Between-within models for survival analysis

Dementia and mortality in older adults: A twin study

INTRODUCTION

Dementia predicts increased mortality. We used case-control and co-twin control models to investigate genetic and shared environmental influences on this association.

METHODS

Case-control design, including 987 twins with dementia and 2938 age- and sex-matched controls in the Swedish Twin Registry. Co-twin control design, including 90 monozygotic (MZ) and 288 dizygotic (DZ) twin pairs discordant for dementia. To test for genetic and environmental confounding, differences were examined in mortality risk between twins with dementia and their matched or co-twin controls.

RESULTS

Twins with dementia showed greater mortality risk than age- and sex-matched controls (HR = 2.02 [1.86, 2.18]). Mortality risk is significantly elevated but attenuated substantially in discordant twin pairs, for example, comparing MZ twins with dementia to their co-twin controls (HR = 1.48 [1.08, 2.04]).

DISCUSSION

Findings suggest that genetic factors partially confound the association between dementia and mortality and provide an alternative hypothesis to increased mortality due to dementia itself. Highlights We studied dementia and mortality in twin pairs discordant for dementia. People without dementia outlived people with dementia. Identical twins with dementia and their co-twin controls had similar survival time. Findings suggest genotype may explain the link between dementia and mortality.

Student loans, mental health, and substance use: A gender comparison among US young adults

Objective: Student loan debt has become a growing crisis. Considering that women are more likely than men to take on student loans and more likely to take on larger amounts, we examine whether the effects of student loans on young adults' mental health and substance use differ by gender. Participants: We used the National Longitudinal Survey of Youth 1997 (NLSY97) data collected from 1997 to 2015. The NLSY97 consists of a nationally representative sample of American youths born between 1980 and 1984. Participants included 2,607 men and 3,004 women who reported college enrollment. Methods: We analyzed data using hybrid regression models. Results: Student loans have more negative effects on young men than young women, in terms of mental health problems, smoking, and heavy drinking. Particularly, young men tend to increase substance use in response to cumulative loan amounts. Conclusions: Borrowing patterns and the health consequences of student loans are gendered.

Impact of COVID-19 on domestic air transportation in China

Assessing the impact of the coronavirus disease 2019 (COVID-19) on air transportation is essential for policymakers and airlines to prevent their widespread shutdown. The panel data observed from January 20, 2020, to April 30, 2020, were used to identify the impact of COVID-19 and the relevant control measures adopted on China's domestic air transportation. Hybrid models within negative binomial models were employed to separate the temporal and spatial effects of COVID-19. Temporal effects show that the number of new confirmed cases and the control measures significantly affect the number of operated flights. Spatial effects show that the network effect of COVID-19 cases in destination cities, lockdown, and adjustment to Level I in the early stages have a negative impact on the operated flights. Adjustment to Level II or Level III both has positive temporal and spatial effects. This indicates that the control measures adopted during the early stage of the pandemic positively impact the restoration of the aviation industry and other industries in the later stage.

Frailty and the risk of dementia: is the association explained by shared environmental and genetic factors?

BACKGROUND

Frailty has been identified as a risk factor for cognitive impairment and dementia. However, it is not known whether familial factors, such as genetics and shared environmental factors, underlie this association. We analyzed the association between frailty and the risk of dementia in a large twin cohort and examined the role of familial factors in the association.

METHODS

The Rockwood frailty index (FI) based on 44 health deficits was used to assess frailty. The population-level association between FI and the risk of all-cause dementia was analyzed in 41,550 participants of the Screening Across the Lifespan Twin (SALT) study (full sample, aged 41-97 years at baseline), using Cox and competing risk models. A subsample of 10,487 SALT participants aged 65 and older who received a cognitive assessment (cognitive sample) was used in a sensitivity analysis to assess the effect of baseline cognitive level on the FI-dementia association. To analyze the influence of familial effects on the FI-dementia association, a within-pair analysis was performed. The within-pair model was also used to assess whether the risk conferred by frailty varies by age at FI assessment.

RESULTS

A total of 3183 individuals were diagnosed with dementia during the 19-year follow-up. A 10% increase in FI was associated with an increased risk of dementia (hazard ratio [HR] 1.17 (95% confidence interval [CI] 1.07, 1.18)) in the full sample adjusted for age, sex, education, and tobacco use. A significant association was likewise found in the cognitive sample, with an HR of 1.13 (95% CI 1.09, 1.20), adjusted for age, sex, and cognitive level at baseline. The associations were not attenuated when adjusted for APOE ɛ4 carrier status or considering the competing risk of death. After adjusting for familial effects, we found no evidence for statistically significant attenuation of the effect. The risk conferred by higher FI on dementia was constant after age 50 until very old age.

CONCLUSIONS

A higher level of frailty predicts the risk of dementia and the association appears independent of familial factors. Targeting frailty might thus contribute to preventing or delaying dementia.

Family size and economic wellbeing following divorce: The United States in comparative perspective

Studies have documented the negative association between divorce and women's economic wellbeing in several countries. Less is known about whether the effects of divorce on women's economic wellbeing, and their persistency, vary by family size. We present the first comprehensive assessment of how the short-term and medium-term economic consequences of divorce vary by family size. Using data from the US (PSID) and between-within random-effects models, we estimate changes in women's gross household income up to six years following divorce, stratified by the number of children in the household in the year of divorce. We add a comparative perspective using a harmonized set of socio-economic panel surveys from Australia (HILDA), Germany (GSOEP), and the UK (BHPS). Our findings demonstrate that the household incomes of women with three or more children decrease most drastically in the US, Germany, and the UK. In these countries, divorce widens the economic gap between child-rich households and those with no or few children. While childless women's incomes do not recover in the medium-term, incomes of mothers in Germany, the UK, and to a lesser extent the US partially recuperate. We demonstrate that differences in labor market attachment, and not remarriage, partially account for the family size differences we observe.

Feeding and Eating Disorder and Risk of Subsequent Neurodevelopmental Disorders: A Population-Based Cohort Study

Background: There are limited data concerning the long-term mental health of children with feeding and eating disorder (FED). We aimed to investigate whether children with FED are at greater risks of developing emotional/behavioral disorders with onset usually during childhood, attention-deficit/hyperactivity disorder (ADHD), autism spectrum disorder (ASD), and intellectual disability (ID). Methods: We conducted a population-based cohort study, including all singleton births in Denmark from January 1, 1995, to December 31, 2015. For each child diagnosed with FED, 10 age- and sex-matched controls who did not meet the criteria for FED were randomly selected from the general population. Associations were estimated with Cox regression modes adjusting for other perinatal and maternal factors, and sibling analyses were performed for controlling potential confounding by shared familial (genetic or environmental) factors. Results: Of the 1,256,989 individuals in the cohort, there were 1967 (53.4% girls) children diagnosed with FED. Children with FED had higher risks for clinically diagnosed emotional/behavioral disorders with onset usually in childhood (hazard ratio [HR], 2.78; 95% CI, 2.34-3.31), ADHD (HR, 1.74; 95% CI, 1.33-2.26), ASD (HR, 3.05; 95% CI, 2.36-3.94), and ID (HR, 6.38; 95% CI, 4.48-9.11), compared with matched controls. Girls with FED are at greater risks for emotional/behavioral disorders and ID, but not ADHD and ASD. Alike, in sibling analysis, increased rates are also observed for other neurodevelopmental disorders, but not for ADHD. Conclusion: Children with FED are associated with substantially increased risks of emotional/behavioral disorders, ADHD, ASD, and ID. This study highlights the importance of carefully monitoring neurodevelopmental disorders in children with FED.

Propensity score methods for time-dependent cluster confounding

In observational studies, subjects are often nested within clusters. In medical studies, patients are often treated by doctors and therefore patients are regarded as nested or clustered within doctors. A concern that arises with clustered data is that cluster-level characteristics (e.g., characteristics of the doctor) are associated with both treatment selection and patient outcomes, resulting in cluster-level confounding. Measuring and modeling cluster attributes can be difficult and statistical methods exist to control for all unmeasured cluster characteristics. An assumption of these methods however is that characteristics of the cluster and the effects of those characteristics on the outcome (as well as probability of treatment assignment when using covariate balancing methods) are constant over time. In this paper, we consider methods that relax this assumption and allow for estimation of treatment effects in the presence of unmeasured time-dependent cluster confounding. The methods are based on matching with the propensity score and incorporate unmeasured time-specific cluster effects by performing matching within clusters or using fixed- or random-cluster effects in the propensity score model. The methods are illustrated using data to compare the effectiveness of two total hip devices with respect to survival of the device and a simulation study is performed that compares the proposed methods. One method that was found to perform well is matching within surgeon clusters partitioned by time. Considerations in implementing the proposed methods are discussed.

Analysis of mortality after hip fracture on patient, hospital, and regional level in Germany

Knowledge about risk factors of mortality after hip fracture might encourage prevention and further improvements in care. This study identified patient risk factors as well as hospital and regional characteristics associated with a decreased risk. Variation of mortality was largest on patient level and modest on hospital and regional level.

INTRODUCTION

Among numerous studies analyzing mortality as worst consequence after hip fracture, the majority focused on patient level and fewer on hospital and regional level. Comprehensive knowledge about contributing factors on all levels might help to reveal relevant inequalities, which would encourage prevention and further improvements in care. This study aimed at investigating variation of mortality after hip fracture on patient, hospital, and regional level in Germany.

METHODS

We performed a retrospective cohort study on hip fracture patients aged 65 and older using statutory health insurance claims data from Jan 2009 through Dec. 2012 and additional information from the Federal Statistical Office Germany. Regions were classified based on two-digit postal code. We applied a multilevel Cox proportional hazard model with random intercepts on hospital and regional level to investigate the risk factors for mortality within 6 and 12 months after hip fracture.

RESULTS

The dataset contained information on 123,119 hip fracture patients in 1014 hospitals in 95 German regions. Within 6/12 months, 20.9%/27.6% of the patients died. On patient level, male sex, increasing age, increased pre-fracture care level, and increasing comorbidity were associated with an increased hazard of mortality. Hospitals with increasing hip fracture volume or with orthogeriatric co-management and regions with increased population density were associated with a decreased hazard. Variation was largest on patient level and rather modest on hospital and regional level.

CONCLUSIONS

The identification of patient-related risk factors enables prognosticating mortality after hip fracture. After adjusting for those, variation seemed to be attributable rather to hospitals than to regions.

Is cesarean section a cause of affective disorders?-A national cohort study using sibling designs

BACKGROUND

The gut microbiota of children delivered by cesarean section differs from that of children delivered vaginally. In light of the gut-brain axis hypothesis, cesarean section may influence risk of affective disorders.

METHODS

Population based prospective cohort study included Danish children born 1982 through 2001, with follow-up until 2015. The effect of delivery mode on the risk of affective disorders was assessed using a standard Cox model and two types of Cox sibling models. Diagnostic codes or prescriptions for antidepressants and lithium were used to define cases of affective disorders.

RESULTS

1,009,444 children were followed for 8,880,794 person-years from the age of 13 years, with relevant covariates available from birth. There are strong calendar time trends in the occurrence of affective disorders with an increasingly younger age at first diagnosis and with a hotspot between the years 2007-2012. Fully adjusted standard Cox models showed an increased risk of affective disorders for both pre-labor (hazard ratio [HR], 1.11; 95% confidence interval [CI], 1.08-1.15) and intrapartum (HR, 1.07; 95% CI, 1.05-1.10) cesarean section, compared to vaginal delivery. This effect disappeared in the between-within sibling model for pre-labor (HR, 1.00; 95% CI, 0.94-1.07) but not intrapartum (HR, 1.05; 95% CI, 1.00-1.12) cesarean section.

LIMITATIONS

Interpretation of results from sibling models may not be relevant to children without siblings.

CONCLUSIONS

These results do not support the hypothesis that a delivery-mode dependent change in gut microbiota is a cause of subsequent affective disorders, despite an apparent association with delivery mode.

Association of Tourette Syndrome and Chronic Tic Disorder With Metabolic and Cardiovascular Disorders

Importance

There are limited data concerning the risk of metabolic and cardiovascular disorders among individuals with Tourette syndrome (TS) or chronic tic disorder (CTD).

Objective

To investigate the risk of metabolic and cardiovascular disorders among individuals with TS or CTD over a period of 40 years.

Design, Settings, and Participants

This longitudinal population-based cohort study included all individuals living in Sweden between January 1, 1973, and December 31, 2013. Families with clusters of full siblings discordant for TS or CTD were further identified. Data analyses were conducted from August 1, 2017, to October 11, 2018.

Exposures

Previously validated International Classification of Diseases diagnoses of TS or CTD in the Swedish National Patient Register.

Main Outcomes and Measures

Registered diagnoses of obesity, dyslipidemia, hypertension, type 2 diabetes, and cardiovascular diseases (including ischemic heart diseases, arrhythmia, cerebrovascular diseases and transient ischemic attack, and arteriosclerosis).

Results

Of the 14 045 026 individuals in the cohort, 7804 individuals (5964 males [76.4%]; median age at first diagnosis, 13.3 years [interquartile range, 9.9-21.3 years]) had a registered diagnosis of TS or CTD in specialist care. Of 2 675 482 families with at least 2 singleton full siblings, 5141 families included siblings who were discordant for these disorders. Individuals with TS or CTD had a higher risk of any metabolic or cardiovascular disorders compared with the general population (hazard ratio adjusted by sex and birth year [aHR], 1.99; 95% CI, 1.90-2.09) and sibling controls (aHR for any disorder, 1.37; 95% CI, 1.24-1.51). Specifically, individuals with TS or CTD had higher risks for obesity (aHR, 2.76; 95% CI, 2.47-3.09), type 2 diabetes (aHR, 1.67; 95% CI, 1.42-1.96), and circulatory system diseases (aHR, 1.76; 95% CI, 1.67-1.86). The risk of any cardiometabolic disorder was significantly greater in males than in females (aHR, 2.13; 95% CI, 2.01-2.26 vs aHR, 1.79; 95% CI, 1.64-1.96), as was the risk of obesity (aHR, 3.24; 95% CI, 2.83-3.70 vs aHR, 1.97; 95% CI, 1.59-2.44). The risks were already evident from childhood (the groups were significantly different by age 8 years) and were significantly reduced with the exclusion of individuals with comorbid attention-deficit/hyperactivity disorder (aHR, 1.52; 95% CI, 1.42-1.62), while excluding other comorbidities did not significantly affect the results. Compared with patients with TS or CTD who were not taking antipsychotics, patients with a longer duration of antipsychotic treatment (>1 year) had significantly lower risks of metabolic and cardiovascular disorders.

Conclusions and Relevance

The findings of this study suggest that TS and CTD are associated with a substantial risk of metabolic and cardiovascular disorders. The results highlight the importance of carefully monitoring cardiometabolic health in patients with TS or CTD across the lifespan, particularly in those with comorbid attention-deficit/hyperactivity disorder.

Familial confounding of the association between maternal smoking in pregnancy and autism spectrum disorder in offspring

Evidence supports no link between maternal smoking in pregnancy and autism spectrum disorder (autism) overall. To address remaining questions about the unexplained heterogeneity between study results and the possibility of risk for specific autism sub-phenotypes, we conducted a whole-population cohort study in Denmark. We followed births 1991-2011 (1,294,906 persons, including 993,301 siblings in 728,271 families), from 1 year of age until an autism diagnosis (13,547), death, emigration, or December 31, 2012. Autism, with and without attention deficit hyperactivity disorder (ADHD) and with and without intellectual disability (ID) were based on ICD-8 and ICD-10 codes from Danish national health registers, including 3,319 autism + ADHD, 10,228 autism - no ADHD, 2,205 autism + ID, and 11,342 autism - no ID. We estimated hazard ratios (HRs) and 95% confidence intervals (95% CIs) between any maternal smoking (from birth records) and autism (or sub-phenotypes) using survival models with robust standard errors, stratifying by birth year and adjusting for child sex, parity, and parental age, education, income, and psychiatric history. To additionally address confounding using family designs, we constructed a maternal cluster model (adjusting for the smoking proportion within the family), and a stratified sibling model. Associations with maternal smoking and autism were elevated in conventional adjusted analyses (HR of 1.17 [1.13-1.22]) but attenuated in the maternal cluster (0.98 [0.88-1.09]) and sibling (0.86 [0.64-1.15]) models. Similarly, risks of autism sub-phenotypes with maternal smoking were attenuated in the family-based models. Together these results support that smoking in pregnancy is not linked with autism or select autism comorbid sub-phenotypes after accounting for familial confounding. Autism Res 2020, 13: 134-144. © 2019 International Society for Autism Research, Wiley Periodicals, Inc. LAY SUMMARY: Smoking during pregnancy has many harmful impacts, which may include harming the baby's developing brain. However, in a study of thousands of families in Denmark, it does not appear that smoking in pregnancy leads to autism or autism in combination with intellectual problems or attention deficits, once you account for the way smoking patterns and developmental disabilities run in families.

The frailty index is a predictor of cause-specific mortality independent of familial effects from midlife onwards: a large cohort study

BACKGROUND

Frailty index (FI) is a well-established predictor of all-cause mortality, but less is known for cause-specific mortality and whether familial effects influence the associations. Middle-aged individuals are also understudied for the association between FI and mortality. Furthermore, the population mortality impact of frailty remains understudied.

METHODS

We estimated the predictive value of FI for all-cause and cause-specific mortality, taking into account familial factors, and tested whether the associations are time-dependent. We also assessed the proportion of all-cause and cause-specific deaths that are attributable to increased levels of frailty. We analyzed 42,953 participants from the Screening Across the Lifespan Twin Study (aged 41-95 years at baseline) with up to 20 years' mortality follow-up. The FI was constructed using 44 health-related items. Deaths due to cardiovascular disease (CVD), respiratory-related causes, and cancer were considered in the cause-specific analysis. Generalized survival models were used in the analysis.

RESULTS

Increased FI was associated with higher risks of all-cause, CVD, and respiratory-related mortality, with the corresponding hazard ratios of 1.28 (1.24, 1.32), 1.31 (1.23, 1.40), and 1.23 (1.11, 1.38) associated with a 10% increase in FI in male single responders, and 1.21 (1.18, 1.25), 1.27 (1.15, 1.34), and 1.26 (1.15, 1.39) in female single responders. No significant associations were observed for cancer mortality. No attenuation of the mortality associations in unrelated individuals was observed when adjusting for familial effects in twin pairs. The associations were time-dependent with relatively greater effects observed in younger ages. Before the age of 80, the proportions of deaths attributable to FI levels > 0.21 were 18.4% of all-cause deaths, 25.4% of CVD deaths, and 20.4% of respiratory-related deaths in men and 19.2% of all-cause deaths, 27.8% of CVD deaths, and 28.5% of respiratory-related deaths in women. After the age of 80, the attributable proportions decreased, most notably for all-cause and CVD mortality.

CONCLUSIONS

Increased FI predicts higher risks of all-cause, CVD, and respiratory-related mortality independent of familial effects. Increased FI presents a relatively greater risk factor at midlife than in old age. Increased FI has a significant population mortality impact that is greatest through midlife until the age of 80.

Relation Between Infant Microbiota and Autism?: Results from a National Cohort Sibling Design Study

BACKGROUND

Hypotheses concerning adverse effects of changes in microbiota have received much recent attention, but unobserved confounding makes them difficult to test. We investigated whether surrogate markers for potential adverse microbiota change in infancy affected autism risk, addressing unobserved confounding using a sibling study design.

METHODS

This is a population-based, prospective cohort study including all singleton live births in Denmark from 1997 to 2010. The exposure variables were cesarean delivery and antibiotic use in the first 2 years of life. The outcome was a subsequent autism diagnosis. We used the between- and within-sibling model and compared it with sibling-stratified Cox models and simpler standard Cox models that ignored sibship.

RESULTS

Of our study population including 671,606 children, who were followed for up to 15 years (7,341,133 person-years), 72% received antibiotics, 17.5% were delivered by cesarean, and 1.2% (8,267) developed autism. The standard Cox models predicted that both cesarean (compared with vaginal) delivery and antibiotics increased the risk of autism. In the sibling-stratified Cox model, only broader spectrum antibiotics were associated with increased risk of autism: hazard ratio (HR) = 1.16 (95% confidence interval = 1.01, 1.36). The between-within model estimated no exposure effects: intrapartum cesarean HR = 1.06 (0.89, 1.26); prelabor cesarean HR = 0.97 (0.83, 1.15); exclusively penicillin HR = 1.05 (0.93, 1.18); and broader spectrum antibiotics HR = 1.05 (0.95, 1.16).

CONCLUSIONS

The between-within model rendered more precise estimates than sibling-stratified Cox models, and we believe that it also provided more valid estimates. Results from these preferred models do not support a causal relation between antibiotic treatment during infancy, cesarean delivery, and autism. See video abstract at, http://links.lww.com/EDE/B432.

Investigating the effects of cesarean delivery and antibiotic use in early childhood on risk of later attention deficit hyperactivity disorder

BACKGROUND

Increasing attention deficit hyperactivity disorder (ADHD) incidence has been proposed to be caused by factors influencing microbiota in early life. We investigated the potential causality between ADHD and two surrogate markers for changes in children's microbiota: birth delivery mode and early childhood antibiotic use.

METHOD

This population-based, prospective cohort study linked nationwide registers of data for native Danish singleton live births in Denmark from 1997 to 2010. Exposure variables were delivery mode and antibiotic use during the first 2 years of life. The main outcome measure was ADHD diagnosis or redeemed ADHD medication prescriptions. For statistical analysis, we used both advanced sibling models and a more traditional approach.

RESULTS

We included 671,592 children, followed from their second birthday in the period 1999-2014 for 7,300,522 person-years. ADHD was diagnosed in 17,971. In total, 17.5% were born by cesarean delivery, and 72% received antibiotic treatment within their first 2 years of life. In the adjusted between-within sibling survival model, mode of delivery or antibiotics had no effect on ADHD when compared with vaginal delivery or no antibiotic treatment as hazard ratios were 1.09 (95% confidence interval 0.97-1.24) for intrapartum cesarean, 1.03 (0.91-1.16) for prelabor cesarean, 0.98 (0.90-1.07) for penicillin, and 0.99 (0.92-1.06) for broader spectrum antibiotics. In a sibling-stratified Cox regression, intrapartum cesarean was associated with increased ADHD risk, but other exposures were not. In a descriptive, nonstratified Cox model, we found increased risk for ADHD for all exposures.

CONCLUSIONS

Detailed family confounder control using the superior between-within model indicates that cesarean delivery or use of antibiotics during the first 2 years of life does not increase ADHD risk. Therefore, our study suggests that changes in children's microbiota related to cesarean delivery or antibiotic use, do not cause ADHD.

The relationship between subjective well-being and mortality within discordant twin pairs from two independent samples

Prior research has shown robust associations between greater subjective well-being (SWB) and reduced mortality. Whether this observed association is causal in nature or due instead to confounding genetic or environmental factors affecting both SWB and mortality is not well understood. We used a combined sample of 6,802 twins drawn from two cohorts: the Longitudinal Study of Middle-Aged Danish Twins (MADT; N = 2,815, baseline age between 45 and 69 years, M = 56.8, SD = 6.4) and the Longitudinal Study of Aging Danish Twins (LSADT; N = 3,987, baseline age between 70 and 97 years, M = 76.6, SD = 4.9). The relationship between SWB, encompassing measures of life satisfaction and affect, and all-cause mortality was evaluated using survival analyses at both the individual level and within twin pairs. Twin difference analyses were completed within 1,053 monozygotic (MZ) twin pairs and 1,143 dizygotic (DZ) twin pairs to control for genetic and shared environmental confounding. As expected, the individual-level results showed that higher levels of SWB were associated with reduced mortality: affect hazard ratio (HR) = .90, 95% confidence interval (CI) [.87, .94]; life satisfaction HR = .88, 95% CI [.84, .92]. The effect of SWB on reduced mortality remained significant within both MZ and DZ pairs, suggesting that the association is independent of genetic and nonshared environmental confounding factors. These findings, which generalized across both younger (MADT) and older (LSADT) cohorts of adults, remained significant when accounting for demographic factors, physical health, and cognitive functioning. (PsycINFO Database Record

Hazard Ratio Estimators after Terminating Observation within Matched Pairs in Sibling and Propensity Score Matched Designs

Similar to unmatched cohort studies, matched cohort studies may suffer from the censoring of events prior to the end of follow-up. Moreover, in some matched-pair cohort studies, observation time is prematurely terminated immediately after the follow-up of his/her matched member is completed by an event or censoring. Although the follow-up termination within matched pairs may or may not change the hazard ratio estimators, when and how the change occurs has not been clarified. We study the change in the estimates of the hazard ratio conditional on matched pairs and/or covariates by considering two types of matched-pair designs in cohort studies-sibling pair matching and propensity score matching-in which termination can be naturally considered. If all possible confounders are shared within the matched pairs, after termination, a wide range of hazard ratio estimators coincides with that obtained from a stratified Cox model. If unshared confounders should be adjusted for in the analysis, however, such coincidence is not observed. Simulation studies on sibling designs with unshared confounders suggested that the pair-stratified covariate-adjusted Cox model for the hazard ratio conditional on matched pairs and covariates is generally preferred, for which termination does not deteriorate the estimation. Conversely, the comparison between stratifying or not stratifying on pair is a more subtle issue in propensity score matching which targets a marginal or covariate-conditional hazard ratio. Based on simulation studies considering Cox models after matching based on estimated propensity scores, we discourage pair-stratified analysis and termination, particularly after data collection.

Leukocyte Telomere Length and All-Cause Mortality: A Between-Within Twin Study With Time-Dependent Effects Using Generalized Survival Models

Although previous studies examining leukocyte telomere length (LTL) and all-cause mortality controlled for several confounders, the observed association could be biased due to unmeasured confounders, including familial factors. We aimed to examine the association of LTL with all-cause mortality in a Swedish twin sample while adjusting for familial factors and allowing for time-dependent effects. A total of 366 participants (174 twin pairs and 18 individuals) were recruited from the Swedish Twin Registry. LTL was assessed using the Southern blot method. All-cause mortality data were obtained through linkage with the Swedish Population Registry, updated through November 15, 2017. To control for familial factors within twin pairs, we applied a between-within shared frailty model based on generalized survival models. Overall, 115 (31.4%) participants were men and 251 (68.6%) were women. The average age of the study participants when blood was drawn was 79.1 years, and follow-up duration ranged from 10 days to 25.7 years (mean = 10.2 years). During the follow-up period, 341 (93.2%) participants died. Shorter LTL was associated with higher mortality rates when controlling for familial factors in the between-within shared frailty model. We found significant time-dependent effects of LTL on all-cause mortality, where the mortality rate ratios were attenuated with increasing age.

A review and empirical comparison of causal inference methods for clustered observational data with application to the evaluation of the effectiveness of medical devices

Observational studies are commonplace in medicine. A frequent concern is confounding bias due to differences in patient characteristics across treatment groups, but other important issues include dependency among observations nested within clusters (e.g. patients clustered within physicians or surgeons) and confounding due to cluster characteristics (e.g. physician or surgeon experience or training). Given the frequency with which these issues arise in medical research, as well as their relative complexity, methods for the analysis of clustered observational data are reviewed. We argue for estimating causal treatment effects using marginal models that either match or weight observations using a suitable distance metric (e.g. the propensity score). Simulation results demonstrated that methods incorporating clustering into calculation of the variance were generally more accurate than those that did not. Moreover, methods that account for cluster confounding when estimating the treatment effect were least biased and most accurate. Throughout the paper we illustrate the proposed methods in a medical device setting that compares the effectiveness of femoral heads used in total hip replacements. Whenever possible the clustered aspect of the data should be considered in the design of the study when constructing the distance measure or in the matching process, as well as in the analysis when estimating the variance of the treatment effect.

Metabolic and Cardiovascular Complications in Obsessive-Compulsive Disorder: A Total Population, Sibling Comparison Study With Long-Term Follow-up

BACKGROUND

Obsessive-compulsive disorder (OCD) is associated with increased mortality, but the causes of this increase are poorly understood. This study examined whether OCD is associated with an increased risk of metabolic and cardiovascular complications.

METHODS

Individuals diagnosed with OCD (n = 25,415) were identified from a cohort of 12,497,002 individuals living in Sweden between 1973 and 2013. Cox proportional hazard regression analyses were used to investigate the risk of metabolic and cardiovascular complications in OCD patients compared with the general population and unaffected full siblings of OCD individuals. Exploratory analyses were used to examine the effect of treatment with serotonin reuptake inhibitors, with or without antipsychotics, on the outcomes of interest.

RESULTS

Individuals with OCD had a higher risk of any metabolic or cardiovascular complications compared with the general population (adjusted hazard ratio = 1.45; 95% confidence interval = 1.42-1.49) and their unaffected full siblings (adjusted hazard ratio = 1.47; 95% confidence interval = 1.40-1.54). In the fully adjusted sibling comparison models, patients had higher risks of obesity, type 2 diabetes mellitus, and circulatory system diseases. The risks were already evident from the beginning of the follow-up period and remained largely unchanged when excluding different groups of psychiatric comorbidities. Compared with patients who were not taking serotonin reuptake inhibitors, patients taking higher doses of serotonin reuptake inhibitors and who had a longer duration of treatment had significantly lower risks of metabolic and cardiovascular complications, regardless of whether they were also taking antipsychotics.

CONCLUSIONS

OCD is associated with an increased risk of metabolic and cardiovascular complications. Our results underscore the importance of carefully monitoring metabolic and cardiovascular health in patients with OCD early in the course of the disorder.

Meta-analysis of individual registry results enhances international registry collaboration

Background and purpose - Although common in medical research, meta-analysis has not been widely adopted in registry collaborations. A meta-analytic approach in which each registry conducts a standardized analysis on its own data followed by a meta-analysis to calculate a weighted average of the estimates allows collaboration without sharing patient-level data. The value of meta-analysis as an alternative to individual patient data analysis is illustrated in this study by comparing the risk of revision of porous tantalum cups versus other uncemented cups in primary total hip arthroplasties from Sweden, Australia, and a US registry (2003-2015). Patients and methods - For both individual patient data analysis and meta-analysis approaches a Cox proportional hazard model was fit for time to revision, comparing porous tantalum (n = 23,201) with other uncemented cups (n = 128,321). Covariates included age, sex, diagnosis, head size, and stem fixation. In the meta-analysis approach, treatment effect size (i.e., Cox model hazard ratio) was calculated within each registry and a weighted average for the individual registries' estimates was calculated. Results - Patient-level data analysis and meta-analytic approaches yielded the same results with the porous tantalum cups having a higher risk of revision than other uncemented cups (HR (95% CI) 1.6 (1.4-1.7) and HR (95% CI) 1.5 (1.4-1.7), respectively). Adding the US cohort to the meta-analysis led to greater generalizability, increased precision of the treatment effect, and similar findings (HR (95% CI) 1.6 (1.4-1.7)) with increased risk of porous tantalum cups. Interpretation - The meta-analytic technique is a viable option to address privacy, security, and data ownership concerns allowing more expansive registry collaboration, greater generalizability, and increased precision of treatment effects.

Associations between childhood ADHD, gender, and adolescent alcohol and marijuana involvement: A causally informative design

BACKGROUND

We report whether the etiology underlying associations of childhood ADHD with adolescent alcohol and marijuana involvement is consistent with causal relationships or shared predispositions, and whether it differs by gender.

METHODS

In three population-based twin samples (N = 3762; 64% monozygotic), including one oversampling females with ADHD, regressions were conducted with childhood inattentive or hyperactive-impulsive symptoms predicting alcohol and marijuana outcomes by age 17. To determine whether ADHD effects were consistent with causality, twin difference analyses divided effects into those shared between twins in the pair and those differing within pairs.

RESULTS

Adolescents with more severe childhood ADHD were more likely to initiate alcohol and marijuana use earlier, escalate to frequent or heavy use, and develop symptoms. While risks were similar across genders, females with more hyperactivity-impulsivity had higher alcohol consumption and progressed further toward daily marijuana use than did males. Monozygotic twins with more severe ADHD than their co-twins did not differ significantly on alcohol or marijuana outcomes, however, suggesting a non-causal relationship. When co-occurring use of other substances and conduct/oppositional defiant disorders were considered, hyperactivity-impulsivity remained significantly associated with both substances, as did inattention with marijuana, but not alcohol.

CONCLUSIONS

Childhood ADHD predicts when alcohol and marijuana use are initiated and how quickly use escalates. Shared familial environment and genetics, rather than causal influences, primarily account for these associations. Stronger relationships between hyperactivity-impulsivity and heavy drinking/frequent marijuana use among adolescent females than males, as well as the greater salience of inattention for marijuana, merit further investigation.

Food insecurity and child behavior problems in fragile families

Food insecurity remains a persistent problem in the United States. Several studies have shown that food insecurity is associated with child externalizing and internalizing behavior problems. However, some potential methodological limitations remain. For example, most studies use a household measure of food insecurity while there is evidence that children, especially younger ones, tend to be shielded by their parents from experiencing food insecurity. In addition, the mechanisms through which food insecurity affects children are not well understood. This study uses longitudinal data from the Fragile Families and Child Wellbeing Study to address these limitations. Fixed-effects models show that the association is even larger using a measure of child food insecurity instead of a household one. Correlated-random effects models show a large difference in child behavior problems between food secure and food insecure children due to unobserved heterogeneity. In addition, the association between child food insecurity and child externalizing behaviors remains largely unexplained while food insecurity among adults explains almost all the variation in the association with child internalizing behaviors. Food insecure children and parents are at risk of micronutrient deficiencies, which may lead to behavior problems in young children. These findings underscore the need for greater focus on reducing the risk of food insecurity, especially for children in fragile families, in order to reduce behavior problems and improve their educational attainment.

Increased Risk of Smoking in Female Adolescents Who Had Childhood ADHD

OBJECTIVE

This study examined the effects of childhood attention deficit hyperactivity disorder (ADHD) symptoms, both inattention and hyperactivity-impulsivity, on the development of smoking in male and female adolescents.

METHOD

Twin difference methods were used to control for shared genetic and environmental confounders in three population-based, same-sex twin samples (N=3,762; 64% monozygotic). One cohort oversampled female adolescents with ADHD beginning in childhood. Regressions of childhood inattentive and hyperactive-impulsive symptoms were conducted to predict smoking outcomes by age 17. ADHD effects were divided into those shared between twins in the pair and those nonshared, or different within pairs.

RESULTS

Adolescents who had more severe ADHD symptoms as children were more likely to initiate smoking and to start smoking younger. The association of ADHD symptoms with daily smoking, number of cigarettes per day, and nicotine dependence was greater in females than in males. Monozygotic female twins with greater attentional problems than their co-twins had greater nicotine involvement, consistent with possible causal influence. These effects remained when co-occurring externalizing behaviors and stimulant medication were considered. Hyperactivity-impulsivity, while also more strongly related to smoking for female adolescents, appeared primarily noncausal.

CONCLUSIONS

Smoking initiation and escalation are affected differentially by ADHD subtype and gender. The association of inattention with smoking in female adolescents may be causal, whereas hyperactivity-impulsivity appears to act indirectly, through shared propensities for both ADHD and smoking.

Regression standardization and attributable fraction estimation with between-within frailty models for clustered survival data

The between-within frailty model has been proposed as a viable analysis tool for clustered survival time outcomes. Previous research has shown that this model gives consistent estimates of the exposure–outcome hazard ratio in the presence of unmeasured cluster-constant confounding, which the ordinary frailty model does not, and that estimates obtained from the between-within frailty model are often more efficient than estimates obtained from the stratified Cox proportional hazards model. In this paper, we derive novel estimation techniques for regression standardization with between-within frailty models. We also show how between-within frailty models can be used to estimate the attributable fraction function, which is a generalization of the attributable fraction for survival time outcomes. We illustrate the proposed methods by analyzing a large cohort on preterm birth and attention deficit hyperactivity disorder. To facilitate use of the proposed methods, we provide R code for all analyses.

Neighborhood Influences on Violent Reoffending Risk in Released Prisoners Diagnosed With Psychotic Disorders

Released prisoners diagnosed with psychotic disorders have elevated rates of violent reoffending risk and their exposure to adverse neighborhood environments may contribute to this risk. We identified all released sentenced prisoners in Sweden between 2003 and 2013 (n = 47226) and followed them up for a median period of 4.4 years. We identified prisoners who had ever been diagnosed with a psychotic disorder (n = 3782) or prescribed antipsychotics (n = 7366). We examined 3 neighborhood characteristics: income, proportion of welfare recipients, and crime rate. By fitting generalized mixed-effects and negative binomial regression models and adopting within-individual designs that controlled for all time-invariant unmeasured confounders within each individual, we estimated neighborhood intraclass correlations (ICCs) and associations between specific neighborhood characteristics and violent reoffending. Neighborhood factors explained 13.5% (95% CI: 10.9%; 16.6%) of the violent reoffending risk among released prisoners diagnosed with psychotic disorders. This contrasted with 4.3% (95% CI: 3.7%; 4.9%) in all released prisoners. However, after controlling for unmeasured confounding, these estimates were not statistically significant (ICCpsychotic disorders = 0.9%; 95% CI: -0.8%; 2.3%; ICCall prisoners = 0.3%; 95% CI: -0.02%; 0.6%). Similarly, none of the within-individual correlations between the specific neighborhood factors and violent reoffending were significantly different from zero. We found consistent results when we investigated prisoners with other psychiatric and substance use disorders. These findings suggest that placing released prisoners with psychotic disorders in less deprived neighborhoods might not reduce their violent reoffending risk, which may also apply to other psychiatric disorders. The assessment, treatment, and community linkage of high-risk prisoners as a strategy to reduce reoffending needs further research.

Evaluating center performance in the competing risks setting: Application to outcomes of wait-listed end-stage renal disease patients

It is often of interest to compare centers or healthcare providers on quality of care delivered. We consider the setting where evaluation of center performance on multiple competing events is of interest. We propose estimating center effects through cause-specific proportional hazards frailty models that allow correlation among a center's cause-specific effects. Estimation of our model proceeds via penalized partial likelihood and is implemented in R. To evaluate center performance, we also propose a directly standardized excess cumulative incidence (ECI) measure. Therefore, based on our proposed methods, practitioners can evaluate centers either through the cause-specific hazards or the cumulative incidence functions. We demonstrate, through simulations, the advantages of the proposed methods to detect outlying centers, by comparing the proposed methods and existing methods which assume uncorrelated random center effects. In addition, we develop a Correlation Score Test to test the null hypothesis that the competing event processes within a center are correlated. Using data from the Scientific Registry of Transplant Recipients, we apply our method to evaluate the performance of Organ Procurement Organizations on two competing risks: (i) receipt of a kidney transplant and (ii) death on the wait-list.

Is There a Difference in Revision Risk Between Metal and Ceramic Heads on Highly Crosslinked Polyethylene Liners?

BACKGROUND

The most common bearing surface used among primary THAs worldwide is a metal or ceramic femoral head that articulates against a highly crosslinked ultrahigh-molecular-weight polyethylene (HXLPE) acetabular liner. Despite their widespread use, relatively little is known about the comparative effectiveness of ceramic versus metal femoral heads with respect to risk of revision and dislocation as well as the role of head size in this relationship.

QUESTIONS/PURPOSES

The purpose of this study was to evaluate the risk of (1) all-cause revision in metal versus ceramic femoral heads when used with an HXLPE liner, including an evaluation of the effect of head size; and (2) dislocation in metal versus ceramic femoral heads when used with an HXLPE liner as well as an assessment of the effect of head size.

METHODS

Data were collected as part of the Kaiser Permanente Total Joint Replacement Registry between 2001 and 2013. Patients in this study were on average overweight (body mass index = 29 kg/m²), 67 years old, mostly female (57%), and had osteoarthritis (93%) as the primary indication for surgery. The material of the femoral head (metal, ceramic) was crossed with head size (< 32, 32, 36, > 36 mm), yielding eight device groupings. Only uncemented devices were evaluated. The primary outcome was all-cause revision (n = 28,772) and the secondary outcome was dislocation within 1 year (n = 19,623). Propensity scores were used to adjust for potential confounding at the implant/patient level using between-within semiparametric survival models that control for surgeon and hospital confounding and adjust estimates for the within-cluster correlation among observations on the response.

RESULTS

For all-cause revision, there was no difference between ceramic versus metal (reference) heads in combination with an HXLPE liner (hazard ratio [HR] = 0.82 [0.65-1.04], p = 0.099). Smaller metal head sizes of < 32 mm were associated with increased risk of revision relative to 36 mm (HR = 1.66 [1.20-2.31], p = 0.002, adjusted p = 0.025). For dislocation, ceramic heads increased risk relative to metal at < 32 mm only (HR = 4.39 [1.72-11.19], p = 0.002, adjusted p = 0.020). Head sizes < 32 mm were associated with increased risk of dislocation relative to 36 mm for metal (HR = 2.99 [1.40-6.39], p = 0.005, adjusted p = 0.047) and ceramic heads (HR = 15.69 [6.07-40.55], p < 0.001, adjusted p < 0.001).

CONCLUSIONS

The results did not provide evidence for use of one femoral head material over another when used with HXLPE liners for the outcome of revision, but for dislocation, metal performed better than ceramic with < 32-mm heads. Overall, the findings suggest increased risk of revision/dislocation with head sizes < 32 mm.

LEVEL OF EVIDENCE

Level III, therapeutic study.

Between-within effects in survival models with cross-classified clustering: Application to the evaluation of the effectiveness of medical devices

In many medical applications involving observational survival data there will be a cross-classification of doctors and hospitals, as well as an interest in controlling for potentially confounding doctor and hospital effects when evaluating the effectiveness of a medical intervention. In this paper, we propose the use of a between-within model with cross-classified random effects and show through simulation that it performs better than alternative models. A real data example illustrates the application of the proposed model in a study of the survival of hip implants. The proposed model has broad utility in determining the effectiveness of medical interventions.

Bilateral Simultaneous vs Staged Total Knee Arthroplasty: A Comparison of Complications and Mortality

BACKGROUND

The purpose of this study was to compare the complications and mortality between bilateral simultaneous total knee arthroplasty (BTKA-Simultaneous) and bilateral staged TKA (BTK-Staged) while adjusting for differences in patient, surgeon, and hospital characteristics.

METHODS

An integrated health care system total joint registry was used to compare patients undergoing BTKA-Simultaneous to BTKA-Staged. For outcomes related to revision and infection, the sample included 11,118 patients, and for outcomes of death, acute myocardial infarction, stroke, and venous thromboembolism, a subsample of 7991 patients with comorbidity data was selected.

RESULTS

Overall death and complications in both groups were rare. The complication rates for BTKA-Simultaneous and BTKA-Staged were comparable: aseptic revision (1.17% vs 0.9%), septic revision/deep infection (0.8% vs 0.7%), death (0.28% vs 0.1%), and adverse events (2.49% vs 1.97%). In the adjusted models, there were no significant differences in any of the outcomes between the 2 groups.

CONCLUSION

There is a lack of evidence to support superiority of either BTKA-Simultaneous or BTKA-Staged.

Urban density and the metabolic reach of metropolitan areas: A panel analysis of per capita transportation emissions at the county-level

We engage a tension in the urban environment literature that positions cities as both drivers of environmental destruction and loci of environmental protection. We argue that the traditional binary view of cities as either harmful or beneficial is too simplistic; we advance a more nuanced understanding of cities to study their internal and external metabolic effects in terms of carbon emissions from on-road transportation at the county-level across the continental United States between 2002 and 2007. First, utilizing satellite imagery from the National Land Cover Database, we create a novel measure of population density by quantifying the number of people per square mile of impervious surface area. Second, we develop a measure of metropolitan adjacency from the rural classifications datasets published by the USDA. In spatial regression models, we find that while higher density reduces emissions, counties that are geographically isolated from metropolitan areas actually have lower per capita emissions, all else equal. We elaborate on the conceptual, methodological, and practical implications of our study in the conclusion.

Antibiotic cement was associated with half the risk of re-revision in 1,154 aseptic revision total knee arthroplasties

BACKGROUND AND PURPOSE

Aseptic revisions comprise 80% of revision total knee arthroplasties (TKAs). We determined the incidence of re-revision TKA, the reasons for re-revision, and risk factors associated with these procedures.

PATIENTS AND METHODS

We conducted a retrospective cohort study of 1,154 patients who underwent aseptic revision TKA between 2002 and 2013 and were followed prospectively by a total joint replacement registry in the USA. Revision was defined as any operation in which an implanted component was replaced. Patient-, surgeon-, and procedure-related risk factors were evaluated. Survival analyses were conducted.

RESULTS

There were 114 re-revisions (10%) with a median time to reoperation of 3.6 years (interquartile range (IQR): 2.6-5.2). The infection rate was 2.9% (34/1,154) and accounted for 30% of re-revisions (34 of 114). In adjusted models, use of antibiotic-loaded cement was associated with a 50% lower risk of all-cause re-revision surgery (hazard ratio (HR) = 0.5, 95% CI: 0.3-0.9), age with a 20% lower risk for every 10-year increase (HR = 0.8, CI: 0.7-1.0), body mass index (BMI) with a 20% lower risk for every 5-unit increase (HR = 0.8, CI: 0.7-1.0), and a surgeon's greater cumulative experience (≥ 20 cases vs. < 20 cases) with a 3 times higher risk of re-revision (HR = 2.8, CI: 1.5-5).

INTERPRETATION

Revised TKAs were at high risk of subsequent failure. The use of antibiotic-loaded cement, higher age, and higher BMI were associated with lower risk of further revision whereas a higher degree of surgeon experience was associated with higher risk.

Permutation tests for centre effect on survival endpoints with application in an acute myeloid leukaemia multicentre study

When analysing multicentre data, it may be of interest to test whether the distribution of the endpoint varies among centres. In a mixed-effect model, testing for such a centre effect consists in testing to zero a random centre effect variance component. It has been shown that the usual asymptotic χ(2) distribution of the likelihood ratio and score statistics under the null does not necessarily hold. In the case of censored data, mixed-effects Cox models have been used to account for random effects, but few works have concentrated on testing to zero the variance component of the random effects. We propose a permutation test, using random permutation of the cluster indices, to test for a centre effect in multilevel censored data. Results from a simulation study indicate that the permutation tests have correct type I error rates, contrary to standard likelihood ratio tests, and are more powerful. The proposed tests are illustrated using data of a multicentre clinical trial of induction therapy in acute myeloid leukaemia patients.