The bidirectional association between premenstrual disorders and perinatal depression: A nationwide register-based study from Sweden.

BACKGROUND: Premenstrual disorders (PMDs) and perinatal depression (PND) share symptomology and the timing of symptoms of both conditions coincide with natural hormonal fluctuations, which may indicate a shared etiology. Yet, there is a notable absence of prospective data on the potential bidirectional association between these conditions, which is crucial for guiding clinical management. Using the Swedish nationwide registers with prospectively collected data, we aimed to investigate the bidirectional association between PMDs and PND. METHODS AND FINDINGS: With 1,803,309 singleton pregnancies of 1,041,419 women recorded in the Swedish Medical Birth Register during 2001 to 2018, we conducted a nested case-control study to examine the risk of PND following PMDs, which is equivalent to a cohort study, and transitioned that design into a matched cohort study with onward follow-up to simulate a prospective study design and examine the risk of PMDs after PND (within the same study population). Incident PND and PMDs were identified through clinical diagnoses or prescribed medications. We randomly selected 10 pregnant women without PND, individually matched to each PND case on maternal age and calendar year using incidence density sampling (N: 84,949: 849,482). We (1) calculated odds ratio (OR) and 95% confidence intervals (CIs) of PMDs using conditional logistic regression in the nested case-control study. Demographic factors (country of birth, educational level, region of residency, and cohabitation status) were adjusted for. We (2) calculated the hazard ratio (HR) and 95% CIs of PMDs subsequent to PND using stratified Cox regression in the matched cohort study. Smoking, BMI, parity, and history of psychiatric disorders were further controlled for, in addition to demographic factors. Pregnancies from full sisters of PND cases were identified for sibling comparison, which contrasts the risk within each set of full sisters discordant on PND. In the nested case-control study, we identified 2,488 PMDs (2.9%) before pregnancy among women with PND and 5,199 (0.6%) among controls. PMDs were associated with a higher risk of subsequent PND (OR 4.76, 95% CI [4.52,5.01]; p < 0.001). In the matched cohort with a mean follow-up of 7.40 years, we identified 4,227 newly diagnosed PMDs among women with PND (incidence rate (IR) 7.6/1,000 person-years) and 21,326 among controls (IR 3.8). Compared to their matched controls, women with PND were at higher risk of subsequent PMDs (HR 1.81, 95% CI [1.74,1.88]; p < 0.001). The bidirectional association was noted for both prenatal and postnatal depression and was stronger among women without history of psychiatric disorders (p for interaction < 0.001). Sibling comparison showed somewhat attenuated, yet statistically significant, bidirectional associations. The main limitation of this study was that our findings, based on clinical diagnoses recorded in registers, may not generalize well to women with mild PMDs or PND. CONCLUSIONS: In this study, we observed a bidirectional association between PMDs and PND. These findings suggest that a history of PMDs can inform PND susceptibility and vice versa and lend support to the shared etiology between both disorders.

Risk of Kidney Failure in Patients With Inflammatory Bowel Disease Undergoing Colectomy: A Nationwide Cohort Study.

BACKGROUND & AIMS: Inflammatory bowel disease (IBD) is frequently accompanied by kidney complications. Potential triggers or subpopulations at high-risk of kidney problems are not well-elucidated. We hypothesized that surgical interventions, specifically colectomy, might in part explain this risk. METHODS: This study was a nationwide Swedish cohort study comprising 82,051 individuals with biopsy-proven IBD diagnosed during 1965 to 2017, with follow-up until 2019. We investigated the association between incident colectomy (time-varying exposure) and future risk of acute kidney injury (AKI) and kidney failure (diagnosis of end-stage kidney disease or death due to chronic kidney disease) using Cox proportional hazard models. We also examined the impact of partial vs total colectomy and the presence/duration of a stoma. Covariates included demographics, education level, and selected comorbidities. RESULTS: Over a median follow-up of 14 years, 16,479 individuals underwent colectomy, and 2556 AKI and 1146 kidney failure events occurred. Colectomy was associated with an increased relative risk of both AKI (adjusted hazard ratio, 2.37; 95% confidence interval, 2.17-2.58) and kidney failure (adjusted hazard ratio, 1.54; 95% confidence interval, 1.34-1.76). Compared with pre-colectomy periods, undergoing total colectomy and colectomy with prolonged stoma showed higher risks of both kidney outcomes versus partial colectomy or colectomy with a temporary stoma, respectively. Subgroup analyses suggested higher risks in patients with ulcerative colitis. CONCLUSIONS: In people with IBD, rates of AKI and kidney failure are higher among those undergoing colectomy, particularly among those following total colectomy, or colectomy with a prolonged stoma. This study identifies a high-risk population that may benefit from established protocols for kidney function monitoring/surveillance and referral to nephrologist care.

Cross-direct effects in settings with two mediators.

When multiple mediators are present, there are additional effects that may be of interest beyond the well-known natural (NDE) and controlled direct effects (CDE). These effects cross the type of control on the mediators, setting one to a constant level and one to its natural level, which differs across subjects. We introduce five such estimands for the cross-CDE and -NDE when two mediators are measured. We consider both the scenario where one mediator is influenced by the other, referred to as sequential mediators, and the scenario where the mediators do not influence each other. Such estimands may be of interest in immunology, as we discuss in relation to measured immunological responses to SARS-CoV-2 vaccination. We provide identifying expressions for the estimands in observational settings where there is no residual confounding, and where intervention, outcome, and mediators are of arbitrary type. We further provide tight symbolic bounds for the estimands in randomized settings where there may be residual confounding of the outcome and mediator relationship and all measured variables are binary.

Absolute and Relative Risks of Kidney and Urological Complications in Patients With Inflammatory Bowel Disease.

INTRODUCTION: The burden of kidney and urological complications in patients with inflammatory bowel disease (IBD) remains poorly characterized. METHODS: We analyzed association between developing IBD (as a time-varying exposure) and relative risks of receiving diagnoses of chronic kidney disease (CKD), acute kidney injury (AKI), or kidney stones, and experiencing a clinically-relevant decline in estimated glomerular filtration rate (eGFR) (CKD progression; composite of kidney failure or an eGFR decline ≥30%) in 1,682,795 individuals seeking healthcare in Stockholm, Sweden, during 2006-2018. We quantified 5- and 10-year absolute risks of these complications in a parallel matched cohort of IBD cases and random controls matched (1:5) on sex, age, and eGFR. RESULTS: During median 9 years, 10,117 participants developed IBD. Incident IBD was associated with higher risks of kidney-related complications compared with non-IBD periods: hazard ratio (HR) (95% confidence interval) was 1.24 (1.10-1.40) for receiving a CKD diagnosis and 1.11 (1.00-1.24) for CKD progression. For absolute risks, 11.8% IBD cases had a CKD event within 10-year. Of these, 6.4% received a CKD diagnosis, and 7.9% reached CKD progression. The risks of AKI (HR 1.97 [1.70-2.29]; 10-year absolute risk 3.6%) and kidney stones (HR 1.69 [1.48-1.93]; 10-year absolute risk 5.6%) were also elevated. Risks were similar in Crohn's disease and ulcerative colitis. DISCUSSION: More than 10% of patients with IBD developed CKD within 10-year from diagnosis, with many not being identified through diagnostic codes. This, together with their elevated AKI and kidney stone risks, highlights the need of established protocols for kidney function monitoring and referral to nephrological/urological care for patients with IBD.

Intensity of and adherence to lipid-lowering therapy as predictors of goal attainment and major adverse cardiovascular events in primary prevention.

BACKGROUND: The effectiveness of lipid-lowering therapy (LLT) for primary prevention of atherosclerotic cardiovascular disease (ASCVD) in routine care may depend on treatment intensity and adherence. METHODS: Observational study of adults with newly initiated LLT for primary prevention of ASCVD in Stockholm, Sweden, during 2017-2021. Study exposures were LLT adherence [proportion of days covered (PDC)], LLT intensity (expected reduction of LDL cholesterol), and the combined measure of adherence and intensity. At each LLT fill, adherence and intensity were calculated during the previous 12 months, and the patients estimated ASCVD risk was categorized. Study outcomes were major adverse cardiovascular events (MACE) and LDL-C goal attainment. RESULTS: Thirty-six thousand two hundred eighty-three individuals (mean age 63 years, 47% women, median follow-up 2 years), with a baseline low-moderate (40%), high (49%), and very-high (11%) ASCVD risk started LLT. Increases in LLT adherence, intensity, or adherence-adjusted intensity of 10% over 1 year were associated with lower risks of MACE (with hazard ratios of 0.95 [95% CI, 0.93-0.98]; 0.93 [0.86-1.00]; and 0.90 [0.85-0.95], respectively) and higher odds of attaining LDL goals. Patients with good adherence (≥80%) had similar risks of MACE and similar odds ratios for LDL-C goal attainment with low-moderate and high-intensity LLT. Treatment discontinuation was associated with increased MACE risk. The relative and absolute benefits of good adherence were greatest in patients with very high ASCVD risk. CONCLUSION: In routine-care primary prevention, better adherence to LLT was associated with a lower risk of MACE across all treatment intensities. Improving adherence is especially important among patients with very high ASCVD risk.

ß2-Adrenoreceptor Agonists, Montelukast, and Parkinson Disease Risk.

OBJECTIVE: This study was undertaken to examine the association between montelukast use, ß2-adrenoreceptor (ß2AR) agonist use, and later Parkinson disease (PD). METHODS: We ascertained use of ß2AR agonists (430,885 individuals) and montelukast (23,315 individuals) from July 1, 2005 to June 30, 2007, and followed 5,186,886 PD-free individuals from July 1, 2007 to December 31, 2013 for incident PD diagnosis. We estimated hazard ratios and 95% confidence intervals using Cox regressions. RESULTS: We observed 16,383 PD cases during on average 6.1 years of follow-up. Overall, use of ß2AR agonists and montelukast were not related to PD incidence. A 38% lower PD incidence was noted among high-dose montelukast users when restricted to PD registered as the primary diagnosis. INTERPRETATION: Overall, our data do not support inverse associations between ß2AR agonists, montelukast, and PD. The prospect of lower PD incidence with high-dose montelukast exposure warrants further investigation, especially with adjustment for high-quality data on smoking. ANN NEUROL 2023;93:1023-1028.

Childhood and adolescence outcomes in offspring to parents with bipolar disorder: the impact of lifetime parental comorbidity, parental sex, and bipolar subtype.

BACKGROUND: Offspring of parents with bipolar disorder have increased risks of their own psychopathology. However, a large-scale survey of psychiatric, somatic, and adverse social outcomes up to adulthood, which could aid in prioritizing and tailoring prevention, is lacking. It also remains to clarify how risks are modified by other parental factors. METHODS: Swedish population registers were linked to compare offspring having (N = 24,788) and not having (N = 247,880) a parent with bipolar disorder with respect to psychiatric diagnoses and psychotropic medication, birth-related and somatic conditions, social outcomes, accidents, suicide attempts, and mortality. Individuals were followed until age 18. We estimated the influence of lifetime parental psychiatric comorbidity, bipolar disorder subtype, and sex on outcomes. RESULTS: Children of parents with bipolar disorder had 2-3 times higher risks of all psychiatric diagnoses, except for bipolar disorder, for which the risk was 11-fold. Significantly increased risks were also found for several somatic conditions, low school grades, criminal behavior, victimization, accidents, and suicidal behavior. Adjusting for lifetime parental psychiatric comorbidity attenuated most associations. Offspring of a parent with bipolar disorder type 2 had statistically significantly higher risks of attention deficit hyperactivity disorder, respiratory tract conditions, and accidents compared with offspring of a parent with bipolar disorder type 1. Offspring of mothers with bipolar disorder had higher risks of several psychiatric diagnoses, respiratory tract conditions, low school grades, and accidents compared with offspring of fathers with bipolar disorder. Having two parents with bipolar disorder entailed the highest risks of psychiatric outcomes in offspring. CONCLUSIONS: Early intervention and family support are particularly warranted for the offspring of a parent with bipolar disorder in the presence of lifetime parental psychiatric comorbidity, when the parent has bipolar disorder type 2, or when the mother or both parents have bipolar disorder.

Combining Mendelian randomization with the sibling comparison design.

Mendelian randomization (MR) is a popular epidemiologic study design that uses genetic variants as instrumental variables (IVs) to estimate causal effects, while accounting for unmeasured confounding. The validity of the MR design hinges on certain IV assumptions, which may sometimes be violated due to dynastic effects, population stratification, or assortative mating. Since these mechanisms act through parental factors it was recently suggested that the bias resulting from violations of the IV assumptions can be reduced by combing the MR design with the sibling comparison design, which implicitly controls for all factors that are constant within families. In this article, we provide a formal discussion of this combined MR-sibling design. We derive conditions under which the MR-sibling design is unbiased, and we relate these to the corresponding conditions for the standard MR and sibling comparison designs. We proceed by considering scenarios where all three designs are biased to some extent, and discuss under which conditions the MR-sibling design can be expected to have less bias than the other two designs. We finally illustrate the theoretical results and conclusions with an application to real data, in a study of low-density lipoprotein and diastolic blood pressure using data from the Swedish Twin Registry.

Inverse probability of treatment weighting with generalized linear outcome models for doubly robust estimation.

There are now many options for doubly robust estimation; however, there is a concerning trend in the applied literature to believe that the combination of a propensity score and an adjusted outcome model automatically results in a doubly robust estimator and/or to misuse more complex established doubly robust estimators. A simple alternative, canonical link generalized linear models (GLM) fit via inverse probability of treatment (propensity score) weighted maximum likelihood estimation followed by standardization (the g $$ g $$ -formula) for the average causal effect, is a doubly robust estimation method. Our aim is for the reader not just to be able to use this method, which we refer to as IPTW GLM, for doubly robust estimation, but to fully understand why it has the doubly robust property. For this reason, we define clearly, and in multiple ways, all concepts needed to understand the method and why it is doubly robust. In addition, we want to make very clear that the mere combination of propensity score weighting and an adjusted outcome model does not generally result in a doubly robust estimator. Finally, we hope to dispel the misconception that one can adjust for residual confounding remaining after propensity score weighting by adjusting in the outcome model for what remains 'unbalanced' even when using doubly robust estimators. We provide R code for our simulations and real open-source data examples that can be followed step-by-step to use and hopefully understand the IPTW GLM method. We also compare to a much better-known but still simple doubly robust estimator.

Confirmatory prediction-driven RCTs in comparative effectiveness settings for cancer treatment.

BACKGROUND: Medical advances in the treatment of cancer have allowed the development of multiple approved treatments and prognostic and predictive biomarkers for many types of cancer. Identifying improved treatment strategies among approved treatment options, the study of which is termed comparative effectiveness, using predictive biomarkers is becoming more common. RCTs that incorporate predictive biomarkers into the study design, called prediction-driven RCTs, are needed to rigorously evaluate these treatment strategies. Although researched extensively in the experimental treatment setting, literature is lacking in providing guidance about prediction-driven RCTs in the comparative effectiveness setting. METHODS: Realistic simulations with time-to-event endpoints are used to compare contrasts of clinical utility and provide examples of simulated prediction-driven RCTs in the comparative effectiveness setting. RESULTS: Our proposed contrast for clinical utility accurately estimates the true clinical utility in the comparative effectiveness setting while in some scenarios, the contrast used in current literature does not. DISCUSSION: It is important to properly define contrasts of interest according to the treatment setting. Realistic simulations should be used to choose and evaluate the RCT design(s) able to directly estimate that contrast. In the comparative effectiveness setting, our proposed contrast for clinical utility should be used.

Long-term Visit-to-Visit Variability in Hemoglobin A1c and Kidney-Related Outcomes in Persons With Diabetes.

RATIONALE & OBJECTIVE: To characterize associations between long-term visit-to-visit variability of hemoglobin A1c (HbA1c) and risk of adverse kidney outcomes in patients with diabetes. STUDY DESIGN: Observational study. SETTING & PARTICIPANTS: 93,598 adults with diabetes undergoing routine care in Stockholm, Sweden. EXPOSURES AND PREDICTORS: Categories of baseline and time-varying HbA1c variability score (HVS, the percentage of total HbA1c measures that vary by>0.5% [5.5mmol/mol] during a 3-year window): 0-20%, 21%-40%, 41%-60%, 61%-80%, and 81%-100%, with 0-20% as the reference group. OUTCOME: Chronic kidney disease (CKD) progression (composite of>50% estimated glomerular filtration rate [eGFR] decline and kidney failure), acute kidney disease (AKI by clinical diagnosis or transient creatinine elevations according to KDIGO criteria), and worsening of albuminuria. ANALYTICAL APPROACH: Multivariable Cox proportional hazards regression. RESULTS: Compared with persons showing low HbA1c variability (HVS 0-20%), any increase in variability was associated with a higher risk of adverse kidney outcomes beyond mean HbA1c. For example, for patients with a baseline HbA1c variability of 81%-100%, the adjusted HR was 1.6 (95% CI, 1.47-1.74) for CKD progression, 1.23 [1.16-1.3] for AKI, and 1.28 [1.21-1.36] for worsening of albuminuria. The results were consistent across subgroups (diabetes subtypes, baseline eGFR, or albuminuria categories), in time-varying analyses and in sensitivity analyses including time-weighted average HbA1c or alternative metrics of variability. LIMITATIONS: Observational study, limitations of claims data, lack of information on diet, body mass index, medication changes, and diabetes duration. CONCLUSIONS: Higher long-term visit-to-visit HbA1c variability is consistently associated with the risks of CKD progression, AKI, and worsening of albuminuria. PLAIN-LANGUAGE SUMMARY: The evidence for current guideline recommendations derives from clinical trials that focus on a single HbA1c as the definitive measure of efficacy of an intervention. However, long-term visit-to-visit fluctuations of HbA1c may provide additional value in the prediction of future kidney complications. We evaluated the long-term fluctuations in glycemic control in almost 100,000 persons with diabetes undergoing routine care in Stockholm, Sweden. We observed that higher long-term HbA1c fluctuation is consistently associated with the risks of chronic kidney disease progression, worsening of albuminuria and acute kidney injury. This finding supports a role for long-term glycemic variability in the development of kidney complications and illustrates the potential usefulness of this metric for risk stratification at the bedside beyond a single HbA1c test.

Selection Bias with Outcome-dependent Sampling.

In a seminal paper, Hernán et al. 2004 provided a systematic classification of selection biases, for scenarios where the selection is a collider between the exposure and the outcome. Hernán 2017 discussed another scenario, where the selection is statistically independent of the exposure, but associated with the outcome through common causes. In this note, we extend the discussion to scenarios where the selection is directly influenced by the outcome, but not by the exposure. We discuss whether these types of outcome-dependent selections preserve the sharp causal null hypothesis, and whether or not they allow for estimation of causal effects in the selected sample and/or in the source population.

Glucocorticoid exposure and the risk of serious infections in rheumatoid arthritis: a marginal structural model application.

OBJECTIVE: Observational studies have reported an increased risk of infections associated with glucocorticoids in RA, not supported by evidence from randomized controlled trials. Inappropriately accommodating time-varying exposure and confounding in observational studies might explain the conflicting results. Therefore, we compared the incidence of serious infections between different oral glucocorticoid dose patterns over three years in a prospective inception cohort, adjusting for time-varying confounders in marginal structural models. METHODS: We included 9654 newly diagnosed RA patients from the Swedish Rheumatology Quality Register between 2007-2018 and followed them for three years after the first rheumatology visit. Follow-up was divided into 90-day periods. A mean oral prednisone daily dose was calculated for each period and categorized into 'no use', 'low' (≤10 mg/day) and 'high' (>10 mg/day) doses. The incidence of serious infections (hospitalization for infection) over follow-up periods was modelled by pooled logistic regression allowing separate effects for recent and past exposure. RESULTS: An increased incidence of serious infections was associated with higher compared with lower doses and with more recent compared with past glucocorticoid exposure. Over 3 years of follow-up, the marginal structural models predicted one additional serious infection for every 83 individuals treated with low GC doses for the first 6 months, and for every 125 individuals treated with high GC doses for the first 3 months, compared with no GC use. CONCLUSION: Our results broadly agree with previous observational studies showing a dose dependent increased risk of infection associated with (recent) use of oral glucocorticoids.

Novel disease associations with schizophrenia genetic risk revealed in ~400,000 UK Biobank participants.

Schizophrenia is a serious mental disorder with considerable somatic and psychiatric morbidity. It is unclear whether comorbid health conditions predominantly arise due to shared genetic risk or consequent to having schizophrenia. To explore the contribution of genetic risk for schizophrenia, we analysed the effect of schizophrenia polygenic risk scores (PRS) on a broad range of health problems in 406 929 individuals with no schizophrenia diagnosis from the UK Biobank. Diagnoses were derived from linked health data including primary care, hospital inpatient records, and registers with information on cancer and deaths. Schizophrenia PRS were generated and tested for associations with general health conditions, 16 ICD10 main chapters, and 603 diseases using linear and logistic regressions. Higher schizophrenia PRS was significantly associated with poorer overall health ratings, more hospital inpatient diagnoses, and more unique illnesses. It was also significantly positively associated with 4 ICD10 chapters: mental disorders; respiratory diseases; digestive diseases; and pregnancy, childbirth and the puerperium, but negatively associated with musculoskeletal disorders. Thirty-one specific phenotypes were significantly associated with schizophrenia PRS, and the 19 novel findings include several musculoskeletal diseases, respiratory diseases, digestive diseases, varicose veins, pituitary hyperfunction, and other peripheral nerve disorders. These findings extend knowledge of the pleiotropic effect of genetic risk for schizophrenia and offer insight into how some conditions often comorbid with schizophrenia arise. Additional studies incorporating the genetic basis of hormone regulation and involvement of immune mechanisms in the pathophysiology of schizophrenia may further elucidate the biological mechanisms underlying schizophrenia and its comorbid conditions.

Associations of parental and perinatal factors with subsequent risk of stress-related disorders: a nationwide cohort study with sibling comparison.

Little is known about the contribution of pregnancy-related parental and perinatal factors to the development of stress-related disorders. We aimed to investigate whether parental/perinatal adversities entail higher risks of stress-related disorders in the offspring, later in life, by accounting for genetic and early environmental factors. Based on the nationwide Swedish registers, we conducted a population-based cohort study of 3,435,747 singleton births (of which 2,554,235 were full siblings), born 1973-2008 and survived through the age of 5 years. Using both population- and sibling designs, we employed Cox regression to assess the association between parental and perinatal factors with subsequent risk of stress-related disorders. We identified 55,511 individuals diagnosed with stress-related disorders in the population analysis and 37,433 in the sibling analysis. In the population-based analysis we observed increased risks of stress-related disorders among offspring of maternal/paternal age <25, single mothers, parity ≥4, mothers with BMI ≥ 25 or maternal smoking in early pregnancy, gestational diabetes, and offspring born moderately preterm (GA 32-36 weeks), or small-for-gestational-age. These associations were significantly attenuated toward null in the sibling analysis. Cesarean-section was weakly associated with offspring stress-related disorders in population [hazard ratio (HR) 1.09, 95% confidence interval (CI) 1.06-1.12] and sibling analyses (HR 1.10, 95% CI 1.02-1.20). Our findings suggest that most of the observed associations between parental and perinatal factors and risk of stress-related disorders in the population analysis are driven by shared familial environment or genetics, and underscore the importance of family designs in epidemiological studies on the etiology of psychiatric disorders.

Sensitivity analysis of G-estimators to invalid instrumental variables.

Instrumental variables regression is a tool that is commonly used in the analysis of observational data. The instrumental variables are used to make causal inference about the effect of a certain exposure in the presence of unmeasured confounders. A valid instrumental variable is a variable that is associated with the exposure, affects the outcome only through the exposure (exclusion), and is not confounded with the outcome (exogeneity). Unlike the first assumption, the other two are generally untestable and rely on subject-matter knowledge. Therefore, a sensitivity analysis is desirable to assess the impact of assumptions' violation on the estimated parameters. In this paper, we propose and demonstrate a new method of sensitivity analysis for G-estimators in causal linear and non-linear models. We introduce two novel aspects of sensitivity analysis in instrumental variables studies. The first is a single sensitivity parameter that captures violations of exclusion and exogeneity assumptions. The second is an application of the method to non-linear models. The introduced framework is theoretically justified and is illustrated via a simulation study. Finally, we illustrate the method by application to real-world data and provide guidelines on conducting sensitivity analysis.

Target parameters and bias in non-causal change-score analyses with measurement errors.

In studies where the outcome is a change-score, it is often debated whether or not the analysis should adjust for the baseline score. When the aim is to make causal inference, it has been argued that the two analyses (adjusted vs. unadjusted) target different causal parameters, which may both be relevant. However, these arguments are not applicable when the aim is to make predictions rather than to estimate causal effects. When the scores are measured with error, there have been attempts to quantify the bias resulting from adjustment for the (mis-)measured baseline score or lack thereof. However, these bias results have been derived under an unrealistically simple model, and assuming that the target parameter is the unadjusted (for the true baseline score) association, thus dismissing the adjusted association as a possibly relevant target parameter. In this paper we address these limitations. We argue that, even if the aim is to make predictions, there are two possibly relevant target parameters; one adjusted for the baseline score and one unadjusted. We consider both the simple case when there are no measurement errors, and the more complex case when the scores are measured with error. For the latter case, we consider a more realistic model than previous authors. Under this model we derive analytic expressions for the biases that arise when adjusting or not adjusting for the (mis-)measured baseline score, with respect to the two possible target parameters. Finally, we use these expressions to discuss when adjustment is warranted in change-score analyses.

Estimation of marginal causal effects in the presence of confounding by cluster.

A popular way to control for unmeasured confounders is to utilize clusters (e.g. sets of siblings), in which a potentially large set of confounders are constant. By estimating the exposure-outcome association within clusters, rather than between unrelated subjects, all cluster-constant confounders are implicitly controlled for. To analyze such clustered data, it is common to use fixed effects models, which absorb all cluster-constant confounders into a cluster-specific intercept. In this article, we show how linear and log-linear fixed effects models can be used to estimate marginal counterfactual means. These counterfactual means can be estimated and presented for each exposure level separately, or contrasted to form a wide range of marginal causal effects. For binary outcomes, we propose to estimate marginal causal effects with marginal logistic between-within models. These models include a constant intercept common for all clusters, and control for unmeasured cluster-constant confounders by adding the mean exposure level in each cluster to the model. We illustrate the proposed methods by re-analyzing data from a co-twin control study on birth weight and Attention-Deficit/Hyperactivity Disorder.

Use of hormonal contraceptives and antidepressants and risks of suicidal behavior and accidents among women with premenstrual disorders: a nationwide cohort study.

BACKGROUND: Women with premenstrual disorders (PMDs) are at increased risks of suicidal behavior and accidents. However, the effect of PMD first-line treatment on such risks have not been assessed. METHODS: To study the association between use of hormonal contraceptives or antidepressants and subsequent risks of suicidal behavior and accidents among women with PMDs. We conducted a nationwide register-based cohort study with between- and within-individual analyses in Sweden. All women with a clinical diagnosis/indication of PMDs recorded in the Patient Register and the Prescribed Drug Register during 1987-2011 were included (n = 23 029, age 15-52 years). Information on hormonal contraceptives and antidepressants prescribed for these women was obtained from the Prescribed Drug Register. Events of suicidal behavior (complete suicide and suicide attempt) and accidents were separately identified through the Patient and the Causes of Death Registers. Incidence rate ratios (IRRs) and 95% confidence intervals (CIs) of suicidal behavior and accidents after use of hormonal contraceptives or antidepressants were estimated in between-individual and within-individual analyses (i.e., comparing the risk between use and no use in the same individual) using Poisson regression. RESULTS: Women with PMDs were followed for a median of 6.2 years. Compared to no use of hormonal contraceptives, use of hormonal contraceptives was associated with a lower risk of suicidal behavior in both between-individual (IRR 0.76, 0.43-1.34) and within-individual analyses (IRR 0.65, 0.51-0.83). These risk reductions were primarily restricted to combined products (IRR 0.18, 0.07-0.47 and 0.19, 0.08-0.42 in between- and within-individual analyses) and observed among women with/without psychiatric comorbidities (p for interaction 0.830 and 0.043 in between- and within-individual analyses). Yet, the use of hormonal contraceptives was not consistently associated with risk of accidents between between-individual (IRR 1.13, 1.01-1.27) and within-individual analyses (IRR 1.01, 0.92-1.11). Use of antidepressants was associated with a higher risk of suicidal behavior and accidents in both between- and within-individual analyses. CONCLUSIONS: Our findings suggest that use of hormonal contraceptives, particularly combined products, is associated with reduced rates of suicidal behaviors, but not accidents, among women with PMDs. The estimates for antidepressants may be biased by indication.

Association between adverse childhood experiences and premenstrual disorders: a cross-sectional analysis of 11,973 women.

BACKGROUND: Childhood abuse and neglect have been associated with premenstrual disorders (PMDs), including premenstrual syndrome (PMS) and premenstrual dysphoric disorder (PMDD). However, the associations of other adverse childhood experiences (ACEs) and the cumulative number of ACEs with PMDs remain to be explored. METHODS: To evaluate the associations of the cumulative number and types of ACEs with PMDs, we conducted a cross-sectional analysis with a subsample of menstruating women within the Stress-And-Gene-Analysis (SAGA) cohort, assessed for PMDs and ACEs (N=11,973). The cumulative and individual exposure of 13 types of ACEs was evaluated by a modified ACE-International Questionnaire. A modified version of the Premenstrual Symptom Screening Tool was used to identify probable cases of PMDs, further sub-grouped into PMS and PMDD. Prevalence ratios (PRs) of PMDs in relation to varying ACEs were estimated using Poisson regression. RESULTS: At a mean age of 34.0 years (standard deviation (SD) 9.1), 3235 (27%) met the criteria of probable PMDs, including 2501 (21%) for PMS and 734 (6%) for PMDD. The number of ACEs was linearly associated with PMDs (fully-adjusted PR 1.12 per ACE, 95% CI 1.11-1.13). Specifically, the PR for PMDs was 2.46 (95% CI 2.21-2.74) for women with 4 or more ACEs compared with women with no ACEs. A stronger association was observed for probable PMDD compared to PMS (p for difference <0.001). The associations between ACEs and PMDs were stronger among women without PTSD, anxiety, or depression, and without childhood deprivation and were stronger among women a lower level of social support (p for interaction<0.001). All types of ACEs were positively associated with PMDs (PRs ranged from 1.11 to 1.51); the associations of sexual abuse, emotional neglect, family violence, mental illness of a household member, and peer and collective violence were independent of other ACEs. CONCLUSIONS: Our findings suggest that childhood adverse experiences are associated with PMDs in a dose-dependent manner. If confirmed by prospective data, our findings support the importance of early intervention for girls exposed to ACEs to minimize risks of PMDs and other morbidities in adulthood.