Good Statistical Practice-development of tailored Good Clinical Practice training for statisticians.

BACKGROUND: Statisticians are fundamental in ensuring clinical research, including clinical trials, are conducted with quality, transparency, reproducibility and integrity. Good Clinical Practice (GCP) is an international quality standard for the conduct of clinical trials research. Statisticians are required to undertake training on GCP but existing training is generic and, crucially, does not cover statistical activities. This results in statisticians undertaking training mostly unrelated to their role and variation in awareness and implementation of relevant regulatory requirements with regards to statistical conduct. The need for role-relevant training is recognised by the UK NHS Health Research Authority and the Medicines and Healthcare products Regulatory Agency (MHRA). METHODS: The Good Statistical Practice (GCP for Statisticians) project was instigated by the UK Clinical Research Collaboration (UKCRC) Registered Clinical Trials Unit (CTU) Statisticians Operational Group and funded by the National Institute for Health and Care Research (NIHR), to develop materials to enable role-specific GCP training tailored to statisticians. Review of current GCP training was undertaken by survey. Development of training materials were based on MHRA GCP. Critical review and piloting was conducted with UKCRC CTU and NIHR researchers with comment from MHRA. Final review was conducted through the UKCRC CTU Statistics group. RESULTS: The survey confirmed the need and desire for the development of dedicated GCP training for statisticians. An accessible, comprehensive, piloted training package was developed tailored to statisticians working in clinical research, particularly the clinical trials arena. The training materials cover legislation and guidance for best practice across all clinical trial processes with statistical involvement, including exercises and real-life scenarios to bridge the gap between theory and practice. Comprehensive feedback was incorporated. The training materials are freely available for national and international adoption. CONCLUSION: All research staff should have training in GCP yet the training undertaken by most academic statisticians does not cover activities related to their role. The Good Statistical Practice (GCP for Statisticians) project has developed and extensively piloted new, role-specific, comprehensive, accessible GCP training tailored to statisticians working in clinical research, particularly the clinical trials arena. This role-specific training will encourage best practice, leading to transparent and reproducible statistical activity, as required by regulatory authorities and funders.

Seven steps toward more transparency in statistical practice.

We argue that statistical practice in the social and behavioural sciences benefits from transparency, a fair acknowledgement of uncertainty and openness to alternative interpretations. Here, to promote such a practice, we recommend seven concrete statistical procedures: (1) visualizing data; (2) quantifying inferential uncertainty; (3) assessing data preprocessing choices; (4) reporting multiple models; (5) involving multiple analysts; (6) interpreting results modestly; and (7) sharing data and code. We discuss their benefits and limitations, and provide guidelines for adoption. Each of the seven procedures finds inspiration in Merton's ethos of science as reflected in the norms of communalism, universalism, disinterestedness and organized scepticism. We believe that these ethical considerations-as well as their statistical consequences-establish common ground among data analysts, despite continuing disagreements about the foundations of statistical inference.

Comparison of Methods to Estimate Low-Density Lipoprotein Cholesterol in Patients With High Triglyceride Levels.

Importance: Low-density lipoprotein cholesterol (LDL-C) is typically estimated with the Friedewald or Martin/Hopkins equation; however, if triglyceride levels are 400 mg/dL or greater, laboratories reflexively perform direct LDL-C (dLDL-C) measurement. The use of direct chemical LDL-C assays and estimation of LDL-C via the National Institutes of Health Sampson equation are not well validated, and data on the accuracy of LDL-C estimation at higher triglyceride levels are limited. Objective: To compare an extended Martin/Hopkins equation for triglyceride values of 400 to 799 mg/dL with the Friedewald and Sampson equations. Design, Setting, and Participants: This cross-sectional study evaluated consecutive patients at clinical sites across the US with patient lipid distributions representative of the US population in the Very Large Database of Lipids from January 1, 2006, to December 31, 2015, with triglyceride levels of 400 to 799 mg/dL. Data analysis was performed from November 9, 2020, to March 23, 2021. Main Outcomes and Measures: Accuracy in LDL-C classification according to guideline-based categories and absolute errors between estimated LDL-C and dLDL-C levels. Patients were randomly assigned 2:1 to derivation and validation data sets. Levels of dLDL-C were measured by vertical spin-density gradient ultracentrifugation. The LDL-C levels were estimated using the Friedewald method, with a fixed ratio of triglycerides to very low-density lipoprotein cholesterol (VLDL-C ratio of 5:1), extended Martin/Hopkins equation with a flexible ratio, and Sampson equation with VLDL-C estimation by multiple least-squares regression. Results: A total of 111 939 patients (mean [SD] age, 52 [13] years; 65.0% male) with triglyceride levels of 400 to 799 mg/dL were included, representing 2.2% of 5 081 680 patients in the database. Across all individual guideline LDL-C classes (<40, 40-69, 70-99, 100-129, 130-159, 160-189, and ≥190), estimation of LDL-C by the extended Martin/Hopkins equation was most accurate (62.1%) compared with the Friedewald (19.3%) and Sampson (40.4%) equations. In classifying LDL-C levels less than 70 mg/dL across all triglyceride strata, the extended Martin/Hopkins equation was most accurate (67.3%) compared with Friedewald (5.1%) and Sampson (26.4%) equations. In addition, for classifying LDL-C levels less than 40 mg/dL across all triglyceride strata, the extended Martin/Hopkins equation was most accurate (57.2%) compared with the Friedewald (4.3%) and Sampson (14.4%) equations. However, considerable underclassification of LDL-C occurred. The magnitude of error between the Martin/Hopkins equation estimation and dLDL-C was also smaller: at LDL-C levels less than 40 mg/dL, 2.7% of patients had 30 mg/dL or greater differences between dLDL-C and estimated LDL-C using the Martin/Hopkins equation compared with the Friedewald (92.5%) and Sampson (38.7%) equations. Conclusions and Relevance: In this cross-sectional study, the extended Martin/Hopkins equation offered greater LDL-C accuracy compared with the Friedewald and Sampson equations in patients with triglyceride levels of 400 to 799 mg/dL. However, regardless of method used, caution is advised with LDL-C estimation in this triglyceride range.

New relevance and significance measures to replace p-values.

The p-value has been debated exorbitantly in the last decades, experiencing fierce critique, but also finding some advocates. The fundamental issue with its misleading interpretation stems from its common use for testing the unrealistic null hypothesis of an effect that is precisely zero. A meaningful question asks instead whether the effect is relevant. It is then unavoidable that a threshold for relevance is chosen. Considerations that can lead to agreeable conventions for this choice are presented for several commonly used statistical situations. Based on the threshold, a simple quantitative measure of relevance emerges naturally. Statistical inference for the effect should be based on the confidence interval for the relevance measure. A classification of results that goes beyond a simple distinction like "significant / non-significant" is proposed. On the other hand, if desired, a single number called the "secured relevance" may summarize the result, like the p-value does it, but with a scientifically meaningful interpretation.

Statistical literacy and scientific reasoning & argumentation in physicians.

Objective: Statistical literacy (SL) of physicians, i.e. the ability to use and interpret statistical numbers in the context of science, is an essential prerequisite for risk estimation and communication. Together with scientific reasoning and argumentation (SRA) skills, SL provides the basis for evidence-based practice. Several studies suggest that in medical students both skills are underdeveloped. The aim of the present study was to investigate these skills in practicing physicians and how these skills were acquired. Methods: Data collection in N=71 physicians was conducted online and as paper pencil. SL was assessed with multiple-choice items. SRA skills evidence evaluation and drawing conclusions were measured with a decision scenario. Results: Study results indicated that physicians have medium levels of SL (M=17.58, SD=6.92, max 30 pts.) and SRA (evidence evaluation: M=7.75, SD=1.85, max 10 pts.; drawing conclusions: M=37.20, SD=5.35, max 60 pts.). Skills development via autodidactic learning activities (M=4.78, SD=1.13, range 1-6) was reported significantly more often than development during formal medical education (M=2.31, SD=1.46), t(71)=-9.915, p<.001, or in extracurricular activities (M=3.34, SD=1.87), t(71)=4.673, p<.001. The active involvement in research seemed decisive: The number of publications and time spent in research significantly correlated with SL, r(71)=.355, p=.002; respectively r(71)=.280, p=.018. SRA skills were predicted by the type of MD-thesis, ß=-.380, p=.016, and working in research, ß=3.355, p=.008. Conclusion: Active involvement in research activities seems to be a very important factor for the development of both SL and SRA skills. The implementation of systematic fostering of these skills during formal medical education seems warranted.

Reporting only relative effect measures was potentially misleading: some good practices for improving the soundness of epidemiological results.

OBJECTIVE: In the medical and epidemiological literature there is a growing tendency to report an excessive number of decimal digits (often three, sometimes four), especially when measures of relative occurrence are small; this can be misleading. STUDY DESIGN AND SETTING: We combined mathematical and statistical reasoning about the precision of relative risks with the meaning of the decimal part of the same measures from biological and public health perspectives. RESULTS: We identified a general rule for minimizing the mathematical error due to rounding of relative risks, depending on the background absolute rate, which justifies the use of one or more decimal digits for estimates close to 1. CONCLUSIONS: We suggest that both relative and absolute risk measures (expressed as a rates) should be reported, and two decimal digits should be used for relative risk close to 1 only if the background rate is at least 1/1,000 py. The use of more than two decimal digits is justified only when the background rate is high (ie, 1/10 py).

Mixed-data deep learning in repeated predictions of general medicine length of stay: a derivation study.

The accurate prediction of likely discharges and estimates of length of stay (LOS) aid in effective hospital administration and help to prevent access block. Machine learning (ML) may be able to help with these tasks. For consecutive patients admitted under General Medicine at the Royal Adelaide Hospital over an 8-month period, daily ward round notes and relevant discrete data fields were collected from the electronic medical record. These data were then split into training and testing sets (7-month/1-month train/test split) prior to use in ML analyses aiming to predict discharge within the next 2 days, discharge within the next 7 days and an estimated date of discharge (EDD). Artificial neural networks and logistic regression were effective at predicting discharge within 48 h of a given ward round note. These models achieved an area under the receiver operator curve (AUC) of 0.80 and 0.78, respectively. Prediction of discharge within 7 days of a given note was less accurate, with artificial neural network returning an AUC of 0.68 and logistic regression an AUC of 0.61. The generation of an exact EDD remains inaccurate. This study has shown that repeated estimates of LOS using daily ward round notes and mixed-data inputs are effective in the prediction of general medicine discharges in the next 48 h. Further research may seek to prospectively and externally validate models for prediction of upcoming discharge, as well as combination human-ML approaches for generating EDDs.

Prospective verification of sonographic fetal weight estimators among term parturients in Uganda.

BACKGROUND: Accuracy of fetal weight estimation by ultrasound is essential in making decisions on the time and mode of delivery. There are many proposed formulas for fetal weight estimation such as Hadlock 1, Hadlock 2, Hadlock 3, Hadlock 4 and Shepard. What best applies to the Ugandan population is not known since no verification of any of the formulas has been done before. The primary aim of this study was to determine the accuracy of sonographic estimation of fetal weight using five most commonly used formulas, and analyze formula variations for different weight ranges. METHODS: This was a hospital based prospective cohort study at Mulago National Referral Hospital, Kampala, Uganda. A total of 356 pregnant women who consented and were within 3 days of birth were enrolled. Prenatal ultrasound fetal weight determined by measuring the biparietal diameter, head circumference, abdominal circumference, femoral length, and then was compared with actual birth weight. RESULTS: The overall accuracy of Hadlock 1, Hadlock 2, Hadlock 3, Hadlock 4 and Shepard formula were 66.9, 73.3, 77.3, 78.4 and 69.7% respectively. All Hadlocks showed significant mean difference between weight estimates and actual birth weight (p < 0.01) whereas Shepard formula did not [p - 0.2], when no stratification of fetal weights was done. However, all Hadlocks showed a none significant (p-values > 0.05) mean difference between weight estimates and actual birth weight when the actual birth weight was ≥4000.0 g. Shepard weight estimates showed a none significant mean difference when actual birth weight was < 4000 g. Bland-Altman graphs also showed a better agreement of weight estimated by Shepard formula and actual birth weights. CONCLUSION: All the five formulas were accurate at estimating actual birth weights within 10% accuracy. However, this accuracy varied with the fetal birth weight. Shepard was more accurate in estimating actual birth weights < 4000 g whereas all Hadlocks were more accurate when the actual birthweight was ≥4000 g.

A CHecklist for statistical Assessment of Medical Papers (the CHAMP statement): explanation and elaboration.

Misuse of statistics in medical and sports science research is common and may lead to detrimental consequences to healthcare. Many authors, editors and peer reviewers of medical papers will not have expert knowledge of statistics or may be unconvinced about the importance of applying correct statistics in medical research. Although there are guidelines on reporting statistics in medical papers, a checklist on the more general and commonly seen aspects of statistics to assess when peer-reviewing an article is needed. In this article, we propose a CHecklist for statistical Assessment of Medical Papers (CHAMP) comprising 30 items related to the design and conduct, data analysis, reporting and presentation, and interpretation of a research paper. While CHAMP is primarily aimed at editors and peer reviewers during the statistical assessment of a medical paper, we believe it will serve as a useful reference to improve authors' and readers' practice in their use of statistics in medical research. We strongly encourage editors and peer reviewers to consult CHAMP when assessing manuscripts for potential publication. Authors also may apply CHAMP to ensure the validity of their statistical approach and reporting of medical research, and readers may consider using CHAMP to enhance their statistical assessment of a paper.

A study of statistics knowledge among nurse faculty in schools with research doctorate programs.

Statistics knowledge is essential for nursing faculty in both teaching and research roles. In the teaching role when discussing nursing research, nursing faculty are confronted with statistical concepts and statistical methods applications. Knowledge of fundamental statistical concepts is needed so that nursing faculty can understand and critically evaluate the literature. The purpose of this study was to assess nursing faculty knowledge of fundamental statistical concepts. A probability sample with a 7.7% response rate yielded participation from 164 nursing faculty from 26 accredited schools. Results showed that most faculty members (91.5%) read peer-reviewed health-related scientific journal articles. On average, nursing faculty answered 5.1 (SD = 1.6) out of 8 statistical knowledge questions correctly. Problematic concepts included randomization (43.3% correct), and interpreting a confidence interval (42.7%) and odds ratio (33.5%). The results of this study may be used to improve statistics education and training for future nursing faculty and strengthen scholarship for nursing faculty conducting research.

Creation and maintenance of a table for assessment of evolving evidence for COVID-19-related treatments.

PURPOSE: This report describes the development and maintenance of a table to present an assessment of evidence for treatments used in patients with coronavirus disease 2019 (COVID-19). SUMMARY: AHFS Drug Information (AHFS DI) (American Society of Health-System Pharmacists, Bethesda, MD) is ASHP's evidence-based drug compendium that contains drug monographs written for pharmacists and other healthcare professionals. The professional editorial and analytical staff of pharmacists critically evaluate published evidence to develop drug monographs for AHFS DI. In response to the global COVID-19 pandemic, these skills were applied to assess emerging evidence for COVID-19-related treatments, and the information was compiled into a new resource for pharmacists and other healthcare professionals to use at the point of care. A list of therapies was developed and prioritized based on review of scientific and public discussions on the use of these therapies in patients with COVID-19; certain therapies used for supportive care and therapies that might theoretically be harmful to patients with COVID-19 also were considered for inclusion. Potential treatments were identified, and the evidence for use in patients with COVID-19 was assessed and summarized in a table format. Information presented for each therapy included the rationale for use, summaries of clinical trials or experience, trial registry numbers, and dosage regimens. Comments on safety and efficacy, including limitations of available data, were presented along with recommendations from recognized authorities. The editorial team continued to add new therapies to the table and update existing entries as new evidence emerged. CONCLUSION: A comprehensive table that summarized available evidence for potential treatments for patients with COVID-19 was developed. The table format enabled the drug information editorial staff to provide ongoing updates as new information emerged during the pandemic.

The Pre-Hospital Initial Fluid Therapy Estimate in Early Nasty Burns (PHIFTEEN B, 15-B) Guideline applied to a retrospective cohort of Intensive Care Unit patients with major burns.

BACKGROUND: Appropriate fluid administration in severe burns is a cornerstone of early burns management. The American Burns Association's (ABA) recommendation is to administer 2 mL-4 mL × burnt Body Surface Area (BSA) × weight in the first 24 h with half administered in the first eight hours. Unfortunately, the calculations involved are complex and clinicians do not estimate the BSA or weight well, which can lead to errors in the amount of fluid administered. To simplify cognitive load to calculate the fluid resuscitation of early burns, the investigators derived the PHIFTEEN B (15-B) guideline. The 15-B guideline estimates the initial hourly fluid for adults ≥ 50 kg to be: 15 mL × BSA (to the nearest 10%) AIMS: To model and determine the accuracy of the 15-B calculated based on the characteristics of a retrospective cohort of patients admitted with ≥ 20% BSA to the Royal Brisbane and Women's Hospital (RBWH) Intensive Care Unit (ICU). METHODS: The 15-B formula was retrospectively calculated on the prehospital BSA estimate on patients admitted to the RBWH ICU. In addition, the 15-B guideline was modelled against a variety of weights and BSAs. The fluid volume was deemed to be clinically significant if it was greater than 250 mL/h outside the ABA's recommendations. RESULTS: The ICU cohort consisted of 107 patients (63.2% male, median age 37 years), with a median ICU estimated BSA of 40% and a median ICU weight estimation of 80 kg. In 43.9% of the cohort, the magnitude of the proportional difference between prehospital and ICU BSA estimate was greater than 25%. The 15-B formula accurately estimated the hourly fluid for all BSA (20%-100%) and weight combinations (50 kg-140 kg) in a BSA- weight matrix. When prehospital BSA estimate was utilized, 15-B guideline accurately estimated the fluid to be given within clinically significant limits for 97.2% of cases. CONCLUSIONS: The 15-B formula is a simple, easy to calculate guideline which approximates the early fluid estimates in severely burned patients despite inaccuracy in prehospital BSA estimates.

How often do leading biomedical journals use statistical experts to evaluate statistical methods? The results of a survey.

Scientific claims in biomedical research are typically derived from statistical analyses. However, misuse or misunderstanding of statistical procedures and results permeate the biomedical literature, affecting the validity of those claims. One approach journals have taken to address this issue is to enlist expert statistical reviewers. How many journals do this, how statistical review is incorporated, and how its value is perceived by editors is of interest. Here we report an expanded version of a survey conducted more than 20 years ago by Goodman and colleagues (1998) with the intention of characterizing contemporary statistical review policies at leading biomedical journals. We received eligible responses from 107 of 364 (28%) journals surveyed, across 57 fields, mostly from editors in chief. 34% (36/107) rarely or never use specialized statistical review, 34% (36/107) used it for 10-50% of their articles and 23% used it for all articles. These numbers have changed little since 1998 in spite of dramatically increased concern about research validity. The vast majority of editors regarded statistical review as having substantial incremental value beyond regular peer review and expressed comparatively little concern about the potential increase in reviewing time, cost, and difficulty identifying suitable statistical reviewers. Improved statistical education of researchers and different ways of employing statistical expertise are needed. Several proposals are discussed.

p de significación: ¿mejor no usarla si se interpreta mal? / P of significance: Is it Better to avoid it if it is poorly understood?

No disponible

Public Health Lessons Learned From Biases in Coronavirus Mortality Overestimation.

In testimony before US Congress on March 11, 2020, members of the House Oversight and Reform Committee were informed that estimated mortality for the novel coronavirus was 10-times higher than for seasonal influenza. Additional evidence, however, suggests the validity of this estimation could benefit from vetting for biases and miscalculations. The main objective of this article is to critically appraise the coronavirus mortality estimation presented to Congress. Informational texts from the World Health Organization and the Centers for Disease Control and Prevention are compared with coronavirus mortality calculations in Congressional testimony. Results of this critical appraisal reveal information bias and selection bias in coronavirus mortality overestimation, most likely caused by misclassifying an influenza infection fatality rate as a case fatality rate. Public health lessons learned for future infectious disease pandemics include: safeguarding against research biases that may underestimate or overestimate an associated risk of disease and mortality; reassessing the ethics of fear-based public health campaigns; and providing full public disclosure of adverse effects from severe mitigation measures to contain viral transmission.

Parametric Statistical Inference for Comparing Means and Variances.

Purpose: The purpose of this tutorial is to provide visual scientists with various approaches for comparing two or more groups of data using parametric statistical tests, which require that the distribution of data within each group is normal (Gaussian). Non-parametric tests are used for inference when the sample data are not normally distributed or the sample is too small to assess its true distribution. Methods: Methods are reviewed using retinal thickness, as measured by optical coherence tomography (OCT), as an example for comparing two or more group means. The following parametric statistical approaches are presented for different situations: two-sample t-test, Analysis of Variance (ANOVA), paired t-test, and the analysis of repeated measures data using a linear mixed-effects model approach. Results: Analyzing differences between means using various approaches is demonstrated, and follow-up procedures to analyze pairwise differences between means when there are more than two comparison groups are discussed. The assumption of equal variance between groups and methods to test for equal variances are examined. Examples of repeated measures analysis for right and left eyes on subjects, across spatial segments within the same eye (e.g. quadrants of each retina), and over time are given. Conclusions: This tutorial outlines parametric inference tests for comparing means of two or more groups and discusses how to interpret the output from statistical software packages. Critical assumptions made by the tests and ways of checking these assumptions are discussed. Efficient study designs increase the likelihood of detecting differences between groups if such differences exist. Situations commonly encountered by vision scientists involve repeated measures from the same subject over time, measurements on both right and left eyes from the same subject, and measurements from different locations within the same eye. Repeated measurements are usually correlated, and the statistical analysis needs to account for the correlation. Doing this the right way helps to ensure rigor so that the results can be repeated and validated.

Statistical Analysis and Reporting Guidelines for CHEST.

Considerable heterogeneity persists in the conduct and reporting of statistical analyses in the medical literature. Authors submitting manuscripts to CHEST are encouraged to adhere to the following guidelines where possible.

External Validation of Equations to Estimate Resting Energy Expenditure in 2037 Children and Adolescents with and 389 without Obesity: A Cross-Sectional Study.

We performed an external cross-validation study of 10 equations to estimate resting energy expenditure (REE) in 2037 children with and 389 without obesity. Inclusion criteria were Caucasian ethnicity, age ≤ 18 years, and availability of REE. REE was measured using indirect calorimetry. The correct classification fraction (CCF) of an equation was defined as the fraction of subjects whose estimated REE was within 10% of measured REE. The Molnár equation was the most accurate REE prediction equation with CCFs of 0.70 (95% CI 0.65 to 0.76) in girls without obesity, 0.64 (95% CI 0.61 to 0.66) in girls with obesity, 0.76 (95% CI 0.67 to 0.83) in boys without obesity, and 0.66 (95% CI 0.63 to 0.69) in boys with obesity. The Mifflin equation was the second most accurate equation with CCFs of 0.67 (95% CI 0.61 to 0.73) in girls without obesity, 0.61 (95% CI 0.58 to 0.64) in girls with obesity, 0.75 (95% CI 0.66 to 0.82) in boys without obesity, and 0.66 (95% CI 0.63 to 0.69) in boys with obesity.