A comparative investigation of methods for logistic regression with separated or nearly separated data

Penalized logistic regression with low prevalence exposures beyond high dimensional settings

Estimating and selecting risk factors with extremely low prevalences of exposure for a binary outcome is a challenge because classical standard techniques, markedly logistic regression, often fail to provide meaningful results in such settings. While penalized regression methods are widely used in high-dimensional settings, we were able to show their usefulness in low-dimensional settings as well. Specifically, we demonstrate that Firth correction, ridge, the lasso and boosting all improve the estimation for low-prevalence risk factors. While the methods themselves are well-established, comparison studies are needed to assess their potential benefits in this context. This is done here using the dataset of a large unmatched case-control study from France (2005-2008) about the relationship between prescription medicines and road traffic accidents and an accompanying simulation study. Results show that the estimation of risk factors with prevalences below 0.1% can be drastically improved by using Firth correction and boosting in particular, especially for ultra-low prevalences. When a moderate number of low prevalence exposures is available, we recommend the use of penalized techniques.

Invasive Pneumococcal Disease in Refugee Children, Germany

Refugee children in Germany are not routinely given a pneumococcal conjugate vaccine. Cases of invasive pneumococcal disease (IPD) in 21 refugee children were compared with those in 405 Germany-born children for 3 pneumococcal seasons. Refugee children had significantly higher odds of vaccine-type IPD and multidrug-resistant IPD than did Germany-born children.

An assessment of pesticide exposures and land use of honey bees in Virginia

Large-scale honey bee colony loss threatens pollination services throughout the United States. An increase in anthropogenic pressure may influence the exposure of hives to household and agricultural pesticides. The objective of this survey was to provide an assessment of the risk of exposure to commonly used pesticides to honey bee colonies in Virginia in relation to land use. Adult honey bee, pollen, and wax samples from colonies throughout Virginia were evaluated for pyrethroid, organophosphate, organochlorine, and triazine pesticides using gas chromatography-mass spectrometry analysis. Of the 11 pesticides analyzed, nine were detected in one or more hive matrices. The probability of detecting a pesticide in pollen was less in forests than in pasture, agriculture, or urban landscapes. Coumaphos and fluvalinate were significantly more likely to be detected across all matrices with concentrations in wax as high as 15500 and 6970 ng/g (dry weight), respectively, indicating the need for further research on the potential effects of miticide accumulation in wax to larval and adult bees.

A study of the relationship between the level of anxiety declared by MRI patients in the STAI questionnaire and their respiratory rate acquired by a fibre-optic sensor system

Magnetic resonance imaging (MRI) patients often experience anxiety-related respiratory disorders, including hyperventilation, but their respiratory indicators are not routinely monitored during scanning. Free from metal parts and immune to electromagnetic radiation, fibre-optic sensors have the potential to better control the patient’s condition by providing continuous non-invasive monitoring of the respiratory rate (RR). The study was purposed to assess the relationship between anxiety in MRI patients and their RR acquired by a fibre-optic sensor system. Forty-four subjects were involved in the study. The mean RR values recorded for 2 minutes immediately after the beginning and immediately before the end of the scanning were assessed relative to the State-Trait Anxiety Inventory (STAI) X-1 scores obtained immediately before and immediately after the scanning, respectively. A growth mixture model analysis was performed to statistically differentiate two groups of subjects according to the trends in repeated measures of RR. A significant lowering of the anxiety state was observed in the group characterised by a decrease in RR, whereas essentially no change in anxiety level was observed in the group with a stable RR. The t-test showed significant differences in changes in anxiety between these groups (t₍₃₉₎ = −2.349, p = 0.012, Cohen’s d = 2.13).

Risk of subsequent colorectal cancers after a solid tumor in childhood: Effects of radiation therapy and chemotherapy

BACKGROUND

Very few previous studies have addressed the question of colorectal cancer (CRC) after childhood cancer treatment. We aimed to quantify the roles of radiation therapy and chemotherapy agents in the occurrence of subsequent CRC.

METHODS

A nested case-control study was conducted using 36 CRC cases and 140 controls selected from 7032 five-year survivors of the French Childhood Cancer Survivor Study (FCCSS) cohort, treated from 1945 to 2000 in France. The radiation dose-distribution metrics at the site of CRC and doses of individual chemotherapeutic agents were calculated. Conditional logistic regressions were performed to calculate odds ratios (ORs).

RESULTS

Overall, patients who received radiotherapy with estimated dose to colon had a 4.3-fold (95% CI, 1.3-17.6) increased risk for CRC compared with patients who did not receive radiotherapy, after adjustment for chemotherapy. This risk increased to 8.9-fold and 19.3-fold among patients who received radiation doses ranging from 20 to 29.99 Gy and ≥30 Gy, respectively. Our data reported a significantly elevated OR for anthracyclines, after controlling for radiotherapy and MOPP regimen. But, restricted analyses excluding patients who had received ≥30 Gy showed that only radiation doses ranging from 20 to 29.99 Gy produced a significant increase in subsequent CRC risk (OR = 7.8; 95% CI, 1.3-56.0), after controlling for anthracyclines and MOPP regimen.

CONCLUSIONS

The risk of subsequent CRC was significantly increased after radiation dose (even < 30 Gy). This novel finding supports the need to update monitoring guidelines for CRC to optimize the long-term follow-up for subsequent CRC in survivors of childhood cancer.

Brain and cord imaging features in neuromyelitis optica spectrum disorders

OBJECTIVES

To validate imaging features able to discriminate neuromyelitis optica spectrum disorders from multiple sclerosis with conventional magnetic resonance imaging (MRI).

METHODS

In this cross-sectional study, brain and spinal cord scans were evaluated from 116 neuromyelitis optica spectrum disorder patients (98 seropositive and 18 seronegative) in chronic disease phase and 65 age-, sex-, and disease duration-matched multiple sclerosis patients. To identify independent predictors of neuromyelitis optica diagnosis, after assessing the prevalence of typical/atypical findings, the original cohort was 2:1 randomized in a training sample (where a multivariate logistic regression analysis was run) and a validation sample (where the performance of the selected variables was tested and validated).

RESULTS

Typical brain lesions occurred in 50.9% of neuromyelitis optica patients (18.1% brainstem periventricular/periaqueductal, 32.7% periependymal along lateral ventricles, 3.4% large hemispheric, 6.0% diencephalic, 4.3% corticospinal tract), 72.2% had spinal cord lesions (46.3% long transverse myelitis, 36.1% short transverse myelitis), 37.1% satisfied 2010 McDonald criteria, and none had cortical lesions. Fulfillment of at least 2 of 5 of absence of juxtacortical/cortical lesions, absence of periventricular lesions, absence of Dawson fingers, presence of long transverse myelitis, and presence of periependymal lesions along lateral ventricles discriminated neuromyelitis optica patients in both training (sensitivity = 0.92, 95% confidence interval [CI] = 0.84-0.97; specificity = 0.91, 95% CI = 0.78-0.97) and validation samples (sensitivity = 0.82, 95% CI = 0.66-0.92; specificity = 0.91, 95% CI = 0.71-0.99). MRI findings and criteria performance were similar irrespective of serostatus.

INTERPRETATION

Although up to 50% of neuromyelitis optica patients have no typical lesions and a relatively high percentage of them satisfy multiple sclerosis criteria, several easily applicable imaging features can help to distinguish neuromyelitis optica from multiple sclerosis. ANN NEUROL 2019;85:371-384.

Regional variations in serotype distribution and vaccination status in children under six years of age with invasive pneumococcal disease in Germany

Overview

The protective effect of infant pneumococcal conjugate vaccine (PCV) recommendation can be seen in Germany as a whole and in smaller regional groups. Comparisons between population-normalized geographic regions of Germany show different serotype distributions after program implementation, particularly in non-vaccine serotypes. The prior distinct differences in serotype distribution in children between the former East and former West German federal states have vanished. Children under six remain a vulnerable group, but the occurrence of vaccine-type (VT) invasive pneumococcal disease (IPD) in children correctly vaccinated (using a three-dose primary series plus one booster dose) with PCV13 was low (9 out of 374 cases, 2.4%). However, only 18.4% of children in Germany with IPD were correctly vaccinated with PCV13 according to the recommended schedule. Continued surveillance and better schedule adherence are essential to definitively establish the most effective PCV administration schedule.

Vaccination effects

For all PCV products used in Germany (PCV7, PCV10, and PCV13), vaccination status was the most common statistically significant predictor of infection with a particular serotype: Unvaccinated children old enough to have received at least one dose of vaccine in the PCV7 group had significantly higher odds (OR: 6.84, 95%CI: 2.66–22.06, adjusted for per capita income and residence in the northeastern federal states) of contracting VT IPD. In the PCV10 group, VT IPD had an OR of 4.52 (95% CI: 1.60–15.62, adjusted for year of infection, median household size, and residence in the southern federal states) in unvaccinated children, and in the PCV13 group, unvaccinated children continued to have higher odds (OR: 6.21, 95%CI: 3.45–11.36, adjusted for year of infection, age of child, per capita income, residence in the southern federal states, and percentage of children using public daycare) of getting vaccine-type IPD. Being unvaccinated was the most frequent significant indicator for infection with vaccine-type serotypes for each analysis group, while geographic groupings showed more limited potential to predict serotype of infection in early childhood IPD in Germany.

A rare missense variant of CASP7 is associated with familial late-onset Alzheimer's disease

INTRODUCTION

The genetic architecture of Alzheimer's disease (AD) is only partially understood.

METHODS

We conducted an association study for AD using whole sequence data from 507 genetically enriched AD cases (i.e., cases having close relatives affected by AD) and 4917 cognitively healthy controls of European ancestry (EA) and 172 enriched cases and 179 controls of Caribbean Hispanic ancestry. Confirmation of top findings from stage 1 was sought in two family-based genome-wide association study data sets and in a whole genome-sequencing data set comprising members from 42 EA and 115 Caribbean Hispanic families.

RESULTS

We identified associations in EAs with variants in 12 novel loci. The most robust finding is a rare CASP7 missense variant (rs116437863; P = 2.44 × 10^-10) which improved when combined with results from stage 2 data sets (P = 1.92 × 10^-10).

DISCUSSION

Our study demonstrated that an enriched case design can strengthen genetic signals, thus allowing detection of associations that would otherwise be missed in a traditional case-control study.

Identification of meteorological factors associated with human infection with avian influenza A H7N9 virus in Zhejiang Province, China

BACKGROUND

Since the first reported human infection with an avian-origin influenza A (H7N9) virus in China in early 2013, there have been recurrent outbreaks of the virus in the country. Previous studies have shown that meteorological factors are associated with the risk of human infection with the virus; however, their possible nonlinear and lagged effects were not commonly taken into account.

METHOD

To quantify the effect of meteorological factors on the risk of human H7N9 infection, daily laboratory-confirmed cases of human H7N9 infection and meteorological factors including total rainfall, average wind speed, average temperature, average relative humidity, and sunshine duration of the 11 sub-provincial/prefecture cities in Zhejiang during the first four outbreaks (13 March 2013-30 June 2016) were analyzed. Separate models were built for the 6 sub-provincial/prefecture cities with the greatest number of reported cases using a combination of logistic generalized additive model and distributed lag nonlinear models, which were then pooled by a multivariate meta-regression model to determine their overall effects.

RESULTS

According to the meta-regression model, for rainfall, the log adjusted overall cumulative odds ratio was statistically significant when log of rainfall was >4.0, peaked at 5.3 with a value of 12.42 (95% confidence intervals (CI): [3.23, 21.62]). On the other hand, when wind speed was 2.1-3.0 m/s or 6.3-7.1 m/s, the log adjusted overall cumulative odds ratio was statistically significant, peaked at 7.1 m/s with a value of 6.75 (95% CI: [0.03, 13.47]). There were signs of nonlinearity and lag effects in their associations with the risk of infection.

CONCLUSION

As rainfall and wind speed were found to be associated with the risk of human H7N9 infection, weather conditions should be taken into account when it comes to disease surveillance, allowing prompt actions when an outbreak takes place.

The Number of Events per Confounder for Valid Estimation of Risk Difference Using Modified Least-Squares Regression

Risk difference is a relevant effect measure in epidemiologic research. Although it is well known that when there are few events per confounder, logistic regression is not suitable for confounding control, it is not clear how many events per confounder are required for valid estimation of risk difference using linear binomial models. Because the maximum likelihood method has a convergence problem, we investigated the number of events per confounder necessary to validly estimate risk difference using modified least-squares regression in a simulation. We simulated 864 scenarios, according to the number of confounders (2-20), the number of events per confounder (2-12), marginal risk (0.5%-40%), exposure proportion (20% and 40%), and 3 sizes of risk difference. Our simulation showed that modified least-squares regression provided unbiased risk difference-regardless of the number of events per confounder-and reliable confidence intervals when more than 5 events were expected in the exposed and in the unexposed, irrespective of the number of events per confounder. We illustrated the modified least-squares regression analysis using perinatal epidemiologic data. Modified least-squares regression is considered to be a useful analytical tool for rare binary outcomes relative to the number of confounders.

Characterization of maternal plasma biomarkers associated with delivery of small and large for gestational age infants in the MIREC study cohort

Objective

Neonatal morbidity and mortality can be influenced by maternal health status. Information on maternal and fetal biomarkers of adverse health outcomes is limited. This work aims at identifying maternal biomarkers associated with low and high birth weight for gestational age groups.

Design and settings

Population-based prospective cohort study of the potential adverse health effects of exposure to environmental contaminants on pregnancy and infant health.

Methods

Third trimester maternal plasma samples (n = 1588) from a pregnancy cohort (Maternal-Infant Research on Environmental Chemicals Study, MIREC) were analyzed for changes in a target spectrum of biomarkers of vascular health (e.g., matrix metalloproteinases MMPs, vascular endothelial cell growth factor VEGF), inflammation (e.g. cellular adhesion molecules CAMs, cytokines, chemokines) by affinity-based multiplex protein array analyses. Multivariate logistic regression analyses were done to examine associations between target plasma biomarkers, maternal-infant characteristics, and birth weight outcomes assessed as small for gestational age (SGA) ≤10^th percentile and large for gestational age (LGA) ≥90^th percentile groups.

Results and outcomes

Our results revealed that maternal plasma biomarkers monocyte chemoattractant protein-1 MCP-1 (p<0.05, +ve) and VEGF (p<0.05, -ve) along with parity = 1 (p<0.01, -ve) and gestational hypertension (p<0.05, +ve) were associated with SGA births. Meanwhile, LGA was associated with maternal plasma VEGF (p<0.05, +ve) and MMP-9 (p<0.05, -ve) and gestational hypertension (p<0.01, +ve), pre-pregnancy body mass index (p<0.01, +ve), parity (p<0.05, +ve) and education (p<0.05, -ve).

Conclusions

Third trimester maternal plasma biomarkers in combination with maternal health and socioeconomic characteristics can be useful in predicting SGA and LGA outcomes. Maternal vascular health and inflammatory status may contribute to both SGA and LGA births through distinct molecular mechanisms.

HLA class II Eplet mismatch predicts De Novo DSA formation post lung transplant

The use of algorithms such as HLAMatchmaker to redefine donor-recipient HLA matching is gaining increasing attention. Our research has previously demonstrated that higher HLA class II eplet mismatches correlated with the development of chronic lung allograft dysfunction (CLAD). In this study of lung transplant recipients we prospectively examined the association between donor-recipient HLA eplet mismatches as defined by HLAMatchmaker (version 2.1) and de-novo anti-HLA donor-specific antibody (DSA) formation, as assessed by single antigen-bead solid phase assay. HLA class II eplet mismatch, when split at the median for the cohort, predicted the development of de-novo HLA class II DSA at 3 months but not at 12 months. Having previously shown that high HLA class II eplet mismatches was associated with CLAD, we now show that the same factors are associated with de-novo HLA class II DSA post-transplant.

Choice of Prophylactic Antibiotics and Surgical Site Infections After Cesarean Delivery

OBJECTIVE

To examine the rate of surgical site infection according to the choice of antibiotics in women undergoing cesarean delivery.

METHODS

We conducted a retrospective cohort study of women undergoing cesarean delivery (labored, unlabored, and scheduled) from 2012 to 2017. Women with chorioamnionitis and those who did not receive any antibiotics were excluded. Our primary outcome was defined a priori as a composite of cellulitis, endometritis, deep wound infection, abdominopelvic abscess, and sepsis. Outcomes were examined according to the choice of antibiotics: cefazolin, a standard alternative (both clindamycin and gentamicin), and inappropriate alternatives (such as clindamycin only). A multivariable logistic regression model was used to calculate the propensity score for each observation, which was the probability of receiving a particular antibiotic regimen. The propensity score-adjusted logistic regression models were conducted to calculate adjusted odds ratios (ORs) and 95% CIs. Firth's penalized likelihood approach was applied to address issues of rare events.

RESULTS

Among 6,584 selected women, 6,163 (93.6%), 274 (4.2%), and 147 (2.2%) received cefazolin, a standard alternative, and inappropriate alternatives, respectively. Use of standard alternative antibiotics compared with cefazolin was not associated with increased odds of the primary outcome (crude OR 1.50 [0.92-2.46]; adjusted OR 1.63 [0.97-2.60]) but was associated with increased odds of cellulitis (crude OR 2.07 [1.16-3.70]; adjusted OR 1.93 [1.03-3.31]). Use of inappropriate alternative antibiotics compared with cefazolin was associated with increased odds of the primary outcome (crude OR 4.37 [2.80-6.83]; adjusted OR 4.13 [2.59-6.36]) as well as some components of the composite outcome such as endometritis before discharge (crude OR 6.85 [3.94-11.90]; adjusted OR 6.68 [3.69-11.44]) and cellulitis (crude OR 3.36 [1.78-6.34]; adjusted OR 3.23 [1.63-5.81]).

CONCLUSION

Both standard alternative and inappropriate alternatives were associated with increased odds of surgical site infections compared with cefazolin.

The role of cEEG as a predictor of patient outcome and survival in patients with intraparenchymal hemorrhages

PURPOSE

The objective of this study was to determine if continuous electroencephalography (cEEG) results are associated with functional outcome and survival in critically ill patients with intraparenchymal hemorrhages (IPH).

METHODS

Patients diagnosed with IPH were selected using a Critical Care EEG Monitoring Consortium Database at Brigham and Women's Hospital in Boston. Functional Outcome in Patients with Primary Intracerebral Hemorrhage (FUNC) scores and Intracerebral Hemorrhage (ICH) scores were calculated as covariates. Outcomes of interest were functional outcome (modified Rankin scale [mRS] <3 vs ≥3) and mortality at hospital discharge. cEEG features, as defined by the American Clinical Neurophysiology Society standard terminology, were assessed for association with outcome after accounting for known clinical covariates.

RESULTS

A total of 94 patients admitted between March 2013 and December 2015 were selected. Multivariate regression analysis revealed that the presence of Stage II Sleep is independently associated with good functional outcome at discharge after correcting for FUNC (p = 0.0080) and ICH (p = 0.0088). The absence of anteroposterior (AP) gradient in an EEG is associated with increased likelihood of mortality at discharge after correcting for FUNC (p = 0.013) and ICH (p = 0.019) scores.

CONCLUSIONS

cEEG measures were significantly associated with functional and mortality outcome measures in patients with IPH even after accounting for known clinical and radiological covariates. Further research is needed to determine whether prediction models are improved by inclusion of cEEG features.

The DSM-5 nonsuicidal self-injury disorder among incoming college students: Prevalence and associations with 12-month mental disorders and suicidal thoughts and behaviors

BACKGROUND

Approximately one in five college students report a history of nonsuicidal self-injury. However, it is unclear how many students meet criteria for the recently proposed DSM-5 nonsuicidal self-injury disorder (NSSI-D). In this study, we used full NSSI-D criteria to identify those students most in need of clinical care.

METHODS

Using data from the Leuven College Surveys (n = 4,565), we examined the 12-month prevalence of DSM-5 NSSI-D in a large and representative sample of incoming college students. We also explored the optimal frequency threshold as a function of interference in functioning due to NSSI, and examined comorbidity patterns with other 12-month mental disorders (i.e., major depressive disorder, broad mania, generalized anxiety disorder, panic disorder, and alcohol dependence) and suicidal thoughts and behaviors (STB).

RESULTS

Twelve-month NSSI-D prevalence was 0.8% and more common among females (1.1%) than males (0.4%). The proposed 5+ diagnostic threshold was confirmed as yielding highest discrimination between threshold and subthreshold cases in terms of distress or disability due to NSSI. A dose-response relationship was observed for NSSI recency-severity (i.e., 12-month NSSI-D, subthreshold 12-month NSSI-D, past NSSI, no history of NSSI) with number of 12-month mental disorders and STB. NSSI-D occurred without comorbid disorders for one in five individuals, and remained associated with severe role impairment when controlling for the number of comorbid disorders.

CONCLUSIONS

These findings offer preliminary evidence that DSM-5 NSSI-D is uncommon among incoming college students, but may help to improve the deployment of targeted resource allocation to those most in need of services. More work examining the validity of NSSI-D is required.

Association of Polygenic Risk Scores for Multiple Cancers in a Phenome-wide Study: Results from The Michigan Genomics Initiative

Health systems are stewards of patient electronic health record (EHR) data with extraordinarily rich depth and breadth, reflecting thousands of diagnoses and exposures. Measures of genomic variation integrated with EHRs offer a potential strategy to accurately stratify patients for risk profiling and discover new relationships between diagnoses and genomes. The objective of this study was to evaluate whether polygenic risk scores (PRS) for common cancers are associated with multiple phenotypes in a phenome-wide association study (PheWAS) conducted in 28,260 unrelated, genotyped patients of recent European ancestry who consented to participate in the Michigan Genomics Initiative, a longitudinal biorepository effort within Michigan Medicine. PRS for 12 cancer traits were calculated using summary statistics from the NHGRI-EBI catalog. A total of 1,711 synthetic case-control studies was used for PheWAS analyses. There were 13,490 (47.7%) patients with at least one cancer diagnosis in this study sample. PRS exhibited strong association for several cancer traits they were designed for, including female breast cancer, prostate cancer, melanoma, basal cell carcinoma, squamous cell carcinoma, and thyroid cancer. Phenome-wide significant associations were observed between PRS and many non-cancer diagnoses. To differentiate PRS associations driven by the primary trait from associations arising through shared genetic risk profiles, the idea of "exclusion PRS PheWAS" was introduced. Further analysis of temporal order of the diagnoses improved our understanding of these secondary associations. This comprehensive PheWAS used PRS instead of a single variant.

Influence of polymorphisms in toll-like receptors (TLRs) on malaria susceptibility in low-endemic area of the Atlantic Forest, São Paulo, Brazil

In low-endemic areas for malaria transmission, asymptomatic individuals play an important role as reservoirs for malarial infection. Understanding the dynamics of asymptomatic malaria is crucial for its efficient control in these regions. Genetic host factors such as Toll-like receptor (TLR) polymorphisms may play a role in the maintenance or elimination of infection. In this study, the effect of TLR polymorphisms on the susceptibility to malaria was investigated among individuals living in the Atlantic Forest of São Paulo, Southern Brazil. A hundred and ninety-five Brazilian individuals were enrolled and actively followed up for malaria for three years. Twenty-four polymorphisms in five toll-like receptor (TLR) genes were genotyped by RFLP, direct sequencing or fragment analysis. The genotypes were analyzed for the risk of malaria. Ongoing Plasmodium vivax or P. malariae infection, was identified by the positive results in PCR tests and previous P. vivax malaria, was assumed when antiplasmodial antibodies against PvMSP1₁₉ were detected by ELISA. An evaluation of genomic ancestry was conducted using biallelic ancestry informative markers and the results were used as correction in the statistical analysis. Nine SNPs and one microsatellite were found polymorphic and three variant alleles in TLR genes were associated to malaria susceptibility. The regression coefficient estimated for SNP TLR9.-1237.T/C indicated that the presence of at least one allele C increased, on average, 2.3 times the malaria odds, compared to individuals with no allele C in this SNP. However, for individuals with the same sex, age and household, the presence of at least one allele C in SNP TLR9.-1486.T/C reduced, on average, 1.9 times the malaria odds, compared to individuals with no allele C. Moreover, this allele C plus an S allele in TLR6.P249S in individuals with same sex, age and ancestry, reduced, on average, 4.4 times the malaria odds. Our findings indicate a significant association of TLR9.-1237.T/C gene polymorphism with malarial infection and contribute to a better knowledge of the role of TLRs in malaria susceptibility in an epidemiological setting different from other settings.

Separation in Logistic Regression: Causes, Consequences, and Control

Separation is encountered in regression models with a discrete outcome (such as logistic regression) where the covariates perfectly predict the outcome. It is most frequent under the same conditions that lead to small-sample and sparse-data bias, such as presence of a rare outcome, rare exposures, highly correlated covariates, or covariates with strong effects. In theory, separation will produce infinite estimates for some coefficients. In practice, however, separation may be unnoticed or mishandled because of software limits in recognizing and handling the problem and in notifying the user. We discuss causes of separation in logistic regression and describe how common software packages deal with it. We then describe methods that remove separation, focusing on the same penalized-likelihood techniques used to address more general sparse-data problems. These methods improve accuracy, avoid software problems, and allow interpretation as Bayesian analyses with weakly informative priors. We discuss likelihood penalties, including some that can be implemented easily with any software package, and their relative advantages and disadvantages. We provide an illustration of ideas and methods using data from a case-control study of contraceptive practices and urinary tract infection.

Predictors of Topiramate Tolerability in Heavy Cannabis-Using Adolescents and Young Adults: A Secondary Analysis of a Randomized, Double-Blind, Placebo-Controlled Trial

PURPOSE/BACKGROUND

Cannabis is the most commonly abused illicit drug and accounts for the greatest number of adolescent substance abuse treatment admissions. Despite urgent need for effective interventions, the best available psychosocial treatment options yield only modest effects. Topiramate showed promise as an adjunctive pharmacotherapy to a psychosocial intervention for cannabis misuse among adolescents and young adults in a recent clinical trial, but it was not well tolerated. This study investigated associations between clinical characteristics and side effects and dropout among adolescents and young adults randomized to topiramate.

METHODS

This study involved secondary data analysis of a randomized placebo-controlled trial of topiramate for treating cannabis misuse (ages, 15-24 years; 50% female). We explored the interaction effects of baseline characteristics and medication condition (topiramate vs placebo) on treatment dropout. We also explored the relationship between side effects and dropout.

FINDINGS/RESULTS

Higher cannabis problems were significantly associated with reduced hazard of dropout in the topiramate group (P = 0.048) and were nonsignificantly associated with increased hazard of dropout in the placebo group (P = 0.062). Results also showed that memory difficulties were an overwhelming predictor of dropout in the topiramate condition; 42% of participants who dropped out experienced memory difficulties, whereas none of those who remained in the study experienced these effects.

IMPLICATIONS/CONCLUSIONS

By identifying who may most benefit from and tolerate this medication, treatment for substance use disorders can become more individualized and positive outcomes may be enhanced.

The role of cuticular hydrocarbons in mate recognition in Drosophila suzukii

Cuticular hydrocarbons (CHCs) play a central role in the chemical communication of many insects. In Drosophila suzukii, an economically important pest insect, very little is known about chemical communication and the possible role of CHCs. In this study, we identified 60 CHCs of Drosophila suzukii and studied their changes in function of age (maturation), sex and interactions with the opposite sex. We demonstrate that age (maturation) is the key factor driving changes in the CHC profiles. We then test the effect on courtship behaviour and mating of six CHCs, five of which were positively associated with maturation and one negatively. The results of these experiments demonstrate that four of the major CHC peaks with a chain length of 23 carbons, namely 9-tricosene (9-C23:1), 7-tricosene (7-C23:1), 5-tricosene (5-C23:1) and tricosane (n-C23), negatively regulated courtship and mating, even though all these compounds were characteristic for sexually mature flies. We then go on to show that this effect on courtship and mating is likely due to the disruption of the natural ratios in which these hydrocarbons occur in Drosophila suzukii. Overall, these results provide key insights into the cuticular hydrocarbon signals that play a role in D. suzukii mate recognition.

Comparison of variance estimators for meta-analysis of instrumental variable estimates

Background: Mendelian randomization studies perform instrumental variable (IV) analysis using genetic IVs. Results of individual Mendelian randomization studies can be pooled through meta-analysis. We explored how different variance estimators influence the meta-analysed IV estimate.

Methods: Two versions of the delta method (IV before or after pooling), four bootstrap estimators, a jack-knife estimator and a heteroscedasticity-consistent (HC) variance estimator were compared using simulation. Two types of meta-analyses were compared, a two-stage meta-analysis pooling results, and a one-stage meta-analysis pooling datasets.

Results: Using a two-stage meta-analysis, coverage of the point estimate using bootstrapped estimators deviated from nominal levels at weak instrument settings and/or outcome probabilities ≤ 0.10. The jack-knife estimator was the least biased resampling method, the HC estimator often failed at outcome probabilities ≤ 0.50 and overall the delta method estimators were the least biased. In the presence of between-study heterogeneity, the delta method before meta-analysis performed best. Using a one-stage meta-analysis all methods performed equally well and better than two-stage meta-analysis of greater or equal size.

Conclusions: In the presence of between-study heterogeneity, two-stage meta-analyses should preferentially use the delta method before meta-analysis. Weak instrument bias can be reduced by performing a one-stage meta-analysis.

Temporal weights in the perception of sound intensity: Effects of sound duration and number of temporal segments

Loudness is a fundamental aspect of auditory perception that is closely related to the physical level of the sound. However, it has been demonstrated that, in contrast to a sound level meter, human listeners do not weight all temporal segments of a sound equally. Instead, the beginning of a sound is more important for loudness estimation than later temporal portions. The present study investigates the mechanism underlying this primacy effect by varying the number of equal-duration temporal segments (5 and 20) and the total duration of the sound (1.0 to 10.0 s) in a factorial design. Pronounced primacy effects were observed for all 20-segment sounds. The temporal weights for the five-segment sounds are similar to those for the 20-segment sounds when the weights of the segments covering the same temporal range as a segment of the five-segment sounds are averaged. The primacy effect can be described by an exponential decay function with a time constant of about 200 ms. Thus, the temporal weight assigned to a specific temporal portion of a sound is determined by the time delay between sound onset and segment onset rather than by the number of segments or the total duration of the sound.

Reducing Bruzzi's Formula to Remove Instability in the Estimation of Population Attributable Fraction for Health Outcomes

The aim of this study was to reconcile 3 approaches to calculating population attributable fractions and attributable burden percentage: the approach of Bruzzi et al. (Am J Epidemiol. 1985;122(5):904-914.), the maximum-likelihood method of Greenland and Drescher (Biometrics. 1993;49(3):865-872.), and the multivariable method of Tanuseputro et al. (Popul Health Metr. 2015;13:5.). Using data from a statewide point prevalence survey (Western Australian Point Prevalence Survey, 2014) linked to an administrative database, we compared estimates of attributable burden percentage obtained using the contrasting methods in 6 logistic models of health outcomes from the survey, estimating 95% confidence intervals using nonparametric and weighted bootstrap approaches. Our results show that instability can arise from the fundamental algebraic construction of Bruzzi's formula, and that this instability may substantially influence the calculation of attributable burden percentage and associated confidence intervals. These observations were confirmed in a simulation study. The algebraic reduction of Bruzzi's formula to the 2 alternative methods resulted in markedly more stable estimates for population attributable fraction and attributable burden percentage in cross-sectional studies and cohort designs with fixed follow-up time. We advocate the widespread implementation of the maximum-likelihood approach and the multivariable method.

Development of a Model for the Prediction of Treatment Response of Uterine Leiomyomas after Uterine Artery Embolization

Background: Uterine artery embolization (UAE) is one of the minimally-invasive alternatives to hysterectomy for treatment of uterine leiomyomas. There are various factors affecting the outcomes of UAE, but these have only been sporadically studied. Study Objective: To identify factors associated with the efficacy of UAE for the treatment of uterine leiomyoma, and to develop a model for the prediction of treatment response of uterine leiomyomas to UAE. Study design: A retrospective cohort study (Canadian Task Force Classification II-2) Patients: One hundred ninety-eight patients with symptomatic uterine leiomyomas. Intervention: UAE Measurements and Main Results: Among 198 leiomyoma patients who were treated with UAE, 104 who underwent pelvic magnetic resonance imaging (MRI) with diffusion-weighted imaging were selected for developing prediction model. Variables that were statistically significant from the univariate analysis were: location of leiomyoma, total number of lesions, sum of leiomyomas diameters, T2 signal intensity of largest leiomyoma, and T2 leiomyoma:muscle ratio. After a logistic regression analysis, leiomyoma location and T2 signal intensity of the largest leiomyoma were found to be statistically significant variables. Using intramural myomas defined as controls, submucosal leiomyomas showed a greater response to UAE with an odds ratio of 7.6904. The odds ratio of T2 signal intensity with an increase in signal intensity of 10 was 1.093. Using these two variables, we developed a prediction model. The AUC in the prediction model was 0.833, and the AUC in the validation set was 0.791. Conclusion: We identified that submucosal leiomyomas and those leiomyomas that show high signal intensity on T2-weighted imaging will exhibit a greater response to UAE. Prediction models are clinically helpful in selecting UAE as an appropriate treatment option for managing uterine leiomyoma.

Seasonality in the adverse outcomes in weight loss surgeries

BACKGROUND

Weight loss surgery is a common procedure in the United States.

OBJECTIVE

As weight loss surgery is largely an elective procedure for which patients and physicians can choose the timing, it could be helpful to explore the seasonality pattern of its perioperative adverse outcomes to help decide the timing of this surgery.

SETTING

United States.

METHODS

We used an obese adult sample (age ≥20 yr) of patients who underwent weight loss surgeries from the Premier Healthcare Database from 2011 to 2014. The International Classification of Diseases, Ninth Revision Clinical Modification procedure codes were used to identify weight loss surgery cases. Binary variables are used for 4 adverse outcomes, including hospital mortality, sepsis, deep vein thrombosis (DVT), and pulmonary embolism. The associations between the adverse outcomes and season of surgery were examined using logistic regressions, adjusting for age, sex, race, marital status, surgery types, body mass index, the Charlson co-morbidity index, and region.

RESULTS

A total of 69,365 weight loss surgeries were identified for the analytic sample. The overall rate was .27% for hospital mortality, .16% for DVT, .10% for pulmonary embolism, and .20% for sepsis. For DVT, adjusted odds ratio for the fall was 2.68 (95% confidence interval: 1.39-5.19) and the odds ratio for the winter was 2.26 (95% confidence interval: 1.09-4.27) compared with the summer. For sepsis, adjusted odds ratio for the spring was 1.83 (95% confidence interval: 1.07-3.12) compared with that of the summer. The seasonality pattern was not statistically significant for hospital mortality and pulmonary embolism.

CONCLUSION

DVT and sepsis are more likely to occur in colder seasons compared with the summer season, although the crude rates of these adverse events were low.

Percutaneous cannulation for cardiopulmonary bypass in minimally invasive surgery is associated with reduced groin complications

OBJECTIVES

Femoral cutdown is standard in most centres if groin cannulation is used for cardiopulmonary bypass in minimally invasive cardiac surgery (MICS). Arterial closure devices (ACDs) allow placement of larger cannulas percutaneously, but its benefit in MICS is unclear. We assessed our results with percutaneous groin cannulation using ACDs in comparison with conventional surgical access in patients undergoing MICS.

METHODS

We reviewed 445 consecutive patients having undergone MICS between October 2010 and March 2015. Of those, 92 (21%) were performed with conventional surgical access to the groin vessels and 353 (79%) with the use of ACDs.

RESULTS

Operative risk was higher in the ACD group [logistic EuroSCORE 7.9% (SD: 8.1) vs 10.6% (SD: 12.3); P = 0.010]. The use of ACDs significantly reduced operation time [193 min (SD: 43.8) vs 173 min (SD: 47.1); P < 0.001] and hospital stay [Cutdown: median 9 days (8, 14); ACD: median 9 days (7, 12), P = 0.040] without affecting the time to full mobilization. The incidence of any complication was significantly lower in the ACD group (2.3% vs 8.7%; P = 0.007). Complications with conventional cannulation consisted of lymphatic fistulae (n = 4), wound infections (n = 2), stenosis (n = 1) and haematoma (n = 1). In the ACD group, there were local dissections (n = 2) and stenoses (n = 3). There was 1 haematoma in both groups. There were 2 vascular injuries in the ACD group (n = 2), leading to conversion to surgical access.

CONCLUSIONS

Percutaneous groin cannulation using ACDs for establishing cardiopulmonary bypass in minimally invasive valve surgery significantly reduces groin complications, operation time and hospital stay. However, the remaining complications are mainly of vascular nature versus wound infection and lymph fistulae with cutdown.

Investigation of parameters predicting the need for diagnostic imaging beyond computed tomography in the evaluation of dogs with thoracolumbar myelopathy: Retrospective evaluation of 555 dogs

Thoracolumbar myelopathy encompasses a number of disease processes such as intervertebral disc disease, discospondylitis, trauma, congenital malformations, neoplasia, and intramedullary spinal cord disease. Compressive disc herniations are most common in dogs and require imaging procedures such as myelography, computed tomography (CT), and/or magnetic resonance imaging (MRI) to determine the need and location for decompressive surgery. The purposes of this retrospective, cross-sectional study were to evaluate all dogs undergoing thoracolumbar CT imaging as the initial diagnostic step between 2010 and 2015 and determine whether any of the imaging characteristics could be used to predict the need for additional imaging in the form of myelography, CT myelography, and/or MRI. A total of 555 dogs were identified in this time frame which underwent CT imaging for myelopathy of the thoracolumbar region. Various parameters including age, gender, sexual status, breed, chronicity, site of lesion, time of study, and contrast administration were evaluated. Findings indicated that 7.6% of dogs needed additional imaging after CT. Dachshunds were less likely to need additional imaging (P = 0.0111) as were patients scanned during normal business hours (P = 0.0075). Increasing age of the patient increased the likelihood of additional imaging (P = 0.0107). Dogs which did not have additional imaging performed were 21.89 times more likely to require surgery than those which did have additional imaging (P < 0.0001). Findings supported the use of CT as a first-line imaging modality for dogs presenting with thoracolumbar myelopathy.

East London Modified-Broset as Decision-Making Tool to Predict Seclusion in Psychiatric Intensive Care Units

Seclusion is a last resort intervention for management of aggressive behavior in psychiatric settings. There is no current objective and practical decision-making instrument for seclusion use on psychiatric wards. Our aim was to test the predictive and discriminatory characteristics of the East London Modified-Broset (ELMB), to delineate its decision-making profile for seclusion of adult psychiatric patients, and second to benchmark it against the psychometric properties of the Broset Violence Checklist (BVC). ELMB, an 8-item modified version of the 6-item BVC, was retrospectively employed to evaluate the seclusion decision-making process on two Psychiatric Intensive Care Units (patients n = 201; incidents n = 2,187). Data analyses were carried out using multivariate regression and Receiver Operating Characteristic (ROC) curves. Predictors of seclusion were: physical violence toward staff/patients OR = 24.2; non-compliance with PRN (pro re nata) medications OR = 9.8; and damage to hospital property OR = 2.9. ROC analyses indicated that ELMB was significantly more accurate that BVC, with higher sensitivity, specificity, and positive likelihood ratio. Results were similar across gender. The ELMB is a sensitive and specific instrument that can be used to guide the decision-making process when implementing seclusion.

Information criteria for Firth's penalized partial likelihood approach in Cox regression models

In the estimation of Cox regression models, maximum partial likelihood estimates might be infinite in a monotone likelihood setting, where partial likelihood converges to a finite value and parameter estimates converge to infinite values. To address monotone likelihood, previous studies have applied Firth's bias correction method to Cox regression models. However, while the model selection criteria for Firth's penalized partial likelihood approach have not yet been studied, a heuristic AIC-type information criterion can be used in a statistical package. Application of the heuristic information criterion to data obtained from a prospective observational study of patients with multiple brain metastases indicated that the heuristic information criterion selects models with many parameters and ignores the adequacy of the model. Moreover, we showed that the heuristic information criterion tends to select models with many regression parameters as the sample size increases. Thereby, in the present study, we propose an alternative AIC-type information criterion based on the risk function. A Bayesian information criterion type was also evaluated. Further, the presented simulation results confirm that the proposed criteria performed well in a monotone likelihood setting. The proposed AIC-type criterion was applied to prospective observational study data. Copyright © 2017 John Wiley & Sons, Ltd.

Only a small proportion of patients with first episode psychosis come via prodromal services: a retrospective survey of a large UK mental health programme

BACKGROUND

Little is known about patients with a first episode of psychosis (FEP) who had first presented to prodromal services with an "at risk mental state" (ARMS) before making the transition to psychosis. We set out to identify the proportion of patients with a FEP who had first presented to prodromal services in the ARMS state, and to compare these FEP patients with FEP patients who did not have prior contact with prodromal services.

METHODS

In this study information on 338 patients aged ≤37 years who presented to mental health services between 2010 and 2012 with a FEP was examined. The data on pathways to care, clinical and socio-demographic characteristics were extracted from the Biomedical Research Council Case Register for the South London and Maudsley NHS Trust.

RESULTS

Over 2 years, 14 (4.1% of n = 338) young adults presented with FEP and had been seen previously by the prodromal services. These ARMS patients were more likely to enter their pathway to psychiatric care via referral from General Practice, be born in the UK and to have had an insidious mode of illness onset than FEP patients without prior contact with the prodromal services.

CONCLUSIONS

In the current pathways to care configuration, prodromal services are likely to prevent only a few at-risk individuals from transitioning to psychosis even if effective preventative treatments become available.

Diclofenac affects kidney histology in the three-spined stickleback (Gasterosteus aculeatus) at low μg/L concentrations

Diclofenac, a commonly used non-steroidal anti-inflammatory drug, is considered for regulation under the European water framework directive. This is because effects on fish have been reported at concentrations around those regularly found in treated sewage effluents (∼1μg/L). However, a recent publication reports no effects on fish at 320μg/L. In this study, three-spined sticklebacks (Gasterosteus aculeatus) were exposed to 0, 4.6, 22, 82 and 271μg/L diclofenac in flow-through systems for 28days using triplicate aquaria per concentration. At the highest concentration, significant mortalities were observed already after 21days (no mortalities found up to 22μg/L). Histological analysis revealed a significant increase in the proportion of renal hematopoietic tissue (renal hematopoietic hyperplasia) after 28days at the lowest concentration and at all higher concentrations, following a clear dose-response pattern. Skin ulcerations of the jaw were noted by macroscopic observations, primarily at the two highest concentrations. No histological changes were observed in the liver. There was an increase in the relative hepatic mRNA levels of c7 (complement component 7), a gene involved in the innate immune system, at 22μg/L and at all higher concentrations, again following a clear dose-response. The bioconcentration factor was stable across concentrations, but lower than reported for rainbow trout, suggesting lower internal exposure to the drug in the stickleback. In conclusion, this study demonstrates that diclofenac causes histological changes in the three-spined stickleback at low μg/L concentrations, which cause concern for fish populations exposed to treated sewage effluents.

Prevalence of sexual dimorphism in mammalian phenotypic traits

The role of sex in biomedical studies has often been overlooked, despite evidence of sexually dimorphic effects in some biological studies. Here, we used high-throughput phenotype data from 14,250 wildtype and 40,192 mutant mice (representing 2,186 knockout lines), analysed for up to 234 traits, and found a large proportion of mammalian traits both in wildtype and mutants are influenced by sex. This result has implications for interpreting disease phenotypes in animal models and humans.

Systemic dissection of sexually dimorphic phenotypes in mice is lacking. Here, Karp and the International Mouse Phenotype Consortium show that approximately 10% of qualitative traits and 56% of quantitative traits in mice as measured in laboratory setting are sexually dimorphic.

Cabergoline and prolactinomas: lack of association between DRD2 polymorphisms and response to treatment

BACKGROUND

About 80% of prolactinomas respond to dopamine agonists (DA) with hormonal normalization and tumor shrinkage. Mechanisms of DA resistance include reduction of dopamine receptor subtype 2 (DRD2) expression, short and long isoform ratio and post-receptor mechanisms. It was suggested that polymorphisms in the gene encoding dopamine receptor subtype 2 gene (DRD2) could be associated with variable effectiveness of cabergoline (CAB).

OBJECTIVE

To assess the influence of DRD2 polymorphisms in responsiveness of CAB treatment in patients with prolactinoma.

STUDY DESIGN AND PATIENTS

Cross-sectional retrospective case-control study analyzing the frequency of five DRD2 polymorphisms in 148 patients with prolactinoma and 349 healthy subjects. The association of genetic variants and clinical characteristics with CAB responsiveness was performed in 118 patients (mean age at diagnosis 29 years; range 11-61 years) with hormonal evaluation. Patients with prolactin (PRL) normalization were considered as responders.

RESULTS

No association in genotypes and allele proportions was found comparing patients and controls. On pharmacogenetic study, 118 patients on CAB were included and 20% were non-responders. No association was found between clinical characteristics (gender, age, PRL level and tumor size at diagnosis) and polymorphisms of DRD2 with CAB responsiveness. Otherwise, there was association between polymorphisms rs1076560 (allele A) and rs1800497 (allele T) and the presence of macroadenomas.

CONCLUSION

No correlation was found between DRD2 polymorphisms and CAB responsiveness in patients with prolactinoma. More data are necessary in order to assess the influence of DRD2 genotyping on DA treatment response.

Changes in J-SOAP-II and SAVRY Scores Over the Course of Residential, Cognitive-Behavioral Treatment for Adolescent Sexual Offending

Although the Juvenile Sex Offender Assessment Protocol-II (J-SOAP-II) and the Structured Assessment of Violence Risk in Youth (SAVRY) include an emphasis on dynamic, or modifiable factors, there has been little research on dynamic changes on these tools. To help address this gap, we compared admission and discharge scores of 163 adolescents who attended a residential, cognitive-behavioral treatment program for sexual offending. Based on reliable change indices, one half of youth showed a reliable decrease on the J-SOAP-II Dynamic Risk Total Score and one third of youth showed a reliable decrease on the SAVRY Dynamic Risk Total Score. Contrary to expectations, decreases in risk factors and increases in protective factors did not predict reduced sexual, violent nonsexual, or any reoffending. In addition, no associations were found between scores on the Psychopathy Checklist:Youth Version and levels of change. Overall, the J-SOAP-II and the SAVRY hold promise in measuring change, but further research is needed.

Tetrodotoxin, Epinephrine, and Chemical Permeation Enhancer Combinations in Peripheral Nerve Blockade

BACKGROUND

Chemical permeation enhancers (CPEs) have the potential to improve nerve blockade by site 1 sodium channel blockers such as tetrodotoxin (TTX). Here, we investigated the efficacy and toxicity of CPE-enhanced nerve blockade across a range of TTX concentrations using 2 CPEs (sodium octyl sulfate and octyl trimethyl ammonium bromide). We also tested the hypothesis that CPEs could be used to reduce the concentrations of TTX and/or of a second adjuvant drug (in this case, epinephrine) needed to achieve prolonged local anesthesia METHODS:: Sprague-Dawley rats were injected at the sciatic nerve with combinations of TTX and CPEs, with and without epinephrine. Sensory and motor nerve blockade were assessed using a modified hot plate test and a weight-bearing test, respectively. Systemic and local toxicities of the different combinations were assessed.

RESULTS

Addition of increasing concentrations of TTX to fixed concentrations of CPEs produced a marked concentration-dependent improvement in the rate of successful nerve blocks and in nerve block duration. CPEs did not affect systemic toxicity. At some concentrations, the addition of sodium octyl sulfate increased the duration of block from TTX plus epinephrine, and epinephrine increased that from TTX plus CPEs. The addition of epinephrine did not cause an increase in local toxicity, and it markedly reduced systemic toxicity.

CONCLUSIONS

CPEs can prolong the duration of nerve blockade across a range of concentrations of TTX. CPEs could also be used to reduce the concentration of epinephrine needed to achieve a given degree of nerve block. CPEs may be useful in enhancing nerve blockade from site 1 sodium channel blockers.

Correcting hazard ratio estimates for outcome misclassification using multiple imputation with internal validation data

OBJECTIVE

Outcome misclassification may occur in observational studies using administrative databases. We evaluated a two-step multiple imputation approach based on complementary internal validation data obtained from two subsamples of study participants to reduce bias in hazard ratio (HR) estimates in Cox regressions.

METHODS

We illustrated this approach using data from a surveyed sample of 6247 individuals in a study of statin-diabetes association in Quebec. We corrected diabetes status and onset assessed from health administrative data against self-reported diabetes and/or elevated fasting blood glucose (FBG) assessed in subsamples. The association between statin use and new onset diabetes was evaluated using administrative data and the corrected data. By simulation, we assessed the performance of this method varying the true HR, sensitivity, specificity, and the size of validation subsamples.

RESULTS

The adjusted HR of new onset diabetes among statin users versus non-users was 1.61 (95% confidence interval: 1.09-2.38) using administrative data only, 1.49 (0.95-2.34) when diabetes status and onset were corrected based on self-report and undiagnosed diabetes (FBG ≥ 7 mmol/L), and 1.36 (0.92-2.01) when corrected for self-report and undiagnosed diabetes/impaired FBG (≥ 6 mmol/L). In simulations, the multiple imputation approach yielded less biased HR estimates and appropriate coverage for both non-differential and differential misclassification. Large variations in the corrected HR estimates were observed using validation subsamples with low participation proportion. The bias correction was sometimes outweighed by the uncertainty introduced by the unknown time of event occurrence.

CONCLUSION

Multiple imputation is useful to correct for outcome misclassification in time-to-event analyses if complementary validation data are available from subsamples. Copyright © 2017 John Wiley & Sons, Ltd.

Autophagy-related polymorphisms predict hypertension in patients with metastatic colorectal cancer treated with FOLFIRI and bevacizumab: Results from TRIBE and FIRE-3 trials

PURPOSE

The most frequent bevacizumab-related side-effects are hypertension, proteinuria, bleeding and thromboembolism. To date, there is no biomarker that predicts anti-VEGF-associated toxicity. As autophagy inhibits angiogenesis, we hypothesised that single-nucleotide polymorphisms (SNPs) within autophagy-related genes may predict bevacizumab-mediated toxicity in patients with metastatic colorectal cancer (mCRC).

PATIENTS AND METHODS

Patients with mCRC treated with first-line FOLFIRI and bevacizumab in two phase III randomised trials, namely the TRIBE trial (n = 219, discovery cohort) and the FIRE-3 trial (n = 234, validation cohort) were included in this study. Patients receiving treatment with FOLFIRI and cetuximab (FIRE-3, n = 204) served as a negative control. 12 SNPs in eight autophagy-related genes (ATG3/5/8/13, beclin 1, FIP200, unc-51-like kinase 1, UVRAG) were analysed by PCR-based direct sequencing.

RESULTS

The FIP200 rs1129660 variant showed significant associations with hypertension in the TRIBE cohort. Patients harbouring any G allele of the FIP200 rs1129660 SNP showed a significantly lower rate of grade 2-3 hypertension compared with the A/A genotype (3% versus 15%, odds ratio [OR] 0.17; 95% confidence interval [CI], 0.02-0.73; P = 0.009). Similarly, G allele carriers of the FIP200 rs1129660 SNP were less likely to develop grade 2-3 hypertension than patients with an A/A genotype in the FIRE-3 validation cohort (9% versus 20%, OR 0.43; 95% CI, 0.14-1.11; P = 0.077), whereas this association could not be observed in the control cohort (12% versus 9%, OR 1.40; 95% CI, 0.45-4.04; P = 0.60).

CONCLUSION

This is the first report demonstrating that polymorphisms in the autophagy-related FIP200 gene may predict hypertension in patients with mCRC treated with FOLFIRI and bevacizumab.

Big Data, Predictive Analytics, and Quality Improvement in Kidney Transplantation: A Proof of Concept

We sought proof of concept of a Big Data Solution incorporating longitudinal structured and unstructured patient-level data from electronic health records (EHR) to predict graft loss (GL) and mortality. For a quality improvement initiative, GL and mortality prediction models were constructed using baseline and follow-up data (0-90 days posttransplant; structured and unstructured for 1-year models; data up to 1 year for 3-year models) on adult solitary kidney transplant recipients transplanted during 2007-2015 as follows: Model 1: United Network for Organ Sharing (UNOS) data; Model 2: UNOS & Transplant Database (Tx Database) data; Model 3: UNOS, Tx Database & EHR comorbidity data; and Model 4: UNOS, Tx Database, EHR data, Posttransplant trajectory data, and unstructured data. A 10% 3-year GL rate was observed among 891 patients (2007-2015). Layering of data sources improved model performance; Model 1: area under the curve (AUC), 0.66; (95% confidence interval [CI]: 0.60, 0.72); Model 2: AUC, 0.68; (95% CI: 0.61-0.74); Model 3: AUC, 0.72; (95% CI: 0.66-077); Model 4: AUC, 0.84, (95 % CI: 0.79-0.89). One-year GL (AUC, 0.87; Model 4) and 3-year mortality (AUC, 0.84; Model 4) models performed similarly. A Big Data approach significantly adds efficacy to GL and mortality prediction models and is EHR deployable to optimize outcomes.

Interaction Landscape of Inherited Polymorphisms with Somatic Events in Cancer

Recent studies have characterized the extensive somatic alterations that arise during cancer. However, the somatic evolution of a tumor may be significantly affected by inherited polymorphisms carried in the germline. Here, we analyze genomic data for 5,954 tumors to reveal and systematically validate 412 genetic interactions between germline polymorphisms and major somatic events, including tumor formation in specific tissues and alteration of specific cancer genes. Among germline-somatic interactions, we found germline variants in RBFOX1 that increased incidence of SF3B1 somatic mutation by 8-fold via functional alterations in RNA splicing. Similarly, 19p13.3 variants were associated with a 4-fold increased likelihood of somatic mutations in PTEN. In support of this association, we found that PTEN knockdown sensitizes the MTOR pathway to high expression of the 19p13.3 gene GNA11 Finally, we observed that stratifying patients by germline polymorphisms exposed distinct somatic mutation landscapes, implicating new cancer genes. This study creates a validated resource of inherited variants that govern where and how cancer develops, opening avenues for prevention research.Significance: This study systematically identifies germline variants that directly affect tumor evolution, either by dramatically increasing alteration frequency of specific cancer genes or by influencing the site where a tumor develops. Cancer Discovery; 7(4); 410-23. ©2017 AACR.See related commentary by Geeleher and Huang, p. 354This article is highlighted in the In This Issue feature, p. 339.

Detecting Differential Item Functioning Using the Logistic Regression Procedure in Small Samples

The logistic regression (LR) procedure for testing differential item functioning (DIF) typically depends on the asymptotic sampling distributions. The likelihood ratio test (LRT) usually relies on the asymptotic chi-square distribution. Also, the Wald test is typically based on the asymptotic normality of the maximum likelihood (ML) estimation, and the Wald statistic is tested using the asymptotic chi-square distribution. However, in small samples, the asymptotic assumptions may not work well. The penalized maximum likelihood (PML) estimation removes the first-order finite sample bias from the ML estimation, and the bootstrap method constructs the empirical sampling distribution. This study compares the performances of the LR procedures based on the LRT, Wald test, penalized likelihood ratio test (PLRT), and bootstrap likelihood ratio test (BLRT) in terms of the statistical power and type I error for testing uniform and non-uniform DIF. The result of the simulation study shows that the LRT with the asymptotic chi-square distribution works well even in small samples.