Steven C. Bagley

Logistic regression in the medical literature : Standards for use and reporting, with particular attention to one medical domain

Logistic regression (LR) is a widely used multivariable method for modeling dichotomous outcomes. This article examines use and reporting of LR in the medical literature by comprehensively assessing its use in a selected area of medical study. Medline, followed by bibliography searches, identified 15 peer-reviewed English-language articles with original data, employing LR, published between 1985 and 1999, pertaining to patient interest in genetic testing for cancer susceptibility. Articles were examined for each of 10 criteria for proper use and reporting of LR models. Substantial shortcomings were found in both use of LR and reporting of results. For many studies, the ratio of the number of outcome events to predictor variables (events per variable) was sufficiently small to call into question the accuracy of the regression model. Additionally, no studies reported validation analysis, regression diagnostics, or goodness-of-fit measures. It is recommended that authors, reviewers, and editors pay greater attention to guidelines concerning the use and reporting of LR models.

Efficacy and Comparative Effectiveness of Atypical Antipsychotic Medications for Off-Label Uses in Adults: A Systematic Review and Meta-analysis

CONTEXT Atypical antipsychotic medications are commonly used for off-label conditions such as agitation in dementia, anxiety, and obsessive-compulsive disorder. OBJECTIVE To perform a systematic review on the efficacy and safety of atypical antipsychotic medications for use in conditions lacking approval for labeling and marketing by the US Food and Drug Administration. DATA SOURCES AND STUDY SELECTION Relevant studies published in the English language were identified by searches of 6 databases (PubMed, EMBASE, CINAHL, PsycInfo, Cochrane DARE, and CENTRAL) from inception through May 2011. Controlled trials comparing an atypical antipsychotic medication (risperidone, olanzapine, quetiapine, aripiprazole, ziprasidone, asenapine, iloperidone, or paliperidone) with placebo, another atypical antipsychotic medication, or other pharmacotherapy for adult off-label conditions were included. Observational studies with sample sizes of greater than 1000 patients were included to assess adverse events. DATA EXTRACTION Independent article review and study quality assessment by 2 investigators. DATA SYNTHESIS Of 12 228 citations identified, 162 contributed data to the efficacy review. Among 14 placebo-controlled trials of elderly patients with dementia reporting a total global outcome score that includes symptoms such as psychosis, mood alterations, and aggression, small but statistically significant effects sizes ranging from 0.12 and 0.20 were observed for aripiprazole, olanzapine, and risperidone. For generalized anxiety disorder, a pooled analysis of 3 trials showed that quetiapine was associated with a 26% greater likelihood of a favorable response (defined as at least 50% improvement on the Hamilton Anxiety Scale) compared with placebo. For obsessive-compulsive disorder, risperidone was associated with a 3.9-fold greater likelihood of a favorable response (defined as a 25% improvement on the Yale-Brown Obsessive Compulsive Scale) compared with placebo. In elderly patients, adverse events included an increased risk of death (number needed to harm [NNH] = 87), stroke (NNH = 53 for risperidone), extrapyramidal symptoms (NNH = 10 for olanzapine; NNH = 20 for risperidone), and urinary tract symptoms (NNH range = 16-36). In nonelderly adults, adverse events included weight gain (particularly with olanzapine), fatigue, sedation, akathisia (for aripiprazole), and extrapyramidal symptoms. CONCLUSIONS Benefits and harms vary among atypical antipsychotic medications for off-label use. For global behavioral symptom scores associated with dementia in elderly patients, small but statistically significant benefits were observed for aripiprazole, olanzapine, and risperidone. Quetiapine was associated with benefits in the treatment of generalized anxiety disorder, and risperidone was associated with benefits in the treatment of obsessive-compulsive disorder; however, adverse events were common.

A Systematic Review of Suicide Prevention Programs for Military or Veterans

Military personnel and veterans have important suicide risk factors. After a systematic review of the literature on suicide prevention, seven (five in the U.S.) studies of military personnel were identified containing interventions that may reduce the risk of suicide. The effectiveness of the individual components was not assessed, and problems in methodology or reporting of data were common. Overall, multifaceted interventions for active duty military personnel are supported by consistent evidence, although of very mixed quality, and in some cases during intervals of declines in suicide rates in the general population. There were insufficient studies of U.S. Veterans to reach conclusions.

Environmental and state-level regulatory factors affect the incidence of autism and intellectual disability.

Many factors affect the risks for neurodevelopmental maladies such as autism spectrum disorders (ASD) and intellectual disability (ID). To compare environmental, phenotypic, socioeconomic and state-policy factors in a unified geospatial framework, we analyzed the spatial incidence patterns of ASD and ID using an insurance claims dataset covering nearly one third of the US population. Following epidemiologic evidence, we used the rate of congenital malformations of the reproductive system as a surrogate for environmental exposure of parents to unmeasured developmental risk factors, including toxins. Adjusted for gender, ethnic, socioeconomic, and geopolitical factors, the ASD incidence rates were strongly linked to population-normalized rates of congenital malformations of the reproductive system in males (an increase in ASD incidence by 283% for every percent increase in incidence of malformations, 95% CI: [91%, 576%], p<6×10−5). Such congenital malformations were barely significant for ID (94% increase, 95% CI: [1%, 250%], p = 0.0384). Other congenital malformations in males (excluding those affecting the reproductive system) appeared to significantly affect both phenotypes: 31.8% ASD rate increase (CI: [12%, 52%], p<6×10−5), and 43% ID rate increase (CI: [23%, 67%], p<6×10−5). Furthermore, the state-mandated rigor of diagnosis of ASD by a pediatrician or clinician for consideration in the special education system was predictive of a considerable decrease in ASD and ID incidence rates (98.6%, CI: [28%, 99.99%], p = 0.02475 and 99% CI: [68%, 99.99%], p = 0.00637 respectively). Thus, the observed spatial variability of both ID and ASD rates is associated with environmental and state-level regulatory factors; the magnitude of influence of compound environmental predictors was approximately three times greater than that of state-level incentives. The estimated county-level random effects exhibited marked spatial clustering, strongly indicating existence of as yet unidentified localized factors driving apparent disease incidence. Finally, we found that the rates of ASD and ID at the county level were weakly but significantly correlated (Pearson product-moment correlation 0.0589, p = 0.00101), while for females the correlation was much stronger (0.197, p<2.26×10−16).

Conserved features in the active site of nonhomologous serine proteases

BACKGROUND Serine protease activity is critical for many biological processes and has arisen independently in a few different protein families. It is not clear, though, the degree to which these protease families share common biochemical and biophysical properties. We have used a computer program to study the properties that are shared by four serine protease active sites with no overall structural or sequence homology. The program systematically compares the region around the catalytic histidines from the four proteins with a set of noncatalytic histidines, used as controls. It reports the three-dimensional locations and level of statistical significance for those properties that distinguish the catalytic histidines from the noncatalytic ones. The method of analysis is general and can be applied easily to other active sites of interest. RESULTS As expected, some of the reported properties correspond to previously known features of the serine protease active site, including the catalytic triad and the oxyanion hole. Novel properties are also found, including the spatial distribution of charged, polar, and hydrophobic groups arranged to stabilize the catalytic residues, and a relative abundance of some residues (Val, Tyr, Leu, and Gly) around the active site. CONCLUSIONS Our findings show that in addition to some properties common to all the proteases examined, there are a set of preferred, but not required, properties that can be reliably observed only by aligning the sites and comparing them with carefully selected statistical controls.

Constraints on Biological Mechanism from Disease Comorbidity Using Electronic Medical Records and Database of Genetic Variants

Patterns of disease co-occurrence that deviate from statistical independence may represent important constraints on biological mechanism, which sometimes can be explained by shared genetics. In this work we study the relationship between disease co-occurrence and commonly shared genetic architecture of disease. Records of pairs of diseases were combined from two different electronic medical systems (Columbia, Stanford), and compared to a large database of published disease-associated genetic variants (VARIMED); data on 35 disorders were available across all three sources, which include medical records for over 1.2 million patients and variants from over 17,000 publications. Based on the sources in which they appeared, disease pairs were categorized as having predominant clinical, genetic, or both kinds of manifestations. Confounding effects of age on disease incidence were controlled for by only comparing diseases when they fall in the same cluster of similarly shaped incidence patterns. We find that disease pairs that are overrepresented in both electronic medical record systems and in VARIMED come from two main disease classes, autoimmune and neuropsychiatric. We furthermore identify specific genes that are shared within these disease groups.

Computing disease incidence, prevalence and comorbidity from electronic medical records

Electronic medical records (EMR) represent a convenient source of coded medical data, but disease patterns found in EMRs may be biased when compared to surveys based on sampling. In this communication we draw attention to complications that arise when using EMR data to calculate disease prevalence, incidence, age of onset, and disease comorbidity. We review known solutions to these problems and identify challenges for future work.

Predicting suicide attempt risk: logistic regression requires large sample sizes

To the Editor: Dr Gilbert and colleagues make an important point about the difficulties in predicting suicide,1 but their conclusions do not follow as strongly from their data as they suppose. They report negative findings based on logistic regressions using clinical, demographic, and cognitive predictor variables. However, their data set is relatively small, with 28 events (suicide attempts) in 67 subjects; their regressions use 12 clinical and demographic predictors and, separately, 7 cognitive and demographic predictors. Simulation experiments2 have shown that logistic regression requires roughly 10 events per predictor, which would limit its use to 2—or, if stretched, 3—predictors for their dataset. The effect for the study in question is not entirely clear, but performing regressions below the advisory event per predictor threshold can bias the coefficients and distort the standard errors and could have been responsible, at least in part, for failure to reach statistical significance. The take-home message is that logistic regression requires relatively large sample sizes for proper statistical inference.3 In passing, I note that the odds ratios listed in their Table 3 are, incorrectly, copies of the β coefficients; the proper odds ratios can be computed by raising e, the base of the natural logarithm, to the power β.