Pairwise Odds Ratios Research Articles

ORTH.Ord: An R package for analyzing correlated ordinal outcomes using alternating logistic regressions with orthogonalized residuals

Background and objectivesMarginal models with generalized estimating equations (GEE) are usually recommended for analyzing correlated ordinal outcomes which are commonly seen in a longitudinal study or clustered randomized trial (CRT). Within-cluster association is often of interest in longitudinal studies or CRTs, and can be estimated with paired estimating equations. However, the estimators for within-cluster association parameters and variances may be subject to finite-sample biases when the number of clusters is small. The objective of this article is to introduce a newly developed R package ORTH.Ord for analyzing correlated ordinal outcomes using GEE models with finite-sample bias corrections. MethodsThe R package ORTH.Ord implements a modified version of alternating logistic regressions with estimation based on orthogonalized residuals (ORTH), which use paired estimating equations to jointly estimate parameters in marginal mean and association models. The within-cluster association between ordinal responses is modeled by global pairwise odds ratios (POR). The R package also provides a finite-sample bias correction to POR parameter estimates based on matrix multiplicative adjusted orthogonalized residuals (MMORTH) for correcting estimating equations, and bias-corrected sandwich estimators with different options for covariance estimation. ResultsA simulation study shows that MMORTH provides less biased global POR estimates and coverage of their 95% confidence intervals closer to the nominal level than uncorrected ORTH. An analysis of patient-reported outcomes from an orthognathic surgery clinical trial illustrates features of ORTH.Ord. ConclusionsThis article provides an overview of the ORTH method with bias-correction on both estimating equations and sandwich estimators for analyzing correlated ordinal data, describes the features of the ORTH.Ord R package, evaluates the performance of the package using a simulation study, and finally illustrates its application in an analysis of a clinical trial.

Transgender Adults From Minoritized Ethnoracial Groups in the U.S. Report Greater Subjective Cognitive Decline.

Investigate subjective cognitive decline (SCD) among 4 study groups consisting of cisgender and transgender adults who are from minoritized ethnoracial groups (i.e., minoritized ethnoracial transgender, minoritized ethnoracial cisgender) and White cisgender and transgender adults aged 45+ (i.e., White transgender, White cisgender) to determine the odds of SCD by group and to test for group differences. Data from the 2015-2020 Behavioral Risk Factor Surveillance System were used in a modified case-control approach to perform an intercategorical intersectional study. Each transgender participant was matched to 2 cisgender men and 2 cisgender women, on state, ethnoracial identity, and age. Multivariable logistic regressions modeled SCD odds by group and post hoc contrasts estimated pairwise odds ratios comparing the SCD odds for each combination of groups. SCD prevalence was highest among minoritized ethnoracial transgender (21.6%), followed by White transgender (15.0%), minoritized ethnoracial cisgender (12.0%), and White cisgender (9.0%). After accounting for age, education, and survey year, the odds of SCD were higher in minoritized ethnoracial transgender when compared to White cisgender (adjusted odds ratio [aOR] = 2.51, 95% confidence interval [CI]: 1.59-3.96) and minoritized ethnoracial cisgender (aOR = 1.89, 95% CI: 1.16-3.09). The odds of SCD were higher in White transgender compared to White cisgender (aOR = 1.66, 95% CI: 1.20-2.30). When considering the intersection of transgender and ethnoracial identities, we found that transgender adults from minoritized ethnoracial groups reported higher odds of SCD when compared to cisgender adults from minoritized ethnoracial groups. Additional studies are needed to understand the relationship between racialized and gendered inequities in cognitive impairment and how specific mechanisms of systemic transphobia and racism may contribute to this inequity.

Evaluation of the Effectiveness of Chemical Control for Chagas Disease Vectors in Loja Province, Ecuador.

The objective of this study was to evaluate the effectiveness of selective and community-wide house insecticide spraying in controlling triatomines in the subtropical areas of Loja Province, Ecuador. We designed a quasi-experimental pre–post-test without a control group to compare entomological levels before and after spraying. The baseline study was conducted in 2008. Second, third, and fourth visits were conducted in 2010, 2011, and 2012 in three rural communities. Out of the 130 domestic units (DU) visited, 41 domestic units were examined in each of the four visits. Selective and community-wide insecticide interventions included spraying with 5% deltamethrin at 25 mg/m2 active ingredient. At each visit, a questionnaire was administered to identify the characteristics of households, and DUs were searched for triatomine bugs. In addition, parasitological analysis was carried out in life triatomines. One and two rounds of selective insecticide spraying decreased the probability of infestation by 62% (pairwise odds ratios [POR] 0.38, 95% confidence interval [CI] 0.17–0.89, p = 0.024) and 51% (POR 0.49, 95% CI 0.23–1.01, p = 0.054), respectively. A similar effect was observed after one round of community-wide insecticide application in Chaquizhca and Guara (POR 0.55, CI 0.24–1.25, p = 0.155) and Bellamaria (POR 0.62, CI 0.22–1.79, p = 0.379); however, it was not statistically significant. Trypanosoma cruzi infection in triatomines (n = 483) increased overtime, from 2008 (42.9% and 8.5% for Rhodnius ecuadoriensis and Panstrongylus chinai, respectively) to 2012 (79.5% and 100%). Neither of the two spraying methodologies was effective for triatomine control in this area and our results point to a high likelihood of reinfestation after insecticide application. This underscores the importance of the implementation of physical barriers that prevent invasion and colonization of triatomines in households, such as home improvement initiatives, accompanied by a concerted effort to address the underlying socioeconomic issues that keep this population at risk of developing Chagas disease.

Efficacy and safety of biologics and small molecule drugs for patients with moderate-to-severe ulcerative colitis: a systematic review and network meta-analysis

There is a growing armamentarium for the treatment of moderate-to-severe ulcerative colitis. We aimed to compare the relative efficacy and safety of biologics and small molecule drugs for the treatment of patients with moderate-to-severe ulcerative colitis. In this systematic review and network meta-analysis, we searched MEDLINE, Embase, and the Cochrane Central Register of Controlled Trials without language restrictions for articles published between Jan 1, 1990, and July 1, 2021. Major congresses' databases from Jan 1, 2018, to July 3, 2021, were reviewed manually. Phase 3, placebo-controlled or head-to-head randomised controlled trials (RCTs) assessing the efficacy and safety of biologics or small molecule drugs as induction or maintenance therapies for patients with moderate-to-severe ulcerative colitis were included. Phase 2 RCTs were excluded because of their small sample sizes and inclusion of doses not further explored in phase 3 RCTs. Summary data from intention-to-treat analyses were extracted from included reports by JSL and PAO. The primary outcome was the induction of clinical remission. A network meta-analysis was done under the frequentist framework, obtaining pairwise odds ratios (ORs) and 95% CIs. The surface under the cumulative ranking (SUCRA) was used to rank the included agents for each outcome. Higher SUCRA scores correlate with better efficacy, whereas lower SUCRA scores correlate with better safety. Maintenance data on efficacy for treat-straight-through and randomised responder trials are also presented. This study is registered with PROSPERO, CRD42021225329. Our search yielded 5904 results, from which 29 studies (four being head-to-head RCTs) fulfilled our inclusion criteria and were included. Of these, 23 studies assessed induction therapy with either a biologic or small molecule drug, comprising 10 061 patients with ulcerative colitis. A risk of bias assessment showed a low risk of bias for most of the included studies. Upadacitinib was significantly superior to all other interventions for the induction of clinical remission (infliximab [OR 2·70, 95% CI 1·18-6·20], adalimumab [4·64, 2·47-8·71], golimumab [3·00, 1·32-6·82], vedolizumab [3·56, 1·84-6·91], ustekinumab [2·92, 1·31-6·51], etrolizumab [4·91, 2·59-9·31], tofacitinib [2·84, 1·28-6·31], filgotinib 100 mg [6·15, 2·98-12·72], filgotinib 200 mg [4·49, 2·18-9·24], and ozanimod (2·70, 1·18-6·20), and ranked highest for the induction of clinical remission (SUCRA 0·996). No differences between active interventions were observed when assessing adverse events and serious adverse events. Vedolizumab ranked lowest for both adverse events (SUCRA 0·184) and serious adverse events (0·139), whereas upadacitinib ranked highest for adverse events (0·843) and ozanimod ranked highest for serious adverse events (0·831). Upadacitinib was the best performing agent for the induction of clinical remission (the primary outcome) but the worst performing agent in terms of adverse events in patients with moderate-to-severe ulcerative colitis. Vedolizumab was the best performing agent for safety outcomes. With the paucity of direct comparisons in the published literature, our results might help clinicians to position drugs in treatment algorithms. None.

Non-steroidal anti-inflammatory drugs, intravenous fluids, pancreatic stents, or their combinations for the prevention of post-endoscopic retrograde cholangiopancreatography pancreatitis: a systematic review and network meta-analysis

Non-steroidal anti-inflammatory drugs (NSAIDs), intravenous fluid, pancreatic stents, or combinations of these have been evaluated in randomised controlled trials (RCTs) for the prevention of post-endoscopic retrograde cholangiopancreatography (ERCP) pancreatitis, but the comparative efficacy of these treatments remains unclear. Our aim was to do an exploratory network meta-analysis of previous RCTs to systematically compare the direct and indirect evidence and rank NSAIDs, intravenous fluids, pancreatic stents, or combinations of these to determine the most efficacious method of prophylaxis for post-ERCP pancreatitis. We searched PubMed, Embase, and the Cochrane Central Register from inception to Nov 15, 2020, for full-text RCTs that evaluated the efficacy of NSAIDs, pancreatic stents, intravenous fluids, or combinations of these for post-ERCP pancreatitis prevention in adult (aged ≥18 years) patients undergoing ERCP. Summary data from intention-to-treat analyses were extracted from published reports. We analysed incidence of post-ERCP pancreatitis across studies using network meta-analysis under the frequentist framework, obtaining pairwise odds ratios (ORs) and 95% CIs. We used the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) system for the confidence rating. This study is registered with PROSPERO, CRD42020172606. We identified 1503 studies, of which 55 RCTs evaluating 20 interventions in 17 062 patients were included in the network meta-analysis. The mean incidence of post-ERCP pancreatitis in the placebo or active control group was 12·2% (95% CI 11·4-13·0). Normal saline plus rectal indometacin (OR 0·02, 95% CI 0·00-0·40), intramuscular diclofenac 75 mg (0·24, 0·09-0·69), intravenous high-volume Ringer's lactate plus rectal diclofenac 100 mg (0·30, 0·16-0·55), intravenous high-volume Ringer's lactate (0·31, 0·12-0·78), 5-7 Fr pancreatic stents (0·35, 0·26-0·48), rectal diclofenac 100 mg (0·36, 0·25-0·52), 3 Fr pancreatic stents (0·47, 0·26-0·87), and rectal indometacin 100 mg (0·60, 0·50-0·73) were all more efficacious than placebo for preventing post-ERCP pancreatitis in pairwise comparisons. 5-7 Fr pancreatic stents (0·59, 0·41-0·84), intravenous high-volume Ringer's lactate plus rectal diclofenac 100 mg (0·49, 0·26-0·94), intravenous standard-volume normal saline plus rectal indometacin 100 mg (0·04, 0·00-0·66), and rectal diclofenac 100 mg (0·59, 0·40-0·89) were more efficacious than rectal indometacin 100 mg. The GRADE confidence rating was low to moderate for 98·3% of the pairwise comparisons. This systematic review and network meta-analysis summarises the available literature on NSAIDs, pancreatic stents, intravenous fluids, or combinations of these for prophylaxis of post-ERCP pancreatitis. Rectal diclofenac 100 mg is the best performing rectal NSAID in this network meta-analysis. Combinations of prophylaxis might be more effective, but there is little evidence. These findings help to establish prophylaxis of post-ERCP pancreatitis for future research and practice, and could reduce costs and increase adoption of prophylaxis. None.

Association of Intensive Lifestyle and Metformin Interventions With Frailty in the Diabetes Prevention Program Outcomes Study.

Frailty is a geriatric syndrome of decreased physiologic reserve and resistance to stressors that results in increased vulnerability to adverse health outcomes with aging. Diabetes and hyperglycemia are established risk factors for frailty. We sought to examine whether the odds of frailty among individuals at high risk of diabetes randomized to treatment with intensive lifestyle (ILS), metformin, or placebo differed after long-term follow-up. The sample comprised participants in the Diabetes Prevention Program (DPP) clinical trial, who continued follow-up in the DPP Outcomes Study (DPPOS) and completed frailty assessments in DPPOS Years 8 (n = 2385) and 10 (n = 2289), approximately 12 and 14 years after DPP randomization. Frailty was classified using Fried Frailty Phenotype criteria. GEE models adjusting for visit year with repeated measures pooled for Years 8 and 10 were used to estimate pairwise odds ratios (ORs) between ILS, metformin, and placebo for the outcomes of frail and prefrail versus nonfrail. Frailty prevalence by treatment group was ILS = 3.0%, metformin = 5.4%, placebo = 5.7% at Year 8, and ILS = 3.6%, metformin = 5.3%, placebo = 5.4% at Year 10. Odds ratios (95% CI) estimated with GEE models were ILS versus placebo, 0.62 (0.42-0.93), p = .022; metformin versus placebo, 0.99 (0.69-1.42), p = .976; and ILS versus metformin, 0.63 (0.42-0.94), p = .022. Odds of being frail versus nonfrail were 37% lower for ILS compared to metformin and placebo. Early ILS intervention, at an average age of about 50 years, in persons at high risk of diabetes may reduce frailty prevalence in later life. Metformin may be ineffective in reducing frailty prevalence. NCT00004992 (DPP) and NCT00038727 (DPPOS).

Synthetic observations from deep generative models and binary omics data with limited sample size.

Deep generative models can be trained to represent the joint distribution of data, such as measurements of single nucleotide polymorphisms (SNPs) from several individuals. Subsequently, synthetic observations are obtained by drawing from this distribution. This has been shown to be useful for several tasks, such as removal of noise, imputation, for better understanding underlying patterns, or even exchanging data under privacy constraints. Yet, it is still unclear how well these approaches work with limited sample size. We investigate such settings specifically for binary data, e.g. as relevant when considering SNP measurements, and evaluate three frequently employed generative modeling approaches, variational autoencoders (VAEs), deep Boltzmann machines (DBMs) and generative adversarial networks (GANs). This includes conditional approaches, such as when considering gene expression conditional on SNPs. Recovery of pair-wise odds ratios (ORs) is considered as a primary performance criterion. For simulated as well as real SNP data, we observe that DBMs generally can recover structure for up to 300 variables, with a tendency of over-estimating ORs when not carefully tuned. VAEs generally get the direction and relative strength of pairwise relations right, yet with considerable under-estimation of ORs. GANs provide stable results only with larger sample sizes and strong pair-wise relations in the data. Taken together, DBMs and VAEs (in contrast to GANs) appear to be well suited for binary omics data, even at rather small sample sizes. This opens the way for many potential applications where synthetic observations from omics data might be useful.

Inequality, Familial Aggregation, and Risk Prediction of Caries in Siblings.

Social and family conditions are likely of great importance to dental health; however, limited evidence of familial aggregation of caries among adolescent siblings exists. Moreover, social and family-level factors have never been evaluated as isolated caries predictors at the individual level. The objectives were to evaluate socioeconomic patterning of caries among siblings, assess sibling-specific aggregation of caries within families, and examine if such aggregation differed by parental socioeconomic position (SEP). We also evaluated the discriminant ability of sibling caries, SEP, and other social and familial factors in predicting caries in cosiblings. This nationwide register-based study included all 15-y-olds in Denmark in 2003 (index siblings) and their biological siblings born within ±3 y (cosiblings). Clinical and sociodemographic data for each subject were compiled from Danish national dental, social, and population registers. Caries was measured by the decayed, missing, and filled tooth surfaces (DMFS) index. Predictors included SEP (parental education, income, and occupational social class), gender, birth order, immigration status, and household type. Adjusted SEP-caries associations were estimated using negative binomial regression. Familial aggregation was evaluated using adjusted pairwise odds ratios from alternating logistic regressions. Caries prediction was based on classification and regression tree (CART) analyses. The study included 23,847 sibling pairs (n = 47,694). Socioeconomic patterning of caries was similar among the index and cosiblings with significant graded SEP-caries associations. Significant sibling-specific aggregation of caries was observed; cosiblings of caries-affected index siblings had odds of having caries 3.9 times (95% confidence interval: 3.65-4.18) as high as that of cosiblings with caries-free index siblings. This sibling similarity was stronger in socioeconomically disadvantaged families (adjusted pairwise odds ratios: 3.08-5.47). CART revealed index sibling caries as the single most important caries predictor, with caries predicted in ≥84% of cosiblings of adolescents with ≥3 carious tooth surfaces. Caries in a sibling should prompt preventive family-based approaches targeting cosiblings. This study revealed significant socioeconomic patterning of caries in adolescent siblings. Prediction modeling indicated that the single most important caries predictor among cosiblings was index sibling caries. Information on sibling caries level should be routinely combined with clinical evaluation to identify children at risk. Moreover, information on social and family conditions should be used to target prevention and health promotion at the school or municipal level. These approaches could possibly contribute to reducing the existing caries inequalities.

Immune thrombocytopenic purpura risk by live, inactivated and simultaneous vaccinations among Japanese adults, children and infants: a matched case-control study.

This case–control study investigated immune thrombocytopenic purpura (ITP) risk following live, inactivated, and simultaneous vaccination, with a focus on infants aged < 2 years. We matched case patients with ITP to one or two control patients with other diseases by institution, hospital visit timing, sex, and age. We calculated McNemar’s pairwise odds ratios (ORs [95% confidence interval]) with 114 case–control pairs. The case group had 27 (44%) males and 22 (35%) infants, and the control group included 49 (43%) males and 42 (37%) infants. For all age groups, the McNemar’s OR for ITP occurrence was 1.80 (0.54–6.84, p = 0.64) for all vaccines. Among infants, these were 1.50 (0.17–18.0, p = 0.50) for all vaccines, 2.00 (0.29–22.1, p = 0.67) for live vaccines, and 1.00 (0.01–78.5, p = 0.50) for inactivated vaccines. Sex-adjusted common ORs for simultaneous vaccination were 1.52 (0.45–5.21, p = 0.71) for all vaccines, 1.83 (0.44–7.59, p = 0.40) for inactivated vaccines only, and 1.36 (0.29–6.30, p = 0.69) for mixed live and inactivated vaccines. In infants, these were 1.95 (0.44–8.72, p = 0.38), 1.41 (0.29–6.94, p = 0.67) and 2.85 (0.43–18.9, p = 0.28), respectively. These limited data suggest no significant ITP risk following vaccinations or simultaneous vaccination in any age group, including infants.Electronic supplementary materialThe online version of this article (10.1007/s12185-020-02866-1) contains supplementary material, which is available to authorized users.

Inequality, familial aggregation, and risk prediction of caries in Siblings. A natiowide study of Danish adolescents

Introduction Sibling designs in assessing familial aggregation and risk prediction of dental caries have been largely unutilized. This large-scale investigation evaluated the robustness of parental socioeconomic position (SEP) as a cardinal determinant of caries experience by appraising it at the family level. Moreover, sibling-specific aggregation of caries within families and the discriminant ability of sibling caries in predicting co-sibling caries was ascertained. Methods A nationwide, register-based cross-sectional study was conducted on Danish adolescent siblings (n = 23,847 sibling pairs; index siblings/probands: all 15-year-olds in 2003; co-siblings: biological siblings born to the same mother within ± 3 years). Clinical (caries) data for each study subject were acquired from the national dental database (Sundhedsstyrelsens Centrale Odontologiske Register [SCOR]). Data on social variables, namely, parental SEP (parental education, income, and occupation), ethnicity, age/birth order, gender, and household type, were compiled from administrative registers at Statistics Denmark (e.g., the Danish population, education, labor market affiliation, and income and transfer payments registers). The social gradient in the magnitude of caries experience among the index- and co-siblings was estimated using negative binomial regression and generalized estimating equations taking into account family level clustering. Familial aggregation of caries was ascertained using probandwise concordance rates and adjusted pairwise odds ratios (PORs) from alternating logistic regressions. In order to determine the discriminant ability of proband caries and that of the other study covariates to effectively predict the absence or presence of caries in a sibling, a classification and regression tree (CART) analysis was undertaken. The predictive power of the CART models was evaluated using the area under the Receiver Operating Characteristic curve (AUROC) statistic. Results The prevalence of dental caries experience in the overall study population was 73.6% (index siblings 74.3%, co-siblings 72.9%). Conspicuous social patterning of caries was observed in both the index- and co-sibling populations even after adjustment for all study covariates. Significant sibling-specific familial aggregation of caries was observed, which varied in a stepwise graded fashion across the social hierarchy. Overall, 70.6% of sibling pairs were concordant in terms of their caries experience (caries 58.9%, caries-free 11.7%). Co-siblings of affected probands had 3.9 times (95% CI: 3.65–4.18) higher (adjusted) odds of having caries compared to those with caries-free siblings. This sibling similarity was further amplified in those with relative socioeconomic disadvantage (e.g., the adjusted PORs varied from 3.39 [95% CI: 3.04–3.77] in the highest to 5.47 [95% CI: 4.36–6.86] in the lowest parental education category). AUROCs from the CART models ranged from 0.7–0.82, indicating useful to excellent overall predictive power of the models. These models revealed sibling (proband) caries experience to be the single-most important risk predictor of caries in individuals (co-siblings), with the results indicating that caries could be expected in ≥ 84% of siblings of adolescents with 3 or more caries-affected tooth surfaces. Conclusions Graded clustering patterns of caries and socioeconomic disadvantage are encapsulated within families in Denmark, contributing to the engendering of health inequalities in society. With sibling caries being such a pertinent marker of increased caries risk, caries in a sibling should elicit preventive family-based approaches targeting co-siblings, particularly in socially disadvantaged households. Since caries shares common risk factors with other non-communicable diseases and conditions (such as diabetes and obesity), such policy constructs could reduce not only caries experience but also wider health inequalities in society.

Association of Intensive Lifestyle Intervention and Metformin with Frailty in the Diabetes Prevention Program (DPP) Outcomes Study (DPPOS)

Objective: To examine whether odds of frailty differ by DPP intervention arms (intensive lifestyle intervention [ILS], metformin [MET], placebo [PLB] after long-term follow-up in the DPPOS. Methods: DPP, a randomized trial of individuals at high risk of diabetes, lasted 2.8 years, after which there was a 13-month bridge period in which 16 sessions of group ILS was offered to all participants. During DPPOS, ILS was offered lifestyle reinforcement semi-annually and MET received unmasked metformin. Frailty was measured initially in DPPOS Year 8 in 2385 participants and repeated in 2264 of these in Year 10 (12 and 14 years after randomization). Using Fried criteria, presence of 3 of 5 characteristics (slow walking speed, weak grip strength, unintentional weight loss, low physical activity, and exhaustion) was classified as frail, 1 or 2 pre-frail, and 0 non-frail.: GEE models adjusting for visit year were employed with repeated measures pooled from Years 8 and 10 to estimate pairwise odds ratios between ILS, MET and PLB for the outcomes of frail and pre-frail vs. non-frail. Results: Odds of frailty vs. non-frailty were 37% lower in ILS compared to PLB (Table) and MET (OR=0.629, CI= 0.420-0.942, p=0.024). There were no significant treatment group differences in odds of pre-frailty. Conclusion: ILS intervention in persons at high risk of diabetes may reduce frailty in later life.12 years Post-randomization14 years post-randomizationOdds of Frailty vs. Non-Frailty Pooled over both examsTreatment GroupPrevalencePrevalence OR (95% CI) compared with placeboP-valuePre-FrailtyFrailtyPre-FrailtyFrailtyPlacebo42.8%5.7%43.7%5.4%1 (referent)n/aMetformin45.8%5.4%42.5%5.3%0.994 (0.694-1.424)0.976Lifestyle46.8%3.0%46.3%3.6%0.625 (0.419-0.934)0.022 Disclosure H.P. Hazuda: None. Q. Pan: None. J.P. Crandall: None. H. Florez: Advisory Panel; Self; Sanofi. J.A. Luchsinger: None. E.M. Venditti: None. A. Kriska: None. S. Golden: None. G. Bray: Advisory Panel; Self; Herbalife International of America, Inc., Novo Nordisk A/S. D. Research Group: None.

Social media for arthritis-related comparative effectiveness and safety research and the impact of direct-to-consumer advertising

BackgroundSocial media may complement traditional data sources to answer comparative effectiveness/safety questions after medication licensure.MethodsThe Treato platform was used to analyze all publicly available social media data including Facebook, blogs, and discussion boards for posts mentioning inflammatory arthritis (e.g. rheumatoid, psoriatic). Safety events were self-reported by patients and mapped to medical ontologies, resolving synonyms. Disease and symptom-related treatment indications were manually redacted. The units of analysis were unique terms in posts. Pre-specified conditions (e.g. herpes zoster (HZ)) were selected based upon safety signals from clinical trials and reported as pairwise odds ratios (ORs); drugs were compared with Fisher’s exact test. Empirically identified events were analyzed using disproportionality analysis and reported as relative reporting ratios (RRRs). The accuracy of a natural language processing (NLP) classifier to identify cases of shingles associated with arthritis medications was assessed.ResultsAs of October 2015, there were 785,656 arthritis-related posts. Posts were predominantly US posts (75%) from patient authors (87%) under 40 years of age (61%). For HZ posts (n = 1815), ORs were significantly increased with tofacitinib versus other rheumatoid arthritis therapies. ORs for mentions of perforated bowel (n = 13) were higher with tocilizumab versus other therapies. RRRs associated with tofacitinib were highest in conditions related to baldness and hair regrowth, infections and cancer. The NLP classifier had a positive predictive value of 91% to identify HZ. There was a threefold increase in posts following television direct-to-consumer advertisement (p = 0.04); posts expressing medication safety concerns were significantly more frequent than favorable posts.ConclusionSocial media is a challenging yet promising data source that may complement traditional approaches for comparative effectiveness research for new medications.

Comparative effectiveness of antiviral treatment for hepatitis B: a systematic review and Bayesian network meta-analysis.

A wide variety of competing drugs are available to patients for the treatment of chronic hepatitis B. We update a recent meta-analysis to include additional trial evidence with the aim of determining which treatment is the most effective. Twelve monotherapy or combination therapy were evaluated in treatment-naive individuals with hepatitis B e antigen (HBeAg) positive or negative patients. Databases were searched for randomized controlled trials in the first year of therapy. Bayesian random effects network meta-analysis was used to calculate the pairwise odds ratios, 95% credible intervals and ranking of six surrogate outcomes. In total, 22 studies were identified (7508 patients): 12 studies analysed HBeAg-positive patients, six analysed HBeAg-negative patients, and four evaluated both. Tenofovir was most effective at increasing efficacy in HBeAg-positive patients, ranking first for three outcomes and increased odds of undetectable levels of hepatitis B virus (HBV) DNA compared with seven other therapies (such as lamivudine: odds ratio 33.0; 95% credible interval 7.0-292.7). For HBeAg-negative patients, the large network (seven therapies) ranked entecavir alone or in combination with tenofovir highly for reduction in HBV DNA and histologic improvement. In the smaller network (three therapies), tenofovir ranked first for undetectable HBV DNA and histologic improvement. No data existed to directly or indirectly compare these treatments. For HBeAg-positive patients tenofovir is the most effective at increasing efficacy, whereas for HBeAg-negative patients, either tenofovir or entecavir is most effective. Further research should focus on strengthening the network connections, in particular comparing tenofovir and entecavir in HBeAg-negative patients.

Profiling phenome-wide associations: a population-based observational study.

To objectively characterize phenome-wide associations observed in the entire Taiwanese population and represent them in a meaningful, interpretable way. In this population-based observational study, we analyzed 782 million outpatient visits and 15 394 unique phenotypes that were observed in the entire Taiwanese population of over 22 million individuals. Our data was obtained from Taiwan's National Health Insurance Research Database.Results We stratified the population into 20 gender-age groups and generated 28.8 million and 31.8 million pairwise odds ratios from male and female subpopulations, respectively. These associations can be accessed online at http://associations.phr.tmu.edu.tw. To demonstrate the database and validate the association estimates obtained, we used correlation analysis to analyze 100 phenotypes that were observed to have the strongest positive association estimates with respect to essential hypertension. The results indicated that association patterns tended to have a strong positive correlation between adjacent age groups, while correlation estimates tended to decline as groups became more distant in age, and they diverged when assessed across gender groups. The correlation analysis of pairwise disease association patterns across different age and gender groups led to outcomes that were broadly predicted before the analysis, thus confirming the validity of the information contained in the presented database. More diverse individual disease-specific analyses would lead to a better understanding of phenome-wide associations and empower physicians to provide personalized care in terms of predicting, preventing, or initiating an early management of concomitant diseases.

Changing The Treatment Paradigm For Previously Treated Chronic Lymphocytic Leukemia Patients With Del(17p) Karyotype

CLL patients (pts) with del(17p13.1) (17p-) karyotype typically have poor response to therapy and dismal clinical outcomes. We describe the success of treatment of 17p- pts on clinical trials at Ohio State University (OSU) with regimens based on 1 of 2 novel agents [ibrutinib (IB; an oral inhibitor of Bruton tyrosine kinase) or one of the cyclin-dependent kinase inhibitors (CDKi; alvocidib, dinaciclib, or TG02)].We retrospectively reviewed records of 174 CLL pts with 17p- who received therapy prior to 1st treatment at OSU (OSU Tx1) from 2002-2013. Progression free survival (PFS) was calculated from date of OSU Tx1 until progression/death, censoring pts at date of second treatment prior to progression or last contact if alive and progression free. Overall survival (OS) was calculated from date of OSU Tx1 until date of death. Pts who received transplant or later IB were censored at that time; otherwise pts were censored at last follow-up or 48 mos if still alive. PFS/OS estimates were calculated using the Kaplan-Meier method. Proportional hazard models were fit with treatment group and adjusted for other prognostic variables (p<0.05).Median time from CLL diagnosis to OSU Tx1 was 5.6 yrs and median number of prior therapies was 3 (range: 1-10). At OSU Tx1, median age was 63 yrs (39-83), 66% were male, and 93% were Caucasian. ECOG performance status (PS) was 0, 1, and 2/3 in 28%, 60%, and 11% of pts, respectively, and 71% had Rai Stage 3/4. In addition to 17p-, 21%, 57%, and 16% of pts harbored 11q-, 13q- and tri(12) aberrations, while 70% had complex karyotype (CK; ≥3 abnormalities including 17p-).At OSU Tx1, 16% (n=27), 33% (n=58), and 51% (n=89) of pts received an IB-based therapy, CDKi-based therapy, or other therapies (O), respectively. Clinical and molecular characteristics were not statistically different across groups except the IB and CDKi groups had more pts with ≥3 prior therapies (70% in IB vs 64% in CDKi vs 42% in O, p=0.005) and the median white blood cell count (WBC) was lowest in the CDKi group [18.8 in IB vs 25.5 in O vs 12.1 in CDKi, p=0.04]. The degree of response to OSU Tx1 was significantly different among groups (p=0.007), where 56%, 45%, and 24% of pts in the IB, CDKi, and O groups, respectively, achieved at least a partial response, and 85%, 64%, and 53% achieved at least stable disease. As shown in Figure 1A, PFS was significantly extended with IB and CDKi compared to O (p<0.0001), and PFS was longer with IB compared to CDKi (p=0.002); 12 mos estimates were 77% (95%CI=0.56-0.89), 38% (95%CI=0.25-0.52), and 17% (95%CI=0.10-0.26) in the IB, CDKi, and O groups, respectively. As shown in Figure 1B, OS was significantly different among the IB, CDKi, and O groups (p=0.02), with 12 mos estimates of 81% (95%CI=0.61-0.92), 78% (95%CI=0.64-0.87), and 58% (95%CI=0.46-0.68). OS was extended with IB and CDKi compared to O (p=0.01 and p=0.04, respectively), although by 48 mos, estimates for CDKi and O were similar. With a median follow-up of 12 mos, no large differences between IB and CDKi were yet observed (p=0.49). [Display omitted] In multivariable analysis, treatment group was significantly associated with achieving at least partial response (p=0.0005) independent of number of prior therapies and presence of 13q-. The odds of response were higher with IB and CDKi compared to O, but not significantly different between IB and CDKi (all pairwise odds ratios (OR): IB vs O=5.66 (95%CI=2.12-15.08); CDKi vs O=3.38 (95%CI=1.57-7.31); IB vs CDKi = 1.67 (95%CI=0.65-4.32)]. Treatment group was significantly associated with PFS (p<0.0001) independent of number of prior therapies, albumin, and CK. The risk of progression/death decreased by 90% and 60%, respectively, with IB or CDKi compared to O (all pairwise hazard ratios (HR): IB vs O=0.10 (95%CI=0.04-0.22); CDKi vs O=0.40 (95%CI=0.27-0.60); IB vs CDKi = 0.24 (95%CI=0.11-0.55)]. Treatment group was significantly associated with OS (p=0.003) independent of number of prior therapies, albumin, CK, ECOG PS, and WBC. The risk of death decreased by 73% and 47%, respectively with IB and CDKi compared to O (all pairwise HR: IB vs O=0.27 (95%CI=0.11-0.66); CDKi vs O=0.53 (95%CI=0.31-0.89); IB vs CDKi=0.52 (95%CI=0.21-1.30)]. Notably, age did not correlate with response, PFS, or OS.Treatment of 17p CLL pts with IB and CDKi at OSU have demonstrated improved response, PFS, and OS. These novel therapies provide hope for patients with a karyotype historically associated with dismal clinical outcomes. Disclosures:Jones:Pharmacyclics: Membership on an entity's Board of Directors or advisory committees, Research Funding. Byrd:Celgene: Consultancy; Johnson and Johnson: Consultancy; Pharmacyclics: Research Funding.

Patterns of cervical coinfection with multiple human papilloma virus types in a screening population in Denmark

Patterns of cervical human papillomavirus (HPV) infection suggest that HPV genotypes are not independent of each other. This may be explained by risk factors common to all HPV infections, but type-specific biological factors may also play a role. This raises the question of whether widespread use of the quadrivalent vaccine (covering HPV6, 11, 16, 18) may indirectly affect the prevalence of any non-vaccine types. Routine screening samples from 5014 Danish women were tested for 35 HPV genotypes (including 13 high-risk) using the Genomica CLART® HPV2 kit, which is a low-density microarray based on PCR amplification. Simulation studies were performed both under independence between genotypes and under a common dependence structure as would arise from common risk factors, and simulation results were compared to observed coinfection patterns. Overall HPV prevalence was 37.4%, with multiple infections in 17.9%. For 15 HPV types of primary interest (13 high-risk plus HPV6, 11), almost all pairs occurred more often than expected under independence; 33/105 (31.4%) were statistically significant (p<0.05 after adjustment for multiple comparisons). The pairwise odds ratios showed significant heterogeneity (Woolf's test p<0.0001). For simulations based on common dependence, three pairs had observed to expected (O/E) ratios significantly different than 1 (31/68, O/E=4.20; 51/68, O/E=2.52; 33/58, O/E=3.27; all p<0.05 after adjustment for multiple comparisons). HPV68 occurred in multiple infections nearly four times as often as expected under common dependence (p<0.005 after adjustment for multiple comparisons). These results indicate some interaction between HPV types, and suggest that common risk factors do not entirely explain the observed HPV coinfection pattern, although no evidence is found that the prevalence of any types not targeted by the quadrivalent vaccine may be indirectly increased or decreased after widespread use of the vaccine.

Obesity Comorbidity in Unipolar Major Depressive Disorder

While a significant body of research has demonstrated high comorbidity rates between depression and obesity, the vast majority of this work has considered depression as a unitary diagnosis. Given that increased appetite and weight gain are highly characteristic of the "atypical" subtype of depression, while classic depression is characterized by decreased appetite and weight loss, it would be important to examine whether increased obesity risk is consistent across the major vegetative subtypes of depression or is limited to the atypical subtype. Using data from the 2001-2002 National Epidemiologic Survey on Alcohol and Related Conditions (NESARC), we identified 5,092 US adults with past or current major depression based on DSM-IV-TR criteria and 1,500 gender-matched controls. Each depressed subject was designated as having classic, atypical, or undifferentiated depression based on core vegetative symptoms. Logistic regression models examined rates of current obesity (defined as a current body mass index [kg/m2] > 30) across the 3 depressive subgroups and nondepressed controls, adjusting for demographic differences. To limit the possible effect of current depressive symptoms on observed obesity rates, secondary analyses were completed in individuals with past depression only. Subjects with atypical depression had markedly elevated obesity rates compared to population controls and to other depressed subjects, with corresponding pairwise odds ratios consistently greater than 2.0 (P < .001). In contrast, obesity rates were not significantly different in subjects with classic depression and nondepressed controls. These results were manifest in individuals with either current or past depression and were independent of gender and age. While many individuals with classic depression will present with obesity due to the high prevalence of both disorders, only atypical depression is associated with an elevated risk of obesity relative to the population at large. Refining the target phenotype(s) for future work on depression and obesity might improve our understanding, prevention, and treatment of this complex clinical problem.

Sample size estimation for alternating logistic regressions analysis of multilevel randomized community trials of under-age drinking.

Under-age drinking is an enormous public health issue in the USA. Evidence that community level structures may impact on under-age drinking has led to a proliferation of efforts to change the environment surrounding the use of alcohol. Although the focus of these efforts is to reduce drinking by individual youths, environmental interventions are typically implemented at the community level with entire communities randomized to the same intervention condition. A distinct feature of these trials is the tendency of the behaviours of individuals residing in the same community to be more alike than that of others residing in different communities, which is herein called 'clustering'. Statistical analyses and sample size calculations must account for this clustering to avoid type I errors and to ensure an appropriately powered trial. Clustering itself may also be of scientific interest. We consider the alternating logistic regressions procedure within the population-averaged modelling framework to estimate the effect of a law enforcement intervention on the prevalence of under-age drinking behaviours while modelling the clustering at multiple levels, e.g. within communities and within neighbourhoods nested within communities, by using pairwise odds ratios. We then derive sample size formulae for estimating intervention effects when planning a post-test-only or repeated cross-sectional community-randomized trial using the alternating logistic regressions procedure.

The Impact of Alcohol Outlet Density on the Geographic Clustering of Underage Drinking Behaviors within Census Tracts

The regulation of alcohol outlet density has been considered as a potential means of reducing alcohol consumption and related harms among underage youth. Whereas prior studies have examined whether alcohol outlet density was associated with an individual's alcohol consumption and related harms, this study examines whether it is related to the co-occurrence, or clustering, of these behaviors within geographic areas, specifically census tracts. The Enforcing Underage Drinking Laws Randomized Community Trial provided cross-sectional telephone survey data in 2006 and 2007 from 10,754 youth aged 14 to 20 from 5 states residing in 1,556 census tracts. The alternating logistic regression approach was used to estimate pairwise odds ratios between responses from youth residing in the same census tract and to model them as a function of alcohol outlet density. Riding with a drinking driver, making an alcohol purchase attempt, and making a successful alcohol purchase attempt clustered significantly within census tracts with the highest off-premise alcohol outlet density while frequent drinking clustered within census tracts with the greatest on-premise density. Driving after drinking and experiencing nonviolent alcohol-related consequences clustered marginally within census tracts with the greatest on-premise and off-premise alcohol outlet density, respectively. Although youth primarily receive alcohol from social sources, commercial alcohol access is geographically concentrated within census tracts with the greatest off-premise outlet density. A potentially greater concern is the clustering of more frequent drinking and drinking and driving within census tracts with the greatest on-premise outlet density which may necessitate alternative census tract level initiatives to reduce these potentially harmful behaviors.

Maternal Effects for Preterm Birth: A Genetic Epidemiologic Study of 630,000 Families

This study was undertaken to disentangle the maternal genetic from the fetal genetic effects for preterm birth and to study the possibility of these effects being explained by known risk factors. By cross-linking of the population-based Swedish Multigeneration and Medical Birth registers, 989,027 births between 1992 and 2004 were identified. Alternating logistic regression was applied to model the familial clustering with pairwise odds ratios (PORs), and covariates were included to evaluate if the familial aggregation was explained by exposure to shared risk factors. Generalized linear mixed models were used to estimate the contribution of genetic and environmental effects. Sisters of women who had a preterm delivery had themselves an increased odds of having a preterm delivery (POR = 1.8, 95% confidence interval: 1.5, 2.1), while there was no corresponding increase in odds in families joined by brothers (POR = 1.1, 95% confidence interval: 0.9, 1.4). Twenty-five percent of the variation in preterm birth was explained by maternal genetic factors, whereas fetal genetic factors only marginally influenced the variation in liability. The increased odds ratio between offspring of sisters was independent of maternal risk factors for preterm birth, suggesting that the relative importance of maternal effects is not explained by these well-known risk factors.