Pooling overdispersed binomial data to estimate event rate

Beta-Binomial Statistical Model for Validation Studies of Analytes with a Binary Response

BACKGROUND

The probability of detection (POD) model has had widespread application for statistically analyzing single and multiple collaborator validations studies with binary outcome data for a wide range of analytes over the last decade.

OBJECTIVE

The POD model is placed on a firm theoretical foundation, and extended to a more generalized beta-binomial model.

METHODS

The POD model is revisited and embedded in the beta-binomial model. This generalization includes collaborator reproducibility as a specific parameter. The new model includes only two distributional parameters: the overall across-collaborator probability of detection (LPOD) and the intraclass correlation of collaborators (ICC), measuring irreproducibility. Differences between methods are measured by the difference in LPOD values, denoted dLPOD.

RESULTS

Accurate statistical estimators and confidence intervals are provided with validation by simulation. This new beta-binomial model will be applicable to a full range of candidate methods giving binary qualitative results, including microbiological, toxin, allergen, biothreat, and botanical analytes.

CONCLUSIONS

The new beta-binomial model provides easy equivalence tests to show the study clearly demonstrates (with 95% confidence) that the method differences and collaborator reproducibility are acceptable.

HIGHLIGHTS

The validation system for qualitative binary methods using probability of detection (POD) of an analyte as the parameter of interest has been modified and further validated.

Prevalence and Etiological Characteristics of Norovirus Infection in China: A Systematic Review and Meta-Analysis

Norovirus is a common cause of sporadic cases and outbreaks of gastroenteritis worldwide, although its prevalence and the dominant genotypes responsible for gastroenteritis outbreaks remain obscure. A systematic review was conducted on norovirus infection in China between January 2009 and March 2021. A meta-analysis and beta-binomial regression model were used to explore the epidemiological and clinical characteristics of norovirus infection and the potential factors contributing to the attack rate of the norovirus outbreaks, respectively. A total of 1132 articles with 155,865 confirmed cases were included, with a pooled positive test rate of 11.54% among 991,786 patients with acute diarrhea and a pooled attack rate of 6.73% in 500 norovirus outbreaks. GII.4 was the predominant genotype in both the etiological surveillance and outbreaks, followed by GII.3 in the etiological surveillance, and GII.17 in the outbreaks, with the proportion of recombinant genotypes increasing in recent years. A higher attack rate in the norovirus outbreaks was associated with age group (older adults), settings (nurseries, primary schools, etc.) and region (North China). The nation-wide pooled positive rate in the etiological surveillance of norovirus is lower than elsewhere in the global population, while the dominant genotypes are similar in both the etiological surveillance and the outbreak investigations. This study contributes to the understanding of norovirus infection with different genotypes in China. The prevention and control of norovirus outbreaks during the cold season should be intensified, with special attention paid to and enhanced surveillance performed in nurseries, schools and nursing homes from November to March.

Benefits and Challenges of Pre-clustered Network-Based Pathway Analysis

Functional analysis of gene sets derived from experiments is typically done by pathway annotation. Although many algorithms exist for analyzing the association between a gene set and a pathway, an issue which is generally ignored is that gene sets often represent multiple pathways. In such cases an association to a pathway is weakened by the presence of genes associated with other pathways. A way to counteract this is to cluster the gene set into more homogenous parts before performing pathway analysis on each module. We explored whether network-based pre-clustering of a query gene set can improve pathway analysis. The methods MCL, Infomap, and MGclus were used to cluster the gene set projected onto the FunCoup network. We characterized how well these methods are able to detect individual pathways in multi-pathway gene sets, and applied each of the clustering methods in combination with four pathway analysis methods: Gene Enrichment Analysis, BinoX, NEAT, and ANUBIX. Using benchmarks constructed from the KEGG pathway database we found that clustering can be beneficial by increasing the sensitivity of pathway analysis methods and by providing deeper insights of biological mechanisms related to the phenotype under study. However, keeping a high specificity is a challenge. For ANUBIX, clustering caused a minor loss of specificity, while for BinoX and NEAT it caused an unacceptable loss of specificity. GEA had very low sensitivity both before and after clustering. The choice of clustering method only had a minor effect on the results. We show examples of this approach and conclude that clustering can improve overall pathway annotation performance, but should only be used if the used enrichment method has a low false positive rate.

Revisiting the analysis pipeline for overdispersed Poisson and binomial data

Overdispersion is a common feature in categorical data analysis and several methods have been developed for detecting and handling it in generalized linear models. The first aim of this study is to clarify the relationships among various score statistics for testing overdispersion and to compare their performances. In addition, we investigate a principled way to correct finite sample bias in the score statistic caused by estimating regression parameters with restricted likelihood. The second aim is to reconsider the current practice for handling overdispersed categorical data. Although the conventional models are based on substantially different mechanisms for generating overdispersion, model selection in practice has not been well studied. We perform an intensive numerical study for determining which method is more robust to various overdispersion mechanisms. In addition, we provide some graphical tools for identifying the better model. The last aim is to reconsider the key assumption for deriving the score statistics. We study the meaning of testing overdispersion when this assumption is violated, and we analytically show the conditions for which it is not appropriate to employ the current statistical practices for analyzing overdispersed data.

Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) Setting-specific Transmission Rates: A Systematic Review and Meta-analysis

BACKGROUND

Understanding the drivers of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) transmission is crucial for control policies, but evidence of transmission rates in different settings remains limited.

METHODS

We conducted a systematic review to estimate secondary attack rates (SARs) and observed reproduction numbers (Robs) in different settings exploring differences by age, symptom status, and duration of exposure. To account for additional study heterogeneity, we employed a beta-binomial model to pool SARs across studies and a negative-binomial model to estimate Robs.

RESULTS

Households showed the highest transmission rates, with a pooled SAR of 21.1% (95% confidence interval [CI]:17.4-24.8). SARs were significantly higher where the duration of household exposure exceeded 5 days compared with exposure of ≤5 days. SARs related to contacts at social events with family and friends were higher than those for low-risk casual contacts (5.9% vs 1.2%). Estimates of SARs and Robs for asymptomatic index cases were approximately one-seventh, and for presymptomatic two-thirds of those for symptomatic index cases. We found some evidence for reduced transmission potential both from and to individuals younger than 20 years of age in the household context, which is more limited when examining all settings.

CONCLUSIONS

Our results suggest that exposure in settings with familiar contacts increases SARS-CoV-2 transmission potential. Additionally, the differences observed in transmissibility by index case symptom status and duration of exposure have important implications for control strategies, such as contact tracing, testing, and rapid isolation of cases. There were limited data to explore transmission patterns in workplaces, schools, and care homes, highlighting the need for further research in such settings.

Pathway-specific model estimation for improved pathway annotation by network crosstalk

Pathway enrichment analysis is the most common approach for understanding which biological processes are affected by altered gene activities under specific conditions. However, it has been challenging to find a method that efficiently avoids false positives while keeping a high sensitivity. We here present a new network-based method ANUBIX based on sampling random gene sets against intact pathway. Benchmarking shows that ANUBIX is considerably more accurate than previous network crosstalk based methods, which have the drawback of modelling pathways as random gene sets. We demonstrate that ANUBIX does not have a bias for finding certain pathways, which previous methods do, and show that ANUBIX finds biologically relevant pathways that are missed by other methods.

Exact inference for the random-effect model for meta-analyses with rare events

Meta-analysis allows for the aggregation of results from multiple studies to improve statistical inference for the parameter of interest. In recent years, random-effect meta-analysis has been employed to synthesize estimates of incidence rates of adverse events across heterogeneous clinical trials to evaluate treatment safety. However, the validity of existing approaches relies on asymptotic approximation as the number of studies becomes large. In practice, a limited number of trials are typically available for analysis. Moreover, adverse events are typically rare; thus, study-specific incidence rate estimates may be unstable or undefined. In this paper, we present a method for construction of an exact confidence interval for the location parameter of the beta-binomial model through inversion of exact tests. The coverage level of the proposed confidence interval is guaranteed to achieve at least the nominal level, regardless of the number of studies or the with-in study sample size, making it particularly applicable to the study of rare-event data.

Overdispersion models for correlated multinomial data: Applications to blinding assessment

Overdispersion models have been extensively studied for correlated normal and binomial data but much less so for correlated multinomial data. In this work, we describe a multinomial overdispersion model that leads to the specification of the first two moments of the outcome and allows the estimation of the global parameters using generalized estimating equations (GEE). We introduce a Global Blinding Index as a target parameter and illustrate the application of the GEE method to its estimation from (1) a clinical trial with clustering by practitioner and (2) a meta-analysis on psychiatric disorders. We examine the impact of a small number of clusters, high variability in cluster sizes, and the magnitude of the intraclass correlation on the performance of the GEE estimators of the Global Blinding Index using the data simulated from different models. We compare these estimators with the inverse-variance weighted estimators and a maximum-likelihood estimator, derived under the Dirichlet-multinomial model. Our results indicate that the performance of the GEE estimators was satisfactory even in situations with a small number of clusters, whereas the inverse-variance weighted estimators performed poorly, especially for larger values of the intraclass correlation coefficient. Our findings and illustrations may be instrumental for practitioners who analyze clustered multinomial data from clinical trials and/or meta-analysis.

Chlamydia, gonorrhoea, trichomoniasis and syphilis: global prevalence and incidence estimates, 2016

OBJECTIVE

To generate estimates of the global prevalence and incidence of urogenital infection with chlamydia, gonorrhoea, trichomoniasis and syphilis in women and men, aged 15-49 years, in 2016.

METHODS

For chlamydia, gonorrhoea and trichomoniasis, we systematically searched for studies conducted between 2009 and 2016 reporting prevalence. We also consulted regional experts. To generate estimates, we used Bayesian meta-analysis. For syphilis, we aggregated the national estimates generated by using Spectrum-STI.

FINDINGS

For chlamydia, gonorrhoea and/or trichomoniasis, 130 studies were eligible. For syphilis, the Spectrum-STI database contained 978 data points for the same period. The 2016 global prevalence estimates in women were: chlamydia 3.8% (95% uncertainty interval, UI: 3.3-4.5); gonorrhoea 0.9% (95% UI: 0.7-1.1); trichomoniasis 5.3% (95% UI:4.0-7.2); and syphilis 0.5% (95% UI: 0.4-0.6). In men prevalence estimates were: chlamydia 2.7% (95% UI: 1.9-3.7); gonorrhoea 0.7% (95% UI: 0.5-1.1); trichomoniasis 0.6% (95% UI: 0.4-0.9); and syphilis 0.5% (95% UI: 0.4-0.6). Total estimated incident cases were 376.4 million: 127.2 million (95% UI: 95.1-165.9 million) chlamydia cases; 86.9 million (95% UI: 58.6-123.4 million) gonorrhoea cases; 156.0 million (95% UI: 103.4-231.2 million) trichomoniasis cases; and 6.3 million (95% UI: 5.5-7.1 million) syphilis cases.

CONCLUSION

Global estimates of prevalence and incidence of these four curable sexually transmitted infections remain high. The study highlights the need to expand data collection efforts at country level and provides an initial baseline for monitoring progress of the World Health Organization global health sector strategy on sexually transmitted infections 2016-2021.

Generalized Empirical Bayes Modeling via Frequentist Goodness of Fit

The two key issues of modern Bayesian statistics are: (i) establishing principled approach for distilling statistical prior that is consistent with the given data from an initial believable scientific prior; and (ii) development of a consolidated Bayes-frequentist data analysis workflow that is more effective than either of the two separately. In this paper, we propose the idea of “Bayes via goodness-of-fit” as a framework for exploring these fundamental questions, in a way that is general enough to embrace almost all of the familiar probability models. Several examples, spanning application areas such as clinical trials, metrology, insurance, medicine, and ecology show the unique benefit of this new point of view as a practical data science tool.

Beta-binomial model for meta-analysis of odds ratios

In meta-analysis of odds ratios (ORs), heterogeneity between the studies is usually modelled via the additive random effects model (REM). An alternative, multiplicative REM for ORs uses overdispersion. The multiplicative factor in this overdispersion model (ODM) can be interpreted as an intra-class correlation (ICC) parameter. This model naturally arises when the probabilities of an event in one or both arms of a comparative study are themselves beta-distributed, resulting in beta-binomial distributions. We propose two new estimators of the ICC for meta-analysis in this setting. One is based on the inverted Breslow-Day test, and the other on the improved gamma approximation by Kulinskaya and Dollinger (2015, p. 26) to the distribution of Cochran's Q. The performance of these and several other estimators of ICC on bias and coverage is studied by simulation. Additionally, the Mantel-Haenszel approach to estimation of ORs is extended to the beta-binomial model, and we study performance of various ICC estimators when used in the Mantel-Haenszel or the inverse-variance method to combine ORs in meta-analysis. The results of the simulations show that the improved gamma-based estimator of ICC is superior for small sample sizes, and the Breslow-Day-based estimator is the best for n⩾100. The Mantel-Haenszel-based estimator of OR is very biased and is not recommended. The inverse-variance approach is also somewhat biased for ORs≠1, but this bias is not very large in practical settings. Developed methods and R programs, provided in the Web Appendix, make the beta-binomial model a feasible alternative to the standard REM for meta-analysis of ORs. © 2017 The Authors. Statistics in Medicine Published by John Wiley & Sons Ltd.

Global Estimates of the Prevalence and Incidence of Four Curable Sexually Transmitted Infections in 2012 Based on Systematic Review and Global Reporting

BACKGROUND

Quantifying sexually transmitted infection (STI) prevalence and incidence is important for planning interventions and advocating for resources. The World Health Organization (WHO) periodically estimates global and regional prevalence and incidence of four curable STIs: chlamydia, gonorrhoea, trichomoniasis and syphilis.

METHODS AND FINDINGS

WHO's 2012 estimates were based upon literature reviews of prevalence data from 2005 through 2012 among general populations for genitourinary infection with chlamydia, gonorrhoea, and trichomoniasis, and nationally reported data on syphilis seroprevalence among antenatal care attendees. Data were standardized for laboratory test type, geography, age, and high risk subpopulations, and combined using a Bayesian meta-analytic approach. Regional incidence estimates were generated from prevalence estimates by adjusting for average duration of infection. In 2012, among women aged 15-49 years, the estimated global prevalence of chlamydia was 4.2% (95% uncertainty interval (UI): 3.7-4.7%), gonorrhoea 0.8% (0.6-1.0%), trichomoniasis 5.0% (4.0-6.4%), and syphilis 0.5% (0.4-0.6%); among men, estimated chlamydia prevalence was 2.7% (2.0-3.6%), gonorrhoea 0.6% (0.4-0.9%), trichomoniasis 0.6% (0.4-0.8%), and syphilis 0.48% (0.3-0.7%). These figures correspond to an estimated 131 million new cases of chlamydia (100-166 million), 78 million of gonorrhoea (53-110 million), 143 million of trichomoniasis (98-202 million), and 6 million of syphilis (4-8 million). Prevalence and incidence estimates varied by region and sex.

CONCLUSIONS

Estimates of the global prevalence and incidence of chlamydia, gonorrhoea, trichomoniasis, and syphilis in adult women and men remain high, with nearly one million new infections with curable STI each day. The estimates highlight the urgent need for the public health community to ensure that well-recognized effective interventions for STI prevention, screening, diagnosis, and treatment are made more widely available. Improved estimation methods are needed to allow use of more varied data and generation of estimates at the national level.

LARVA: an integrative framework for large-scale analysis of recurrent variants in noncoding annotations

In cancer research, background models for mutation rates have been extensively calibrated in coding regions, leading to the identification of many driver genes, recurrently mutated more than expected. Noncoding regions are also associated with disease; however, background models for them have not been investigated in as much detail. This is partially due to limited noncoding functional annotation. Also, great mutation heterogeneity and potential correlations between neighboring sites give rise to substantial overdispersion in mutation count, resulting in problematic background rate estimation. Here, we address these issues with a new computational framework called LARVA. It integrates variants with a comprehensive set of noncoding functional elements, modeling the mutation counts of the elements with a β-binomial distribution to handle overdispersion. LARVA, moreover, uses regional genomic features such as replication timing to better estimate local mutation rates and mutational hotspots. We demonstrate LARVA's effectiveness on 760 whole-genome tumor sequences, showing that it identifies well-known noncoding drivers, such as mutations in the TERT promoter. Furthermore, LARVA highlights several novel highly mutated regulatory sites that could potentially be noncoding drivers. We make LARVA available as a software tool and release our highly mutated annotations as an online resource (larva.gersteinlab.org).

Fosmidomycin as an antimalarial drug: a meta-analysis of clinical trials

With first indications of resistance against artemisinin compounds, the development of novel alternative antimalarials remains an urgent need. One candidate is fosmidomycin (Fos), a phosphonic acid derivative. This PRISMA guideline-adhering and PROSPERO-registered systematic review and meta-analysis provides an overview of the state-of-the-art of the clinical development of Fos as an antimalarial. Pooling six clinical trials of Fos against uncomplicated malaria in African children yielded an overall day 28 cure rate of 85% (95% CI: 71-98%); a parasite clearance time of 39 h; and a fever clearance time of 30 h. In four adult cohorts, the corresponding values were 70% (95% CI: 40-100%), 49 and 42 h, respectively. Data suggest that besides the partner drug, formulation determines efficacy. We advocate further clinical development of Fos-combinations. PROSPERO registration number: CRD42014013688.

Estimating influenza disease burden from population-based surveillance data in the United States

Annual estimates of the influenza disease burden provide information to evaluate programs and allocate resources. We used a multiplier method with routine population-based surveillance data on influenza hospitalization in the United States to correct for under-reporting and estimate the burden of influenza for seasons after the 2009 pandemic. Five sites of the Influenza Hospitalization Surveillance Network (FluSurv-NET) collected data on the frequency and sensitivity of influenza testing during two seasons to estimate under-detection. Population-based rates of influenza-associated hospitalization and Intensive Care Unit admission from 2010-2013 were extrapolated to the U.S. population from FluSurv-NET and corrected for under-detection. Influenza deaths were calculated using a ratio of deaths to hospitalizations. We estimated that influenza-related hospitalizations were under-detected during 2010-11 by a factor of 2.1 (95%CI 1.7-2.9) for age < 18 years, 3.1 (2.4-4.5) for ages 18-64 years, and 5.2 (95%CI 3.8-8.3) for age 65+. Results were similar in 2011-12. Extrapolated estimates for 3 seasons from 2010-2013 included: 114,192-624,435 hospitalizations, 18,491-95,390 ICU admissions, and 4,915-27,174 deaths per year; 54-70% of hospitalizations and 71-85% of deaths occurred among adults aged 65+. Influenza causes a substantial disease burden in the U.S. that varies by age and season. Periodic estimation of multipliers across multiple sites and age groups improves our understanding of influenza detection in sentinel surveillance systems. Adjusting surveillance data using a multiplier method is a relatively simple means to estimate the impact of influenza and the subsequent value of interventions to prevent influenza.

O father where art thou? Paternity analyses in a natural population of the haploid-diploid seaweed Chondrus crispus

The link between life history traits and mating systems in diploid organisms has been extensively addressed in the literature, whereas the degree of selfing and/or inbreeding in natural populations of haploid-diploid organisms, in which haploid gametophytes alternate with diploid sporophytes, has been rarely measured. Dioecy has often been used as a proxy for the mating system in these organisms. Yet, dioecy does not prevent the fusion of gametes from male and female gametophytes originating from the same sporophyte. This is likely a common occurrence when spores from the same parent are dispersed in clumps and recruit together. This pattern of clumped spore dispersal has been hypothesized to explain significant heterozygote deficiency in the dioecious haploid-diploid seaweed Chondrus crispus. Fronds and cystocarps (structures in which zygotes are mitotically amplified) were sampled in two 25 m(2) plots located within a high and a low intertidal zone and genotyped at 5 polymorphic microsatellite loci in order to explore the mating system directly using paternity analyses. Multiple males sired cystocarps on each female, but only one of the 423 paternal genotypes corresponded to a field-sampled gametophyte. Nevertheless, larger kinship coefficients were detected between males siring cystocarps on the same female in comparison with males in the entire population, confirming restricted spermatial and clumped spore dispersal. Such dispersal mechanisms may be a mode of reproductive assurance due to nonmotile gametes associated with putatively reduced effects of inbreeding depression because of the free-living haploid stage in C. crispus.

Meta-analysis of Proportions of Rare Events-A Comparison of Exact Likelihood Methods with Robust Variance Estimation

The conventional random effects model for meta-analysis of proportions approximates within-study variation using a normal distribution. Due to potential approximation bias, particularly for the estimation of rare events such as some adverse drug reactions, the conventional method is considered inferior to the exact methods based on binomial distributions. In this paper, we compare two existing exact approaches-beta binomial (B-B) and normal-binomial (N-B)-through an extensive simulation study with focus on the case of rare events that are commonly encountered in medical research. In addition, we implement the empirical ("sandwich") estimator of variance into the two models to improve the robustness of the statistical inferences. To our knowledge, it is the first such application of sandwich estimator of variance to meta-analysis of proportions. The simulation study shows that the B-B approach tends to have substantially smaller bias and mean squared error than N-B for rare events with occurrences under five percent, while N-B outperforms B-B for relatively common events. Use of the sandwich estimator of variance improves the precision of estimation for both models. We illustrate the two approaches by applying them to two published meta-analysis from the fields of orthopedic surgery and prevention of adverse drug reactions.

Understanding epidemiological transition in India

Background

Omran's theory explains changing disease patterns over time predominantly from infectious to chronic noncommunicable diseases (NCDs). India's epidemiological transition is characterized by dual burden of diseases. Kumar addressed low mortality and high morbidity in Kerala, which seems also to be true for India as a country in the current demographic scenario.

Methods

NSS data (1986–1987, 1995–1996, 2004) and aggregated data on causes of death provided by Registrar General India (RGI) were used to examine the structural changes in morbidity and causes of death. A zero-inflated poisson (ZIP) regression model and a beta-binomial model were used to corroborate the mounting age pattern of morbidity. Measures, namely the 25th and 75th percentiles of age-at-death and modal age-at-death, were used to examine the advances in mortality transition.

Objective

This study addressed the advances in epidemiological transition via exploring the structural changes in pattern of diseases and progress in mortality transition.

Results

The burden of NCDs has been increasing in old age without replacing the burden of communicable diseases. The manifold rise of chronic diseases in recent decades justifies the death toll and is responsible for transformation in the age pattern of morbidity. Over time, deaths have been concentrated near the modal age-at-death. Modal age-at-death increased linearly by 5 years for females (r²=0.9515) and males (r²=0.9020). Significant increase in modal age-at-death ascertained the dominance of old age mortality over the childhood/adult age mortality.

Conclusions

India experiences a dual burden of diseases associated with a remarkable transformation in the age pattern of morbidity and mortality, contemporaneous with structural changes in disease patterns. Continued progress in the pattern of diseases and mortality transition, accompanied by a linear rise in e_x, unravels a compelling variation in advances found so far in epidemiological transition witnessed by the developed nations, with similar matrices for India.

Intravenous tissue-type plasminogen activator therapy is an independent risk factor for symptomatic intracerebral hemorrhage after carotid endarterectomy

BACKGROUND

Carotid endarterectomy (CEA) for symptomatic carotid artery stenosis and intravenous tissue-type plasminogen activator (IV-tPA) for acute ischemic stroke are proven therapies; however, the safety of CEA in stroke patients who recently received IV-tPA has not been established.

OBJECTIVE

To evaluate the safety of CEA in stroke patients who recently received IV-tPA.

METHODS

A retrospective review of patients who underwent CEA for symptomatic carotid artery stenosis was performed. The primary end point was postoperative symptomatic intracerebral hemorrhage (sICH). A univariate analysis of potential risk factors for sICH, including IV-tPA therapy, timing of CEA, degree of stenosis, and stroke severity, was performed. Factors with a value of P < .1 on univariate analysis were tested further.

RESULTS

Among 142 patients, 3 suffered sICH after CEA: 2 of 11 patients treated with IV-tPA (18.2%) and 1 of 131 patients not treated with IV-tPA (0.8%). Both IV-tPA patients suffering sICH underwent CEA within 3 days of tPA administration. On univariate analysis, IV-tPA (P = .02), female sex (P = .09), shorter time between ischemic event and CEA (P = .06), and lower mean arterial pressure during the first 48 hours of admission (P = .08) were identified as risk factors for sICH. On multivariate analysis, IV-tPA was the only significant risk factor (P = .002 by stepwise logistic regression; P = .03 by nominal logistic regression).

CONCLUSION

This study indicates that IV-tPA is an independent risk factor for sICH after CEA. This suggests that CEA should be pursued cautiously in patients who recently received IV-tPA. Early surgery may be associated with an increased risk for sICH.

ABBREVIATIONS

CEA, carotid endarterectomyIV-tPA, intravenous recombinant tissue-type plasminogen activatorMAP, mean arterial pressureNASCET, North American Symptomatic Carotid Endarterectomy TrialNIHSS, National Institutes of Health Stroke ScaleNINDS, National Institute of Neurological Disorders and StrokesICH, symptomatic intracerebral hemorrhageTIA, transient ischemic attack.

An overdispersion model in meta-analysis

We propose a general approach based on the concept of overdispersion for specification of a random effects model (REM) in meta-analysis. This approach is similar to that used in generalized linear models, and includes the traditional REM as a particular case. A key feature of the model is the interpretation of the multiplicative factor as an intra-class correlation parameter. We provide several motivating examples, discuss statistical inference, and compare the new and standard methods on two examples of published meta-analyses. Estimation of the overdispersion parameter in the proposed model is compared in simulations to that of the traditional between-studies variance in the case of normal means. For small values of heterogeneity, the coverage of the confidence intervals for the overdispersion parameter is more stable.

Using a bayesian latent class model to evaluate the utility of investigating persons with negative polymerase chain reaction results for pertussis

Pertussis remains difficult to control. Imperfect sensitivity of diagnostic tests and lack of specific guidance regarding interpretation of negative test results among patients with compatible symptoms may contribute to its spread. In this study, we examined whether additional pertussis cases could be identified if persons with negative pertussis test results were routinely investigated. We conducted interviews among 250 subjects aged ≤18 years with pertussis polymerase chain reaction (PCR) results reported from 2 reference laboratories in Wisconsin during July-September 2010 to determine whether their illnesses met the Centers for Disease Control and Prevention's clinical case definition (CCD) for pertussis. PCR validity measures were calculated using the CCD as the standard for pertussis disease. Two Bayesian latent class models were used to adjust the validity measures for pertussis detectable by 1) culture alone and 2) culture and/or more sensitive measures such as serology. Among 190 PCR-negative subjects, 54 (28%) had illnesses meeting the CCD. In adjusted analyses, PCR sensitivity and the negative predictive value were 1) 94% and 99% and 2) 43% and 87% in the 2 types of models, respectively. The models suggested that public health follow-up of reported pertussis patients with PCR-negative results leads to the detection of more true pertussis cases than follow-up of PCR-positive persons alone. The results also suggest a need for a more specific pertussis CCD.

Suspected extracolonic neoplasms detected on CT colonography: literature review and possible outcomes

RATIONALE AND OBJECTIVES

This study summarizes the literature on the detection of cancer among indeterminate extracolonic findings on computed tomographic (CT) colonography in five targeted organs.

MATERIALS AND METHODS

We searched PubMed for English-language literature published between January 1, 1994, and December 31, 2010. We describe extracolonic findings in the kidney, lung, liver, pancreas, and ovary suspect for malignancy as they are associated with high mortality. For each organ, we calculated the median prevalence, positive predictive value (PPV), and false positive rate of malignancy and a pooled false-positive rate across studies.

RESULTS

Of 91 publications initially identified, 24 were eligible for review. Indeterminate renal masses on CT colonography had 20.5% median PPV and low pooled false positive rate of 1.3% (95% confidence interval 0.6-2.0). In contrast, indeterminate masses of the lung, liver, pancreas, and ovary had low PPV (median values ranged from 0% to 3.8%). Indeterminate masses of the ovary resulted in the highest pooled false-positive rate of 2.2%. Results were similar in studies of both screening and nonscreening populations. We estimated the probability of false positive results through the detection of significant extracolonic findings as 46 per 1000 for men and 68 per 1000 for women.

CONCLUSIONS

Indeterminate renal masses newly detected on CT colonography have an estimated one in five chance of malignancy and therefore warrant further follow-up to provide a definitive diagnosis. Conversely, indeterminate masses of the lung, liver, pancreas, and ovary are associated with high false positive rates and merit more conservative clinical follow-up.

Safety diagnosis: are we doing a good job?

Collision diagnosis is the second step in the six-step road safety management process described in the AASHTO Highway Safety Manual (HSM). Diagnosis is designed to identify a dominant or abnormally high proportion of particular collision configurations (e.g., rear end, right angle, etc.) at a target location. The primary diagnosis method suggested in the HSM is descriptive data analysis. This type of analysis relies on, for example, pie charts, histograms, and/or collision diagrams. Using location specific collision data (e.g., collision frequency per collision configuration for a target location), safety engineers identify (the most) frequent collision configurations. Safety countermeasures are then likely to concentrate on preventing the selected collision configurations. Although its real-world application in engineering practice is limited, an additional collision diagnosis method, known as the beta-binomial (BB) test, is also presented as the secondary diagnosis tool in the HSM. The BB test compares the proportion of a particular collision configuration observed at one location with the proportion of the same collision configuration found at other reference locations which are similar to the target location in terms of selected traffic and roadway characteristics (e.g., traffic volume, traffic control, and number of lanes). This study compared the outcomes obtained from descriptive data analysis and the BB test, and investigates two questions: (1) Do descriptive data analysis and the BB tests produce the same results (i.e., do they select the same collision configurations at the same locations)? and (2) If the tests produce different results, which result should be adopted in engineering practice? This study's analysis was based on a sample of the most recent five years (2005-2009) of collision and roadway configuration data for 143 signalized intersections in the City of Saskatoon, Saskatchewan. The study results show that the BB test's role in diagnosing safety concerns in road safety engineering projects such as safety review projects for existing roadways may be just as important as the descriptive data analysis method.

Safety network screening for municipalities with incomplete traffic volume data

Safety network screening is used to identify road locations, such as intersections and roadway segments that exhibit an unusually high number of expected collisions or an abnormally high proportion of a certain configuration of collisions. Current state-of-the-art network screening methods rely on safety performance functions (SPFs) that require traffic volume as an input, but many cities in North America, including the city of Saskatoon, do not collect traffic volume for every intersection and segment within the city limits. Lack of traffic volume data severely restricts the applicability of a SPF-based network screening method. The binomial and beta-binomial tests, however, are formal collision diagnosis tests that can be used as a supplementary tool for screening roadway networks that include roadway segments for which traffic volume data are not available. However, previous studies have applied these two collision diagnosis tests without explicitly defining the particular circumstances that indicate which test is preferable. This study introduces a formal statistical test known as the "C(α) test" to determine when there is a need to apply the beta-binomial test instead of the binomial test to screen a roadway network. The study targeted major arterial uncontrolled access segments in Saskatoon using five years (2005-2009) of collision data for the two most frequent collision configurations: (1) rear end, and (2) side swipe same direction collisions. ArcGIS was used to develop collision maps that visually display the screening results. The collision maps are designed to facilitate the governing agencies' decision-making processes when selecting appropriate safety countermeasures to reduce target collision configurations at screened locations.

Benchmarking inappropriate empirical antibiotic treatment

Inappropriate empirical antibiotic treatment for severe infections is associated with increased mortality. Superfluous treatment is associated with resistance induction. We aimed to define acceptable rates of inappropriate empirical antibiotic treatment. We included all prospective cohort studies published between 1975 and 2009 reporting the proportion of appropriate and inappropriate empirical antibiotic treatment of microbiologically documented infections. Studies were identified in PubMed and in reference lists of included studies. Funnel plots were drawn using the proportion of inappropriate empirical treatment as the effect size. A pooled estimate of inappropriate empirical antibiotic treatment was calculated using a β-binomial model. Control limits were calculated with the overdispersion factor technique and 20% winsorized data. Heterogeneity was assessed through subgroup analysis for categorical moderators and meta-regression for continuous variables. Eighty-seven studies, comprising 92 study groups, with 27 628 patients met inclusion criteria. The pooled rate of inappropriate empirical antibiotic treatment was 28.6% (95% CI 25.4-31.8). Funnel plot analysis yielded a dispersed graph with only 37 (40%) studies falling within the control limits. Using the overdispersion factor technique with 20% winsorizing, 79 (86%) studies fell within the control limits. None of the clinical or methodological factors could explain the large heterogeneity observed. The funnel plot presented can be used to benchmark rates of inappropriate empirical antibiotic treatment. Based on the control limits found, at least 500 patients should be evaluated before establishing a local rate. Lower and higher than expected rates might indicate overly aggressive treatment or poor performance, respectively.

Hepatocellular Carcinomas and Primary Liver Tumors as Predictive Factors for Postoperative Mortality after Liver Resection: A Meta-Analysis of More than 35,000 Hepatic Resections

Liver resection is considered the therapeutic gold standard for primary and metastatic liver neoplasms. The reduction of postoperative complications and mortality has resulted in a more aggressive approach to hepatic malignancies. For the most part, results of liver surgery have been published by highly experienced institutions, but the observations of highly specialized units results may not reflect the current status of hepatic surgery, underestimating mortality and complications. The objective of this study is to evaluate morbidity and mortality as a result of liver resection for primary and metastatic lesions, analyzing a large number of studies with a meta-analytic process taking into account the overdispersion of data. An extensive literature search has been conducted, and 148 papers published between January 2000 and April 2008, including a total of 36,629 patients from both high-volume and low volume institutions, were included in the meta-analysis. A beta binomial model was used to provide a robust estimate of the summary event rate by pooling overdispersion binomial data from different studies. Overall morbidity and mortality after liver surgery were 29.32 per cent and 3.15 per cent, respectively. Significantly higher postoperative mortality was observed after liver resection for hepatocellular carcinomas and primary hepatic tumors. The application of a beta binomial model to correct for overdispersion of liver surgery data showed significantly higher postoperative mortality rates in patients with hepatocellular carcinomas or primary hepatic tumors after liver resection.

Investigation of a supplementary tool to assist in the prioritization of emphasis areas in North American strategic highway safety plans

An important potential benefit of a jurisdiction developing an upper-level traffic safety policy statement, such as a strategic highway safety plan (SHSP) or a traffic safety action plan, is the creation of a manageable number of focus areas, known as emphasis areas. The responsible agencies in the jurisdiction can then direct their finite resources in a systematic and strategic way designed to maximize the effort to reduce the number and severity of roadway collisions. In the United States, the federal government through AASHTO has suggested 22 potential emphasis areas. In Canada, CCMTA's 10 potential emphasis areas have been listed for consideration. This study reviewed the SHSP and traffic safety action plan of 53 jurisdictions in North America, and conducted descriptive data analyses to clarify the issues that currently affect the selection and prioritization process of jurisdiction-specific emphasis areas. We found that the current process relies heavily on high-level collision data analysis and communication among the SHSP stakeholders, but may not be the most efficient and effective way of selecting and prioritizing the emphasis areas and allocating safety improvement resources. This study then formulated a formal collision diagnosis test, known as the beta-binomial test, to clarify and illuminate the selection and the prioritization of jurisdiction-specific emphasis areas. We developed numerical examples to demonstrate how engineers can apply the proposed diagnosis test to improve the selection and prioritization of individual jurisdictions' emphasis areas.

Outcomes and risks of operative treatment of rupture of the anterior cruciate ligament in children and adolescents

PURPOSE

The aim of this meta-analysis was to evaluate the clinical outcomes and risks of anterior cruciate ligament (ACL) surgery in children and adolescents.

METHODS

We electronically searched Medline, the Cochrane Controlled Trial Database, Embase, and Medpilot for studies on surgical treatment for ACL ruptures in skeletally immature patients. We extracted baseline demographics, follow-up intervals, surgical details (i.e., ligament suture or reconstruction, physeal-sparing or transphyseal techniques, type of transplant, and methods of fixation). Endpoints comprised rates of growth disturbances and reruptures, as well as knee function (measured by the International Knee Documentation Committee's documentation system and the Lysholm score). Unweighted overall effect sizes (risks, risk ratios [RRs], and means of functional scores) were estimated by use of crude nominators and denominators, and random-effects meta-regression analysis was used for weighted data synthesis.

RESULTS

A total of 55 articles reporting on 935 patients (median age, 13 years; range, 1.5 to 16 years) were suitable for the study. After a median follow-up of 40 months (range, 14 to 89 months), the weighted rate of leg-length differences or axis deviations was 1.8% (95% confidence interval [CI], 0% to 3.9%] and that of reruptures was 4.8% (95% CI, 2.9% to 6.7%). Excellent or good function (International Knee Documentation Committee grade A or B) was achieved in 84.2% (95% CI, 75.8% to 92.6%) of all knees, and Lysholm scores averaged 96.3 (95% CI, 95.5 to 97.2). Transphyseal reconstruction was associated with a significantly lower risk of leg-length differences or axis deviations compared with physeal-sparing techniques (1.9% v 5.8%; RR, 0.34; 95% CI, 0.14 to 0.81) but had a higher risk of rerupture (4.2% v 1.4%; RR, 2.91; 95% CI, 0.70 to 12.12). Sutures did not result in any growth disturbances, with a weighted rerupture rate of 4.6% (95% CI, 2.6 to 6.7). Fixation far from the joint line fared better than close fixation with regard to this endpoint (1.4% v 3.2%; RR, 0.42; 95% CI, 0.09 to 1.93). Bone-patellar tendon-bone grafts, which are also less likely to fail, were associated with higher risks of leg-length differences or axis deviations than were hamstrings (3.6% v 2.0%; RR, 1.82; 95% CI, 0.66 to 5.03). Meta-regression did not show a significant impact of the publication year on event rates.

CONCLUSIONS

This meta-analysis showed low rates of leg-length differences or axis deviations and graft failures after ACL reconstruction in skeletally immature patients. Hamstring transplants may lower the risk of leg-length differences or axis deviations, and physeal-sparing techniques may increase the risk. Randomized controlled trials are needed to clarify important issues in managing ACL ruptures in children and adolescents.

LEVEL OF EVIDENCE

Level IV, meta-analysis of case series.

Performance of low-risk criteria in the evaluation of young infants with fever: review of the literature

OBJECTIVE

The goal was to determine the performance of low-risk criteria for serious bacterial illnesses (SBIs) in febrile infants in prospective studies in which empiric antibiotic treatment was withheld, compared with studies (prospective and retrospective) in which empiric antibiotic treatment was administered.

METHODS

A search of the English-language literature was undertaken by using a PubMed database and reference lists of relevant studies of fever, low-risk criteria, and SBIs. Studies of infants >90 days of age, infants with specific infections, or infants with additional risk factors for infection were excluded. Publications were categorized as retrospective, prospective with empiric antibiotic treatment for all patients, or prospective with antibiotics withheld. The relative risk of SBI in high-risk versus low-risk patients was determined for pooled data in each category. The rates of SBIs in low-risk patients in each category were compared.

RESULTS

Twenty-one studies met the inclusion criteria. In prospective studies in which patients were cared for without empiric antibiotic treatment, 6 patients assigned to the low-risk category had SBIs; all recovered uneventfully. The rate of SBIs in these low-risk patients was 0.67%. The relative risk of SBIs in high-risk versus low-risk patients in these studies was 30.56 (95% confidence interval: 7.0-68.13). The rate of SBIs in low-risk patients in all studies was 2.23%. The rate of SBIs in low-risk patients in the prospective studies without empiric antibiotic treatment was significantly different from the rate in all other studies (0.67% vs 2.71%; P = .01).

CONCLUSIONS

Low-risk criteria perform well in prospective studies in which empiric antibiotic treatment is withheld. These criteria allow approximately 30% of young febrile infants to be observed without antibiotic treatment, thus avoiding unnecessary hospitalization, nosocomial infection, injudicious use of antibiotics, and adverse effects of antibiotics.