Missing data occur frequently in empirical studies in health and social sciences, often compromising our ability to make accurate inferences. An outcome is said to be missing not at random (MNAR) if, conditional on the observed variables, the missing data mechanism still depends on the unobserved outcome. In such settings, identification is generally not possible without imposing additional assumptions. Identification is sometimes possible, however, if an instrumental variable (IV) is observed for all subjects which satisfies the exclusion restriction that the IV affects the missingness process without directly influencing the outcome. In this paper, we provide necessary and sufficient conditions for nonparametric identification of the full data distribution under MNAR with the aid of an IV. In addition, we give sufficient identification conditions that are more straightforward to verify in practice. For inference, we focus on estimation of a population outcome mean, for which we develop a suite of semiparametric estimators that extend methods previously developed for data missing at random. Specifically, we propose inverse probability weighted estimation, outcome regression-based estimation and doubly robust estimation of the mean of an outcome subject to MNAR. For illustration, the methods are used to account for selection bias induced by HIV testing refusal in the evaluation of HIV seroprevalence in Mochudi, Botswana, using interviewer characteristics such as gender, age and years of experience as IVs.

In observational studies, treatments are typically not randomized and therefore estimated treatment effects may be subject to confounding bias. The instrumental variable (IV) design plays the role of a quasi-experimental handle since the IV is associated with the treatment and only affects the outcome through the treatment. In this paper, we present a novel framework for identification and inference using an IV for the marginal average treatment effect amongst the treated (ETT) in the presence of unmeasured confounding. For inference, we propose three different semiparametric approaches: (i) inverse probability weighting (IPW), (ii) outcome regression (OR), and (iii) doubly robust (DR) estimation, which is consistent if either (i) or (ii) is consistent, but not necessarily both. A closed-form locally semiparametric efficient estimator is obtained in the simple case of binary IV and outcome and the efficiency bound is derived for the more general case.

Because Helicobacter pylori persist for decades in the human stomach, the aim of this study was to examine the long-term course of H. pylori-specific serum immunoglobulin G (IgG) responses with respect to subclass and antigenic target. We studied paired serum samples obtained in 1973 and in 1994 in Vammala, Finland, from 64 healthy H. pylori-positive adults and from other healthy control subjects. H. pylori serum immunoglobulin A, IgG, and IgG subclass responses were determined by antigen-specific enzyme-linked immunosorbent assays. H. pylori-specific IgG1 and IgG4 subtype responses from 47 subjects were similar in 1973 and 1994, but not when compared with unrelated persons. H. pylori-specific IgG1:IgG4 ratios among the participants varied >1000-fold; however, 57 (89.1%) of 64 subjects had an IgG1:IgG4 ratio >1.0, consistent with a predominant IgG1 (Th1) response. Furthermore, ratios in individual hosts were stable over the 21-year period (r = 0.56; P < .001). The immune response to heat shock protein HspA was unchanged in 49 (77%) of the 64 subjects tested; of the 15 whose serostatus changed, all seroconverted and were significantly younger than those whose status did not change. These findings indicate that H. pylori-specific antibody responses are host-specific with IgG1:IgG4 ratios stable over 21 years, IgG1 responses predominating, and HspA seroconversion with aging.

Genetic factors explain a major proportion of human height variation, but differences in mean stature have also been found between socioeconomic categories suggesting a possible effect of environment. By utilizing a classical twin design which allows decomposing the variation of height into genetic and environmental components, we tested the hypothesis that environmental variation in height is greater in offspring of lower educated parents. Twin data from 29 cohorts including 65,978 complete twin pairs with information on height at ages 1 to 69 years and on parental education were pooled allowing the analyses at different ages and in three geographic-cultural regions (Europe, North America and Australia, and East Asia). Parental education mostly showed a positive association with offspring height, with significant associations in mid-childhood and from adolescence onwards. In variance decomposition modeling, the genetic and environmental variance components of height did not show a consistent relation to parental education. A random-effects meta-regression analysis of the aggregate-level data showed a trend towards greater shared environmental variation of height in low parental education families. In conclusion, in our very large dataset from twin cohorts around the globe, these results provide only weak evidence for the study hypothesis.

Shorter gestational age (GA) is a risk factor of developmental delay. GA is usually estimated clinically from last menstrual period and ultrasound. DNA methylation (DNAm) estimates GA using sets of cytosine-guanine-sites coupled with a clock algorithm. Therefore, DNAm-estimated GA may better ref lect biological maturation. A DNAm GA greater than clinical GA, known as gestational age acceleration (GAA), may indicate epigenetic maturity and holds potential as an early biomarker for developmental delay risk. We used data from the Upstate KIDS Study to examine associations of DNAm GA and developmental delay within the first 3 years based on the Ages & Stages Questionnaire ® (n = 1010). We estimated DNAm GA using two clocks specific to the Illumina Methylation EPIC 850K, the Haftorn clock and one developed from the Effects of Aspirin in Gestation and Reproduction study, in which women were followed to detect pregnancy at the earliest time possible. Among singletons, each week increase in DNAm GA was protective for overall delay (odds ratio:0.74; 95% confidence interval:0.61-0.90) and delay in all domains except for problem-solving skills. Among twins, we observed similar point estimates but lower precision. Results were similar for clinical GA. GAA was largely not associated with developmental delays. In summary, either DNAm GA or clinical GA at birth, but not epigenetic maturity (i.e. GAA), was associated with decreased odds of developmental delay in early childhood. Our study does not support using DNAm GA or GAA as separate risk factors for future risk of developmental delay within the first 3 years of age.