to evaluate a subset of the outcomes in more detail, they can construct reduced “outcome-specific” models that exclude IVs for the particular study. Accurate estimates of PSs that do not condition on IVs, however, can be compromised when simply excluding instruments from the full PS model. This misspecification may have a negligible impact on effect estimates in many settings, but is likely to be more pronounced for situations where instruments modify the effects of covariates on treatment (instrument–confounder interactions). In studies evaluating drugs during early dissemination, the effects of covariates on treatment are likely modified over calendar time and IV–confounder interaction effects on treatment are likely to exist. In these settings, refitting more flexible PS models after excluding IVs and IV–confounder interactions can work well. The authors propose an alternative method based on the concept of marginalization that can be used to remove the negative effects of controlling for IVs and IV–confounder interactions without having to refit the full PS model. This method fits the full PS model, including IVs and IV–confounder interactions, but marginalizes over values of the instruments. Fitting more flexible PS models after excluding IVs or using the full model to marginalize over IVs can prevent model misspecification along with the negative effects of balancing instruments in certain settings.

scholarly journals Cardiorenal outcomes with sodium/glucose cotransporter-2 inhibitors in patients with type 2 diabetes and low kidney risk: real world evidence

2021 ◽  
Vol 20 (1) ◽  
Author(s):  
Meir Schechter ◽  
Cheli Melzer-Cohen ◽  
Aliza Rozenberg ◽  
Ilan Yanuv ◽  
Gabriel Chodick ◽  
...  
Keyword(s):  
Type 2 Diabetes ◽  
Kidney Disease ◽  
Propensity Score ◽  
Lower Risk ◽  
Urinary Albumin ◽  
Sodium Glucose Cotransporter ◽  
Myocardial Infraction ◽  
Specific Outcome ◽  
End Stage

Abstract Background Randomized controlled trials showed that sodium/glucose cotransporter-2 inhibitors (SGLT2i) protect the heart and kidney in an array of populations with type 2 diabetes (T2D) and increased cardiorenal risk. However, the extent of these benefits also in lower kidney-risk T2D populations needs further investigation. Methods Members of Maccabi Healthcare Systems listed in their T2D registry who initiated new glucose lowering agents (GLA), were divided into SGLT2i initiators and other GLAs (oGLAs). Groups were propensity score-matched by baseline demographic and medical characteristics. Two composite cardiovascular outcomes were defined: all-cause mortality (ACM) or hospitalization for heart failure (hHF); and ACM, myocardial infraction (MI) or stroke. The cardiorenal outcome was: ACM, new end-stage kidney disease (ESKD) or  ≥  40% reduction from baseline estimated glomerular filtration rate (eGFR). Renal-specific outcome was new ESKD or  ≥  40% eGFR reduction. Single components of cardiovascular and kidney outcomes were also assessed. Three subgroup definitions of low baseline kidney-risk were used: eGFR  >  90 ml/min/1.73 m2; urinary albumin below detectable levels; and low risk according to Kidney Disease: Improving Global Outcomes (KDIGO) classification. Analyses were performed utilizing an unadjusted model, and a model adjusted to baseline eGFR and urinary albumin-to-creatinine ratio. Results Between April 1, 2015 and June 30, 2018; 68,187 patients initiated new GLAs — 11,321 SGLT2i initiators and 42,077 oGLAs initiators were eligible. Propensity score-matching yielded two comparable cohorts; each included 9219 participants. Median follow-up was 1.7 years. Compared to oGLAs, SGLT2i initiators had lower incidence of ACM or hHF [HR95%CI  =  0.62(0.51–0.75)]; ACM, MI or stroke [0.67(0.57–0.80)]; the cardiorenal outcome [0.65(0.56–0.76)]; and the renal-specific outcome [0.70(0.57–0.85)]. SGLT2i initiators also had lower risk for ACM, hHF and  ≥  30%,  ≥  40%,  ≥  50%,  ≥  57% eGFR reduction. No difference between groups was observed for MI or stroke. In the low baseline kidney-risk subgroups, SGLT2i initiation was generally associated with lower risk of the cardiovascular and cardiorenal outcomes, driven mainly by lower ACM incidence. Conclusions Our findings in the general population of patients with T2D demonstrates lower risk of cardiorenal outcomes associated with initiation of SGLT2i compared with oGLAs, including specifically in patients with low baseline kidney-risk.

scholarly journals Geographical surveillance of COVID-19: Diagnosed cases and death in the United States

2020 ◽  
Author(s):  
Raid Amin ◽  
Terri Hall ◽  
Jacob Church ◽  
Daniela Schlierf ◽  
Martin Kulldorff
Keyword(s):  
Public Health ◽  
New York ◽  
New England ◽  
Disease Surveillance ◽  
Multiple Testing ◽  
The United States ◽  
Space Time ◽  
Scan Statistics ◽  
Public Health Agencies ◽  
Incidence And Mortality

AbstractBackgroundCOVID-19 is a new coronavirus that has spread from person to person throughout the world. Geographical disease surveillance is a powerful tool to monitor the spread of epidemics and pandemic, providing important information on the location of new hot-spots, assisting public health agencies to implement targeted approaches to minimize mortality.MethodsCounty level data from January 22-April 28 was downloaded from USAfacts.org to create heat maps with ArcMap™ for diagnosed COVID-19 cases and mortality. The data was analyzed using spatial and space-time scan statistics and the SaTScan™ software, to detect geographical cluster with high incidence and mortality, adjusting for multiple testing. Analyses were adjusted for age. While the spatial clusters represent counties with unusually high counts of COVID-19 when averaged over the time period January 22-April 20, the space-time clusters allow us to identify groups of counties in which there exists a significant change over time.ResultsThere were several statistically significant COVID-19 clusters for both incidence and mortality. Top clusters with high rates included the areas in and around New York City, New Orleans and Chicago, but there were also several small rural clusters. Top clusters for a recent surge in incidence and mortality included large parts of the Midwest, the Mid-Atlantic Region, and several smaller areas in and around New York and New England.ConclusionsSpatial and space-time surveillance of COVID-19 can be useful for public health departments in their efforts to minimize mortality from the disease. It can also be applied to smaller regions with more granular data.

scholarly journals Propensity score matched mortality comparisons of peritoneal and in-centre haemodialysis: systematic review and meta-analysis

2020 ◽  
Vol 35 (12) ◽  
pp. 2172-2182 ◽  
Author(s):  
Mohamed E Elsayed ◽  
Adam D Morris ◽  
Xia Li ◽  
Leonard D Browne ◽  
Austin G Stack
Keyword(s):  
Systematic Review ◽  
Propensity Score ◽  
Meta Analysis ◽  
Mortality Risks ◽  
Hazard Ratios ◽  
Patients With Diabetes ◽  
Meta Analyses ◽  
Study Type ◽  
Residual Confounding ◽  
Substantial Degree

Abstract Background Accurate comparisons of haemodialysis (HD) and peritoneal dialysis (PD) survival based on observational studies are difficult due to substantial residual confounding that arises from imbalances between treatments. Propensity score matching (PSM) comparisons confer additional advantages over conventional methods of adjustment by further reducing selection bias between treatments. We conducted a systematic review of studies that compared mortality between in-centre HD with PD using a PSM-based approach. Methods A sensitive search strategy identified all citations in the PubMed, Cochrane and EMBASE databases from inception through November 2018. Pooled PD versus HD mortality hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated through random-effects meta-analysis. A subsequent meta-regression explored factors to account for between-study variation. Results The systematic review yielded 214 citations with 17 cohort studies and 113 578 PSM incident dialysis patients. Cohort periods spanned the period 1993–2014. The pooled HR for PD versus HD was 1.06 (95% CI 0.99–1.14). There was considerable variation by country, however, mortality risks for PD versus HD remained virtually unchanged when stratified by geographical region with HRs of 1.04 (95% CI 0.94–1.15), 1.14 (95% CI 0.99–1.32) and 0.98 (0.87–1.10) for European, Asian and American cohorts, respectively. Subgroup meta-analyses revealed similar risks for patients with diabetes [HR 1.09 (95% CI 0.98–1.21)] and without diabetes [HR 0.99 (95% CI 0.90–1.09)]. Heterogeneity was substantial (I2 = 87%) and was largely accounted for by differences in cohort period, study type and country of origin. Together these factors explained a substantial degree of between-studies variance (R2 = 90.6%). Conclusions This meta-analysis suggests that PD and in-centre HD carry equivalent survival benefits. Reported differences in survival between treatments largely reflect a combination of factors that are unrelated to clinical efficacy.

scholarly journals Renin-angiotensin system inhibitors and susceptibility to COVID-19 in patients with hypertension: a propensity score-matched cohort study in primary care

2020 ◽  
Author(s):  
Shamil Haroon ◽  
Anuradhaa Subramanian ◽  
Jennifer Cooper ◽  
Astha Anand ◽  
Krishna Gokhale ◽  
...  
Keyword(s):  
Propensity Score ◽  
Incidence Rate ◽  
Ace Inhibitors ◽  
Renin Angiotensin System ◽  
Negative Control ◽  
Matched Cohort ◽  
Angiotensin System ◽  
All Cause Mortality ◽  
Residual Confounding ◽  
Ras Inhibitors

Introduction A significant proportion of patients with Coronavirus Disease-19 (COVID-19) have hypertension and are treated with renin-angiotensin system (RAS) inhibitors, namely angiotensin-converting enzyme I inhibitors (ACE inhibitors) or angiotensin II type-1 receptor blockers (ARBs). These medications have been postulated to influence susceptibility to Severe Acute Respiratory Syndrome Coronavirus-2 (SARS-CoV-2). The objective of this study was to assess a possible association between prescription of RAS inhibitors and the incidence of COVID-19 and all-cause mortality. Methods We conducted a propensity-score matched cohort study to assess the incidence of COVID-19 among patients with hypertension who were prescribed ACE inhibitors or ARBs compared to patients treated with calcium channel blockers (CCBs) in a large UK-based primary care database (The Health Improvement Network). We estimated crude incidence rates for confirmed/suspected COVID-19 among those prescribed ACE inhibitors, ARBs and CCBs. We used a Cox proportional hazards model to produce adjusted hazard ratios for COVID-19 comparing patients prescribed ACE inhibitors or ARBs to those prescribed CCBs. We further assessed all-cause mortality as a secondary outcome and a composite of accidents, trauma or fractures as a negative control outcome to assess for residual confounding. Results In the propensity score matched analysis, 83 of 18,895 users (0.44%) of ACE inhibitors developed COVID-19 over 8,923 person-years, an incidence rate of 9.3 per 1000 person-years. 85 of 18,895 (0.45%) users of CCBs developed COVID-19 over 8,932 person-years, an incidence rate of 9.5 per 1000 person-years. The adjusted hazard ratio for suspected/confirmed COVID-19 for users of ACE inhibitors compared to CCBs was 0.92 (95% CI 0.68 to 1.26). 79 out of 10,623 users (0.74%) of ARBs developed COVID-19 over 5010 person-years, an incidence rate of 15.8 per 1000 person-years, compared to 11.6 per 1000 person-years among users of CCBs. The adjusted hazard ratio for suspected/confirmed COVID-19 for users of ARBs compared to CCBs was 1.38 (95% CI 0.98 to 1.95). There were no significant associations between use of ACE inhibitors or ARBs and all-cause mortality, compared to use of CCBs. We found no evidence of significant residual confounding with the negative control analysis. Conclusion Current use of ACE inhibitors was not associated with the risk of suspected or confirmed COVID-19 whereas use of ARBs was associated with a statistically non-significant 38% relative increase in risk compared to use of CCBs. However, no significant associations were observed between prescription of either ACE inhibitors or ARBs and all-cause mortality during the peak of the pandemic.

Sign in / Sign up

Export Citation Format

Share Document