State of the art in selection of variables and functional forms in multivariable analysis-outstanding issues

Diagnostic and prognostic value of CD44v9 and TIM3 expression in CK‑ and CK+ regions in gastric cancer tissues

The specificity and sensitivity of the current diagnostic and prognostic biomarkers for gastric cancer (GC) are limited. The present study aimed to evaluate the diagnostic and prognostic significance of cluster-of-differentiation gene 44 variant isoform 9 (CD44v9) and T cell immunoglobulin and mucin domain-containing protein 3 (TIM3) expression levels alone or combined in the tumor tissues of patients with GC and reveal the roles of CD44v9 and TIM3 in the cytokeratin (CK)+ and CK- regions. Multiplex immunofluorescence staining was performed for CD44v9, TIM3 and CK using a tissue microarray. The tissues were divided into three regions based on CK expression: Total, CK+, and CK- regions. The diagnostic and prognostic value was evaluated using receiver operating characteristic curves, Kaplan-Meier and Cox regression analyses. The results demonstrated that the density of cells expressing CD44v9, TIM3 and co-expressing CD44v9 and TIM3 (CD44v9/TIM3) in both the CK+ and CK- regions of tumor tissues was significantly higher than those in normal tissues (P<0.001). Moreover, the expression of CD44v9 in the CK- region was significantly positively correlated with age and tumor grade (P<0.05), and the expression of CD44v9/TIM3 in the CK- region of tumor tissues was significantly positively correlated with age, tumor grade and metastasis (P<0.05). Furthermore, the area under the curve for TIM3 expression in the CK+ region was 0.709, with a sensitivity of 45.83% and a specificity of 85.54% (P<0.001). High expression of CD44v9 in the CK- region was also significantly associated with poor survival and independently predicted a poor prognosis in patients with GC (hazard ratio, 2.387; 95% confidence interval, 1.384-4.118; P<0.01). In conclusion, dividing tissue regions based on CK expression is important for the diagnosis of GC. The expression of TIM3 in the CK+ region demonstrated diagnostic potential for GC, and high expression of CD44v9 in the CK- region was an independent prognostic risk factor for patients with GC.

Prognostic value of total, free and lipoprotein fraction-bound plasma mitotane levels in advanced adrenocortical carcinoma: a prospective study of the ENDOCAN-COMETE-Cancer network

PURPOSE

Mitotane is the only approved treatment for metastatic adrenocortical carcinoma (ACC). Monitoring plasma levels is recommended, but its predictive value is insufficient.

METHODS

This prospective study of the French ENDOCAN-COMETE network aimed to investigate the prognostic role of plasma mitotane levels pharmacokinetics and free or bound to lipoprotein fraction measurements during six consecutive months. Lipoprotein fractions were isolated by ultracentrifugation, and mitotane level was determined by HPLC-UV. Total, free, and lipoprotein fraction bound plasma mitotane were monitored every two months for six months with morphological assessment. The primary endpoint was overall survival (OS).

RESULTS

21 patients with metastatic ACC were included. Median overall survival was 23 months. The median free mitotane level per patient was 12% (± 7%), and the majority (88%) was bound to lipoprotein fractions. Several pharmacokinetics measures of total mitotane were related to OS: first level at one month (p = 0.026), mean level (p = 0.055), and area under the curve (AUC) (p = 0.048), with higher exposure associated to longer OS. Free mitotane (not bounded) and mitotane bounded to lipoprotein subfraction added no prognostic values. The relationship between the mitotane level and OS suggested a minimum "effective" threshold of 10-15 mg/L or an area under the curve above 100 mg/L/month with no individualized maximum value.

CONCLUSION

This prospective study did not identify any added prognostic value of free mitotane level over the total level. Early total mitotane level measurements (before 3-6 months) were related to OS with a higher and faster exposure related to more prolonged survival.

Baseline Dual-Energy Computed Tomography Urate Volume Predicts Fulfillment of Gout Remission After Two Years of Urate-Lowering Therapy

OBJECTIVE

This study aimed to identify variables that predict gout remission in people with erosive gout receiving urate-lowering therapy.

METHODS

We analyzed data from a two-year, double-masked randomized-controlled trial of people with erosive gout, randomized to a serum urate target of <0.20 mmol/L or <0.30 mmol/L using oral urate-lowering therapies. All participants had dual-energy computed tomography (DECT) scans of the feet and ankles at baseline. The proportion of participants achieving gout remission according to the 2016 preliminary gout remission criteria and simplified gout remission criteria (without the patient reported outcomes) was analyzed. Logistic regression models were used to evaluate predictors of gout remission in year 2.

RESULTS

The preliminary gout remission criteria were fulfilled in 11 of 97 participants (11%) at year 1 and 21 of 92 participants (23%) at year 2. The simplified criteria were fulfilled in 26 of 97 participants (27%) in year 1 and 40 of 92 participants (44%) in year 2. In multivariable regression models, baseline DECT monosodium urate crystal volume was the only significant independent predictor of gout remission at year 2, using either criteria. Each 1-cm3 increase in the baseline DECT monosodium urate crystal volume decreased the odds of fulfilling the 2016 preliminary gout remission criteria (odds ratio [OR] 0.65, 95% confidence interval [CI] 0.46-0.93; P = 0.02) and the simplified gout remission criteria (OR 0.57, 95% CI 0.41-0.78; P < 0.001).

CONCLUSION

In people with erosive gout on urate-lowering therapy, higher baseline DECT monosodium urate crystal volume is associated with lower odds of gout remission after two years of treatment, defined by either the preliminary gout remission criteria or simplified gout remission criteria.

Regression without regrets -initial data analysis is a prerequisite for multivariable regression

Statistical regression models are used for predicting outcomes based on the values of some predictor variables or for describing the association of an outcome with predictors. With a data set at hand, a regression model can be easily fit with standard software packages. This bears the risk that data analysts may rush to perform sophisticated analyses without sufficient knowledge of basic properties, associations in and errors of their data, leading to wrong interpretation and presentation of the modeling results that lacks clarity. Ignorance about special features of the data such as redundancies or particular distributions may even invalidate the chosen analysis strategy. Initial data analysis (IDA) is prerequisite to regression analyses as it provides knowledge about the data needed to confirm the appropriateness of or to refine a chosen model building strategy, to interpret the modeling results correctly, and to guide the presentation of modeling results. In order to facilitate reproducibility, IDA needs to be preplanned, an IDA plan should be included in the general statistical analysis plan of a research project, and results should be well documented. Biased statistical inference of the final regression model can be minimized if IDA abstains from evaluating associations of outcome and predictors, a key principle of IDA. We give advice on which aspects to consider in an IDA plan for data screening in the context of regression modeling to supplement the statistical analysis plan. We illustrate this IDA plan for data screening in an example of a typical diagnostic modeling project and give recommendations for data visualizations.

Multivariate modeling and prediction of cerebral physiology in acute traumatic neural injury: A scoping review

Traumatic brain injury (TBI) poses a significant global public health challenge necessitating a profound understanding of cerebral physiology. The dynamic nature of TBI demands sophisticated methodologies for modeling and predicting cerebral signals to unravel intricate pathophysiology and predict secondary injury mechanisms prior to their occurrence. In this comprehensive scoping review, we focus specifically on multivariate cerebral physiologic signal analysis in the context of multi-modal monitoring (MMM) in TBI, exploring a range of techniques including multivariate statistical time-series models and machine learning algorithms. Conducting a comprehensive search across databases yielded 7 studies for evaluation, encompassing diverse cerebral physiologic signals and parameters from TBI patients. Among these, five studies concentrated on modeling cerebral physiologic signals using statistical time-series models, while the remaining two studies primarily delved into intracranial pressure (ICP) prediction through machine learning models. Autoregressive models were predominantly utilized in the modeling studies. In the context of prediction studies, logistic regression and Gaussian processes (GP) emerged as the predominant choice in both research endeavors, with their performance being evaluated against each other in one study and other models such as random forest, and decision tree in the other study. Notably among these models, random forest model, an ensemble learning approach, demonstrated superior performance across various metrics. Additionally, a notable gap was identified concerning the absence of studies focusing on prediction for multivariate outcomes. This review addresses existing knowledge gaps and sets the stage for future research in advancing cerebral physiologic signal analysis for neurocritical care improvement.

Comparing secondary prevention for patients with coronary heart disease and stroke attending Australian general practices: a cross-sectional study using nationwide electronic database

OBJECTIVES

To compare secondary prevention care for patients with coronary heart disease (CHD) and stroke, exploring particularly the influences due to frequency and regularity of primary care visits.

SETTING

Secondary prevention for patients (≥18 years) in the National Prescription Service administrative electronic health record database collated from 458 Australian general practice sites across all states and territories.

DESIGN

Retrospective cross-sectional and panel study. Patient and care-level characteristics were compared for differing CHD/stroke diagnoses. Associations between the type of cardiovascular diagnosis and medication prescription as well as risk factor assessment were examined using multivariable logistic regression.

PARTICIPANTS

Patients with three or more general practice encounters within 2 years of their latest visit during 2016-2020.

OUTCOME MEASURES

Proportions and odds ratios (ORs) for (1) prescription of antihypertensives, antilipidaemics and antiplatelets and (2) assessment of blood pressure (BP) and low-density lipoprotein cholesterol (LDL-C) in patients with stroke only compared against those with CHD only and those with both conditions.

RESULTS

There were 111 892 patients with CHD only, 27 863 with stroke only and 9791 with both conditions. Relative to patients with CHD, patients with stroke were underprescribed antihypertensives (70.8% vs 82.8%), antilipidaemics (63.1% vs 78.7%) and antiplatelets (42.2% vs 45.7%). With sociodemographic factors, comorbidities and level of care considered as covariates, the odds of non-prescription of any recommended secondary prevention medications were higher in patients with stroke only (adjusted OR 1.37; 95% CI (1.31, 1.44)) compared with patients with CHD only. Patients with stroke only were also more likely to have neither BP nor LDL-C monitored (adjusted OR 1.26; 95% CI (1.18, 1.34)). Frequent and regular general practitioner encounters were independently associated with the prescription of secondary prevention medications (p<0.001).

CONCLUSIONS

Secondary prevention management is suboptimal in cardiovascular disease patients and worse post-stroke compared with post-CHD. More frequent and regular primary care encounters were associated with improved secondary prevention.

Bootstrap Approximation of Model Selection Probabilities for Multimodel Inference Frameworks

Most statistical modeling applications involve the consideration of a candidate collection of models based on various sets of explanatory variables. The candidate models may also differ in terms of the structural formulations for the systematic component and the posited probability distributions for the random component. A common practice is to use an information criterion to select a model from the collection that provides an optimal balance between fidelity to the data and parsimony. The analyst then typically proceeds as if the chosen model was the only model ever considered. However, such a practice fails to account for the variability inherent in the model selection process, which can lead to inappropriate inferential results and conclusions. In recent years, inferential methods have been proposed for multimodel frameworks that attempt to provide an appropriate accounting of modeling uncertainty. In the frequentist paradigm, such methods should ideally involve model selection probabilities, i.e., the relative frequencies of selection for each candidate model based on repeated sampling. Model selection probabilities can be conveniently approximated through bootstrapping. When the Akaike information criterion is employed, Akaike weights are also commonly used as a surrogate for selection probabilities. In this work, we show that the conventional bootstrap approach for approximating model selection probabilities is impacted by bias. We propose a simple correction to adjust for this bias. We also argue that Akaike weights do not provide adequate approximations for selection probabilities, although they do provide a crude gauge of model plausibility.

What role do socioeconomic and clinical factors play in disease activity states in rheumatoid arthritis? Data from a large UK early inflammatory arthritis audit

BACKGROUND

Persistently active rheumatoid arthritis (pactiveRA) may be due to the interplay between biological and non-biological factors. The role of socioeconomic factors remains unclear.

OBJECTIVES

To explore which biological and non-biological factors associate with pactiveRA.

METHODS

Adults with early RA in the National Early Inflammatory Arthritis Audit, recruited from May 2018 to October 2022, were included if having pactiveRA or persistently low RA (plowRA). The pactiveRA was defined as three consecutive Disease Activity Score-28 joints (DAS28) of >3.2 at baseline, 3 and 12 months. The plowRA was defined as DAS28 ≤3.2 at 3 and 12 months. Stepwise forward logistic regression was used to explore associations with pactiveRA (outcome). Age and gender were included a priori, with socioeconomic factors and comorbidities as exposure variables.

RESULTS

682 patients with pactiveRA and 1026 plowRA were included. Compared with plowRA, patients with pactiveRA were younger (58, IQR: 49-67) versus (62, IQR: 52-72), and included more women (69% vs 59%). The pactiveRA was associated with worse scores in patient-reported outcomes at baseline, and anxiety and depression screens. Overall, there was clear social patterning in pactiveRA, with age-by-gender interaction. Logistic regression indicated age, gender, social deprivation and previous or current smoking, were independently associated with pactiveRA, after controlling for disease severity markers (seropositivity). Depression, lung disease, gastric ulcers and baseline corticosteroid use, were also associated with pactiveRA (p<0.05 for all).

CONCLUSION

Socioeconomic factors and deprivation were associated with pactiveRA, independent of clinical and disease characteristics. Identifying 'adverse' socioeconomic drivers of pactiveRA can help tailor interventions according to individual need.

Categorization of continuous covariates and complex regression models-friends or foes in intersectionality research

OBJECTIVES

To reduce health inequities, it is important to identify intersections in characteristics of individuals subject to privilege or disadvantage. Different proposals for that have recently been published. One approach (1) considers models specified with first- and all second-order effects and another (2) the stratification based on multiple covariates; both categorize continuous covariates. A simulation study was conducted in order to review both methods with regard to identification of intersections showing true differences, rate of false-positive results, and generalizability to independent data compared to an established approach (3) of backward variable elimination according to Bayesian information criterion (BE-BIC) combined with splines.

STUDY DESIGN AND SETTING

R software has been used to simulate the covariates age, sex, body mass index, education, and diabetes to examine their association with a continuous frailty score for osteoporosis using multiple linear regression. In setting 1, none of the covariates was associated with the frailty score, that is, only noise is present in the data. In setting 2, the covariates age, sex, and their interaction were associated with the frailty score, such that only females above 55 years formed an intersection associated with an increased frailty score. All approaches were compared under varying sample sizes (N = 200-3000) and signal-to-noise ratios (SNRs, 0.5-4) in 1000 replications. For model evaluation, bootstrap resampling was used. The models were fitted in internal learning data and then used to predict outcomes in the internal validation data. The mean squared error (MSE) was used for comparison and the frequency of false-positive findings calculated.

RESULTS

In setting 1, approaches 1 and 2 generated spurious effects in more than 90% of simulations across all sample sizes. In a smaller sample size, approach 3 (BE-BIC) selected 36.5% of the correct model, in larger sample size in 89.8% and always had a lower number of spurious effects. MSE in independent data was generally higher for approaches 1 and 2 when compared to 3. In setting 2, approach 1 selected most frequently the correct interaction but frequently showed spurious effects (>75%). Across all sample sizes and SNR, approach 3 generated least often spurious results and had lowest MSE in independent data.

CONCLUSION

Categorization of continuous covariates is detrimental to studies on intersectionality. Due to high and unrestricted model complexity, such approaches are prone to spurious effects and often lack interpretability. Approach 3 (BE-BIC) is considerably more robust against spurious findings, showed better generalizability to independent data, and can be used with most statistical software. For intersectionality research, we consider it most important to describe relevant differences between intersections and to avoid nonreproducible and spurious findings.

Global patterns of rebound to normal RSV dynamics following COVID-19 suppression

BACKGROUND

Annual epidemics of respiratory syncytial virus (RSV) had consistent timing and intensity between seasons prior to the SARS-CoV-2 pandemic (COVID-19). However, starting in April 2020, RSV seasonal activity declined due to COVID-19 non-pharmaceutical interventions (NPIs) before re-emerging after relaxation of NPIs. We described the unusual patterns of RSV epidemics that occurred in multiple subsequent waves following COVID-19 in different countries and explored factors associated with these patterns.

METHODS

Weekly cases of RSV from twenty-eight countries were obtained from the World Health Organisation and combined with data on country-level characteristics and the stringency of the COVID-19 response. Dynamic time warping and regression were used to cluster time series patterns and describe epidemic characteristics before and after COVID-19 pandemic, and identify related factors.

RESULTS

While the first wave of RSV epidemics following pandemic suppression exhibited unusual patterns, the second and third waves more closely resembled typical RSV patterns in many countries. Post-pandemic RSV patterns differed in their intensity and/or timing, with several broad patterns across the countries. The onset and peak timings of the first and second waves of RSV epidemics following COVID-19 suppression were earlier in the Southern than Northern Hemisphere. The second wave of RSV epidemics was also earlier with higher population density, and delayed if the intensity of the first wave was higher. More stringent NPIs were associated with lower RSV growth rate and intensity and a shorter gap between the first and second waves.

CONCLUSION

Patterns of RSV activity have largely returned to normal following successive waves in the post-pandemic era. Onset and peak timings of future epidemics following disruption of normal RSV dynamics need close monitoring to inform the delivery of preventive and control measures.

Evaluation of health care utilisation and mortality in medical hospitalisations with multimorbidity and kidney disease, according to frailty: a nationwide cohort study

INTRODUCTION

The impact of impaired kidney function on healthcare use among medical hospitalisations with multimorbidity and frailty is incompletely understood. In this study, we assessed the prevalence of acute kidney injury (AKI) and chronic kidney disease (CKD) among multimorbid medical hospitalisations in Switzerland and explored the associations of kidney disease with in-hospital outcomes across different frailty strata.

METHODS

This observational study analysed nationwide hospitalisation records from 1 January 2012 to 31 December 2020. We included adults (age ≥18 years) with underlying multimorbidity hospitalised in a medical ward. The study population consisted of hospitalisations with AKI, CKD or no kidney disease (reference group), and was stratified by three frailty levels (non-frail, pre-frail, frail). Main outcomes were in-hospital mortality, intensive care unit (ICU) treatment, length of stay (LOS) and all-cause 30-day readmission. We estimated multivariable adjusted odds ratios (OR) and changes in percentage of log-transformed continuous outcomes with 95% confidence intervals (CI).

RESULTS

Among 2,651,501 medical hospitalisations with multimorbidity, 198,870 had a diagnosis of AKI (7.5%), 452,990 a diagnosis of CKD (17.1%) and 1,999,641 (75.4%) no kidney disease. For the reference group, the risk of in-hospital mortality was 4.4%, for the AKI group 14.4% (adjusted odds ratio [aOR] 2.56 [95% CI 2.52-2.61]) and for the CKD group 5.9% (aOR 0.98 [95% CI 0.96-0.99]), while prevalence of ICU treatment was, respectively, 10.5%, 21.8% (aOR 2.39 [95% CI 2.36-2.43]) and 9.3% (aOR 1.01 [95% CI 1.00-1.02]). Median LOS was 5 days (interquartile range [IQR] 2.0-9.0) in hospitalisations without kidney disease, 9 days (IQR 5.0-15.0) (adjusted change [%] 67.13% [95% CI 66.18-68.08%]) in those with AKI and 7 days (IQR 4.0-12.0) (adjusted change [%] 18.94% [95% CI 18.52-19.36%]) in those with CKD. The prevalence of 30-day readmission was, respectively, 13.3%, 13.7% (aOR 1.21 [95% CI 1.19-1.23]) and 14.8% (aOR 1.26 [95% CI 1.25-1.28]). In general, the frequency of adverse outcomes increased with the severity of frailty.

CONCLUSION

In medical hospitalisations with multimorbidity, the presence of AKI or CKD was associated with substantial additional hospitalisations and healthcare utilisation across all frailty strata. This information is of major importance for cost estimates and should stimulate discussion on reimbursement.

Flexible parametrization of graph-theoretical features from individual-specific networks for prediction

Statistical techniques are needed to analyze data structures with complex dependencies such that clinically useful information can be extracted. Individual-specific networks, which capture dependencies in complex biological systems, are often summarized by graph-theoretical features. These features, which lend themselves to outcome modeling, can be subject to high variability due to arbitrary decisions in network inference and noise. Correlation-based adjacency matrices often need to be sparsified before meaningful graph-theoretical features can be extracted, requiring the data analysts to determine an optimal threshold. To address this issue, we propose to incorporate a flexible weighting function over the full range of possible thresholds to capture the variability of graph-theoretical features over the threshold domain. The potential of this approach, which extends concepts from functional data analysis to a graph-theoretical setting, is explored in a plasmode simulation study using real functional magnetic resonance imaging (fMRI) data from the Autism Brain Imaging Data Exchange (ABIDE) Preprocessed initiative. The simulations show that our modeling approach yields accurate estimates of the functional form of the weight function, improves inference efficiency, and achieves a comparable or reduced root mean square prediction error compared to competitor modeling approaches. This assertion holds true in settings where both complex functional forms underlie the outcome-generating process and a universal threshold value is employed. We demonstrate the practical utility of our approach by using resting-state fMRI data to predict biological age in children. Our study establishes the flexible modeling approach as a statistically principled, serious competitor to ad-hoc methods with superior performance.

Prognostic model of in-hospital ischemic stroke mortality based on an electronic health record cohort in Indonesia

Stroke patients rarely have satisfactory survival, which worsens further if comorbidities develop in such patients. Limited data availability from Southeast Asian countries, especially Indonesia, has impeded the disentanglement of post-stroke mortality determinants. This study aimed to investigate predictors of in-hospital mortality in patients with ischemic stroke (IS). This retrospective observational study used IS medical records from the National Brain Centre Hospital, Jakarta, Indonesia. A theoretically driven Cox's regression and Fine-Gray models were established by controlling for age and sex to calculate the hazard ratio of each plausible risk factor for predicting in-hospital stroke mortality and addressing competing risks if they existed. This study finally included 3,278 patients with IS, 917 (28%) of whom had cardiovascular disease and 376 (11.5%) suffered renal disease. Bivariate exploratory analysis revealed lower blood levels of triglycerides, low density lipoprotein, and total cholesterol associated with in-hospital-stroke mortality. The average age of patients with post-stroke mortality was 64.06 ± 11.32 years, with a mean body mass index (BMI) of 23.77 kg/m2 and a median Glasgow Coma Scale (GCS) score of 12 and an IQR of 5. Cardiovascular disease was significantly associated with IS mortality risk. NIHSS score at admission (hazard ratio [HR] = 1.04; 95% confidence interval [CI]: 1.00-1.07), male sex (HR = 1.51[1.01-2.26] and uric acid level (HR = 1.02 [1.00-1.03]) predicted survivability. Comorbidities, such as cardiovascular disease (HR = 2.16 [1.37-3.40], pneumonia (HR = 2.43 [1.42-4.15] and sepsis (HR = 2.07 [1.09-3.94, had higher hazards for post-stroke mortality. Contrarily, the factors contributing to a lower hazard of mortality were BMI (HR = 0.94 [0.89-0.99]) and GCS (HReye = 0.66 [0.48-0.89]. In summary, our study reported that male sex, NIHSS, uric acid level, cardiovascular diseases, pneumonia, sepsis. BMI, and GCS on admission were strong determinants of in-hospital mortality in patients with IS.

What question are we trying to answer? Embracing causal inference

This study summarizes a presentation at the symposium for the Calvin Schwabe Award for Lifetime Achievement in Veterinary Epidemiology and Preventive Medicine, which was awarded to the first author. As epidemiologists, we are taught that "correlation does not imply causation." While true, identifying causes is a key objective for much of the research that we conduct. There is empirical evidence that veterinary epidemiologists are conducting observational research with the intent to identify causes; many studies include control for confounding variables, and causal language is often used when interpreting study results. Frameworks for studying causes include the articulation of specific hypotheses to be tested, approaches for the selection of variables, methods for statistical estimation of the relationship between the exposure and the outcome, and interpretation of that relationship as causal. When comparing observational studies in veterinary populations to those conducted in human populations, the application of each of these steps differs substantially. The a priori identification of exposure-outcome pairs of interest are less common in observational studies in the veterinary literature compared to the human literature, and prior knowledge is used to select confounding variables in most observational studies in human populations, whereas data-driven approaches are the norm in veterinary populations. The consequences of not having a defined exposure-outcome hypotheses of interest and using data-driven analytical approaches include an increased probability of biased results and poor replicability of results. A discussion by the community of researchers on current approaches to studying causes in observational studies in veterinary populations is warranted.

Misstatements, misperceptions, and mistakes in controlling for covariates in observational research

We discuss 12 misperceptions, misstatements, or mistakes concerning the use of covariates in observational or nonrandomized research. Additionally, we offer advice to help investigators, editors, reviewers, and readers make more informed decisions about conducting and interpreting research where the influence of covariates may be at issue. We primarily address misperceptions in the context of statistical management of the covariates through various forms of modeling, although we also emphasize design and model or variable selection. Other approaches to addressing the effects of covariates, including matching, have logical extensions from what we discuss here but are not dwelled upon heavily. The misperceptions, misstatements, or mistakes we discuss include accurate representation of covariates, effects of measurement error, overreliance on covariate categorization, underestimation of power loss when controlling for covariates, misinterpretation of significance in statistical models, and misconceptions about confounding variables, selecting on a collider, and p value interpretations in covariate-inclusive analyses. This condensed overview serves to correct common errors and improve research quality in general and in nutrition research specifically.

Investigating the uses of machine learning algorithms to inform risk factor analyses: The example of avian infectious bronchitis virus (IBV) in broiler chickens

Infectious bronchitis virus (IBV) is a contagious coronavirus causing respiratory and urogenital disease in chickens and is responsible for significant economic losses for both the broiler and table egg layer industries. Despite IBV being regularly monitored using standard epidemiologic surveillance practices, knowledge and evidence of risk factors associated with IBV transmission remain limited. The study objective was to compare risk factor modeling outcomes between a traditional stepwise variable selection approach and a machine learning-based random forest Boruta algorithm using routinely collected IBV antibody titer data from broiler flocks. IBV antibody sampling events (n = 1111) from 166 broiler sites between 2016 and 2021 were accessed. Ninety-two geospatial-related and poultry-density variables were obtained using a geographic information system and data sets from publicly available sources. Seventeen and 27 candidate variables were screened to potentially have an association with elevated IBV antibody titers according to the manual selection and machine learning algorithm, respectively. Selected variables from both methods were further investigated by construction of multivariable generalized mixed logistic regression models. Six variables were shortlisted by both screening methods, which included year, distance to urban areas, main roads, landcover, density of layer sites and year, however, final models for both approaches only shared year as an important predictor. Despite limited significance of clinical outcomes, this work showcases the potential of a novel explorative modeling approach in combination with often unutilized resources such as publicly available geospatial data, surveillance health data and machine learning as potential supplementary tools to investigate risk factors related to infectious diseases.

Mortality rates among adult critical care patients with unusual or extreme values of vital signs and other physiological parameters: a retrospective study

BACKGROUND

We evaluated relationships of vital signs and laboratory-tested physiological parameters with in-hospital mortality, focusing on values that are unusual or extreme even in critical care settings.

METHODS

We retrospectively studied Philips Healthcare-MIT eICU data (207 U.S. hospitals, 20142015), including 166,959 adult-patient critical care admissions. Analyzing most-deranged (worst) value measured in the first admission day, we investigated vital signs (body temperature, heart rate, mean arterial pressure, and respiratory rate) as well as albumin, bilirubin, blood pH via arterial blood gas (ABG), blood urea nitrogen, creatinine, FiO2 ABG, glucose, hematocrit, PaO2 ABG, PaCO2 ABG, sodium, 24-hour urine output, and white blood cell count (WBC).

RESULTS

In-hospital mortality was ≥50% at extremes of low blood pH, low and high body temperature, low albumin, low glucose, and low heart rate. Near extremes of blood pH, temperature, glucose, heart rate, PaO2 , and WBC, relatively. Small changes in measured values correlated with several-fold mortality rate increases. However, high mortality rates and abrupt mortality increases were often hidden by the common practice of thresholding or binning physiological parameters. The best predictors of in-hospital mortality were blood pH, temperature, and FiO2 (scaled Brier scores: 0.084, 0.063, and 0.049, respectively).

CONCLUSIONS

In-hospital mortality is high and sharply increasing at extremes of blood pH, body temperature, and other parameters. Common-practice thresholding obscures these associations. In practice, vital signs are sometimes treated more casually than laboratory-tested parameters. Yet, vitals are easier to obtain and we found they are often the best mortality predictors, supporting perspectives that vitals are undervalued.

The underlying mechanisms by which Post-Traumatic Growth is associated with cardiovascular health in male UK military personnel: The ADVANCE cohort study

Post-Traumatic Growth (PTG) is associated with good cardiovascular health, but the mechanisms of this are poorly understood. This cross-sectional analysis assessed whether factors of PTG (Appreciation of Life (AOL), New Possibilities (NP), Personal Strength (PS), Relating to Others (RTO) and Spiritual Change (SC)) are associated with cardiovascular health in a cohort of 1006 male UK military personnel (median age 34). The findings suggest AOL, PS and RTO are associated with better cardiovascular health through cardiometabolic effects (lower levels of triglycerides, and total cholesterol) and haemodynamic functioning (lower diastolic blood pressure), but not inflammation. However, NP and SC were associated with poorer cardiovascular health through cardiometabolic effects (lower levels of high-density lipoproteins and higher levels of total cholesterol) and AOL had a non-linear association with low-density lipoproteins. These findings suggest that the relationship between PTG and cardiovascular functioning is complex and in need of further scrutiny.

Addition of inflammation-related biomarkers to the CAIDE model for risk prediction of all-cause dementia, Alzheimer's disease and vascular dementia in a prospective study

BACKGROUND

It is of interest whether inflammatory biomarkers can improve dementia prediction models, such as the widely used Cardiovascular Risk Factors, Aging and Dementia (CAIDE) model.

METHODS

The Olink Target 96 Inflammation panel was assessed in a nested case-cohort design within a large, population-based German cohort study (n = 9940; age-range: 50-75 years). All study participants who developed dementia over 20 years of follow-up and had complete CAIDE variable data (n = 562, including 173 Alzheimer's disease (AD) and 199 vascular dementia (VD) cases) as well as n = 1,356 controls were selected for measurements. 69 inflammation-related biomarkers were eligible for use. LASSO logistic regression and bootstrapping were utilized to select relevant biomarkers and determine areas under the curve (AUCs).

RESULTS

The CAIDE model 2 (including Apolipoprotein E (APOE) ε4 carrier status) predicted all-cause dementia, AD, and VD better than CAIDE model 1 (without APOE ε4) with AUCs of 0.725, 0.752 and 0.707, respectively. Although 20, 7, and 4 inflammation-related biomarkers were selected by LASSO regression to improve CAIDE model 2, the AUCs did not increase markedly. CAIDE models 1 and 2 generally performed better in mid-life (50-64 years) than in late-life (65-75 years) sub-samples of our cohort, but again, inflammation-related biomarkers did not improve their predictive abilities.

CONCLUSIONS

Despite a lack of improvement in dementia risk prediction, the selected inflammation-related biomarkers were significantly associated with dementia outcomes and may serve as a starting point to further elucidate the pathogenesis of dementia.

External validation of six COVID-19 prognostic models for predicting mortality risk in older populations in a hospital, primary care, and nursing home setting

OBJECTIVES

To systematically evaluate the performance of COVID-19 prognostic models and scores for mortality risk in older populations across three health-care settings: hospitals, primary care, and nursing homes.

STUDY DESIGN AND SETTING

This retrospective external validation study included 14,092 older individuals of ≥70 years of age with a clinical or polymerase chain reaction-confirmed COVID-19 diagnosis from March 2020 to December 2020. The six validation cohorts include three hospital-based (CliniCo, COVID-OLD, COVID-PREDICT), two primary care-based (Julius General Practitioners Network/Academisch network huisartsgeneeskunde/Network of Academic general Practitioners, PHARMO), and one nursing home cohort (YSIS) in the Netherlands. Based on a living systematic review of COVID-19 prediction models using Prediction model Risk Of Bias ASsessment Tool for quality and risk of bias assessment and considering predictor availability in validation cohorts, we selected six prognostic models predicting mortality risk in adults with COVID-19 infection (GAL-COVID-19 mortality, 4C Mortality Score, National Early Warning Score 2-extended model, Xie model, Wang clinical model, and CURB65 score). All six prognostic models were validated in the hospital cohorts and the GAL-COVID-19 mortality model was validated in all three healthcare settings. The primary outcome was in-hospital mortality for hospitals and 28-day mortality for primary care and nursing home settings. Model performance was evaluated in each validation cohort separately in terms of discrimination, calibration, and decision curves. An intercept update was performed in models indicating miscalibration followed by predictive performance re-evaluation.

MAIN OUTCOME MEASURE

In-hospital mortality for hospitals and 28-day mortality for primary care and nursing home setting.

RESULTS

All six prognostic models performed poorly and showed miscalibration in the older population cohorts. In the hospital settings, model performance ranged from calibration-in-the-large -1.45 to 7.46, calibration slopes 0.24-0.81, and C-statistic 0.55-0.71 with 4C Mortality Score performing as the most discriminative and well-calibrated model. Performance across health-care settings was similar for the GAL-COVID-19 model, with a calibration-in-the-large in the range of -2.35 to -0.15 indicating overestimation, calibration slopes of 0.24-0.81 indicating signs of overfitting, and C-statistic of 0.55-0.71.

CONCLUSION

Our results show that most prognostic models for predicting mortality risk performed poorly in the older population with COVID-19, in each health-care setting: hospital, primary care, and nursing home settings. Insights into factors influencing predictive model performance in the older population are needed for pandemic preparedness and reliable prognostication of health-related outcomes in this demographic.

Laboratory biomarkers associated with COVID-19 mortality among inpatients in a Peruvian referral hospital

Aim

To evaluate the biochemical and hematological markers associated with the risk of death due to COVID-19 in a clinical cohort with a severe clinical profile.

Methods

A retrospective study was conducted among 215 anonymized inpatient records from the Hospital Nacional Almanzor Aguinaga Asenjo, Peru, between April and June 2020. The association between biomarkers and death due to COVID-19 was assessed using Cox regression, with a multivariable modeling of 1) biochemical and 2) hematological markers. Kaplan-Meier analyses and time-dependent receiver operating characteristic curves were calculated for each associated biomarker (p < 0.05).

Results

Data analysis of 215 inpatient records revealed an overall mortality rate of 51.30% (95% CI 44.70-58.50), a mean age of 63.90 ± 14.10 years, and a median oxygen saturation of 88% (interquartile range 82-92%). The best-fitted biochemical model included higher levels of C-reactive protein (CRP), D-dimer, fibrinogen, urea, and lactate dehydrogenase. Similarly, the best-fitted hematological model included higher absolute neutrophil and prothrombin time, and lower absolute platelet counts. The best area under the curve values in both models were found to be CRP and D-dimer values (>0.74) and the absolute neutrophil count (0.63).

Conclusions

Some specific biochemical markers outperformed hematological markers. Evaluated hematological counts analyzed in multivariable models proved to be better markers and could be useful to discriminate COVID-19 patients at high risk of death.

Comparison of mortality and outcomes of four respiratory viruses in the intensive care unit: a multicenter retrospective study

This retrospective study aimed to compare the mortality and burden of respiratory syncytial virus (RSV group), SARS-CoV-2 (COVID-19 group), non-H1N1 (Seasonal influenza group) and H1N1 influenza (H1N1 group) in adult patients admitted to intensive care unit (ICU) with respiratory failure. A total of 807 patients were included. Mortality was compared between the four following groups: RSV, COVID-19, seasonal influenza, and H1N1 groups. Patients in the RSV group had significantly more comorbidities than the other patients. At admission, patients in the COVID-19 group were significantly less severe than the others according to the simplified acute physiology score-2 (SAPS-II) and sepsis-related organ failure assessment (SOFA) scores. Using competing risk regression, COVID-19 (sHR = 1.61; 95% CI 1.10; 2.36) and H1N1 (sHR = 1.87; 95% CI 1.20; 2.93) were associated with a statistically significant higher mortality while seasonal influenza was not (sHR = 0.93; 95% CI 0.65; 1.31), when compared to RSV. Despite occurring in more severe patients, RSV and seasonal influenza group appear to be associated with a more favorable outcome than COVID-19 and H1N1 groups.

Outcomes after laparoscopic cholecystectomy in patients older than 80 years: two-years follow-up

BACKGROUND

The laparoscopic cholecystectomy is the treatment of choice for patients with benign biliary disease. It is necessary to evaluate survival after laparoscopic cholecystectomy in patients over 80 years old to determine whether the long-term mortality rate is higher than the reported recurrence rate. If so, this age group could benefit from a more conservative approach, such as antibiotic treatment or cholecystostomy. Therefore, the aim of this study was to evaluate the factors associated with 2 years survival after laparoscopic cholecystectomy in patients over 80 years old.

METHODS

We conducted a retrospective observational cohort study. We included all patients over 80 years old who underwent laparoscopic cholecystectomy. Survival analysis was conducted using the Kaplan‒Meier method. Cox regression analysis was implemented to determine potential factors associated with mortality at 24 months.

RESULTS

A total of 144 patients were included in the study, of whom 37 (25.69%) died at the two-year follow-up. Survival curves were compared for different ASA groups, showing a higher proportion of survivors at two years among patients classified as ASA 1-2 at 87.50% compared to ASA 3-4 at 63.75% (p = 0.001). An ASA score of 3-4 was identified as a statistically significant factor associated with mortality, indicating a higher risk (HR: 2.71, CI95%:1.20-6.14).

CONCLUSIONS

ASA 3-4 patients may benefit from conservative management due to their higher risk of mortality at 2 years and a lower probability of disease recurrence.

Investigation of the involvement of platelet-activating factor in the control of hypertension by aerobic training. A randomized controlled trial

Although studies have demonstrated the effectiveness of exercise in controlling systemic arterial hypertension (SAH), the mechanisms involved in this effect are still poorly understood. Thus, this study investigated the impact of aerobic training on the relationship between platelet-activating factor (PAF) circulating levels and blood pressure in hypertensives. Seventy-seven hypertensive subjects were enrolled in this randomized controlled trial (age 66.51 ± 7.53 years, body mass 76.17 ± 14.19 kg). Participants were randomized to two groups: the intervention group (IG, n = 36), composed of hypertensive individuals submitted to an aerobic training protocol, and the control group (CG, n = 41), composed of non-exercised hypertensives. Body mass index, arterial blood pressure, quality of life, respiratory muscle strength, and functional capacity were assessed before and after 12 weeks. PAF and plasma cytokine levels were also evaluated respectively by liquid chromatography coupled with mass spectrometry and enzyme-linked immunosorbent assay. Aerobic training promoted a significant reduction in blood pressure while functional capacity, expiratory muscle strength, and quality of life, PAFC16:0 and PAFC18:1 plasma levels were increased in comparison to the CG (p < 0.05). In addition, multiple correlation analysis indicated a positive correlation [F (3.19) = 6.322; p = 0.001; R2adjusted = 0.499] between PAFC16:0 levels and expiratory muscle strength after aerobic training. Taken together, our findings indicate that PAF may be involved in the indirect mechanisms that control SAH, being mainly associated with increased respiratory muscle strength in hypertensive subjects undergoing aerobic training.

Raising awareness of uncertain choices in empirical data analysis: A teaching concept toward replicable research practices

Throughout their education and when reading the scientific literature, students may get the impression that there is a unique and correct analysis strategy for every data analysis task and that this analysis strategy will always yield a significant and noteworthy result. This expectation conflicts with a growing realization that there is a multiplicity of possible analysis strategies in empirical research, which will lead to overoptimism and nonreplicable research findings if it is combined with result-dependent selective reporting. Here, we argue that students are often ill-equipped for real-world data analysis tasks and unprepared for the dangers of selectively reporting the most promising results. We present a seminar course intended for advanced undergraduates and beginning graduate students of data analysis fields such as statistics, data science, or bioinformatics that aims to increase the awareness of uncertain choices in the analysis of empirical data and present ways to deal with these choices through theoretical modules and practical hands-on sessions.

Recent Methodological Trends in Epidemiology: No Need for Data-Driven Variable Selection?

Variable selection in regression models is a particularly important issue in epidemiology, where one usually encounters observational studies. In contrast to randomized trials or experiments, confounding is often not controlled by the study design, but has to be accounted for by suitable statistical methods. For instance, when risk factors should be identified with unconfounded effect estimates, multivariable regression techniques can help to adjust for confounders. We investigated the current practice of variable selection in 4 major epidemiologic journals in 2019 and found that the majority of articles used subject-matter knowledge to determine a priori the set of included variables. In comparison with previous reviews from 2008 and 2015, fewer articles applied data-driven variable selection. Furthermore, for most articles the main aim of analysis was hypothesis-driven effect estimation in rather low-dimensional data situations (i.e., large sample size compared with the number of variables). Based on our results, we discuss the role of data-driven variable selection in epidemiology.

Application of an in-hospital, surgeon-led anti-osteoporotic medication algorithm in patients with hip fractures improves persistence to medication and can prevent the second fragility fracture

INTRODUCTION

Secondary fracture prevention is an essential part of hip fracture treatment. Despite this, many patients are discharged without the appropriate anti-osteoporotic medication. The aim of this study is to report the outcomes of the application of an in-hospital, surgeon-led anti-osteoporotic medication algorithm to patients with hip fractures.

MATERIALS AND METHODS

This prospective cohort study followed patients with hip fractures who were treated at a tertiary referral hospital between 2020 and 2022. At discharge, anti-osteoporotic medication according to the Arbeitsgemeinschaft für Osteosynthesefragen (AO) Foundation algorithm was prescribed to all patients. Multivariate Cox regression analysis was used to investigate the risks of non-persistence to medication and of secondary fracture.

RESULTS

Two hundred thirteen consecutive patients were prospectively followed. Mean follow-up was 17.2 ± 7.1 months. Persistence to medication at 2 years was 58% (95%CI 51-65%). A secondary osteoporotic fracture occurred in 1/126 (0.8%) persistent patients and 9/87 (11.4%) non-persistent patients. Multivariable Cox regression analysis confirmed that persistence to medication was significantly associated with a lower risk of secondary fracture (cause-specific hazard ratio [csHR] 0.05; 95%CI 0.01-0.45; p = 0.007).

CONCLUSION

The application of the surgeon-led AO Foundation algorithm enables the in-hospital initiation of anti-osteoporotic treatment, leading to better persistence to medication and decreased incidence of secondary osteoporotic fractures.

Predicting direct healthcare costs of general practitioner-guided care in patients with musculoskeletal complaints

ABSTRACT

Information on healthcare utilization and costs of general practitioner (GP)-guided care in patients with musculoskeletal complaints is important for keeping healthcare affordable and accessible. A registry-based study was performed to describe healthcare utilization and costs of GP-guided care in patients with musculoskeletal complaints and to predict having higher direct healthcare costs. Healthcare costs of GP-guided care included all healthcare resources used by patients due to a musculoskeletal condition in 2018. Data were extracted from the database with a 1-year follow-up and descriptively analyzed. A general linear model was developed to predict having higher direct healthcare costs. In total, 403,719 patients were included, of whom 92% only received a single consultation. The number of referrals varied across the different types of complaints. Total annual direct healthcare costs amounted to €39,180,531, of which a key cost driver was referrals. Primary care consultations accounted for the largest part of referral-related costs. For all musculoskeletal conditions combined, the mean annual direct healthcare cost per patient was €97 (SEM = €0.18). Older age, being a woman, low socioeconomic status, spine complaints, high number of musculoskeletal diagnoses, and a high comorbidity score were predictive of having higher direct healthcare costs and explained 0.7% of the variance. This study showed that mean annual direct healthcare costs of GP-guided care in patients with musculoskeletal conditions were relatively low and did not differ considerably across conditions. The predictive model explained a negligible part of the variance in costs. Thus, it is unclear which factors do predict high direct healthcare costs in this population.

Impact of antiseizure medication with a very long half-life on long term video-EEG monitoring in focal epilepsy

PURPOSE

To assess the impact of antiseizure medications (ASMs) with a very long half-life on long term video-EEG monitoring (LTM) in people with focal epilepsy (FE).

METHODS

In this retrospective cohort study, we searched our local database for people with FE who underwent ASM reduction during LTM at the University Hospital of 'La Fe', Valencia, from January 2013 to December 2019. Taking into account the half-life of the ASM, people with FE were divided into two groups: Group A contained individuals who were taking at least one ASM with a very long half-life at admission, and Group B consisted of those not taking very long half-life ASMs. Using multivariable analysis to control for important confounders, we compared the following outcomes between both groups: seizure rates per day, time to first seizure, and LTM duration.

RESULTS

Three hundred seventy individuals were included in the study (154 in Group A and 216 in Group B). The median recorded seizure rates (1.3 seizures/day, range 0-15.3 vs.1.3 seizures/day, range 0-9.3, p-value=0.68), median time to the first seizure (24 h, range 2-119 vs. 24 h, range 2-100, p-value=0.92), and median LTM duration (4 days, range 2-5 vs. 4 days, range 2-5, p-value=0.94) were similar in both groups. Multivariable analysis did not reveal any significant differences in the three outcomes between the two groups (all p-values>0.05).

CONCLUSION

ASMs with a very long half-life taken as co-medication do not significantly affect important LTM outcomes, including recorded seizure rates, time to the first seizure, or LTM duration. Therefore, in general, there is no need to discontinue ASMs with a very long half-life prior to LTM.

A comparison of full model specification and backward elimination of potential confounders when estimating marginal and conditional causal effects on binary outcomes from observational data

A common view in epidemiology is that automated confounder selection methods, such as backward elimination, should be avoided as they can lead to biased effect estimates and underestimation of their variance. Nevertheless, backward elimination remains regularly applied. We investigated if and under which conditions causal effect estimation in observational studies can improve by using backward elimination on a prespecified set of potential confounders. An expression was derived that quantifies how variable omission relates to bias and variance of effect estimators. Additionally, 3960 scenarios were defined and investigated by simulations comparing bias and mean squared error (MSE) of the conditional log odds ratio, log(cOR), and the marginal log risk ratio, log(mRR), between full models including all prespecified covariates and backward elimination of these covariates. Applying backward elimination resulted in a mean bias of 0.03 for log(cOR) and 0.02 for log(mRR), compared to 0.56 and 0.52 for log(cOR) and log(mRR), respectively, for a model without any covariate adjustment, and no bias for the full model. In less than 3% of the scenarios considered, the MSE of the log(cOR) or log(mRR) was slightly lower (max 3%) when backward elimination was used compared to the full model. When an initial set of potential confounders can be specified based on background knowledge, there is minimal added value of backward elimination. We advise not to use it and otherwise to provide ample arguments supporting its use.

A neutral comparison of algorithms to minimize L0 penalties for high-dimensional variable selection

Variable selection methods based on L0 penalties have excellent theoretical properties to select sparse models in a high-dimensional setting. There exist modifications of the Bayesian Information Criterion (BIC) which either control the familywise error rate (mBIC) or the false discovery rate (mBIC2) in terms of which regressors are selected to enter a model. However, the minimization of L0 penalties comprises a mixed-integer problem which is known to be NP-hard and therefore becomes computationally challenging with increasing numbers of regressor variables. This is one reason why alternatives like the LASSO have become so popular, which involve convex optimization problems that are easier to solve. The last few years have seen some real progress in developing new algorithms to minimize L0 penalties. The aim of this article is to compare the performance of these algorithms in terms of minimizing L0 -based selection criteria. Simulation studies covering a wide range of scenarios that are inspired by genetic association studies are used to compare the values of selection criteria obtained with different algorithms. In addition, some statistical characteristics of the selected models and the runtime of algorithms are compared. Finally, the performance of the algorithms is illustrated in a real data example concerned with expression quantitative trait loci (eQTL) mapping.

Replicability and reproducibility of predictive models for diagnosis of depression among young adults using Electronic Health Records

BACKGROUND

Recent advances in machine learning combined with the growing availability of digitized health records offer new opportunities for improving early diagnosis of depression. An emerging body of research shows that Electronic Health Records can be used to accurately predict cases of depression on the basis of individual's primary care records. The successes of these studies are undeniable, but there is a growing concern that their results may not be replicable, which could cast doubt on their clinical usefulness.

METHODS

To address this issue in the present paper, we set out to reproduce and replicate the work by Nichols et al. (2018), who trained predictive models of depression among young adults using Electronic Healthcare Records. Our contribution consists of three parts. First, we attempt to replicate the methodology used by the original authors, acquiring a more up-to-date set of primary health care records to the same specification and reproducing their data processing and analysis. Second, we test models presented in the original paper on our own data, thus providing out-of-sample prediction of the predictive models. Third, we extend past work by considering several novel machine-learning approaches in an attempt to improve the predictive accuracy achieved in the original work.

RESULTS

In summary, our results demonstrate that the work of Nichols et al. is largely reproducible and replicable. This was the case both for the replication of the original model and the out-of-sample replication applying NRCBM coefficients to our new EHRs data. Although alternative predictive models did not improve model performance over standard logistic regression, our results indicate that stepwise variable selection is not stable even in the case of large data sets.

CONCLUSION

We discuss the challenges associated with the research on mental health and Electronic Health Records, including the need to produce interpretable and robust models. We demonstrated some potential issues associated with the reliance on EHRs, including changes in the regulations and guidelines (such as the QOF guidelines in the UK) and reliance on visits to GP as a predictor of specific disorders.

Comparison of Olympic-Style Weightlifting Performances of Elite Athletes: Scaling Models Account for Body Mass

PURPOSE

We developed a scale for comparison of performances by weightlifters of different body mass and compare this scaling formula to current systems.

METHODS

Data from Olympics and World and Continental Championships from 2017 to 2021 were obtained; results from athletes with doping violations were excluded, resulting in performances from 1900 athletes from 150 countries for use in analysis. Functional relationships between performance and body mass were explored by testing various transformations of body mass in the form of fractional polynomials that include a wide range of nonlinear relationships. These transformations were evaluated in quantile regression models to determine the best fit, examine sex differences, and distinguish fits for different performance levels (90th, 75th, and 50th percentiles).

RESULTS

The resulting model used a transformation of body mass with powers -2 and 2 for males and females and was used to specify a scaling formula. The small percentage deviations between modeled and actual performances confirm the high accuracy of the model. In the subset of medalists, scaled performances were comparable across different body masses, whereas both Sinclair and Robi scalings, currently used in competitions, were more variable. The curves had similar shapes for the 90th and 75th percentile levels but were less steep for the 50th percentile.

CONCLUSIONS

The scaling formula we derived to compare weightlifting performances across a range of body mass can easily be implemented in the competition software to determine the overall best lifters. This is an improvement over current methods that do not accurately account for differences in body mass and result in bias or yield large variations even with small differences in body mass despite identical performances.

Neurodevelopmental Correlates of Brain Magnetic Resonance Imaging Abnormalities in Extremely Low-birth-weight Infants

OBJECTIVE

To evaluate the relationship between impaired brain growth and structural brain abnormalities at term-equivalent age (TEA) and neurodevelopment in extremely low-birth-weight (ELBW) infants over the first 2 years.

METHODS

ELBW infants born from 2009 through 2018 and undergoing brain magnetic resonance imaging (MRI) at TEA were enrolled in this retrospective cohort study. MRI scans were reviewed using a validated quali-quantitative score, including several white and gray matter items. Neurodevelopment was assessed at 6, 12, 18, and 24 months using the Griffiths scales. The independent associations between MRI subscores and the trajectories of general and specific neurodevelopmental functions were analyzed by generalized estimating equations.

RESULTS

One hundred-nine ELBW infants were included. White matter volume reduction and delayed myelination were associated with worse general development (b = -2.33, P = .040; b = -6.88, P = .049 respectively), social skills (b = -3.13, P = .019; b = -4.79, P = .049), and eye-hand coordination (b = -3.48, P = .009; b = -7.21, P = .045). Cystic white matter lesions were associated with poorer motor outcomes (b = -4.99, P = .027), while white matter signal abnormalities and corpus callosum thinning were associated with worse nonverbal cognitive performances (b = -6.42, P = .010; b = -6.72, P = .021, respectively). Deep gray matter volume reduction correlated with worse developmental trajectories.

CONCLUSIONS

Distinctive MRI abnormalities correlate with specific later developmental skills. This finding may suggest that TEA brain MRI may assist with neurodevelopmental prediction, counseling of families, and development of targeted supportive interventions to improve neurodevelopment in ELBW neonates.

Prediction models in prehospital and emergency medicine research: How to derive and internally validate a clinical prediction model

Clinical prediction models are created to help clinicians with medical decision making, aid in risk stratification, and improve diagnosis and/or prognosis. With growing availability of both prehospital and in-hospital observational registries and electronic health records, there is an opportunity to develop, validate, and incorporate prediction models into clinical practice. However, many prediction models have high risk of bias due to poor methodology. Given that there are no methodological standards aimed at developing prediction models specifically in the prehospital setting, the objective of this paper is to describe the appropriate methodology for the derivation and validation of clinical prediction models in this setting. What follows can also be applied to the emergency medicine (EM) setting. There are eight steps that should be followed when developing and internally validating a prediction model: (1) problem definition, (2) coding of predictors, (3) addressing missing data, (4) ensuring adequate sample size, (5) variable selection, (6) evaluating model performance, (7) internal validation, and (8) model presentation. Subsequent steps include external validation, assessment of impact, and cost-effectiveness. By following these steps, researchers can develop a prediction model with the methodological rigor and quality required for prehospital and EM research.

Confounding adjustment in observational studies on cardiothoracic interventions: a systematic review of methodological practice

OBJECTIVES

It is unknown which confounding adjustment methods are currently used in the field of cardiothoracic surgery and whether these are appropriately applied. The aim of this study was to systematically evaluate the quality of conduct and reporting of confounding adjustment methods in observational studies on cardiothoracic interventions.

METHODS

A systematic review was performed, which included all observational studies that compared different interventions and were published between 1 January and 1 July 2022, in 3 European and American cardiothoracic surgery journals. Detailed information on confounding adjustment methods was extracted and subsequently described.

RESULTS

Ninety-two articles were included in the analysis. Outcome regression (n = 49, 53%) and propensity score (PS) matching (n = 44, 48%) were most popular (sometimes used in combination), whereas 11 (12%) studies applied no method at all. The way of selecting confounders was not reported in 42 (46%) of the studies, solely based on previous literature or clinical knowledge in 14 (16%), and (partly) data-driven in 25 (27%). For the studies that applied PS matching, the matched cohorts comprised on average 46% of the entire study population (range 9-82%).

CONCLUSIONS

Current reporting of confounding adjustment methods is insufficient in a large part of observational studies on cardiothoracic interventions, which makes quality judgement difficult. Appropriate application of confounding adjustment methods is crucial for causal inference on optimal treatment strategies for clinical practice. Reporting on these methods is an important aspect of this, which can be improved.

An enhanced version of FREM (Fracture Risk Evaluation Model) using national administrative health data: analysis protocol for development and validation of a multivariable prediction model

BACKGROUND

Osteoporosis poses a growing healthcare challenge owing to its rising prevalence and a significant treatment gap, as patients are widely underdiagnosed and consequently undertreated, leaving them at high risk of osteoporotic fracture. Several tools aim to improve case-finding in osteoporosis. One such tool is the Fracture Risk Evaluation Model (FREM), which in contrast to other tools focuses on imminent fracture risk and holds potential for automation as it relies solely on data that is routinely collected via the Danish healthcare registers. The present article is an analysis protocol for a prediction model that is to be used as a modified version of FREM, with the intention of improving the identification of subjects at high imminent risk of fracture by including pharmacological exposures and using more advanced statistical methods compared to the original FREM. Its main purposes are to document and motivate various aspects and choices of data management and statistical analyses.

METHODS

The model will be developed by employing logistic regression with grouped LASSO regularization as the primary statistical approach and gradient-boosted classification trees as a secondary statistical modality. Hyperparameter choices as well as computational considerations on these two approaches are investigated by an unsupervised data review (i.e., blinded to the outcome), which also investigates and handles multicollinarity among the included exposures. Further, we present an unsupervised review of the data and testing of analysis code with respect to speed and robustness on a remote analysis environment. The data review and code tests are used to adjust the analysis plans in a blinded manner, so as not to increase the risk of overfitting in the proposed methods.

DISCUSSION

This protocol specifies the planned tool development to ensure transparency in the modeling approach, hence improving the validity of the enhanced tool to be developed. Through an unsupervised data review, it is further documented that the planned statistical approaches are feasible and compatible with the data employed.

Expert elicitation of risk factors for progression to dementia in individuals with mild cognitive impairment

INTRODUCTION

This study assesses experts' beliefs about important predictors of developing dementia in persons with mild cognitive impairment (MCI).

METHODS

Structured expert elicitation, a methodology to quantify expert knowledge, was used to elicit the most important risk factors for developing dementia. We recruited 11 experts (6 neurologists, 3 geriatricians, and 2 psychiatrists). Ten experts fully participated in introductory meetings, two rounds of surveys, and discussion meetings. The data from these ten experts were utilized for this study.

RESULTS

The expert elicitation identified age, CSF analysis, fluorodeoxyglucose-positron emission tomography (FDG-PET) findings, hippocampal atrophy, MoCA (or MMSE) score, parkinsonism, apathy, psychosis, informant report of cognitive symptoms, and global atrophy as the ten most important predictors of progressing to dementia in persons with MCI.

DISCUSSION

Several dementia predictors are not routinely collected in existing registries, observational studies, or usual care. This might partially explain the low uptake of existing published dementia risk scores in clinical practice.

Combining glucose and high-sensitivity cardiac troponin in the early diagnosis of acute myocardial infarction

Glucose is a universally available inexpensive biomarker, which is increased as part of the physiological stress response to acute myocardial infarction (AMI) and may therefore help in its early diagnosis. To test this hypothesis, glucose, high-sensitivity cardiac troponin (hs-cTn) T, and hs-cTnI were measured in consecutive patients presenting with acute chest discomfort to the emergency department (ED) and enrolled in a large international diagnostic study (NCT00470587). Two independent cardiologists centrally adjudicated the final diagnosis using all clinical data, including serial hs-cTnT measurements, cardiac imaging and clinical follow-up. The primary diagnostic endpoint was index non-ST-segment elevation MI (NSTEMI). Prognostic endpoints were all-cause death, and cardiovascular (CV) death or future AMI, all within 730-days. Among 5639 eligible patients, NSTEMI was the adjudicated final diagnosis in 1051 (18.6%) patients. Diagnostic accuracy quantified using the area under the receiver-operating characteristics curve (AUC) for the combination of glucose with hs-cTnT and glucose with hs-cTnI was very high, but not higher versus that of hs-cTn alone (glucose/hs-cTnT 0.930 [95% CI 0.922-0.937] versus hs-cTnT 0.929 [95% CI 0.922-0.937]; glucose/hs-cTnI 0.944 [95% CI 0.937-0.951] versus hs-cTnI 0.944 [95% CI 0.937-0.951]). In early-presenters, a dual-marker strategy (glucose < 7 mmol/L and hs-cTnT < 5/hs-cTnI < 4 ng/L) provided very high and comparable sensitivity to slightly lower hs-cTn concentrations (cTnT/I < 4/3 ng/L) alone, and possibly even higher efficacy. Glucose was an independent predictor of 730-days endpoints. Our results showed that a dual marker strategy of glucose and hs-cTn did not increase the diagnostic accuracy when used continuously. However, a cutoff approach combining glucose and hs-cTn may provide diagnostic utility for patients presenting ≤ 3 h after onset of symptoms, also providing important prognostic information.

Poor handling of continuous predictors in clinical prediction models using logistic regression: a systematic review

BACKGROUND AND OBJECTIVES

When developing a clinical prediction model, assuming a linear relationship between the continuous predictors and outcome is not recommended. Incorrect specification of the functional form of continuous predictors could reduce predictive accuracy. We examine how continuous predictors are handled in studies developing a clinical prediction model.

METHODS

We searched PubMed for clinical prediction model studies developing a logistic regression model for a binary outcome, published between July 01, 2020, and July 30, 2020.

RESULTS

In total, 118 studies were included in the review (18 studies (15%) assessed the linearity assumption or used methods to handle nonlinearity, and 100 studies (85%) did not). Transformation and splines were commonly used to handle nonlinearity, used in 7 (n = 7/18, 39%) and 6 (n = 6/18, 33%) studies, respectively. Categorization was most often used method to handle continuous predictors (n = 67/118, 56.8%) where most studies used dichotomization (n = 40/67, 60%). Only ten models included nonlinear terms in the final model (n = 10/18, 56%).

CONCLUSION

Though widely recommended not to categorize continuous predictors or assume a linear relationship between outcome and continuous predictors, most studies categorize continuous predictors, few studies assess the linearity assumption, and even fewer use methodology to account for nonlinearity. Methodological guidance is provided to guide researchers on how to handle continuous predictors when developing a clinical prediction model.

Pediatric prognostic models predicting inhospital child mortality in resource-limited settings: An external validation study

Background and Aims

Prognostic models provide evidence-based predictions and estimates of future outcomes, facilitating decision-making, patient care, and research. A few of these models have been externally validated, leading to uncertain reliability and generalizability. This study aims to externally validate four models to assess their transferability and usefulness in clinical practice. The models include the respiratory index of severity in children (RISC)-Malawi model and three other models by Lowlavaar et al.

Methods

The study used data from the Clinical Information Network (CIN) to validate the four models where the primary outcome was in-hospital mortality. 163,329 patients met eligibility criteria. Missing data were imputed, and the logistic function was used to compute predicted risk of in-hospital mortality. Models' discriminatory ability and calibration were determined using area under the curve (AUC), calibration slope, and intercept.

Results

The RISC-Malawi model had 50,669 pneumonia patients who met the eligibility criteria, of which the case-fatality ratio was 4406 (8.7%). Its AUC was 0.77 (95% CI: 0.77-0.78), whereas the calibration slope was 1.04 (95% CI: 1.00 -1.06), and calibration intercept was 0.81 (95% CI: 0.77-0.84). Regarding the external validation of Lowlavaar et al. models, 10,782 eligible patients  were included, with an in-hospital mortality rate of 5.3%. The primary model's AUC was 0.75 (95% CI: 0.72-0.77), the calibration slope was 0.78 (95% CI: 0.71-0.84), and the calibration intercept was 0.37 (95% CI: 0.28-0.46). All models markedly underestimated the risk of mortality.

Conclusion

All externally validated models exhibited either underestimation or overestimation of the risk as judged from calibration statistics. Hence, applying these models with confidence in settings other than their original development context may not be advisable. Our findings strongly suggest the need for recalibrating these model to enhance their generalizability.

Statistical analysis of high-dimensional biomedical data: a gentle introduction to analytical goals, common approaches and challenges

BACKGROUND

In high-dimensional data (HDD) settings, the number of variables associated with each observation is very large. Prominent examples of HDD in biomedical research include omics data with a large number of variables such as many measurements across the genome, proteome, or metabolome, as well as electronic health records data that have large numbers of variables recorded for each patient. The statistical analysis of such data requires knowledge and experience, sometimes of complex methods adapted to the respective research questions.

METHODS

Advances in statistical methodology and machine learning methods offer new opportunities for innovative analyses of HDD, but at the same time require a deeper understanding of some fundamental statistical concepts. Topic group TG9 "High-dimensional data" of the STRATOS (STRengthening Analytical Thinking for Observational Studies) initiative provides guidance for the analysis of observational studies, addressing particular statistical challenges and opportunities for the analysis of studies involving HDD. In this overview, we discuss key aspects of HDD analysis to provide a gentle introduction for non-statisticians and for classically trained statisticians with little experience specific to HDD.

RESULTS

The paper is organized with respect to subtopics that are most relevant for the analysis of HDD, in particular initial data analysis, exploratory data analysis, multiple testing, and prediction. For each subtopic, main analytical goals in HDD settings are outlined. For each of these goals, basic explanations for some commonly used analysis methods are provided. Situations are identified where traditional statistical methods cannot, or should not, be used in the HDD setting, or where adequate analytic tools are still lacking. Many key references are provided.

CONCLUSIONS

This review aims to provide a solid statistical foundation for researchers, including statisticians and non-statisticians, who are new to research with HDD or simply want to better evaluate and understand the results of HDD analyses.

Development and internal-external validation of statistical and machine learning models for breast cancer prognostication: cohort study

OBJECTIVE

To develop a clinically useful model that estimates the 10 year risk of breast cancer related mortality in women (self-reported female sex) with breast cancer of any stage, comparing results from regression and machine learning approaches.

DESIGN

Population based cohort study.

SETTING

QResearch primary care database in England, with individual level linkage to the national cancer registry, Hospital Episodes Statistics, and national mortality registers.

PARTICIPANTS

141 765 women aged 20 years and older with a diagnosis of invasive breast cancer between 1 January 2000 and 31 December 2020.

MAIN OUTCOME MEASURES

Four model building strategies comprising two regression (Cox proportional hazards and competing risks regression) and two machine learning (XGBoost and an artificial neural network) approaches. Internal-external cross validation was used for model evaluation. Random effects meta-analysis that pooled estimates of discrimination and calibration metrics, calibration plots, and decision curve analysis were used to assess model performance, transportability, and clinical utility.

RESULTS

During a median 4.16 years (interquartile range 1.76-8.26) of follow-up, 21 688 breast cancer related deaths and 11 454 deaths from other causes occurred. Restricting to 10 years maximum follow-up from breast cancer diagnosis, 20 367 breast cancer related deaths occurred during a total of 688 564.81 person years. The crude breast cancer mortality rate was 295.79 per 10 000 person years (95% confidence interval 291.75 to 299.88). Predictors varied for each regression model, but both Cox and competing risks models included age at diagnosis, body mass index, smoking status, route to diagnosis, hormone receptor status, cancer stage, and grade of breast cancer. The Cox model's random effects meta-analysis pooled estimate for Harrell's C index was the highest of any model at 0.858 (95% confidence interval 0.853 to 0.864, and 95% prediction interval 0.843 to 0.873). It appeared acceptably calibrated on calibration plots. The competing risks regression model had good discrimination: pooled Harrell's C index 0.849 (0.839 to 0.859, and 0.821 to 0.876, and evidence of systematic miscalibration on summary metrics was lacking. The machine learning models had acceptable discrimination overall (Harrell's C index: XGBoost 0.821 (0.813 to 0.828, and 0.805 to 0.837); neural network 0.847 (0.835 to 0.858, and 0.816 to 0.878)), but had more complex patterns of miscalibration and more variable regional and stage specific performance. Decision curve analysis suggested that the Cox and competing risks regression models tested may have higher clinical utility than the two machine learning approaches.

CONCLUSION

In women with breast cancer of any stage, using the predictors available in this dataset, regression based methods had better and more consistent performance compared with machine learning approaches and may be worthy of further evaluation for potential clinical use, such as for stratified follow-up.

Decisional needs assessment for patient-centred pain care in Canada: the DECIDE-PAIN study protocol

INTRODUCTION

The 2021 Action Plan for Pain from the Canadian Pain Task Force advocates for patient-centred pain care at all levels of healthcare across provinces. Shared decision-making is the crux of patient-centred care. Implementing the action plan will require innovative shared decision-making interventions, specifically following the disruption of chronic pain care during the COVID-19 pandemic. The first step in this endeavour is to assess current decisional needs (ie, decisions most important to them) of Canadians with chronic pain across their care pathways.

METHODS AND ANALYSIS

DesignGrounded in patient-oriented research approaches, we will perform an online population-based survey across the ten Canadian provinces. We will report methods and data following the CROSS reporting guidelines.SamplingThe Léger Marketing company will administer the online population-based survey to its representative panel of 500 000 Canadians to recruit 1646 adults (age ≥18 years old) with chronic pain according to the definition by the International Association for the Study of Pain (eg, pain ≥12 weeks). ContentBased on the Ottawa Decision Support Framework, the self-administered survey has been codesigned with patients and contain six core domains: (1) healthcare services, consultation and postpandemic needs, (2) difficult decisions experienced, (3) decisional conflict, (4) decisional regret, (5) decisional needs and (6) sociodemographic characteristics. We will use several strategies such as random sampling to improve survey quality. AnalysisWe will perform descriptive statistical analysis. We will identify factors associated with clinically significant decisional conflict and decision regret using multivariate analyses.

ETHICS AND DISSEMINATION

Ethics was approved by the Research Ethics Board at the Research Centre of the Centre Hospitalier Universitaire de Sherbrooke (project #2022-4645). We will codesign knowledge mobilisation products with research patient partners (eg, graphical summaries and videos). Results will be disseminated via peer-reviewed journals and national and international conferences to inform the development of innovative shared decision-making interventions for Canadians with chronic pain.

A systematic review of methodological considerations in time to diagnosis and treatment in colorectal cancer research

Research focusing on timely diagnosis and treatment of colorectal cancer is necessary to improve outcomes for people with cancer. Previous attempts to consolidate research on time to diagnosis and treatment have noted varied methodological approaches and quality, limiting the comparability of findings. This systematic review was conducted to comprehensively assess the scope of methodological issues in this field and provide recommendations for future research. Eligible articles had to assess the role of any interval up to treatment, on any outcome in colorectal cancer, in English, with no limits on publication time. Four databases were searched (Ovid Medline, EMBASE, EMCARE and PsycInfo). Papers were screened by two independent reviewers using a two-stage process of title and abstract followed by full text review. In total, 130 papers were included and had data extracted on specific methodological and statistical features. Several methodological problems were identified across the evidence base. Common issues included arbitrary categorisation of intervals (n = 107, 83%), no adjustment for potential confounders (n = 65, 50%), and lack of justification for included covariates where there was adjustment (n = 40 of 65 papers that performed an adjusted analysis, 62%). Many articles introduced epidemiological biases such as immortal time bias (n = 37 of 80 papers that used survival as an outcome, 46%) and confounding by indication (n = 73, 56%), as well as other biases arising from inclusion of factors outside of their temporal sequence. However, determination of the full extent of these problems was hampered by insufficient reporting. Recommendations include avoiding artificial categorisation of intervals, ensuring bias has not been introduced due to out-of-sequence use of key events and increased use of theoretical frameworks to detect and reduce bias. The development of reporting guidelines and domain-specific risk of bias tools may aid in ensuring future research can reliably contribute to recommendations regarding optimal timing and strengthen the evidence base.

Risk factors associated with sow mortality in breeding herds under one production system in the Midwestern United States

Sow mortality has significantly increased throughout the world over the past several years, and it is a growing concern to the global swine industry. Sow mortality increases economic losses, including higher replacement rates, affects employees' morale, and raises concerns about animal well-being and sustainability. This study aimed to assess herd-level risk factors associated with sow mortality in a large swine production system in the Midwestern United States. This retrospective observational study used available production, health, nutritional, and management information between July 2019 and December 2021. A Poisson mixed regression model was used to identify the risk factors and to build a multivariate model using the weekly mortality rate per 1000 sows as the outcome. Different models were used to identify the risk factors according to this study's main reasons for sow mortality (total death, sudden death, lameness, and prolapse). The main reported causes of sow mortality were sudden death (31.22 %), lameness (28.78 %), prolapse (28.02 %), and other causes (11.99 %). The median (25th-75th percentile) distribution of the crude sow mortality rate/1000 sows was 3.37 (2.19 - 4.16). Breeding herds classified as epidemic for porcine reproductive and respiratory syndrome virus (PRRSV) were associated with higher total death, sudden death, and lameness death. Open pen gestation was associated with a higher total death and lameness compared with stalls. Pulses of feed medication was associated with lower sow mortality rate for all outcomes. Farms not performing bump feeding were associated with higher sow mortality due to lameness and prolapses, while Senecavirus A (SVA)-positive herds were associated with a higher mortality rate for total deaths and deaths due to lameness. Disease interactions (herds Mycoplasma hyopneumoniae positive and epidemic for PRRSV; SVA positive herds and epidemic for PRRSV) were associated with higher mortality rates compared to farms with single disease status. This study identified and measured the major risk factors associated with total sow mortality rate, sudden deaths, lameness deaths, and prolapse deaths in breeding herds under field conditions.

External validation and updating of prognostic models for predicting recovery of disability in people with (sub)acute neck pain was successful: broad external validation in a new prospective cohort

QUESTION

Can existing post-treatment prognostic models for predicting neck pain recovery (primarily in terms of disability and secondarily in terms of pain intensity and perceived improvement) be externally validated and updated at the end of the treatment period and at 6 and 12 weeks of follow-up in a new Dutch cohort of people with neck pain treated with guideline-based usual care physiotherapy?

DESIGN

External validation and model updating in a new prospective cohort of three previously developed prognostic models.

PARTICIPANTS

People with (sub)acute neck pain and registered for primary care physiotherapy treatment.

OUTCOME MEASURES

Recovery of disability, pain intensity, and perceived recovery at 6 and 12 weeks and at the end of the treatment period.

RESULTS

Discriminative performance (c-statistic) of the disability model at 6 weeks was 0.73 (95% CI 0.69 to 0.77) and reasonably well calibrated after intercept recalibration. The disability model at 12 weeks and at the end of the treatment period showed discriminative c-statistic performance values of 0.69 (95% CI 0.64 to 0.73) and 0.68 (95% CI 0.63 to 0.72), respectively, and was well calibrated. Pain models and perceived recovery models did not reach acceptable performance. Cervical mobility added value to the disability models and pain catastrophising to the disability and pain models at 6 weeks.

DISCUSSION

Broad external validation of the disability model was successful in people with (sub)acute neck pain and clinicians may use this model in clinical practice with reasonable accuracy. Further research is required to assess the disability model's clinical impact and generalisability, and to identify additional valuable model predictors.

REGISTRATION

https://osf.io/a6r3k/.

Flexible modeling of longitudinal health-related quality of life data accounting for informative dropout in a cancer clinical trial

PURPOSE

A joint modeling approach is recommended for analysis of longitudinal health-related quality of life (HRQoL) data in the presence of potentially informative dropouts. However, the linear mixed model modeling the longitudinal HRQoL outcome in a joint model often assumes a linear trajectory over time, an oversimplification that can lead to incorrect results. Our aim was to demonstrate that a more flexible model gives more reliable and complete results without complicating their interpretation.

METHODS

Five dimensions of HRQoL in patients with esophageal cancer from the randomized clinical trial PRODIGE 5/ACCORD 17 were analyzed. Joint models assuming linear or spline-based HRQoL trajectories were applied and compared in terms of interpretation of results, graphical representation, and goodness of fit.

RESULTS

Spline-based models allowed arm-by-time interaction effects to be highlighted and led to a more precise and consistent representation of the HRQoL over time; this was supported by the martingale residuals and the Akaike information criterion.

CONCLUSION

Linear relationships between continuous outcomes (such as HRQoL scores) and time are usually the default choice. However, the functional form turns out to be important by affecting both the validity of the model and the statistical significance.

TRIAL REGISTRATION

This study is registered with ClinicalTrials.gov, number NCT00861094.

Development and internal validation of a model to predict long-term survival of ANCA associated vasculitis

Objectives

Risk stratification and prognosis prediction are critical for appropriate management of anti-neutrophil cytoplasmic antibody (ANCA) associated vasculitis (AAV). Herein, we aim to develop and internally validate a prediction model specifically for long-term survival of patients with AAV.

Methods

We thoroughly reviewed the medical charts of patients with AAV admitted to Peking Union Medical College Hospital from January 1999 to July 2019. The Least Absolute Shrinkage and Selection Operator method and the COX proportional hazard regression was used to develop the prediction model. The Harrell's concordance index (C-index), calibration curves and Brier scores were calculated to evaluate the model performance. The model was internally validated by bootstrap resampling methods.

Results

A total of 653 patients were included in the study, including 303 patients with microscopic polyangiitis, 245 patients with granulomatosis with polyangiitis and 105 patients with eosinophilic granulomatosis with polyangiitis, respectively. During a median follow-up of 33 months (interquartile range 15-60 months), 120 deaths occurred. Age at admission, chest and cardiovascular involvement, serum creatinine grade, hemoglobin levels at baseline and AAV sub-types were selected as predictive parameters in the final model. The optimism-corrected C-index and integrated Brier score of our prediction model were 0.728 and 0.109. The calibration plots showed fine agreement between observed and predicted probability of all-cause death. The decision curve analysis (DCA) showed that in a wide range of threshold probabilities, our prediction model had higher net benefits compared with the revised five factor score (rFFSand) and the birmingham vasculitis activity score (BVAS) system.

Conclusion

Our model performs well in predicting outcomes of AAV patients. Patients with moderate-to-high probability of death should be followed closely and personalized monitoring plan should be scheduled.