Maternal pre-pregnancy body mass index and risk of preterm birth: a collaboration using large routine health datasets

BACKGROUND

Preterm birth (PTB) is a leading cause of child morbidity and mortality. Evidence suggests an increased risk with both maternal underweight and obesity, with some studies suggesting underweight might be a greater factor in spontaneous PTB (SPTB) and that the relationship might vary by parity. Previous studies have largely explored established body mass index (BMI) categories. Our aim was to compare associations of maternal pre-pregnancy BMI with any PTB, SPTB and medically indicated PTB (MPTB) among nulliparous and parous women across populations with differing characteristics, and to identify the optimal BMI with lowest risk for these outcomes.

METHODS

We used three UK datasets, two USA datasets and one each from South Australia, Norway and Denmark, together including just under 29 million pregnancies resulting in a live birth or stillbirth after 24 completed weeks gestation. Fractional polynomial multivariable logistic regression was used to examine the relationship of maternal BMI with any PTB, SPTB and MPTB, among nulliparous and parous women separately. The results were combined using a random effects meta-analysis. The estimated BMI at which risk was lowest was calculated via differentiation and a 95% confidence interval (CI) obtained using bootstrapping.

RESULTS

We found non-linear associations between BMI and all three outcomes, across all datasets. The adjusted risk of any PTB and MPTB was elevated at both low and high BMIs, whereas the risk of SPTB was increased at lower levels of BMI but remained low or increased only slightly with higher BMI. In the meta-analysed data, the lowest risk of any PTB was at a BMI of 22.5 kg/m2 (95% CI 21.5, 23.5) among nulliparous women and 25.9 kg/m2 (95% CI 24.1, 31.7) among multiparous women, with values of 20.4 kg/m2 (20.0, 21.1) and 22.2 kg/m2 (21.1, 24.3), respectively, for MPTB; for SPTB, the risk remained roughly largely constant above a BMI of around 25-30 kg/m2 regardless of parity.

CONCLUSIONS

Consistency of findings across different populations, despite differences between them in terms of the time period covered, the BMI distribution, missing data and control for key confounders, suggests that severe under- and overweight may play a role in PTB risk.

The dynamic interplay between sleep and mood: an intensive longitudinal study of individuals with bipolar disorder

BACKGROUND

Sleep disturbances are important symptoms to monitor in people with bipolar disorder (BD) but the precise longitudinal relationships between sleep and mood remain unclear. We aimed to examine associations between stable and dynamic aspects of sleep and mood in people with BD, and assess individual differences in the strength of these associations.

METHODS

Participants (N = 649) with BD-I (N = 400) and BD-II (N = 249) provided weekly self-reports of insomnia, depression and (hypo)mania symptoms using the True Colours online monitoring tool for 21 months. Dynamic structural equation models were used to examine the interplay between weekly reports of insomnia and mood. The effects of clinical and demographic characteristics on associations were also assessed.

RESULTS

Increased variability in insomnia symptoms was associated with increased mood variability. In the sample as a whole, we found strong evidence of bidirectional relationships between insomnia and depressive symptoms but only weak support for bidirectional relationships between insomnia and (hypo)manic symptoms. We found substantial variability between participants in the strength of prospective associations between insomnia and mood, which depended on age, gender, bipolar subtype, and a history of rapid cycling.

CONCLUSIONS

Our results highlight the importance of monitoring sleep in people with BD. However, researchers and clinicians investigating the association between sleep and mood should consider subgroup differences in this relationship. Advances in digital technology mean that intensive longitudinal data on sleep and mood are becoming increasingly available. Novel methods to analyse these data present an exciting opportunity for furthering our understanding of BD.

Improved two-stage estimation to adjust for treatment switching in randomised trials: g-estimation to address time-dependent confounding

In oncology trials, control group patients often switch onto the experimental treatment during follow-up, usually after disease progression. In this case, an intention-to-treat analysis will not address the policy question of interest - that of whether the new treatment represents an effective and cost-effective use of health care resources, compared to the standard treatment. Rank preserving structural failure time models (RPSFTM), inverse probability of censoring weights (IPCW) and two-stage estimation (TSE) have often been used to adjust for switching to inform treatment reimbursement policy decisions. TSE has been applied using a simple approach (TSEsimp), assuming no time-dependent confounding between the time of disease progression and the time of switch. This is problematic if there is a delay between progression and switch. In this paper we introduce TSEgest, which uses structural nested models and g-estimation to account for time-dependent confounding, and compare it to TSEsimp, RPSFTM and IPCW. We simulated scenarios where control group patients could switch onto the experimental treatment with and without time-dependent confounding being present. We varied switching proportions, treatment effects and censoring proportions. We assessed adjustment methods according to their estimation of control group restricted mean survival times that would have been observed in the absence of switching. All methods performed well in scenarios with no time-dependent confounding. TSEgest and RPSFTM continued to perform well in scenarios with time-dependent confounding, but TSEsimp resulted in substantial bias. IPCW also performed well in scenarios with time-dependent confounding, except when inverse probability weights were high in relation to the size of the group being subjected to weighting, which occurred when there was a combination of modest sample size and high switching proportions. TSEgest represents a useful addition to the collection of methods that may be used to adjust for treatment switching in trials in order to address policy-relevant questions.

DNA methylation from birth to late adolescence and development of multiple-risk behaviours

BACKGROUND

Risk behaviours in adolescence are linked to poor educational attainment and health and other outcomes in young adulthood. We explored whether there are molecular mechanisms associated with the development, or the result, of multiple risk behaviours (MRBs).

METHODS

MRBs (antisocial behaviour and delinquency, traffic-related risk behaviour, risky sexual behaviour, lack of exercise) and their sumscore were characterized based on self-reported questions at age 7 and 17 within the ARIES subsample of the ALSPAC birth cohort, and were linked to DNA methylation at over 485,000 CpG sites at ages 0,7 and 17. Associations were determined for participants with complete data (n = 227-575).

RESULTS

There was weak evidence of associations between cumulative MRBs and methylation at cg01783492 and cg16720578 at age 17. DNA methylation at age 17 was associated with risky sexual behaviour (cg22883332), lack of exercise (cg03152353, cg20056908, cg20571116) and substance use (cg02188400, cg13906377). No associations between DNA methylation and individual risk behaviours at age 7 were observed. DNA methylation at age 7 might predispose for traffic-related risk behaviour (cg24683561) and substance use (cg08761410) at age 17.

LIMITATIONS

Main limitations are absence of information on directly measured blood cell type proportions and tissue specificity, and a modest sample size.

CONCLUSIONS

Cumulative MRB in late adolescence was associated with effects on DNA methylation. More specifically, risky sexual behaviour and sedentary behaviour are associated with changes in DNA methylation, while DNA methylation in childhood may predict later traffic-related risky behaviour. For substance use effects in both temporal directions were observed.

Multiple imputation using linked proxy outcome data resulted in important bias reduction and efficiency gains: a simulation study

Background

When an outcome variable is missing not at random (MNAR: probability of missingness depends on outcome values), estimates of the effect of an exposure on this outcome are often biased. We investigated the extent of this bias and examined whether the bias can be reduced through incorporating proxy outcomes obtained through linkage to administrative data as auxiliary variables in multiple imputation (MI).

Methods

Using data from the Avon Longitudinal Study of Parents and Children (ALSPAC) we estimated the association between breastfeeding and IQ (continuous outcome), incorporating linked attainment data (proxies for IQ) as auxiliary variables in MI models. Simulation studies explored the impact of varying the proportion of missing data (from 20 to 80%), the correlation between the outcome and its proxy (0.1-0.9), the strength of the missing data mechanism, and having a proxy variable that was incomplete.

Results

Incorporating a linked proxy for the missing outcome as an auxiliary variable reduced bias and increased efficiency in all scenarios, even when 80% of the outcome was missing. Using an incomplete proxy was similarly beneficial. High correlations (> 0.5) between the outcome and its proxy substantially reduced the missing information. Consistent with this, ALSPAC analysis showed inclusion of a proxy reduced bias and improved efficiency. Gains with additional proxies were modest.

Conclusions

In longitudinal studies with loss to follow-up, incorporating proxies for this study outcome obtained via linkage to external sources of data as auxiliary variables in MI models can give practically important bias reduction and efficiency gains when the study outcome is MNAR.

Body mass index trajectories from 2 to 18 years - exploring differences between European cohorts

BACKGROUND

In recent decades, there has been an increase in the prevalence of childhood overweight in most high-income countries. Within northern Europe, prevalence tends to be higher in the UK compared with the Scandinavian countries. We aimed to study differences in body mass index (BMI) trajectories between large cohorts of children from UK and Scandinavian populations.

METHODS

We compared BMI trajectories in participants from the English Avon Longitudinal Study of Parents and Children born in 1991-1993 (ALSPAC) (N = 6517), the Northern Finland Birth Cohorts born in 1966 (NFBC1966) (N = 3321) and 1986 (NFBC1986) (N = 4764), and the Danish Aarhus Birth Cohort born in 1990-1992 (ABC) (N = 1920). We used multilevel models to estimate BMI trajectories from 2 to 18 years. We explored whether cohort differences were explained by maternal BMI, height, education or smoking during pregnancy and whether differences were attributable to changes in the degree of skew in the BMI distribution.

RESULTS

Differences in mean BMI between the cohorts were small but emerged early and persisted in most cases across childhood. Girls in ALSPAC had a higher BMI than all other cohorts throughout childhood, e.g. compared with the NFBC1986 BMI was 2.2-3.5% higher. For boys, the difference emerging over time (comparing the two NFBC's) exceeded the differences across populations (comparing NFBC1986, ABC and ALSPAC). BMI distribution demonstrated increasing right skew with age.

CONCLUSION

Population-level differences between cohorts were small, tended to emerge very early, persisted across childhood, and demonstrated an increase in the right-hand tail of the BMI distribution.

Comparison of imputation variance estimators

Appropriate imputation inference requires both an unbiased imputation estimator and an unbiased variance estimator. The commonly used variance estimator, proposed by Rubin, can be biased when the imputation and analysis models are misspecified and/or incompatible. Robins and Wang proposed an alternative approach, which allows for such misspecification and incompatibility, but it is considerably more complex. It is unknown whether in practice Robins and Wang's multiple imputation procedure is an improvement over Rubin's multiple imputation. We conducted a critical review of these two multiple imputation approaches, a re-sampling method called full mechanism bootstrapping and our modified Rubin's multiple imputation procedure via simulations and an application to data. We explored four common scenarios of misspecification and incompatibility. In general, for a moderate sample size (n = 1000), Robins and Wang's multiple imputation produced the narrowest confidence intervals, with acceptable coverage. For a small sample size (n = 100) Rubin's multiple imputation, overall, outperformed the other methods. Full mechanism bootstrapping was inefficient relative to the other methods and required modelling of the missing data mechanism under the missing at random assumption. Our proposed modification showed an improvement over Rubin's multiple imputation in the presence of misspecification. Overall, Rubin's multiple imputation variance estimator can fail in the presence of incompatibility and/or misspecification. For unavoidable incompatibility and/or misspecification, Robins and Wang's multiple imputation could provide more robust inferences.

Challenges and novel approaches for investigating molecular mediation

Understanding mediation is useful for identifying intermediates lying between an exposure and an outcome which, when intervened upon, will block (some or all of) the causal pathway between the exposure and outcome. Mediation approaches used in conventional epidemiology have been adapted to understanding the role of molecular intermediates in situations of high-dimensional omics data with varying degrees of success. In particular, the limitations of observational epidemiological study including confounding, reverse causation and measurement error can afflict conventional mediation approaches and may lead to incorrect conclusions regarding causal effects. Solutions to analysing mediation which overcome these problems include the use of instrumental variable methods such as Mendelian randomization, which may be applied to evaluate causality in increasingly complex networks of omics data.

Joint modelling compared with two stage methods for analysing longitudinal data and prospective outcomes: A simulation study of childhood growth and BP

There is a growing debate with regards to the appropriate methods of analysis of growth trajectories and their association with prospective dependent outcomes. Using the example of childhood growth and adult BP, we conducted an extensive simulation study to explore four two-stage and two joint modelling methods, and compared their bias and coverage in estimation of the (unconditional) association between birth length and later BP, and the association between growth rate and later BP (conditional on birth length). We show that the two-stage method of using multilevel models to estimate growth parameters and relating these to outcome gives unbiased estimates of the conditional associations between growth and outcome. Using simulations, we demonstrate that the simple methods resulted in bias in the presence of measurement error, as did the two-stage multilevel method when looking at the total (unconditional) association of birth length with outcome. The two joint modelling methods gave unbiased results, but using the re-inflated residuals led to undercoverage of the confidence intervals. We conclude that either joint modelling or the simpler two-stage multilevel approach can be used to estimate conditional associations between growth and later outcomes, but that only joint modelling is unbiased with nominal coverage for unconditional associations.

Reliability and validity of a postal version of the Reintegration to Normal Living Index, modified for use with stroke patients

Objective: To adapt the Reintegration to Normal Living Index (RNLI) for postal use with stroke patients.

Design: Reliability was examined using test–retest on 26 stroke patients. Construct validity was tested on 76 patients by examining correlations between the modified RNLI and related scales.

Subjects: Patients at three months to one year post stroke.

Results: All items demonstrated better than chance agreement between test and retest and seven items substantial agreement (kappa= >0.61). The modified RNLI correlated positively with related scales. Patients with stronger reintegration to normal living had better outcomes in anxiety, depression, daily activity and quality of life.

Conclusion: This postal instrument appears reliable and valid and may be a useful outcome measure in stroke studies.

Effect of recommended positioning on stroke outcome at six months: a randomized controlled trial

Objective: To evaluate the effect on patient outcome of a teaching package for nurses designed to improve the positioning of stroke patients.

Design: Cluster randomized controlled trial with six-month follow-up.

Setting: Ten stroke rehabilitation hospital units located within one UK inner city region. These were randomized to control or intervention group.

Subjects: A sample of 120 patients admitted within four weeks of a first stroke and with a hemiplegia. No eligible patient refused to participate. Eighty-three (69%) completed the study.

Intervention: All nursing staff on the intervention units received a group teaching package to improve their clinical practice in patient positioning.

Main outcome measure: Rivermead Mobility Index (RMI) at six months post stroke. Patient's position was recorded using an established observational tool.

Results: After the teaching there was some evidence of better positioning in the intervention than the control group (difference in percentage of correct positions per patient 4.9%, 95% confidence interval (CI-0.1% to 9.9%, p-0.055). There was no evidence of differences between the two groups in any of the outcome measures at six months although there was a trend towards increased elbow flexor tone in the control group.

Conclusions: A teaching intervention to improve patient positioning made no significant impact on outcome at six months post stroke. However, following the teaching there was only a slightly higher incidence of recommended patient positioning within the intervention group. Thus, a teaching package may not be powerful enough to enable any effect on patient outcome to be measured.

Agreement between routine and research measurement of infant height and weight

In many countries, routine data relating to growth of infants are collected as a means of tracking health and illness up to school age. These have potential to be used in research. For health monitoring and research, data should be accurate and reliable. This study aimed to determine the agreement between length/height and weight measurements from routine infant records and researcher-collected data.

METHODS

Height/length and weight at ages 6, 12 and 24 months from the longitudinal UK birth cohort (born in Bradford; n=836-1280) were compared with routine data collected by health visitors within 2 months of the research data (n=104-573 for different comparisons). Data were age adjusted and compared using Bland Altman plots.

RESULTS

There was agreement between data sources, albeit weaker for height than for weight. Routine data tended to underestimate length/height at 6 months (0.5 cm (95% CI -4.0 to 4.9)) and overestimate it at 12 (-0.3 cm (95% CI -0.5 to 4.0)) and 24 months (0.3 cm (95% CI -4.0 to 3.4)). Routine data slightly overestimated weight at all three ages (range -0.04 kg (95% CI -1.2 to 0.9) to -0.04 (95% CI -0.7 to 0.6)). Limits of agreement were wide, particularly for height. Differences were generally random, although routine data tended to underestimate length in taller infants and underestimate weight in lighter infants.

CONCLUSIONS

Routine data can provide an accurate and feasible method of data collection for research, though wide limits of agreement between data sources may be observed. Differences could be due to methodological issues; but may relate to variability in clinical practice. Continued provision of appropriate training and assessment is essential for health professionals responsible for collecting routine data.

Breast-feeding and health consequences in early childhood: is there an impact of time-dependent confounding?

BACKGROUND

Estimated effects of breast-feeding on childhood health vary between studies, possibly due to confounding by baseline maternal and child characteristics. Possible time-dependent confounding has received little consideration. Our aim was to evaluate the impact of such confounding.

METHODS

We estimated the relationship between cumulative exclusive breast-feeding up to 6 months and wheezing, rash and body mass index (BMI) at 12 months [in the Whistler cohort (n = 494) and PROBIT (n = 11,463)], and wheezing, rash, asthma, hay fever, eczema, allergy and BMI at age 6.5 years (PROBIT). We adjusted for time-dependent confounding by weight, length, rash, respiratory illness and day care attendance using marginal structural models (MSMs).

RESULTS

Weight and day care attendance appeared potential time-dependent confounders, since these predicted breast-feeding status and were influenced by previous breast-feeding. However, adjustment for time-dependent confounders did not markedly change the estimated associations. For example, in PROBIT the adjusted increase in BMI at 12 months per 1-month increase in exclusive breast-feeding was 0.04 (95% CI -0.09 to 0.01) using logistic regression and -0.06 (95% CI -0.11 to -0.01) using MSM. In Whistler, these estimates were each -0.05 (95% CI -0.10 to 0.00).

CONCLUSIONS

In two cohort studies, there was little evidence of time-dependent confounding by weight, length, rash, respiratory illness or day care attendance of the effects of breast-feeding on early childhood health.

Accuracy of height and weight data from child health records

BACKGROUND

There is limited knowledge of the accuracy of height and weight measurements from child health records, despite widespread use for research and clinical care. We assess the accuracy of such measurements, using research measurements as the gold standard.

METHODS

We compare height/length and weight measurements from clinics of the Avon Longitudinal Study of Parents and Children with routinely collected child health records within 2 months of the clinic date at age 4 (n = 345), 8 (n = 1051), 12 (n = 139), 18 (n = 649), 25 (n = 183), and 43 months (n = 123). To adjust for age differences at measurement, growth data were converted into standard deviation scores using the UK 1990 growth reference.

RESULTS

Mean weight standard deviation score (SDS) differences were < or =0.08, with mean predicted differences < or =0.1 kg (eg, mean predicted difference at 8 months -0.011 kg, 95% level of agreement -0.64 to 0.62 kg). Mean height SDS differences were < or =0.45, with mean predicted differences < or =0.9 cm (eg, mean predicted difference at 8 months -0.59 cm, 95% level of agreement -3.84 to 2.66 cm). There was indication of lower accuracy at 4 months old (mean predicted height difference at 4 months -0.91 cm, 95% level of agreement -4.61 to 2.79 cm), but this decreased when the age difference between measurements was reduced. Routine measurements slightly overestimated heights of tall children and underestimated those of short children, but otherwise differences were not associated with sex, social class, birth weight, birth length, or maternal anthropometry.

CONCLUSION

Routinely collected child health record height/length and weight data are compatible with no systematic bias, at least in children over 8 months old, supporting their use in clinical practice and research.

Current strategies for monitoring men with localised prostate cancer lack a strong evidence base: observational longitudinal study

Background:

The UK National Institute for Health and Clinical Excellence (NICE) guidance recommends conservative management of men with ‘low-risk’ localised prostate cancer, monitoring the disease using prostate-specific antigen (PSA) kinetics and re-biopsy. However, there is little evidence of the changes in PSA level that should alert to the need for clinical re-assessment.

Methods:

This study compares the alerts resulting from PSA kinetics and a novel longitudinal reference range approach, which incorporates age-related changes, during the monitoring of 408 men with localised prostate cancer. Men were monitored by regular PSA tests over a mean of 2.9 years, recording when a man's PSA doubling time fell below 2 years, PSA velocity exceeded 2 ng ml^–1 per year, or when his upper 10% reference range was exceeded.

Results:

Prostate-specific antigen doubling time and PSA velocity alerted a high proportion of men initially but became unresponsive to changes with successive tests. Calculating doubling time using recent PSA measurements reduced the decline in response. The reference range method maintained responsiveness to changes in PSA level throughout the monitoring.

Conclusion:

The increasing unresponsiveness of PSA kinetics is a consequence of the underlying regression model. Novel methods are needed for evaluation in cohorts currently being managed by monitoring. Meanwhile, the NICE guidance should be cautious.