Computational tools for exact conditional logistic regression

Cerebral autoregulation in anoxic brain injury patients treated with targeted temperature management

Background

Little is known about the prevalence of altered CAR in anoxic brain injury and the association with patients’ outcome. We aimed at investigating CAR in cardiac arrest survivors treated by targeted temperature management and its association to outcome.

Methods

Retrospective analysis of prospectively collected data. Inclusion criteria: adult cardiac arrest survivors treated by targeted temperature management (TTM). Exclusion criteria: trauma; sepsis, intoxication; acute intra-cranial disease; history of supra-aortic vascular disease; severe hemodynamic instability; cardiac output mechanical support; arterial carbon dioxide partial pressure (PaCO₂) > 60 mmHg; arrhythmias; lack of acoustic window. Middle cerebral artery flow velocitiy (FV) was assessed by transcranial Doppler (TCD) once during hypothermia (HT) and once during normothermia (NT). FV and blood pressure (BP) were recorded simultaneously and Mxa calculated (MATLAB). Mxa is the Pearson correlation coefficient between FV and BP. Mxa > 0.3 defined altered CAR. Survival was assessed at hospital discharge. Cerebral Performance Category (CPC) 3–5 assessed 3 months after CA defined unfavorable neurological outcome (UO).

Results

We included 50 patients (Jan 2015–Dec 2018). All patients had out-of-hospital cardiac arrest, 24 (48%) had initial shockable rhythm. Time to return of spontaneous circulation was 20 [10–35] min. HT (core body temperature 33.7 [33.2–34] °C) lasted for 24 [23–28] h, followed by rewarming and NT (core body temperature: 36.9 [36.6–37.4] °C). Thirty-one (62%) patients did not survive at hospital discharge and 36 (72%) had UO. Mxa was lower during HT than during NT (0.33 [0.11–0.58] vs. 0.58 [0.30–0.83]; p = 0.03). During HT, Mxa did not differ between outcome groups. During NT, Mxa was higher in patients with UO than others (0.63 [0.43–0.83] vs. 0.31 [− 0.01–0.67]; p = 0.03). Mxa differed among CPC values at NT (p = 0.03). Specifically, CPC 2 group had lower Mxa than CPC 3 and 5 groups. At multivariate analysis, initial non-shockable rhythm, high Mxa during NT and highly malignant electroencephalography pattern (HMp) were associated with in-hospital mortality; high Mxa during NT and HMp were associated with UO.

Conclusions

CAR is frequently altered in cardiac arrest survivors treated by TTM. Altered CAR during normothermia was independently associated with poor outcome.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40560-021-00579-z.

Artificial neural network approach for acute poisoning mortality prediction in emergency departments

OBJECTIVE

The number of deaths due to acute poisoning (AP) is on the increase. It is crucial to predict AP patient mortality to identify those requiring intensive care for providing appropriate patient care as well as preserving medical resources. The aim of this study is to predict the risk of in-hospital mortality associated with AP using an artificial neural network (ANN) model.

METHODS

In this multicenter retrospective study, ANN and logistic regression models were constructed using the clinical and laboratory data of 1,304 patients seeking emergency treatment for AP. The ANN model was first trained on 912/1,304 (70%) randomly selected patients and then tested on the remaining 392/1,304 (30%). Receiver operating characteristic curve analysis was used to evaluate the mortality prediction of the two models.

RESULTS

Age, endotracheal intubation status, and intensive care unit admission were significant predictors of mortality in patients with AP in the multivariate logistic regression model. The ANN model indicated age, Glasgow Coma Scale, intensive care unit admission, and endotracheal intubation status were critical factors among the 12 independent variables related to in-hospital mortality. The area under the receiver operating characteristic curve for mortality prediction was significantly higher in the ANN model compared to the logistic regression model.

CONCLUSION

This study establishes that the ANN model could be a valuable tool for predicting the risk of death following AP. Thus, it may facilitate effective patient triage and improve the outcomes.

Risk of Stevens-Johnson syndrome and toxic epidermal necrolysis associated with anticonvulsants in a Japanese population: Matched case-control and cohort studies

BACKGROUND

Evidence for the risk and incidence of anticonvulsant-induced Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) in Japan is scarce.

METHODS

We conducted a matched case-control study using a large-scale Japanese claims database. SJS/TEN cases were identified using a claims-based algorithm developed in a previous study (sensitivity 76.9%, specificity 99.0%). Conditional logistic regression with Firth's bias correction to address an issue of rare events was used to estimate odds ratios (ORs) for SJS/TEN for each anticonvulsant use (90 days before the index date) versus non-use. 90-day cumulative incidence of SJS/TEN per 100,000 new users was calculated for 33 anticonvulsants. Causality between anticonvulsant use and SJS/TEN in each exposed case was assessed using the algorithm of drug causality for epidermal necrolysis (ALDEN) score.

RESULTS

From 5,114,492 subjects, we selected 71 SJS/TEN cases and 284 controls. We observed significantly increased ORs for SJS/TEN among new users of carbamazepine (OR 68.00) and lamotrigine (OR 36.00) with ALDEN scores of "probable" or higher. Cumulative incidence of SJS/TEN was 93.83 for carbamazepine and 84.33 for lamotrigine. One case newly exposed to phenytoin which developed SJS/TEN was rated "unlikely" in ALDEN causality, resulting in cumulative incidence of 66.27. Cumulative incidence of SJS/TEN was 25.23 for levetiracetam, 7.52 for clonazepam, and 1.23 for diazepam, but their ALDEN scores were "very unlikely".

CONCLUSIONS

This study is the first to document the differential risk of SJS/TEN for anticonvulsants in a real-world setting in Japan. Exposure to carbamazepine and lamotrigine was associated with an increased risk of SJS/TEN.

Inferential tools in penalized logistic regression for small and sparse data: A comparative study

This paper focuses on inferential tools in the logistic regression model fitted by the Firth penalized likelihood. In this context, the Likelihood Ratio statistic is often reported to be the preferred choice as compared to the 'traditional' Wald statistic. In this work, we consider and discuss a wider range of test statistics, including the robust Wald, the Score, and the recently proposed Gradient statistic. We compare all these asymptotically equivalent statistics in terms of interval estimation and hypothesis testing via simulation experiments and analyses of two real datasets. We find out that the Likelihood Ratio statistic does not appear the best inferential device in the Firth penalized logistic regression.

Sparse conditional logistic regression for analyzing large-scale matched data from epidemiological studies: a simple algorithm

This paper considers the problem of estimation and variable selection for large high-dimensional data (high number of predictors p and large sample size N, without excluding the possibility that N < p) resulting from an individually matched case-control study. We develop a simple algorithm for the adaptation of the Lasso and related methods to the conditional logistic regression model. Our proposal relies on the simplification of the calculations involved in the likelihood function. Then, the proposed algorithm iteratively solves reweighted Lasso problems using cyclical coordinate descent, computed along a regularization path. This method can handle large problems and deal with sparse features efficiently. We discuss benefits and drawbacks with respect to the existing available implementations. We also illustrate the interest and use of these techniques on a pharmacoepidemiological study of medication use and traffic safety.

Hypothesis testing at the extremes: fast and robust association for high-throughput data

A number of biomedical problems require performing many hypothesis tests, with an attendant need to apply stringent thresholds. Often the data take the form of a series of predictor vectors, each of which must be compared with a single response vector, perhaps with nuisance covariates. Parametric tests of association are often used, but can result in inaccurate type I error at the extreme thresholds, even for large sample sizes. Furthermore, standard two-sided testing can reduce power compared with the doubled [Formula: see text]-value, due to asymmetry in the null distribution. Exact (permutation) testing is attractive, but can be computationally intensive and cumbersome. We present an approximation to exact association tests of trend that is accurate and fast enough for standard use in high-throughput settings, and can easily provide standard two-sided or doubled [Formula: see text]-values. The approach is shown to be equivalent under permutation to likelihood ratio tests for the most commonly used generalized linear models (GLMs). For linear regression, covariates are handled by working with covariate-residualized responses and predictors. For GLMs, stratified covariates can be handled in a manner similar to exact conditional testing. Simulations and examples illustrate the wide applicability of the approach. The accompanying mcc package is available on CRAN http://cran.r-project.org/web/packages/mcc/index.html.

Riskier sexual partners contribute to the increased rate of sexually transmitted diseases among youth with substance use disorders

BACKGROUND

Youth with alcohol or marijuana dependence or disorders (substance use disorders [SUDs]) are at increased risk of acquiring sexually transmitted diseases (STDs). Sexual partner characteristics may explain the relationship between SUD and STD.

METHODS

Clinical criteria for SUD, clinical STD diagnosis, and sexual partner characteristics were assessed among 15- to 24-year-old STD clinic attendees between 1999 and 2002 (n = 412). We used exact logistic regression and path analysis to examine the mediation effect of sexual partner characteristics (age discordance, incarceration, STD diagnosis, other partners, perceived alcohol problem, perceived marijuana problem, and a calculated composite measure) on the relationship between SUD and STD, adjusting for important demographics and condom use.

RESULTS

We found evidence of mediation by partner characteristics on the relationship between SUD and STD. For the logistic regression analysis, adding the partner characteristic composite reduced the strength of the association between SUD and STD from a statistically significant odds ratio of 1.7 (95% confidence interval = 1.0-2.7) to a statistically nonsignificant odds ratio of 1.5 (95% confidence interval = 0.9-2.5). In the path analysis, adding the partner characteristic composite reduced the significant direct effect of SUD on STD (β = 0.5, P = 0.04) to statistically nonsignificant effect (β = 0.1, P = 0.2). We estimated that 31% of the total effect of SUD on STD was attributable to the indirect path through the composite partner characteristic measure.

CONCLUSIONS

Even when controlling for demographics and condom use, partner characteristics partially explained the association between SUD and STD.

Diffusion tensor MRI: preliminary anisotropy measures and mapping of breast tumors

PURPOSE

To investigate whether diffusion tensor imaging (DTI) measures of anisotropy in breast tumors are different from normal breast tissue and can improve the discrimination between benign and malignant lesions.

MATERIALS AND METHODS

The study included 81 women with 105 breast lesions (76 malignant, 29 benign). DTI was performed during breast MRI examinations, and fractional anisotropy (FA) and apparent diffusion coefficient (ADC) values were measured for breast lesions and normal tissue in each subject. FA and ADC were compared between cancers, benign lesions, and normal tissue by univariate and multivariate analyses.

RESULTS

The FA of carcinomas (mean +/- SD: 0.24 +/- 0.07) was significantly lower than normal breast tissue in the same subjects (0.29 +/- 0.07; P < 0.0001). Multiple logistic regression showed that FA and ADC were each independent discriminators of malignancy (P < 0.0001), and that FA improved discrimination between cancer and normal tissue over ADC alone. However, there was no difference in FA between malignant and benign lesions (P = 0.98).

CONCLUSION

Diffusion anisotropy is significantly lower in breast cancers than normal tissue, which may reflect alterations in tissue organization. Our preliminary results suggest that FA adds incremental value over ADC alone for discriminating malignant from normal tissue but does not help with distinguishing benign from malignant lesions.

Sexual partner characteristics and sexually transmitted diseases among adolescents and young adults

BACKGROUND

Among adolescents and young adults, the extent that partner characteristics account for sexually transmitted diseases (STDs) in context of individual sexual activities and demographic characteristics is unclear.

METHODS

Sexual partner characteristics, individual sexual activities, and STD diagnosis were assessed among 15 to 24-year-old STD clinic attendees from 1999 to 2002 (n = 412). We used exact logistic regression to calculate odds ratios (OR) for several sexual partner characteristics (age discordance, incarceration, STD diagnosis, other partners, alcohol problem, marijuana problem, and a calculated composite variable) adjusting for demographics and individual sexual activities, including condom use.

RESULTS

Sexual partner characteristics associated with STD diagnosis were >or=5 years age discordance [OR = 2.6 (95% confidence interval (CI) = 1.6, 4.5)] and STD in the past year [OR = 3.4 (95% CI = 2.0, 5.7)]. Even when considering individual sexual activities, composite partner risk was associated with STD diagnosis [intermediate to low OR = 2.1 (95% CI = 1.0, 4.2) and high to low OR = 3.4 (95% CI = 1.6, 7.0)]. Composite individual sexual activities was associated with STD diagnosis when considering demographics [intermediate to low OR = 1.8 (95% CI = 1.0, 3.2), high to low OR = 2.3 (95% CI = 1.2, 4.5)], but not when also considering partner characteristics [intermediate to low OR = 1.6 (95% CI = 0.9, 2.8), high to low OR = 1.8 (95% CI = 0.8, 3.9)].

CONCLUSIONS

Among this sample of 15 to 24-year-olds, sexual partner characteristics identified individuals at increased risk of prevalent STDs and were more predictive of STDs than an individual's sexual activities.

Assessing the quality of pharmacological treatments from administrative databases: the case of low-molecular-weight heparin after major orthopaedic surgery

RATIONALE, AIMS AND OBJECTIVES

'Real world data' are needed to assess the quality of pharmacological treatments in clinical practice. The aim of this study was to determine whether administrative databases can be used to assess the quality of prophylaxis with low-molecular-weight heparin after major orthopaedic surgery.

METHODS

The study was performed in a Belgian university hospital. Patients undergoing total hip replacement (THR), total knee replacement (TKR) or hip fracture surgery (HFS) were selected retrospectively from the hospital's 2002 and 2003 administrative databases. Readmissions during the same year as the procedure were also analysed. Three quality indicators were assessed: incidence of venous thromboembolism (VTE), major bleeding and death; adherence to guidelines; and the costs of care.

RESULTS

Although 70% of data were collected from administrative databases, patients' records also had to be examined. During the period studied, VTE and major bleeding events were rare. Patients undergoing HFS were at greater risk of having a pulmonary embolism [Exact odds ratio (OR)=3.78; 95% confidence interval (CI)=1.13-16.22; P=0.03] or of death from any cause (Exact OR=2.15; 95% CI=1.52-infinity; P<0.01) than patients undergoing THR or TKR. The hospital's prophylaxis protocol was not always followed. Half the patients received higher prophylaxis doses than recommended and 11% received lower doses but no impact on adverse events was demonstrated.

CONCLUSION

Results show that administrative databases contain enough information to measure the frequency of adverse events but complementary data on patient weight and on non-reimbursed drugs must be extracted from the patients' records to evaluate adherence to guidelines. Our findings stress the need for better integration of information systems.