Bias and efficiency of multiple imputation compared with complete-case analysis for missing covariate values

Improving Regression Analysis with Imputation in a Longitudinal Study of Alzheimer's Disease

Background

Missing data is prevalent in the Alzheimer's Disease Neuroimaging Initiative (ADNI). It is common to deal with missingness by removing subjects with missing entries prior to statistical analysis; however, this can lead to significant efficiency loss and sometimes bias. It has yet to be demonstrated that the imputation approach to handling this issue can be valuable in some longitudinal regression settings.

Objective

The purpose of this study is to demonstrate the importance of imputation and how imputation is correctly done in ADNI by analyzing longitudinal Alzheimer's Disease Assessment Scale -Cognitive Subscale 13 (ADAS-Cog 13) scores and their association with baseline patient characteristics.

Methods

We studied 1,063 subjects in ADNI with mild cognitive impairment. Longitudinal ADAS-Cog 13 scores were modeled with a linear mixed-effects model with baseline clinical and demographic characteristics as predictors. The model estimates obtained without imputation were compared with those obtained after imputation with Multiple Imputation by Chained Equations (MICE). We justify application of MICE by investigating the missing data mechanism and model assumptions. We also assess robustness of the results to the choice of imputation method.

Results

The fixed-effects estimates of the linear mixed-effects model after imputation with MICE yield valid, tighter confidence intervals, thus improving the efficiency of the analysis when compared to the analysis done without imputation.

Conclusions

Our study demonstrates the importance of accounting for missing data in ADNI. When deciding to perform imputation, care should be taken in choosing the approach, as an invalid one can compromise the statistical analyses.

An inflexible adherence to food rules mediates the longitudinal association between shape/weight overvaluation and binge eating

Understanding factors that account for the persistence of binge eating is necessary for developing effective early intervention and treatment approaches. The cognitive-behavioral model of eating disorders proposes that judgements of self-worth contingent upon weight and shape encourage adoption of multiple demanding rules related to food, eating and dieting, in turn increasing vulnerability to binge eating. However, there is limited longitudinal evidence supporting this hypothesized sequence of events. This study tested whether the longitudinal association between shape/weight overvaluation and binge eating is mediated by an inflexible adherence to food rules. Participants were 1760 adult women recruited from an online eating disorder platform who were invited to complete study measures at baseline (T1), 3-month follow-up (T2), and 6-month follow-up (T3). Path analysis was conducted to test the hypothesized associations. Missing data were handled using multiple imputation techniques. A significant indirect effect emerged, in that higher overvaluation scores at T1 predicted increased inflexible adherence to food rules at T2, which in turn predicted increased binge eating symptoms at T3. Findings offer support for the predictions outlined by the cognitive-behavioral model of eating disorders and highlight clear targets for early intervention and treatment.

Finding predictors for successful opioid response in cancer patients: An analysis of data from four randomized controlled trials

CONTEXT

There is no consensus on which "strong" (or step 3 WHO analgesic ladder) opioid to prescribe to a particular patient with cancer-related pain. A better understanding of opioid and patient characteristics on treatment response will contribute to a more personalized opioid treatment.

OBJECTIVES

Assessment of potential predictors for successful opioid treatment response in patients with cancer pain.

METHODS

An international partnership between four cancer pain research groups resulted in a combined individual-level database from four relevant randomized controlled trials (RCTs; n = 881). Together, these RCTs investigated the short-term (1 week) and medium-term (4 or 5 weeks) treatment responses for morphine, buprenorphine, methadone, oxycodone, and fentanyl. Candidate predictors for treatment response were sex, age, pain type, pain duration, depression, anxiety, Karnofsky performance score, opioid type, and use of anti-neuropathic drug.

RESULTS

Opioid type and pain type were found statistically significant predictors of short-term treatment success. Sex, age, pain type, anxiety, and opioid type were statistically, significantly associated with medium-term treatment success. However, these models showed low discriminative power.

CONCLUSION

Fentanyl and methadone, and mixed pain were found to be statistically significant predictors of treatment success in patients with cancer-related pain. With the predictors currently assessed our data did not allow for the creation of a clinical prediction model with good discriminative power. Additional - unrevealed - predictors are necessary to develop a future prediction model.

Understanding the implications of a complete case analysis for regression models with a right-censored covariate

Despite its drawbacks, the complete case analysis is commonly used in regression models with incomplete covariates. Understanding when the complete case analysis will lead to consistent parameter estimation is vital before use. Our aim here is to demonstrate when a complete case analysis is consistent for randomly right-censored covariates and to discuss the implications of its use even when consistent. Across the censored covariate literature, different assumptions are made to ensure a complete case analysis produces a consistent estimator, which leads to confusion in practice. We make several contributions to dispel this confusion. First, we summarize the language surrounding the assumptions that lead to a consistent complete case estimator. Then, we show a unidirectional hierarchical relationship between these assumptions, which leads us to one sufficient assumption to consider before using a complete case analysis. Lastly, we conduct a simulation study to illustrate the performance of a complete case analysis with a right-censored covariate under different censoring mechanism assumptions, and we demonstrate its use with a Huntington disease data example.

Predicting child development and school readiness, at age 5, for Aboriginal and non-Aboriginal children in Australia's Northern Territory

BACKGROUND

Positive early development is critical in shaping children's lifelong health and wellbeing. Identifying children at risk of poor development is important in targeting early interventions to children and families most in need of support. We aimed to develop a predictive model that could inform early support for vulnerable children.

METHODS

We analysed linked administrative records for a birth cohort of 2,380 Northern Territory children (including 1,222 Aboriginal children) who were in their first year of school in 2015 and had a completed record from the Australian Early Development Census (AEDC). The AEDC measures early child development (school readiness) across five domains of development. We fitted prediction models, for AEDC weighted summary scores, using a Partial Least Square Structural Equation Model (PLS-SEM) considering four groups of factors-pre-pregnancy, pregnancy, known at birth, and child-related factors. We first assessed the models' internal validity and then the out-of-sample predictive power (external validity) using the PLSpredict procedure.

RESULT

We identified separate predictive models, with a good fit, for Aboriginal and non-Aboriginal children. For Aboriginal children, a significant pre-pregnancy predictor of better outcomes was higher socioeconomic status (direct, β = 0.22 and indirect, β = 0.16). Pregnancy factors (gestational diabetes and maternal smoking (indirect, β = -0.09) and child-related factors (English as a second language and not attending preschool (direct, β = -0.28) predicted poorer outcomes. Further, pregnancy and child-related factors partially mediated the effects of pre-pregnancy factors; and child-related factors fully mediated the effects of pregnancy factors on AEDC weighted scores. For non-Aboriginal children, pre-pregnancy factors (increasing maternal age, socioeconomic status, parity, and occupation of the primary carer) directly predicted better outcomes (β = 0.29). A technical observation was that variance in AEDC weighted scores was not equally captured across all five AEDC domains; for Aboriginal children results were based on only three domains (emotional maturity; social competence, and language and cognitive skills (school-based)) and for non-Aboriginal children, on a single domain (language and cognitive skills (school-based)).

CONCLUSION

The models give insight into the interplay of multiple factors at different stages of a child's development and inform service and policy responses. Recruiting children and their families for early support programs should consider both the direct effects of the predictors and their interactions. The content and application of the AEDC measurement need to be strengthened to ensure all domains of a child's development are captured equally.

Trends in the long-term impact of childhood visual impairment on health and social outcomes in the UK: a cross-cohort study across three decades of disability-related legislation and policy implementation

BACKGROUND

Childhood vision impairment (VI) can adversely impact health and social outcomes and limit life chances. We investigated whether its adverse impacts into adult life changed during a period in which legislation, policy and services to address inequalities relating to disability were implemented.

METHODS

Cross-cohort study comprising 14 247 participants from the 1946, 1958 and 1970 British birth cohorts (BC). Participants dichotomized as VI at age 15/16 (distance visual acuity was 6/12 or worse in the better-seeing eye) or normally sighted. Associations of childhood VI with health, well-being, socioeconomic and social participation outcomes in mid-adult life were investigated using regression models adjusted for participants' early life socioeconomic markers and sex. Change in adjusted odds ratios of >10% in the same direction in successive cohorts, or a > 20% difference between 1970BC and one older cohort were considered meaningful.

RESULTS

Trends over time in impacts of childhood onset VI into mid-adult life were complex. This included worsening of odds of poorer physical health (odds ratio 1.47; 95% confidence interval 1.02-2.14), living in unsatisfactory (1.54; 1.03-2.29) or overcrowded (2.34; 1.26-4.06) households, being unemployed (2.19; 1.19-3.97) and not gaining additional educational qualifications during mid-adult life (1.61; 1.08-2.47). By contrast the odds of not participating in some social activities (e.g. seeing friends) improved over time. Associations with other outcomes were unchanged.

CONCLUSIONS

Many adverse impacts of childhood VI do not appear ameliorated over time by legislation, policies and provision that would have been expected to reduce inequalities. Moreover, some were increased. Childhood VI continues to cast a life-long shadow.

Trial arm outcome variance difference after dropout as an indicator of missing-not-at-random bias in randomized controlled trials

Randomized controlled trials (RCTs) are vulnerable to bias from missing data. When outcomes are missing not at random (MNAR), estimates from complete case analysis (CCA) and multiple imputation (MI) may be biased. There is no statistical test for distinguishing between outcomes missing at random (MAR) and MNAR. Current strategies rely on comparing dropout proportions and covariate distributions, and using auxiliary information to assess the likelihood of dropout being associated with the outcome. We propose using the observed variance difference across trial arms as a tool for assessing the risk of dropout being MNAR in RCTs with continuous outcomes. In an RCT, at randomization, the distributions of all covariates should be equal in the populations randomized to the intervention and control arms. Under the assumption of homogeneous treatment effects and homoskedastic outcome errors, the variance of the outcome will also be equal in the two populations over the course of follow-up. We show that under MAR dropout, the observed outcome variances, conditional on the variables included in the model, are equal across trial arms, whereas MNAR dropout may result in unequal variances. Consequently, unequal observed conditional trial arm variances are an indicator of MNAR dropout and possible bias of the estimated treatment effect. Heterogeneous treatment effects or heteroskedastic outcome errors are another potential cause of observing different outcome variances. We show that for longitudinal data, we can isolate the effect of MNAR outcome-dependent dropout by considering the variance difference at baseline in the same set of patients who are observed at final follow-up. We illustrate our method in simulation for CCA and MI, and in applications using individual-level data and summary data.

Consumption of ultra-processed foods and risk of multimorbidity of cancer and cardiometabolic diseases: a multinational cohort study

Background

It is currently unknown whether ultra-processed foods (UPFs) consumption is associated with a higher incidence of multimorbidity. We examined the relationship of total and subgroup consumption of UPFs with the risk of multimorbidity defined as the co-occurrence of at least two chronic diseases in an individual among first cancer at any site, cardiovascular disease, and type 2 diabetes.

Methods

This was a prospective cohort study including 266,666 participants (60% women) free of cancer, cardiovascular disease, and type 2 diabetes at recruitment from seven European countries in the European Prospective Investigation into Cancer and Nutrition (EPIC) study. Foods and drinks consumed over the previous 12 months were assessed at baseline by food-frequency questionnaires and classified according to their degree of processing using Nova classification. We used multistate modelling based on Cox regression to estimate cause-specific hazard ratios (HR) and their 95% confidence intervals (CI) for associations of total and subgroups of UPFs with the risk of multimorbidity of cancer and cardiometabolic diseases.

Findings

After a median of 11.2 years of follow-up, 4461 participants (39% women) developed multimorbidity of cancer and cardiometabolic diseases. Higher UPF consumption (per 1 standard deviation increment, ∼260 g/day without alcoholic drinks) was associated with an increased risk of multimorbidity of cancer and cardiometabolic diseases (HR: 1.09, 95% CI: 1.05, 1.12). Among UPF subgroups, associations were most notable for animal-based products (HR: 1.09, 95% CI: 1.05, 1.12), and artificially and sugar-sweetened beverages (HR: 1.09, 95% CI: 1.06, 1.12). Other subgroups such as ultra-processed breads and cereals (HR: 0.97, 95% CI: 0.94, 1.00) or plant-based alternatives (HR: 0.97, 95% CI: 0.91, 1.02) were not associated with risk.

Interpretation

Our findings suggest that higher consumption of UPFs increases the risk of cancer and cardiometabolic multimorbidity.

Funding

Austrian Academy of Sciences, Fondation de France, Cancer Research UK, World Cancer Research Fund International, and the Institut National du Cancer.

Does eating with others promote happiness among older adults living alone? A 3-year longitudinal study of the Japan gerontological evaluation study

OBJECTIVE

Living a happy life is an essential issue for old adults. However, how eating with others contributes to happiness and whether this association is different by living arrangements or not is unknown. The current study examined the relationship between the frequency of eating with others and happiness among older adults according to their living arrangements using 3-year longitudinal data.

METHODS

The analyzed sample comprised 18,727 people (10,920 males and 7807 females) with low happiness (0-7 points on score of 0-10 points) from Japan Gerontological Evaluation Study (JAGES) in 2016. Our exposure was the frequency of eating with others: rarely, a few times a year, a few times a month, and a few times a week or more. We performed Modified Poisson Regression to examine the association between the frequency of eating with others and high happiness (8-10 points) in 2019 stratified by living arrangement (living alone/with others).

RESULTS

A total of 4352 (23.2%) people showed high happiness in 2019. After adjusting for age, sex, marital status, education, household income, social participation, illnesses under treatment, and depressive symptoms in 2016, the cumulative incidence ratio (CIR) for high happiness in 2019 among people living alone was more significant, that is, 1.28 (95% confidence intervals: 0.88-1.87), 1.50 (1.05-2.14), and 1.82 (1.26-2.63), than 1.28 (1.11-1.48), 1.30 (1.12-1.50), and 1.33 (1.16-1.52) among people living with others for those who ate with others a few times a year, a few times a month, and a few times a week or more compared to those who rarely ate with others, respectively. The interaction between the frequency of eating with others and living arrangements was statistically significant. The trend test showed that higher frequency of eating with others was significantly associated with high happiness.

CONCLUSIONS

Eating with others was associated with improved happiness among older adults, with such an association being stronger among people living alone.

Racial and ethnic disparities in the usage and outcomes of ischemic stroke treatment in the United States

OBJECTIVES

This study explores racial and ethnic differences in 1) receiving tissue plasminogen activator (tPA) and endovascular thrombectomy (EVT) as treatment for ischemic stroke and 2) outcomes and quality of care after use of tPA or EVT in the US.

MATERIALS AND METHODS

An observational analysis of 89,035 ischemic stroke patients from the 2019 National Inpatient Sample was conducted. We performed weighted logistic regressions between race and ethnicity and 1) tPA and EVT utilization and 2) in-hospital mortality. We also performed a weighted Poisson regression between race and ethnicity and length of stay (LOS) after tPA or EVT.

RESULTS

Non-Hispanic (NH) Black patients had significantly lower odds of receiving tPA (Adjusted odds ratio [AOR] = 0.85, 95 % Confidence Internal [C.I.]: 0.80-0.91) and EVT (AOR = 0.75, 95 % CI: 0.70-0.82) than NH White patients. Minority populations (including but not limited to NH Black, Hispanic, Pacific Islander, Native American, and Asian) had significantly longer hospital LOS after treatment with tPA or EVT. We did not find a significant difference between race/ethnicity and in-hospital mortality post-tPA or EVT.

CONCLUSIONS

While we failed to find a difference in in-hospital mortality, racial and ethnic disparities are still evident in the decreased usage of tPA and EVT and longer LOSs for racial and ethnic minority patients. This study calls for interventions to expand the utilization of tPA and EVT and advance quality of care post-tPA or EVT in order to improve stroke care for minority patients.

DHA, nutrient intake, and maternal characteristics as predictors of pregnancy outcomes in a randomised clinical trial of DHA supplementation

PURPOSE

To investigate the relationships among docosahexaenoic acid (DHA) intake, nutrient intake, and maternal characteristics on pregnancy outcomes in a phase III randomised clinical trial designed to determine the effect of a DHA dose of 1000 mg/day compared to 200 mg/day on early preterm birth (<34 weeks gestation).

METHODS

A secondary aim of the phase III randomised trial was to explore the relationships among pregnancy outcomes (maternal red blood cell phospholipid (RBC-PL) DHA at delivery, preterm birth, gestational age at delivery, labor type, birth anthropometric measures, low birth weight, gestational diabetes, pre-eclampsia, and admission to a neonatal intensive care unit) in participants (n = 1100). We used Bayesian multiple imputation and linear and logistic regression models to conduct an analysis of five general classes of predictor variables collected during the trial: a) DHA intake, b) nutrient intake from food and supplements, c) environmental exposure to tobacco and alcohol, d) maternal demographics, and e) maternal medical history.

RESULTS

DHA supplementation lowered the risk of preterm birth and NICU admission, and increased gestation and birth weight as observed in the primary analysis. Higher maternal RBC-PL-DHA at delivery was associated with DHA supplementation and formal education of a bachelor's degree or higher. DHA supplementation and maternal age were associated with a higher risk of gestational diabetes. Total vitamin A intake was associated with longer gestation, while fructose and intake of the long chain omega-6 fatty acid, arachidonic acid, were associated with shorter gestation. Risk of preterm birth was associated with a history of low birth weight, preterm birth, pre-eclampsia, and NICU admission.

CONCLUSION

Bayesian models provide a comprehensive approach to relationships among DHA intake, nutrient intake, maternal characteristics, and pregnancy outcomes. We observed previously unreported relationships between gestation duration and fructose, vitamin A, and arachidonic acid that could be the basis for future research.

TRIAL REGISTRATION NUMBER AND DATE

ClinicalTrials.gov (NCT02626299); December 10, 2015.

Statistics for the hand surgeon. Part 1: principles and application to hand surgery research

Hand surgeons have the potential to improve patient care, both with their own research and by using evidenced-based practice. In this first part of a two-part article, we describe key steps for the analysis of clinical data using quantitative methodology. We aim to describe the principles of medical statistics and their relevance and use in hand surgery, with contemporaneous examples. Hand surgeons seek expertise and guidance in the clinical domain to improve their practice and patient care. Part of this process involves the critical analysis and appraisal of the research of others.

Association between urate-lowering therapy and cardiovascular events in patients with asymptomatic hyperuricemia

INTRODUCTION/OBJECTIVES

To investigate the role of urate-lowering therapy (ULT) in the prevention of cardiovascular disease (CVD) in patients with asymptomatic hyperuricemia using the Japanese healthcare record database.

METHODS

This retrospective cohort study used data from the JMDC Claims Database, which includes records of medical check-ups and Japanese health insurance claims. Subjects aged at least 18 years with a serum uric acid (sUA) level ≥ 7.0 mg/dL and at least one medical check-up from January 2007 to August 2021 were included in this study. The exposure was any ULT prescription, and the primary outcome included composite CVD outcomes, including coronary artery disease, stroke, and atrial fibrillation. Analysis was performed with a new-user design and overlap weighting to balance the baseline characteristics of the subjects. Cox proportional hazards models were used to investigate the association between ULT and the development of CVD.

RESULTS

In total, 152,166 patients were included in the main analysis before overlap weighting in this retrospective cohort study. The number of subjects in the ULT group was 5,270, and there were 146,896 subjects in the control group. Composite CVD outcomes were observed in a total of 7,703 patients. The risk of developing composite CVD outcomes was not different between the ULT group and the control group (HR: 1.01, 95% CI: 0.89 to 1.13).

CONCLUSIONS

ULT for patients with asymptomatic hyperuricemia did not prevent the development of CVD based on the Japanese claims database. Key points • Among subjects with asymptomatic hyperuricemia, ULT was not associated with a lower risk of CVD • There was no appropriate cutoff for initiating ULT in patients with asymptomatic hyperuricemia • There was no appropriate cutoff as the therapeutic goal of ULT in patients with asymptomatic hyperuricemia.

Cost-Effectiveness of SGLT2 Inhibitors in a Real-World Population: A MICADO Model-Based Analysis Using Routine Data from a GP Registry

OBJECTIVE

Sodium-glucose cotransporter 2 inhibitors (SGLT2i) have been shown to reduce the risk of cardiovascular complications, which largely drive diabetes' health and economic burdens. Trial results indicated that SGLT2i are cost effective. However, these findings may not be generalizable to the real-world target population. This study aims to evaluate the cost effectiveness of SGLT2i in a routine care type 2 diabetes population that meets Dutch reimbursement criteria using the MICADO model.

METHODS

Individuals from the Hoorn Diabetes Care System cohort (N = 15,392) were filtered to satisfy trial inclusion criteria (including EMPA-REG, CANVAS, and DECLARE-TIMI58) or satisfy the current Dutch reimbursement criteria for SGLT2i. We validated a health economic model (MICADO) by comparing simulated and observed outcomes regarding the relative risks of events in the intervention and comparator arm from three trials, and used the validated model to evaluate the long-term health outcomes using the filtered cohorts' baseline characteristics and treatment effects from trials and a review of observational studies. The incremental cost-effectiveness ratio (ICER) of SGLT2i, compared with care-as-usual, was assessed from a third-party payer perspective, measured in euros (2021 price level), using a discount rate of 4% for costs and 1.5% for effects.

RESULTS

From Dutch individuals with diabetes in routine care, 15.8% qualify for the current Dutch reimbursement criteria for SGLT2i. Their characteristics were significantly different (lower HbA1c, higher age, and generally more preexisting complications) than trial populations. After validating the MICADO model, we found that lifetime ICERs of SGLT2i, when compared with usual care, were favorable (< €20,000/QALY) for all filtered cohorts, resulting in an ICER of €5440/QALY using trial-based treatment effect estimates in reimbursed population. Several pragmatic scenarios were tested, the ICERs remained favorable.

CONCLUSIONS

Although the Dutch reimbursement indications led to a target group that deviates from trial populations, SGLT2i are likely to be cost effective when compared with usual care.

Understanding the Broader Impact of Stuttering: Suicidal Ideation

PURPOSE

Suicide and suicidal ideation are topics that have a long but limited history in stuttering research. Clinicians and clinical researchers have discussed personal and therapeutic experiences with clients who have attempted suicide, died by suicide, or struggled with suicidal thoughts. This study sought to (a) explore the occurrence of suicidal ideation in a sample of adults who stutter; (b) evaluate the relationship between adverse impact related to stuttering and suicidal ideation; and (c) document respondents' thoughts related to suicide, stuttering, and their intersection.

METHOD

One hundred forty adults who stutter completed the Suicide Behavior Questionnaire-Revised (SBQ-R). Of these, 70 participants completed the Perseverative Thinking Questionnaire (PTQ), and 67 completed the Overall Assessment of the Speaker's Experience of Stuttering (OASES). Participants who indicated at least some tendency for suicidal thoughts on the SBQ-R (n = 95) were then asked a set of follow-up questions to explore their experiences of suicidal ideation related to stuttering.

RESULTS

Quantitative data indicated that the majority of adults who reported experiencing suicidal ideation associated these experiences with stuttering (61.2%, n = 59). Individuals with higher Total Scores on the PTQ and OASES were predicted to experience significantly higher rates of suicidal ideation and, in particular, a higher likelihood of having more frequent suicidal ideation in the past year. Qualitative analyses revealed that suicidal ideation intersects meaningfully with the experience of stuttering.

CONCLUSIONS

Data from this study highlight the importance of considering broader life consequences of stuttering that some adults may experience, including suicidal ideation. By being cognizant that clients may develop such thoughts, speech-language pathologists can play a valuable role in identifying and providing necessary support for at-risk individuals.

SUPPLEMENTAL MATERIAL

https://doi.org/10.23641/asha.23699688.

Deriving and validating an asthma diagnosis prediction model for children and young people in primary care

Introduction: Accurately diagnosing asthma can be challenging. We aimed to derive and validate a prediction model to support primary care clinicians assess the probability of an asthma diagnosis in children and young people. Methods: The derivation dataset was created from the Avon Longitudinal Study of Parents and Children (ALSPAC) linked to electronic health records. Participants with at least three inhaled corticosteroid prescriptions in 12-months and a coded asthma diagnosis were designated as having asthma. Demographics, symptoms, past medical/family history, exposures, investigations, and prescriptions were considered as candidate predictors. Potential candidate predictors were included if data were available in ≥60% of participants. Multiple imputation was used to handle remaining missing data. The prediction model was derived using logistic regression. Internal validation was completed using bootstrap re-sampling. External validation was conducted using health records from the Optimum Patient Care Research Database (OPCRD). Results: Predictors included in the final model were wheeze, cough, breathlessness, hay-fever, eczema, food allergy, social class, maternal asthma, childhood exposure to cigarette smoke, prescription of a short acting beta agonist and the past recording of lung function/reversibility testing. In the derivation dataset, which comprised 11,972 participants aged <25 years (49% female, 8% asthma), model performance as indicated by the C-statistic and calibration slope was 0.86, 95% confidence interval (CI) 0.85-0.87 and 1.00, 95% CI 0.95-1.05 respectively. In the external validation dataset, which included 2,670 participants aged <25 years (50% female, 10% asthma), the C-statistic was 0.85, 95% CI 0.83-0.88, and calibration slope 1.22, 95% CI 1.09-1.35. Conclusions: We derived and validated a prediction model for clinicians to calculate the probability of asthma diagnosis for a child or young person up to 25 years of age presenting to primary care. Following further evaluation of clinical effectiveness, the prediction model could be implemented as a decision support software.

Deriving and validating an asthma diagnosis prediction model for children and young people in primary care

Introduction: Accurately diagnosing asthma can be challenging. We aimed to derive and validate a prediction model to support primary care clinicians assess the probability of an asthma diagnosis in children and young people. Methods: The derivation dataset was created from the Avon Longitudinal Study of Parents and Children (ALSPAC) linked to electronic health records. Participants with at least three inhaled corticosteroid prescriptions in 12-months and a coded asthma diagnosis were designated as having asthma. Demographics, symptoms, past medical/family history, exposures, investigations, and prescriptions were considered as candidate predictors. Potential candidate predictors were included if data were available in ≥60% of participants. Multiple imputation was used to handle remaining missing data. The prediction model was derived using logistic regression. Internal validation was completed using bootstrap re-sampling. External validation was conducted using health records from the Optimum Patient Care Research Database (OPCRD). Results: Predictors included in the final model were wheeze, cough, breathlessness, hay-fever, eczema, food allergy, social class, maternal asthma, childhood exposure to cigarette smoke, prescription of a short acting beta agonist and the past recording of lung function/reversibility testing. In the derivation dataset, which comprised 11,972 participants aged <25 years (49% female, 8% asthma), model performance as indicated by the C-statistic and calibration slope was 0.86, 95% confidence interval (CI) 0.85-0.87 and 1.00, 95% CI 0.95-1.05 respectively. In the external validation dataset, which included 2,670 participants aged <25 years (50% female, 10% asthma), the C-statistic was 0.85, 95% CI 0.83-0.88, and calibration slope 1.22, 95% CI 1.09-1.35. Conclusions: We derived and validated a prediction model for clinicians to calculate the probability of asthma diagnosis for a child or young person up to 25 years of age presenting to primary care. Following further evaluation of clinical effectiveness, the prediction model could be implemented as a decision support software.

The risk of acute kidney injury in colorectal cancer survivors: an english population-based matched cohort study

BACKGROUND

Colorectal cancer survival has improved in recent decades but there are concerns that survivors may develop kidney problems due to adverse effects of cancer treatment or complications of the cancer itself. We quantified the risk of acute kidney injury (AKI) in colorectal cancer survivors compared to people with no prior cancer.

METHODS

Retrospective matched cohort study using electronic health record primary care data from the Clinical Practice Research Datalink GOLD linked to hospital data in England (HES-APC). Individuals with colorectal cancer between 1997-2018 were individually matched on age, sex, and GP practice to people with no prior cancer. We used Cox models to estimate hazard ratios for an incident hospital diagnosis of AKI in colorectal cancer survivors compared to individuals without cancer, overall and stratified by time since diagnosis adjusted for other individual-level factors (adj-HR).

RESULTS

Twenty thousand three hundred forty colorectal cancer survivors were matched to 100,058 cancer-free individuals. Colorectal cancer survivors were at increased risk of developing AKI compared to people without cancer (adj-HR = 2.16; 95%CI 2.05-2.27). The HR was highest in the year after diagnosis (adj-HR 7.47, 6.66-8.37), and attenuated over time, but there was still increased AKI risk > 5 years after diagnosis (adj-HR = 1.26, 1.17-1.37). The association between colorectal cancer and AKI was greater for younger people, men, and those with pre-existing chronic kidney disease.

CONCLUSIONS

Colorectal cancer survivors were at increased risk of AKI for several years after cancer diagnosis, suggesting a need to prioritise monitoring, prevention, and management of kidney problems in this group of cancer survivors.

Alcohol Use and Sustained Virologic Response to Hepatitis C Virus Direct-Acting Antiviral Therapy

Importance

Some payers and clinicians require alcohol abstinence to receive direct-acting antiviral (DAA) therapy for chronic hepatitis C virus (HCV) infection.

Objective

To evaluate whether alcohol use at DAA treatment initiation is associated with decreased likelihood of sustained virologic response (SVR).

Design, Setting, and Participants

This retrospective cohort study used electronic health records from the US Department of Veterans Affairs (VA), the largest integrated national health care system that provides unrestricted access to HCV treatment. Participants included all patients born between 1945 and 1965 who were dispensed DAA therapy between January 1, 2014, and June 30, 2018. Data analysis was completed in November 2020 with updated sensitivity analyses performed in 2023.

Exposure

Alcohol use categories were generated using responses to the Alcohol Use Disorders Identification Test-Consumption (AUDIT-C) questionnaire and International Classification of Diseases, Ninth Revision and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision diagnoses for alcohol use disorder (AUD): abstinent without history of AUD, abstinent with history of AUD, lower-risk consumption, moderate-risk consumption, and high-risk consumption or AUD.

Main Outcomes and Measures

The primary outcome was SVR, which was defined as undetectable HCV RNA for 12 weeks or longer after completion of DAA therapy. Multivariable logistic regression was used to estimate odds ratios (ORs) and 95% CIs of SVR associated with alcohol category.

Results

Among 69 229 patients who initiated DAA therapy (mean [SD] age, 62.6 [4.5] years; 67 150 men [97.0%]; 34 655 non-Hispanic White individuals [50.1%]; 28 094 non-Hispanic Black individuals [40.6%]; 58 477 individuals [84.5%] with HCV genotype 1), 65 355 (94.4%) achieved SVR. A total of 32 290 individuals (46.6%) were abstinent without AUD, 9192 (13.3%) were abstinent with AUD, 13 415 (19.4%) had lower-risk consumption, 3117 (4.5%) had moderate-risk consumption, and 11 215 (16.2%) had high-risk consumption or AUD. After adjustment for potential confounding variables, there was no difference in SVR across alcohol use categories, even for patients with high-risk consumption or AUD (OR, 0.95; 95% CI, 0.85-1.07). There was no evidence of interaction by stage of hepatic fibrosis measured by fibrosis-4 score (P for interaction = .30).

Conclusions and Relevance

In this cohort study, alcohol use and AUD were not associated with lower odds of SVR. Restricting access to DAA therapy according to alcohol use creates an unnecessary barrier to patients and challenges HCV elimination goals.

Exogenous female sex steroid hormones and new-onset asthma in women: a matched case-control study

BACKGROUND

Evidence on the role of exogenous female sex steroid hormones in asthma development in women remains conflicting. We sought to quantify the potential causal role of hormonal contraceptives and menopausal hormone therapy (MHT) in the development of asthma in women.

METHODS

We conducted a matched case-control study based on the West Sweden Asthma Study, nested in a representative cohort of 15,003 women aged 16-75 years, with 8-year follow-up (2008-2016). Data were analyzed using Frequentist and Bayesian conditional logistic regression models.

RESULTS

We included 114 cases and 717 controls. In Frequentist analysis, the odds ratio (OR) for new-onset asthma with ever use of hormonal contraceptives was 2.13 (95% confidence interval [CI] 1.03-4.38). Subgroup analyses showed that the OR increased consistently with older baseline age. The OR for new-onset asthma with ever MHT use among menopausal women was 1.17 (95% CI 0.49-2.82). In Bayesian analysis, the ORs for ever use of hormonal contraceptives and MHT were, respectively, 1.11 (95% posterior interval [PI] 0.79-1.55) and 1.18 (95% PI 0.92-1.52). The respective probability of each OR being larger than 1 was 72.3% and 90.6%.

CONCLUSIONS

Although use of hormonal contraceptives was associated with an increased risk of asthma, this may be explained by selection of women by baseline asthma status, given the upward trend in the effect estimate with older age. This indicates that use of hormonal contraceptives may in fact decrease asthma risk in women. Use of MHT may increase asthma risk in menopausal women.

Investigating the role of perceived ingroup and outgroup colourism on body image and wellbeing among Black, Asian, and other racialised/ethnic minority groups living in the UK

Colourism is a form of prejudice and discrimination based on skin shade, disadvantaging people of colour with darker skin. This study investigates the relationship between perceived colourism, body image, and psychological wellbeing, considering perceived colourism from the ingroup (people of the same racialised group) and the outgroup (white people). A total of 516 Black, Asian, and other racialised/ethnic minority adults living in the UK (56.8 % women) completed an online survey. Using structural equation modelling, we tested a theoretically informed model: ingroup and outgroup colourism were predictors, body image and psychological distress were outcomes, and skin shade satisfaction and surveillance were hypothesised mediators. The model provided a good fit to the data. Ingroup colourism was related to lower skin shade satisfaction and higher skin shade surveillance, which in turn related to worse body image and greater psychological distress. Outgroup colourism was related to higher skin shade surveillance, which in turn was associated with worse body image. Outgroup colourism was directly associated with greater psychological distress. Results showed perceived colourism was associated with worse body image and psychological distress among people of colour in the UK. Therefore, colourism should be considered in the development of societal-, community-, and individual-level body image interventions.

Cannabis Use Disorder and Perioperative Complications

Importance

Cannabis use is growing in the US and is increasingly perceived as harmless. However, the perioperative impact of cannabis use remains uncertain.

Objective

To assess whether cannabis use disorder is associated with increased morbidity and mortality after major elective, inpatient, noncardiac surgery.

Design, Setting, and Participants

This retrospective, population-based, matched cohort study used data from the National Inpatient Sample for adult patients aged 18 to 65 years who underwent major elective inpatient surgery (including cholecystectomy, colectomy, inguinal hernia repair, femoral hernia repair, mastectomy, lumpectomy, hip arthroplasty, knee arthroplasty, hysterectomy, spinal fusion, and vertebral discectomy) from January 2016 to December 2019. Data were analyzed from February to August 2022.

Exposure

Cannabis use disorder, as defined by the presence of specific International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) diagnostic codes.

Main Outcome and Measures

The primary composite outcome was in-hospital mortality and 7 major perioperative complications (myocardial ischemia, acute kidney injury, stroke, respiratory failure, venous thromboembolism, hospital-acquired infection, and surgical procedure-related complications) based on ICD-10 discharge diagnosis codes. Propensity score matching was performed to create a 1:1 matched cohort that was well balanced with respect to covariates, which included patient comorbidities, sociodemographic factors, and procedure type.

Results

Among 12 422 hospitalizations, a cohort of 6211 patients with cannabis use disorder (median age, 53 years [IQR, 44-59 years]; 3498 [56.32%] male) were matched with 6211 patients without cannabis use disorder for analysis. Cannabis use disorder was associated with an increased risk of perioperative morbidity and mortality compared with hospitalizations without cannabis use disorder in adjusted analysis (adjusted odds ratio, 1.19; 95% CI, 1.04-1.37; P = .01). The outcome occurred more frequently in the group with cannabis use disorder (480 [7.73%]) compared with the unexposed group (408 [6.57%]).

Conclusions and Relevance

In this cohort study, cannabis use disorder was associated with a modest increased risk of perioperative morbidity and mortality after major elective, inpatient, noncardiac surgery. In the context of increasing cannabis use rates, our findings support preoperative screening for cannabis use disorder as a component of perioperative risk stratification. However, further research is needed to quantify the perioperative impact of cannabis use by route and dosage and to inform recommendations for preoperative cannabis cessation.

Sex-specific risks for cardiovascular disease across the glycaemic spectrum: a population-based cohort study using the UK Biobank

Background

We sought to examine sex-specific risks for incident cardiovascular disease (CVD) across the full glycaemic spectrum.

Methods

Using data from UK Biobank, we categorised participants' glycated haemoglobin (HbA1c) at baseline as low-normal (<35 mmol/mol), normal (35-41 mmol/mol), pre-diabetes (42-47 mmol/mol), undiagnosed diabetes (≥48 mmol/mol), or diagnosed diabetes. Our outcomes were coronary artery disease (CAD), atrial fibrillation, deep vein thrombosis (DVT), pulmonary embolism (PE), stroke, heart failure, and a composite outcome of any CVD. Cox regression estimated sex-specific associations between HbA1c and each outcome, sequentially adjusting for socio-demographic, lifestyle, and clinical characteristics.

Findings

Among 427,435 people, CVD rates were 16.9 and 9.1 events/1000 person-years for men and women, respectively. Both men and women with pre-diabetes, undiagnosed diabetes, and, more markedly, diagnosed diabetes were at higher risks of CVD than those with normal HbA1c, with relative increases more pronounced in women than men. Age-adjusted HRs for pre-diabetes and undiagnosed diabetes ranged from 1.30 to 1.47; HRs for diagnosed diabetes were 1.55 (1.49-1.61) in men and 2.00 (1.89-2.12) in women (p-interaction <0.0001). Excess risks attenuated and were more similar between men and women after adjusting for clinical and lifestyle factors particularly obesity and antihypertensive or statin use (fully adjusted HRs for diagnosed diabetes: 1.06 [1.02-1.11] and 1.17 [1.10-1.24], respectively).

Interpretation

Excess risks in men and women were largely explained by modifiable factors, and could be ameliorated by attention to weight reduction strategies and greater use of antihypertensive and statin medications. Addressing these risk factors could reduce sex disparities in risk of CVD among people with and without diabetes.

Funding

Diabetes UK (#15/0005250) and British Heart Foundation (SP/16/6/32726).

Combining meta-analysis with multiple imputation for one-step, privacy-protecting estimation of causal treatment effects in multi-site studies

Missing data complicates statistical analyses in multi-site studies, especially when it is not feasible to centrally pool individual-level data across sites. We combined meta-analysis with within-site multiple imputation for one-step estimation of the average causal effect (ACE) of a target population comprised of all individuals from all data-contributing sites within a multi-site distributed data network, without the need for sharing individual-level data to handle missing data. We considered two orders of combination and three choices of weights for meta-analysis, resulting in six approaches. The first three approaches, denoted as RR + metaF, RR + metaR and RR + std, first combined results from imputed data sets within each site using Rubin's rules and then meta-analyzed the combined results across sites using fixed-effect, random-effects and sample-standardization weights, respectively. The last three approaches, denoted as metaF + RR, metaR + RR and std + RR, first meta-analyzed results across sites separately for each imputation and then combined the meta-analysis results using Rubin's rules. Simulation results confirmed very good performance of RR + std and std + RR under various missing completely at random and missing at random settings. A direct application of the inverse-variance weighted meta-analysis based on site-specific ACEs can lead to biased results for the targeted network-wide ACE in the presence of treatment effect heterogeneity by site, demonstrating the need to clearly specify the target population and estimand and properly account for potential site heterogeneity in meta-analyses seeking to draw causal interpretations. An illustration using a large administrative claims database is presented.

Predicting incident dementia in cerebral small vessel disease: comparison of machine learning and traditional statistical models

Background

Cerebral small vessel disease (SVD) contributes to 45% of dementia cases worldwide, yet we lack a reliable model for predicting dementia in SVD. Past attempts largely relied on traditional statistical approaches. Here, we investigated whether machine learning (ML) methods improved prediction of incident dementia in SVD from baseline SVD-related features over traditional statistical methods.

Methods

We included three cohorts with varying SVD severity (RUN DMC, n = 503; SCANS, n = 121; HARMONISATION, n = 265). Baseline demographics, vascular risk factors, cognitive scores, and magnetic resonance imaging (MRI) features of SVD were used for prediction. We conducted both survival analysis and classification analysis predicting 3-year dementia risk. For each analysis, several ML methods were evaluated against standard Cox or logistic regression. Finally, we compared the feature importance ranked by different models.

Results

We included 789 participants without missing data in the survival analysis, amongst whom 108 (13.7%) developed dementia during a median follow-up of 5.4 years. Excluding those censored before three years, we included 750 participants in the classification analysis, amongst whom 48 (6.4%) developed dementia by year 3. Comparing statistical and ML models, only regularised Cox/logistic regression outperformed their statistical counterparts overall, but not significantly so in survival analysis. Baseline cognition was highly predictive, and global cognition was the most important feature.

Conclusions

When using baseline SVD-related features to predict dementia in SVD, the ML survival or classification models we evaluated brought little improvement over traditional statistical approaches. The benefits of ML should be evaluated with caution, especially given limited sample size and features.

Assumptions and analysis planning in studies with missing data in multiple variables: moving beyond the MCAR/MAR/MNAR classification

Researchers faced with incomplete data are encouraged to consider whether their data are 'missing completely at random' (MCAR), 'missing at random' (MAR) or 'missing not at random' (MNAR) when planning their analysis. However, there are two major problems with this classification as originally defined by Rubin in the 1970s. First, when there are missing data in multiple variables, the plausibility of the MAR assumption is difficult to assess using substantive knowledge and is more stringent than is generally appreciated. Second, although MCAR and MAR are sufficient conditions for consistent estimation with specific methods, they are not necessary conditions and therefore this categorization does not directly determine the best approach for handling the missing data in an analysis. How best to handle missing data depends on the assumed causal relationships between variables and their missingness, and what these relationships imply in terms of the 'recoverability' of the target estimand (the population parameter that encodes the answer to the underlying research question). Recoverability is defined as whether the estimand can be consistently estimated from the patterns and associations in the observed data without needing to invoke external information on the extent to which the distribution of missing values might differ from that of observed values. In this manuscript we outline an approach for deciding which method to use to handle multivariable missing data in an analysis, using directed acyclic graphs to depict missingness assumptions and determining the implications in terms of recoverability of the target estimand.

Distinct cardiovascular and cancer burdens associated with social position, work environment and unemployment: a cross-sectional and retrospective study in a large population-based French cohort

OBJECTIVES

Distinguish the respective effects of social position, work environment and unemployment on cardiovascular and cancer risks.

DESIGN

A cross-sectional and retrospective observational study.

SETTING

A population-based French cohort (CONSTANCES).

PARTICIPANTS

130 197 adults enrolled between 2012 and 2021 without missing values.

PRIMARY OUTCOME MEASURES

The associations of social position, work environment and unemployment exposure with the prevalence of cardiovascular events and cancers simultaneously tested using logistic regression models adjusting for common risk factors.

RESULTS

While social position, work environment and unemployment exposure are strongly inter-related with each other, they are not linked to the same cardiovascular and cancer outcomes. Low social position and long unemployment duration are significantly associated with an increased prevalence of angina pectoris, myocardial infarction and peripheral arterial disease (OR=1.22 to 1.90, p<0.04 to p<0.0001) but not of stroke. In contrast, a bad work environment is associated with an increased prevalence of stroke (OR=1.29, p<0.01) but not of angina pectoris, myocardial infarction and peripheral arterial disease. Low social position is associated with an increased prevalence of cervical and lung cancers (OR=1.73 and 1.95, p<0.002 and p<0.03) and a decreased prevalence of skin cancer (OR=0.70, p<0.0001) while a bad work environment is associated with an increased prevalence of breast, skin, prostate and colon cancers (OR=1.31 to 2.91, p<0.0002 to p<0.0001). Unemployment exposure is not associated with the prevalence of any type of cancers.

CONCLUSIONS

Social position, work environment and unemployment are associated with distinct cardiovascular and cancerous diseases that could add up during lifetime, they should therefore be considered all together in any preventive strategy.

Cardiovascular burden and unemployment: A retrospective study in a large population-based French cohort

The specific effect of unemployment on cardiovascular health relatively to the effects of social position and work environment is still unclear. To clarify this effect, the associations between current or past unemployment and the prevalence of common cardiovascular risk factor and events were tested using multiple logistic regression models with adjustment for both social position and prior work environment. The analyses were performed in a population-based French cohort (CONSTANCES) that included 131,186 adults enrolled between 2012 and 2021. Participants who were unemployed at inclusion (n = 8278) were overexposed to non-moderate alcohol consumption, smoking, leisure-time physical inactivity and depression (odds ratios (ORs) from 1.19 to 1.58) whereas those who have been unemployed at least once in the past (n = 19,015) were additionally overexposed not only to the previous risk factors but also to obesity, diabetes and sleep disorders (ORs from 1.10 to 1.35). These latter were also more exposed to non-fatal myocardial infarction and peripheral arterial disease (ORs of 1.44 and 1.47 respectively), overexposures that persisted after further adjustment for cardiovascular risk factors (ORs of 1.36 and 1.33). The overexposures to risk factors and cardiovascular events were both dependent on the duration of past unemployment. They were equally observed in participants with low social position or bad work environment. These results suggest that unemployment increases cardiovascular risk independently from social position and work environment with a cumulative effect over time. The effect of unemployment could add up to those of low social position and bad work environment during lifetime to further increase cardiovascular risk. They also suggest that long-term unemployment increases the prevalence of cardiovascular events through pathways including but not limited to overexposure to common risk factors.

Differences between Kidney Transplant Recipients from Deceased Donors with Diabetes Mellitus as Identified by Machine Learning Consensus Clustering

Clinical outcomes of deceased donor kidney transplants coming from diabetic donors currently remain inconsistent, possibly due to high heterogeneities in this population. Our study aimed to cluster recipients of diabetic deceased donor kidney transplants using an unsupervised machine learning approach in order to identify subgroups with high risk of inferior outcomes and potential variables associated with these outcomes. Consensus cluster analysis was performed based on recipient-, donor-, and transplant-related characteristics in 7876 recipients of diabetic deceased donor kidney transplants from 2010 to 2019 in the OPTN/UNOS database. We determined the important characteristics of each assigned cluster and compared the post-transplant outcomes between the clusters. Consensus cluster analysis identified three clinically distinct clusters. Recipients in cluster 1 (n = 2903) were characterized by oldest age (64 ± 8 years), highest rate of comorbid diabetes mellitus (55%). They were more likely to receive kidney allografts from donors that were older (58 ± 6.3 years), had hypertension (89%), met expanded criteria donor (ECD) status (78%), had a high rate of cerebrovascular death (63%), and carried a high kidney donor profile index (KDPI). Recipients in cluster 2 (n = 687) were younger (49 ± 13 years) and all were re-transplant patients with higher panel reactive antibodies (PRA) (88 [IQR 46, 98]) who received kidneys from younger (44 ± 11 years), non-ECD deceased donors (88%) with low numbers of HLA mismatch (4 [IQR 2, 5]). The cluster 3 cohort was characterized by first-time kidney transplant recipients (100%) who received kidney allografts from younger (42 ± 11 years), non-ECD deceased donors (98%). Compared to cluster 3, cluster 1 had higher incidence of primary non-function, delayed graft function, patient death and death-censored graft failure, whereas cluster 2 had higher incidence of delayed graft function and death-censored graft failure but comparable primary non-function and patient death. An unsupervised machine learning approach characterized diabetic donor kidney transplant patients into three clinically distinct clusters with differing outcomes. Our data highlight opportunities to improve utilization of high KDPI kidneys coming from diabetic donors in recipients with survival-limiting comorbidities such as those observed in cluster 1.

A Spatio-Temporal Analysis of OECD Member Countries' Health Care Systems: Effects of Data Missingness and Geographically and Temporally Weighted Regression on Inference

Determinants of health care quality and efficiency are of importance to researchers, policy-makers, and public health officials as they allow for improved human capital and resource allocation as well as long-term fiscal planning. Statistical analyses used to understand determinants have neglected to explicitly discuss how missing data are handled, and consequently, previous research has been limited in inferential capability. We study OECD health care data and highlight the importance of transparency in the assumptions grounding the treatment of data missingness. Attention is drawn to the variation in ordinary least squares coefficient estimates and performance resulting from different imputation methods, and how this variation can undermine statistical inference. We also suggest that parametric regression models used previously are limited and potentially ill-suited for analysis of OECD data due to the inability to deal with both spatial and temporal autocorrelation. We propose the use of an alternative method in geographically and temporally weighted regression. A spatio-temporal analysis of health care system efficiency and quality of care across OECD member countries is performed using four proxy variables. Through a forward selection procedure, medical imaging equipment in a country is identified as a key determinant of quality of care and health outcomes, while government and compulsory health insurance expenditure per capita is identified as a key determinant of health care system efficiency.

Habitual dietary nitrate intake and cognition in the Australian Imaging, Biomarkers and Lifestyle Study of ageing: A prospective cohort study

BACKGROUND & AIMS

Dietary nitrate improves cardiovascular health via a nitric oxide (NO) pathway. NO is key to both cardiovascular and brain health. There is also a strong association between vascular risk factors and brain health. Dietary nitrate intake could therefore be associated with better cognitive function and reduced risk of cognitive decline. This is yet to be investigated. The aim of this study was to investigate the association between habitual intake of dietary nitrate from sources where nitrate is naturally present, and cognitive function, and cognitive decline, in the presence or absence of the apolipoprotein E (APOE) ε4 allele.

METHODS

The study included 1254 older adult participants of the Australian Imaging, Biomarkers and Lifestyle Study of Ageing who were cognitively normal at baseline. Plant-derived, vegetable-derived, animal derived nitrate (not including meat where nitrate is an allowed additive), and total nitrate intakes were calculated from baseline food frequency questionnaires using comprehensive nitrate databases. Cognition was assessed at baseline and every 18 months over a follow-up period of 126 months using a comprehensive neuropsychological test battery. Multivariable-adjusted linear mixed effect models were used to examine the association between baseline nitrate intake and cognition over the 126 months (median [IQR] follow-up time of 36 [18-72] months), stratified by APOE ε4 carrier status.

RESULTS

In non APOE ε4 carriers, for every 60 mg/day higher intake of plant-derived nitrate at baseline there was an associated higher language score [β (95% CI): 0.10 (0.01, 0.19)] over 126 months, after multivariable adjustments. In APOE ε4 carriers, there was an associated better episodic recall memory [0.24 (0.08, 0.41)] and recognition memory [0.15 (0.01, 0.30)] scores. Similar associations were seen for the intakes of vegetable-derived and total nitrate. Additionally, in APOE ε4 carriers, for every 6 mg/day higher intake of animal-derived nitrate (excluding meat with nitrate as an allowed additive) at baseline there was an associated higher executive function score [β (95% CI): 1.41 (0.42, 2.39)]. We did not find any evidence of an association between dietary nitrate intake and rate of cognitive decline.

CONCLUSION

Our results suggest that habitual intake of dietary nitrate from sources where nitrate is naturally present impacts cognitive performance in an APOE genotype contingent manner. Further work is needed to validate our findings and understand potential mechanisms underlying the observed effects.

Handling missing covariate data in clinical studies in haematology

Missing data are frequently encountered across studies in clinical haematology. Failure to handle these missing values in an appropriate manner can complicate the interpretation of a study's findings, as estimates presented may be biased and/or imprecise. In the present work, we first provide an overview of current methods for handling missing covariate data, along with their advantages and disadvantages. Furthermore, a systematic review is presented, exploring both contemporary reporting of missing values in major haematological journals, and the methods used for handling them. A principal finding was that the method of handling missing data was explicitly specified in a minority of articles (in 76 out of 195 articles reporting missing values, 39%). Among these, complete case analysis and the missing indicator method were the most common approaches to dealing with missing values, with more complex methods such as multiple imputation being extremely rare (in 7 out of 195 articles). An example analysis (with associated code) is also provided using hematopoietic stem cell transplantation data, illustrating the different approaches to handling missing values. We conclude with various recommendations regarding the reporting and handling of missing values for future studies in clinical haematology.

Increasing efficiency of SVMp+ for handling missing values in healthcare prediction

Missing data presents a challenge for machine learning applications specifically when utilizing electronic health records to develop clinical decision support systems. The lack of these values is due in part to the complex nature of clinical data in which the content is personalized to each patient. Several methods have been developed to handle this issue, such as imputation or complete case analysis, but their limitations restrict the solidity of findings. However, recent studies have explored how using some features as fully available privileged information can increase model performance including in SVM. Building on this insight, we propose a computationally efficient kernel SVM-based framework (l2-SVMp+) that leverages partially available privileged information to guide model construction. Our experiments validated the superiority of l2-SVMp+ over common approaches for handling missingness and previous implementations of SVMp+ in both digit recognition, disease classification and patient readmission prediction tasks. The performance improves as the percentage of available privileged information increases. Our results showcase the capability of l2-SVMp+ to handle incomplete but important features in real-world medical applications, surpassing traditional SVMs that lack privileged information. Additionally, l2-SVMp+ achieves comparable or superior model performance compared to imputed privileged features.

Antibiotics for lower respiratory tract infection in children presenting in primary care: ARTIC-PC RCT

Background

Antimicrobial resistance is a global health threat. Antibiotics are commonly prescribed for children with uncomplicated lower respiratory tract infections, but there is little randomised evidence to support the effectiveness of antibiotics in treating these infections, either overall or relating to key clinical subgroups in which antibiotic prescribing is common (chest signs; fever; physician rating of unwell; sputum/rattly chest; shortness of breath).

Objectives

To estimate the clinical effectiveness and cost-effectiveness of amoxicillin for uncomplicated lower respiratory tract infections in children both overall and in clinical subgroups.

Design

Placebo-controlled trial with qualitative, observational and cost-effectiveness studies.

Setting

UK general practices.

Participants

Children aged 1-12 years with acute uncomplicated lower respiratory tract infections.

Outcomes

The primary outcome was the duration in days of symptoms rated moderately bad or worse (measured using a validated diary). Secondary outcomes were symptom severity on days 2-4 (0 = no problem to 6 = as bad as it could be); symptom duration until very little/no problem; reconsultations for new or worsening symptoms; complications; side effects; and resource use.

Methods

Children were randomised to receive 50 mg/kg/day of oral amoxicillin in divided doses for 7 days, or placebo using pre-prepared packs, using computer-generated random numbers by an independent statistician. Children who were not randomised could participate in a parallel observational study. Semistructured telephone interviews explored the views of 16 parents and 14 clinicians, and the data were analysed using thematic analysis. Throat swabs were analysed using multiplex polymerase chain reaction.

Results

A total of 432 children were randomised (antibiotics, n = 221; placebo, n = 211). The primary analysis imputed missing data for 115 children. The duration of moderately bad symptoms was similar in the antibiotic and placebo groups overall (median of 5 and 6 days, respectively; hazard ratio 1.13, 95% confidence interval 0.90 to 1.42), with similar results for subgroups, and when including antibiotic prescription data from the 326 children in the observational study. Reconsultations for new or worsening symptoms (29.7% and 38.2%, respectively; risk ratio 0.80, 95% confidence interval 0.58 to 1.05), illness progression requiring hospital assessment or admission (2.4% vs. 2.0%) and side effects (38% vs. 34%) were similar in the two groups. Complete-case (n = 317) and per-protocol (n = 185) analyses were similar, and the presence of bacteria did not mediate antibiotic effectiveness. NHS costs per child were slightly higher (antibiotics, £29; placebo, £26), with no difference in non-NHS costs (antibiotics, £33; placebo, £33). A model predicting complications (with seven variables: baseline severity, difference in respiratory rate from normal for age, duration of prior illness, oxygen saturation, sputum/rattly chest, passing urine less often, and diarrhoea) had good discrimination (bootstrapped area under the receiver operator curve 0.83) and calibration. Parents found it difficult to interpret symptoms and signs, used the sounds of the child's cough to judge the severity of illness, and commonly consulted to receive a clinical examination and reassurance. Parents acknowledged that antibiotics should be used only when 'necessary', and clinicians noted a reduction in parents' expectations for antibiotics.

Limitations

The study was underpowered to detect small benefits in key subgroups.

Conclusion

Amoxicillin for uncomplicated lower respiratory tract infections in children is unlikely to be clinically effective or to reduce health or societal costs. Parents need better access to information, as well as clear communication about the self-management of their child's illness and safety-netting.

Future work

The data can be incorporated in the Cochrane review and individual patient data meta-analysis.

Trial registration

This trial is registered as ISRCTN79914298.

Funding

This project was funded by the National Institute for Health and Care Research (NIHR) Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 27, No. 9. See the NIHR Journals Library website for further project information.

Lung Cancer Screening Among U.S. Military Veterans by Health Status and Race and Ethnicity, 2017-2020: A Cross-Sectional Population-Based Study

Introduction

Veterans are at high risk for lung cancer and are an important group for lung cancer screening. Previous research suggests that lung cancer screening may not be reaching healthier and/or non-White individuals, who stand to benefit most from lung cancer screening. We sought to test whether lung cancer screening is associated with poor health and/or race and ethnicity among veterans.

Methods

This cross-sectional, population-based study included veterans eligible for lung cancer screening (aged 55-79 years, ≥30 pack-year smoking history, current smokers or quit within 15 years, no previous lung cancer) in the 2017-2020 Behavioral Risk Factor Surveillance System surveys. Exposures were (1) poor health, defined as fair/poor health status and difficulty walking or climbing stairs, aligning with eligibility criteria for a pivotal lung cancer screening trial, and (2) race/ethnicity. The outcome was a receipt of lung cancer screening. All variables were self-reported.

Results

Of 3,376 lung cancer screening-eligible veterans representing an underlying population of 866,000 individuals, 20.3% (95% CI=17.3, 23.6) had poor health, and 13.7% (95% CI=10.6, 17.5) identified as non-White. Poor health was strongly associated with lung cancer screening (adjusted RR=1.64, 95% CI=1.06, 2.27); one third of veterans screened for lung cancer would not qualify for a pivotal lung cancer screening trial in terms of health. Marked racial disparities were observed among veterans: after adjustment, non-White veterans were 67% less likely to report lung cancer screening than White veterans (adjusted RR=0.33, 95% CI=0.11, 0.66).

Conclusions

Lung cancer screening is correlated with poorer health and White race/ethnicity among veterans, which may undermine its population-level effectiveness. These results highlight the need to promote lung cancer screening, especially for healthier and/or non-White veterans, an important group of Americans for lung cancer screening.

Factors Associated with Hospital Length of Stay and Adverse Events following Percutaneous Ablation of Lung Tumors

PURPOSE

To explore the association between risk factors established in the surgical literature and hospital length of stay (HLOS), adverse events, and hospital readmission within 30 days after percutaneous image-guided thermal ablation of lung tumors.

MATERIALS AND METHODS

This bi-institutional retrospective cohort study included 131 consecutive adult patients (67 men [51%]; median age, 65 years) with 180 primary or metastatic lung tumors treated in 131 sessions (74 cryoablation and 57 microwave ablation) from 2006 to 2019. Age-adjusted Charlson Comorbidity Index, sex, performance status, smoking status, chronic obstructive pulmonary disease (COPD), primary lung cancer versus pulmonary metastases, number of tumors treated per session, maximum axial tumor diameter, ablation modality, number of pleural punctures, anesthesia type, pulmonary artery-to-aorta ratio, lung densitometry, sarcopenia, and adipopenia were evaluated. Associations between risk factors and outcomes were assessed using univariable and multivariable generalized linear models.

RESULTS

In univariable analysis, HLOS was associated with current smoking (incidence rate ratio [IRR], 4.54 [1.23-16.8]; P = .02), COPD (IRR, 3.56 [1.40-9.04]; P = .01), cryoablations with ≥3 pleural punctures (IRR, 3.13 [1.07-9.14]; P = .04), general anesthesia (IRR, 10.8 [4.18-27.8]; P < .001), and sarcopenia (IRR, 2.66 [1.10-6.44]; P = .03). After multivariable adjustment, COPD (IRR, 3.56 [1.57-8.11]; P = .003) and general anesthesia (IRR, 12.1 [4.39-33.5]; P < .001) were the only risk factors associated with longer HLOS. No associations were observed between risk factors and adverse events in multivariable analysis. Tumors treated per session were associated with risk of hospital readmission (P = .03).

CONCLUSIONS

Identified preprocedural risk factors from the surgical literature may aid in risk stratification for HLOS after percutaneous ablation of lung tumors, but were not associated with adverse events.

COVID-19 after two years: trajectories of different components of mental health in the Spanish population

AIMS

Our study aimed to (1) identify trajectories on different mental health components during a two-year follow-up of the COVID-19 pandemic and contextualise them according to pandemic periods; (2) investigate the associations between mental health trajectories and several exposures, and determine whether there were differences among the different mental health outcomes regarding these associations.

METHODS

We included 5535 healthy individuals, aged 40-65 years old, from the Barcelona Brain Health Initiative (BBHI). Growth mixture models (GMM) were fitted to classify individuals into different trajectories for three mental health-related outcomes (psychological distress, personal growth and loneliness). Moreover, we fitted a multinomial regression model for each outcome considering class membership as the independent variable to assess the association with the predictors.

RESULTS

For the outcomes studied we identified three latent trajectories, differentiating two major trends, a large proportion of participants was classified into 'resilient' trajectories, and a smaller proportion into 'chronic-worsening' trajectories. For the former, we observed a lower susceptibility to the changes, whereas, for the latter, we noticed greater heterogeneity and susceptibility to different periods of the pandemic. From the multinomial regression models, we found global and cognitive health, and coping strategies as common protective factors among the studied mental health components. Nevertheless, some differences were found regarding the risk factors. Living alone was only significant for those classified into 'chronic' trajectories of loneliness, but not for the other outcomes. Similarly, secondary or higher education was only a risk factor for the 'worsening' trajectory of personal growth. Finally, smoking and sleeping problems were risk factors which were associated with the 'chronic' trajectory of psychological distress.

CONCLUSIONS

Our results support heterogeneity in reactions to the pandemic and the need to study different mental health-related components over a longer follow-up period, as each one evolves differently depending on the pandemic period. In addition, the understanding of modifiable protective and risk factors associated with these trajectories would allow the characterisation of these segments of the population to create targeted interventions.

The M-Value: A Simple Sensitivity Analysis for Bias Due to Missing Data in Treatment Effect Estimates

Complete-case analyses can be biased if missing data are not missing completely at random. We propose simple sensitivity analyses that apply to complete-case estimates of treatment effects; these analyses use only simple summary data and obviate specifying the precise mechanism of missingness and making distributional assumptions. Bias arises when treatment effects differ between retained and nonretained participants or, among retained participants, the estimate is biased because conditioning on retention has induced a noncausal path between the treatment and outcome. We thus bound the overall treatment effect on the difference scale by specifying: 1) the unobserved treatment effect among nonretained participants; and 2) the strengths of association that unobserved variables have with the exposure and with the outcome among retained participants ("induced confounding associations"). Working with the former sensitivity parameter subsumes certain existing methods of worst-case imputation while also accommodating less-conservative assumptions (e.g., that the treatment is not detrimental on average even among nonretained participants). As an analog to the E-value for confounding, we propose the M-value, which represents, for a specified treatment effect among nonretained participants, the strength of induced confounding associations required to reduce the treatment effect to the null or to any other value. These methods could help characterize the robustness of complete-case analyses to potential bias due to missing data.

Is a more selective exit exam related to shadow education use? An analysis of two cohorts of final-year secondary school students in the Netherlands

The prevalence of private supplementary tutoring (i.e. shadow education) is growing, particularly in nations with selective school exams. The hypothesis that tutoring attendance rises as pressure to perform increases has not yet been tested. Therefore, our research question is: does the likelihood of attending shadow education increase with an increase in educational system selectivity (in this case: stricter requirements to graduate)? Our study used an opportunity to study this question in the Dutch context, where performance standards on the nationwide secondary education exit exam were raised in 2011. We used data from two cohorts of Dutch exam year students: one before and one after the raised selectivity, hypothesising that the latter group had a higher likelihood of attending shadow education than the former. Results from Propensity Score Matching (PSM), applied to obtain cohorts as similar as possible on observable characteristics, provide support for our hypothesis. Our results on the Dutch policy change confirm that shadow education use might emerge at key moments of selection. In doing so, our study suggests that policymakers and researchers may want to include shadow education use in discussions on the merits and pitfalls of introducing more stringent performance standards into the education system.

Distinct Phenotypes of Non-Citizen Kidney Transplant Recipients in the United States by Machine Learning Consensus Clustering

Better understanding of the different phenotypes/subgroups of non-U.S. citizen kidney transplant recipients may help the transplant community to identify strategies that improve outcomes among non-U.S. citizen kidney transplant recipients. This study aimed to cluster non-U.S. citizen kidney transplant recipients using an unsupervised machine learning approach; Methods: We conducted a consensus cluster analysis based on recipient-, donor-, and transplant- related characteristics in non-U.S. citizen kidney transplant recipients in the United States from 2010 to 2019 in the OPTN/UNOS database using recipient, donor, and transplant-related characteristics. Each cluster's key characteristics were identified using the standardized mean difference. Post-transplant outcomes were compared among the clusters; Results: Consensus cluster analysis was performed in 11,300 non-U.S. citizen kidney transplant recipients and identified two distinct clusters best representing clinical characteristics. Cluster 1 patients were notable for young age, preemptive kidney transplant or dialysis duration of less than 1 year, working income, private insurance, non-hypertensive donors, and Hispanic living donors with a low number of HLA mismatch. In contrast, cluster 2 patients were characterized by non-ECD deceased donors with KDPI <85%. Consequently, cluster 1 patients had reduced cold ischemia time, lower proportion of machine-perfused kidneys, and lower incidence of delayed graft function after kidney transplant. Cluster 2 had higher 5-year death-censored graft failure (5.2% vs. 9.8%; p < 0.001), patient death (3.4% vs. 11.4%; p < 0.001), but similar one-year acute rejection (4.7% vs. 4.9%; p = 0.63), compared to cluster 1; Conclusions: Machine learning clustering approach successfully identified two clusters among non-U.S. citizen kidney transplant recipients with distinct phenotypes that were associated with different outcomes, including allograft loss and patient survival. These findings underscore the need for individualized care for non-U.S. citizen kidney transplant recipients.

Ignoring Non-ignorable Missingness

The classical missing at random (MAR) assumption, as defined by Rubin (Biometrika 63:581-592, 1976), is often not required for valid inference ignoring the missingness process. Neither are other assumptions sometimes believed to be necessary that result from misunderstandings of MAR. We discuss three strategies that allow us to use standard estimators (i.e., ignore missingness) in cases where missingness is usually considered to be non-ignorable: (1) conditioning on variables, (2) discarding more data, and (3) being protective of parameters.

Should multiple imputation be stratified by exposure group when estimating causal effects via outcome regression in observational studies?

BACKGROUND

Despite recent advances in causal inference methods, outcome regression remains the most widely used approach for estimating causal effects in epidemiological studies with a single-point exposure and outcome. Missing data are common in these studies, and complete-case analysis (CCA) and multiple imputation (MI) are two frequently used methods for handling them. In randomised controlled trials (RCTs), it has been shown that MI should be conducted separately by treatment group. In observational studies, causal inference is now understood as the task of emulating an RCT, which raises the question of whether MI should be conducted by exposure group in such studies.

METHODS

We addressed this question by evaluating the performance of seven methods for handling missing data when estimating causal effects with outcome regression. We conducted an extensive simulation study based on an illustrative case study from the Victorian Adolescent Health Cohort Study, assessing a range of scenarios, including seven outcome generation models with exposure-confounder interactions of differing strength.

RESULTS

The simulation results showed that MI by exposure group led to the least bias when the size of the smallest exposure group was relatively large, followed by MI approaches that included the exposure-confounder interactions.

CONCLUSIONS

The findings from our simulation study, which was designed based on a real case study, suggest that current practice for the conduct of MI in causal inference may need to shift to stratifying by exposure group where feasible, or otherwise including exposure-confounder interactions in the imputation model.

Ethnic differences in hypertension management, medication use and blood pressure control in UK primary care, 2006-2019: a retrospective cohort study

Background

In the UK, previous work suggests ethnic inequalities in hypertension management. We studied ethnic differences in hypertension management and their contribution to blood pressure (BP) control.

Methods

We conducted a cohort study of antihypertensive-naïve individuals of European, South Asian and African/African Caribbean ethnicity with a new raised BP reading in UK primary care from 2006 to 2019, using the Clinical Practice Research Datalink (CPRD). We studied differences in: BP re-measurement after an initial hypertensive BP, antihypertensive initiation, BP monitoring, antihypertensive intensification, antihypertensive persistence/adherence and BP control one year after antihypertensive initiation. Models adjusted for socio-demographics, BP, comorbidity, healthcare usage and polypharmacy (plus antihypertensive class, BP monitoring, intensification, persistence and adherence for BP control models).

Findings

A total of 731,506 (93.5%), 30,379 (3.9%) and 20,256 (2.6%) people of European, South Asian and African/African Caribbean ethnicity were studied. Hypertension management indicators were similar or more favourable for South Asian than European groups (OR/HR [95% CI] in fully-adjusted models of BP re-measurement: 1.16 [1.09, 1.24]), antihypertensive initiation: 1.49 [1.37, 1.62], BP monitoring: 0.97 [0.94, 1.00] and antihypertensive intensification: 1.10 [1.04, 1.16]). For people of African/African Caribbean ethnicity, BP re-measurement rates were similar to those of European ethnicity (0.98 [0.91, 1.05]), and antihypertensive initiation rates greater (1.48 [1.32, 1.66]), but BP monitoring (0.91 [0.87, 0.95]) and intensification rates lower (0.93 [0.87, 1.00]). Persistence and adherence were lower in South Asian (0.48 [0.45, 0.51] and 0.51 [0.47, 0.56]) and African/African Caribbean (0.38 [0.35, 0.42] and 0.39 [0.36, 0.43]) than European groups. BP control was similar in South Asian and less likely in African/African Caribbean than European groups (0.98 [0.90, 1.06] and 0.81 [0.74, 0.89] in age, gender and BP adjusted models). The latter difference attenuated after adjustment for persistence (0.91 [0.82, 0.99]) or adherence (0.92 [0.83, 1.01]), and was absent for antihypertensive-adherent people (0.99 [0.88, 1.10]).

Interpretation

We demonstrate that antihypertensive initiation does not vary by ethnicity, but subsequent BP control was notably lower among people of African/African Caribbean ethnicity, potentially associated with being less likely to remain on regular treatment. A nationwide strategy to understand and address differences in ongoing management of people on antihypertensives is imperative.

Funding

Diabetes UK.

rox: A Statistical Model for Regression with Missing Values

High-dimensional omics datasets frequently contain missing data points, which typically occur due to concentrations below the limit of detection (LOD) of the profiling platform. The presence of such missing values significantly limits downstream statistical analysis and result interpretation. Two common techniques to deal with this issue include the removal of samples with missing values and imputation approaches that substitute the missing measurements with reasonable estimates. Both approaches, however, suffer from various shortcomings and pitfalls. In this paper, we present "rox", a novel statistical model for the analysis of omics data with missing values without the need for imputation. The model directly incorporates missing values as "low" concentrations into the calculation. We show the superiority of rox over common approaches on simulated data and on six metabolomics datasets. Fully leveraging the information contained in LOD-based missing values, rox provides a powerful tool for the statistical analysis of omics data.

Determinants of verbal fluency trajectories among older adults from the English Longitudinal Study of Aging

BACKGROUND

Prevalence of dementia and cognitive impairment increase creating the need for identifying modifiable risk factors to reduce their burden. The aim of this study was to identify latent groups following similar trajectories in cognitive performance assessed with the verbal fluency test, as well as their determinants.

METHODS

Data from English Longitudinal Study of Aging (ELSA) were studied. Latent groups of similar course through a 6-year period in the outcome variable (verbal fluency) were investigated, along with their determinants, using Group Based Trajectory Modeling (GBTM).

RESULTS

Four latent groups of verbal fluency trajectories were revealed. Education was the strongest predictor for a favorable trajectory, while cardiovascular disease and depression symptoms were associated with lower within each trajectory.

CONCLUSION

Cardiovascular diseases and depressive symptoms are associated with a worse course of verbal fluency through aging, implying that they might serve as targets for interventions to prevent cognitive decline in the aging population. Contrarily, higher level of education is associated with a more favorable course through aging.

Sociodemographic disparities in preterm birth and low birthweight in the State of Georgia: Results from the 2017-2018 Pregnancy Risk Assessment Monitoring System

PURPOSE

To update the overall prevalence of preterm birth (PTB) (<37 weeks gestation) and low birthweight (LBW) (<2,500 g) in the State of Georgia, including rural and urban counties.

METHODS

A sample was drawn from the 2017-2018 Georgia Pregnancy Risk Assessment Monitoring System (PRAMS). In the complete-case data of singleton births (n=1,258), we estimated the weighted percentage prevalence of PTB, LBW, early/late PTB, and moderately/very LBW subcategories in association with maternal sociodemographic characteristics, and the prevalence stratified by rural/urban county of residence. Univariate and multivariate logistic regression models were fitted to estimate the odds ratios (ORs) of PTB and LBW adjusting for selected covariates. Logistic regression results from multiple imputation by chained equations (MICE) were used for comparison.

FINDINGS

The overall rate for PTB was 9.3% and 6.8% for LBW and among them, 2.3% were early PTB, 7.0% were late PTB, 5.4% were moderately LBW (MLBW), and 1.3% were very LBW (VLBW). Non-Hispanic Black women had the highest prevalence of PTB, LBW, early PTB, MLBW, and VLBW, as well as PTB and LBW in urban counties and LBW in rural counties. The odds of PTB (aOR 1.38; 95% CI: 0.81, 2.35) and LBW (aOR 2.68; 95% CI: 1.32, 5.43) were also higher among non-Hispanic Black relative to non-Hispanic White women and among women who received adequate-plus prenatal care compared to inadequate prenatal care.

CONCLUSIONS

Socioeconomic and health disparities created by disadvantage should be a focus of state policy to improve neonatal outcomes in the State of Georgia.

The importance of plausible data-generating mechanisms in simulation studies: A response to 'Comparing methods for handling missing covariates in meta-regression' by Lee and Beretvas (doi: 10.1002/jrsm.1585)

The paper by Lee and Beretvas (doi:10.1002/jrsm.1585) described a well-executed simulation study comparing 'modern' with 'ad hoc' methods for performing meta-regression when some covariates are incomplete. However, they drew practical conclusions after simulating data under a single missing data mechanism which favoured the 'modern' methods, while other missing data mechanisms would have favoured the 'ad hoc' methods. Broad recommendations about methods to use in practice should instead be based on simulation studies using a range of plausible data-generating mechanisms. This range must represent what is believed likely to occur in practice, and not what is convenient for statistical analysis.

Prenatal cocaine exposure and substance use disorder in emerging adulthood at age 21

BACKGROUND

Prenatal cocaine exposure (PCE) has been associated with child and adolescent externalizing behaviors and early substance use, yet few studies investigated its association with substance use disorder (SUD) in emerging adults. The present study examined the association of PCE with SUD in emerging adulthood, and whether childhood externalizing behaviors and adolescent substance use mediated the relationship.

METHODS

Participants were 367 (187 PCE; 53% female) adults at age 21, primarily urban African American who were recruited at birth. PCE and exposure to alcohol, tobacco, and marijuana were determined using biologic assays for drug metabolites and/or maternal self-report at birth. Offspring externalizing problems were assessed using the Youth Self-Report at age 12, substance use and substance use-related problems via biologic assays and/or self-report at age 15, and SUD determined using DSM-5 diagnostic criteria at age 21.

RESULTS

About 32.3% of the emerging adults were determined to have marijuana use disorder, 30.3% tobacco use disorder, and 15.5% alcohol use disorder. PCE was related to greater externalizing behaviors at age 12 (β = 0.12, p = .042), which in turn was related to SUD (β = 0.22, p = .008). PCE was also related to substance use, mainly marijuana, at age 15 (β = 0.22, p = .011), which was related to SUD (β = 0.51, p < .001). Total indirect effects including these two pathways were significant (β = 0.19, p = .002).

CONCLUSIONS

PCE may increase risk for SUD in emerging adulthood through childhood externalizing behaviors and adolescent substance use.

External validation of a CT-based radiomics signature in oropharyngeal cancer: Assessing sources of variation

BACKGROUND AND PURPOSE

Radiomics is a high-throughput approach that allows for quantitative analysis of imaging data for prognostic applications. Medical images are used in oropharyngeal cancer (OPC) diagnosis and treatment planning and these images may contain prognostic information allowing for treatment personalization. However, the lack of validated models has been a barrier to the translation of radiomic research to the clinic. We hypothesize that a previously developed radiomics model for risk stratification in OPC can be validated in a local dataset.

MATERIALS AND METHODS

The radiomics signature predicting overall survival incorporates features derived from the primary gross tumor volume of OPC patients treated with radiation +/- chemotherapy at a single institution (n = 343). Model fit, calibration, discrimination, and utility were evaluated. The signature was compared with a clinical model using overall stage and a model incorporating both radiomics and clinical data. A model detecting dental artifacts on computed tomography images was also validated.

RESULTS

The radiomics signature had a Concordance index (C-index) of 0.66 comparable to the clinical model's C-index of 0.65. The combined model significantly outperformed (C-index of 0.69, p = 0.024) the clinical model, suggesting that radiomics provides added value. The dental artifact model demonstrated strong ability in detecting dental artifacts with an area under the curve of 0.87.

CONCLUSION

This work demonstrates model performance comparable to previous validation work and provides a framework for future independent and multi-center validation efforts. With sufficient validation, radiomic models have the potential to improve traditional systems of risk stratification, treatment personalization and patient outcomes.

Missingness mechanism messiness in meta-analysis: A response to "The importance of plausible data-generating mechanisms in simulation studies"

The current paper responds to the commentary on the article (doi:10.1002/jrsm.1605). We discuss our perspectives on the missing data mechanisms and models assumed and used in our simulation study while acknowledging the inherent generalizability limitations of any (simulation) study. The plausibility of the exact missing data mechanism is challenging to definitively identify in any applied dataset. We describe and justify our assumed scenario in meta-regression that we investigated. We also revisit the performance of the deletion method and how it is tied into the assumed missingness model. Lastly, we reiterate the importance of conducting sensitivity analyses assessing different ways of handling missing data given different assumptions and offer this study as a starting point for future research.