Two sample Mendelian Randomisation using an outcome from a multilevel model of disease progression

Identifying factors that are causes of disease progression, especially in neurodegenerative diseases, is of considerable interest. Disease progression can be described as a trajectory of outcome over time-for example, a linear trajectory having both an intercept (severity at time zero) and a slope (rate of change). A technique for identifying causal relationships between one exposure and one outcome in observational data whilst avoiding bias due to confounding is two sample Mendelian Randomisation (2SMR). We consider a multivariate approach to 2SMR using a multilevel model for disease progression to estimate the causal effect an exposure has on the intercept and slope. We carry out a simulation study comparing a naïve univariate 2SMR approach to a multivariate 2SMR approach with one exposure that effects both the intercept and slope of an outcome that changes linearly with time since diagnosis. The simulation study results, across six different scenarios, for both approaches were similar with no evidence against a non-zero bias and appropriate coverage of the 95% confidence intervals (for intercept 93.4-96.2% and the slope 94.5-96.0%). The multivariate approach gives a better joint coverage of both the intercept and slope effects. We also apply our method to two Parkinson's cohorts to examine the effect body mass index has on disease progression. There was no strong evidence that BMI affects disease progression, however the confidence intervals for both intercept and slope were wide.

Are physical therapy interventions effective in improving sleep in people with chronic pain? A systematic review and multivariate meta-analysis

Chronic pain exerts an enormous personal and economic burden, with sleep disturbances being one of the most reported problems by adults with chronic pain. The aim of this study was to analyse whether different physical therapy interventions could lead to improvements in sleep quality and pain intensity in individuals with chronic pain, as well as if there is any association. A systematic review and a univariate and multivariate meta-analysis were carried out according to the PRISMA guidelines. A search in PubMed, Scopus and Web of Science databases was performed. Six randomised controlled trials were included in the review and four of them were included in the meta-analysis; all of them with a moderate to high methodological quality. Data from adult participants with chronic pain after different physical therapy interventions was extracted. For the meta-analysis, the Insomnia Severity Index and the Numerical Rating Scale were considered. Results from the qualitative and quantitative analysis showed that most of the physical therapy interventions included had higher improvements in the intervention group than in the control group, although the effect size was not statistically significant (univariate for sleep quality: -0.08 [-0.34, 0.18], p = 0.46; univariate for pain intensity: -0.47 [-1.24, 0.30], p = 0.18; multivariate for both outcomes: -0.27). More studies are still needed to determine which physical therapy interventions are effective to improve sleep in people with chronic pain and if there are patients with specific characteristics who may benefit more than others.

Comparing asexual with heterosexual, bisexual, and gay/lesbian individuals in common mental health problems: A multivariate meta-analysis

We aimed to test whether asexual individuals were at increased risk of higher levels of depressive symptoms, self-harm attempts, and suicide attempts compared with heterosexual, bisexual, or gay/lesbian individuals using multivariate meta-analysis. Seventeen, five, and eight samples were included for depressive symptoms, self-harm attempts, and suicide attempts, respectively, reaching a total sample size of 125,675, 30,116, and 73,366, respectively. Asexual individuals reported higher levels of depressive symptoms than heterosexual individuals (Hedges' g = -0.44, 95%CI = [-0.61, -0.26]) but did not differ from heterosexual individuals in the risk of self-harm (odds ratio = 1.11, 95%CI = [0.88, 1.39]) and suicide attempts (odds ratio = 0.76, 95%CI = [0.56, 1.04]). Asexual individuals were at lower risk of self-harm and suicide attempts than bisexual and gay/lesbian individuals but did not differ from bisexual and gay/lesbian individuals in the levels of depressive symptoms. The greatest risk of higher levels of depressive symptoms was found in bisexual and asexual, followed by gay/lesbian individuals; the greatest risk of self-harm and suicide attempts was found in bisexual, followed by gay/lesbian individuals, and the lowest risk was found in asexual individuals. The magnitude of the disparities in the risk of poorer mental health among heterosexual, bisexual, gay/lesbian, and asexual individuals depended on the type of mental health outcomes.

Daytime and nighttime high temperatures differentially increased the risk of cardiovascular disease: A nationwide hospital-based study in China

Short-term exposure to ambient high temperature (heat) could increase the risk of cardiovascular disease (CVD). However, available evidence on the burden of daytime and nighttime heat on CVD is limited and vulnerable populations remain unknown so far. We aimed to examine and differentiate the impact of daytime and nighttime heat on CVD in China. Data on daily outpatient visits for CVD were collected from 15 Chinese cities spanning multiple geographical regions, climates, and socio-economic conditions. The population-weighted temperature was used to calculate excess heat exposure in warm seasons (June-September) from 2011 to 2015. Hot day excess (HDE) and hot night excess (HNE), the sum of temperature above the heat threshold during daytime and nighttime respectively, were used to represent daytime and nighttime excess heat. A distributed lag non-linear model was employed to estimate the city-level association between HDE/HNE and daily CVD cases. The city-level association was then pooled by multivariate meta-analysis. We further estimated the disease burden of CVD attributable to HDE and HNE by geographical regions, gender, and age. A total of 729,409 cases of CVD were included in this study. Both HDE and HNE were associated with an increased risk of CVD, with greater effects from nighttime heat (relative risk (RR): 1.38; 95% confidence interval (CI): 1.18-1.61) than daytime heat (RR: 1.10; 95% CI: 1.05-1.15). The proportion of CVD cases attributable to HNE was 15.7%, which was almost three times as high as HDE (4.6%, p for difference <0.05). Males, people living in northern cities, and those aged less than 45 years were more vulnerable to HNE. Our findings for the first time revealed an intra-day difference in the heat effect on CVD, with a greater impact from nighttime heat exposure, which should be considered to protect vulnerable populations on hot days.

The prenatal weekly temperature exposure and neonatal congenital heart disease: a large population-based observational study in China

We aim to explore the link between maternal weekly temperature exposure and CHD in offspring and identify the relative contributions from heat and cold and from moderate and extreme atmospheric temperature. From January 2019 to December 2020, newborns who were diagnosed with CHD by echocardiography in the Network Platform for Congenital Heart Disease (NPCHD) from 11 cities in eastern China were enrolled in the present study. We appraised the exposure lag response relationship between temperature and CHDs in the distributed lag nonlinear model and further probed the pooled estimates by multivariate meta-analysis. We further performed the exposure-response curves in extreme temperature (5th percentile for cold and 95th for hot events). We also delve into the cumulative risk ratios (CRRs) of temperature on CHDs in general and subgroups. In this study, 5904 of 983, 523 infants were diagnosed with CHDs. The temperature-CHD combination performed positive significance in two exposure windows, gestational weeks 10-16 and 26-31, and reached the maximum effect in the 28th week. Compared with extreme cold (5th, 6.14℃), these effects were higher in extreme heat (95th, 29.26℃). The cumulative exposure-response curve showed a steep nonlinear rise in the hot tail but showed non-significance at low temperatures. In this range, the CRRs of temperature showed an increment to a ceiling of 3.781 (95% CI: 1.460-10.723). The temperature- CHD curves for both sex groups showed a general growth trend. No statistical significance was observed between these two groups (P = 0.106). The cumulative effect of the temperature related CHD was significant in regions with lower education levels (maximum CRR was 9.282 (3.019-28.535)). A degree centigrade increase in temperature exposure was associated with the increment of CHD risk in the first and second trimesters, especially in extreme heat. Neonates born in lower education regions were more vulnerable to temperature-related CHDs.

Advanced methods and implementations for the meta-analyses of animal models: Current practices and future recommendations

Meta-analytic techniques have been widely used to synthesize data from animal models of human diseases and conditions, but these analyses often face two statistical challenges due to complex nature of animal data (e.g., multiple effect sizes and multiple species): statistical dependency and confounding heterogeneity. These challenges can lead to unreliable and less informative evidence, which hinders the translation of findings from animal to human studies. We present a literature survey of meta-analysis using animal models (animal meta-analysis), showing that these issues are not adequately addressed in current practice. To address these challenges, we propose a meta-analytic framework based on multilevel (linear mixed-effects) models. Through conceptualization, formulations, and worked examples, we illustrate how this framework can appropriately address these issues while allowing for testing new questions. Additionally, we introduce other advanced techniques such as multivariate models, robust variance estimation, and meta-analysis of emergent effect sizes, which can deliver robust inferences and novel biological insights. We also provide a tutorial with annotated R code to demonstrate the implementation of these techniques.

Mortality burden attributable to temperature variability in China

BACKGROUND

Several studies have investigated the associations between temperature variability (TV) and death counts. However, evidence of TV-attributable years of life lost (YLL) is scarce.

OBJECTIVES

To investigate the associations between TV and YLL rates (/100,000 population), and quantify average life loss per death (LLD) caused by TV in China.

METHODS

We calculated daily YLL rates (/100,000 population) of non-accidental causes and cardiorespiratory diseases by using death data from 364 counties of China during 2006-2017, and collected meteorological data during the same period. A distributed lag non-linear model (DLNM) and multivariate meta-analysis were used to estimate the effects of TV at national or regional levels. Then, we calculated the LLD to quantify the mortality burden of TV.

RESULTS

U-shaped curves were observed in the associations of YLL rates with TV in China. The minimum YLL TV (MYTV) was 2.5 °C nationwide. An average of 0.89 LLD was attributable to TV in total, most of which was from high TV (0.86, 95% CI: 0.56, 1.16). However, TV caused more LLD in the young (<65 years old) (1.87, 95% CI: 1.03, 2.71) than 65-74 years old (0.85, 95% CI: 0.40-1.31) and ≥75 years old (0.40, 95% CI: 0.21-0.59), cerebrovascular disease (0.74, 95% CI: 0.36, 1.11) than respiratory disease (0.54, 95% CI: 0.21, 0.87), South (1.23, 95% CI: 0.77, 1.68) than North (0.41, 95% CI: -0.7, 1.52) and Central China (0.40, 95% CI: -0.02, 0.81). TV-attributed LLD was modified by annual mean temperature, annual mean relative humidity, altitude, latitude, longitude, and education attainment.

SIGNIFICANCE

Our findings indicate that high and low TVs are both associated with increases in premature death, however the majority of LLD was attributable to high TV. TV-related LLD was modified by county level characteristics. TV should be considered in planning adaptation to climate change or variability.

IMPACT

(1) We estimated the associations of TV with YLL rates, and quantified the life loss per death (LLD) caused by TV. (2) An average of 0.89 years of LLD were attributable to TV, most of which were from high TVs. (3) TV caused more LLD in the young, cerebrovascular disease, and southern China. (4) The mortality burdens were modified by county level characteristics.

Investigation of the Utility of Multivariate Meta-Analysis Methods in Estimating the Summary Dose Response Curve

BACKGROUND

Traditional meta-analyses often assess the effectiveness of different doses of the same intervention separately or examine the overall differences between intervention and placebo groups. The present study aimed to model the effect sizes obtained from different doses in multiple studies using a two-stage dose-response meta-analytic approach while taking dose variations into account.

METHODS

Different dose-response meta-analysis models using linear, quadratic, and restricted cubic spline (RCS) functions were fitted. A two-stage approach utilizing multivariate meta-analysis was performed and the obtained results were compared with those of the univariate meta-analysis. A random effect dose-response meta-analysis was performed using data from an existing systematic review on combination therapy with zonisamide and anti-Parkinson drugs for Parkinson's disease. The effective or optimum dose for producing maximum response was also investigated. Moreover, a sensitivity analysis was performed by changing the knots of the RCS model.

RESULTS

Dose-response meta-analysis was performed using data from four double-blinded randomized controlled trials with 724 and 309 patients with Parkinson's disease in dose and placebo arms, respectively. The quadratic model yielded the smallest Akaike information criterion (AIC), compared to the linear and RCS models, indicating it to be the best fit for the data.

CONCLUSION

Compared to the traditional approach, the two-stage approach could model the dose-dependent effect of zonisamide on the Unified Parkinson's Disease Rating Scale (UPRDS) part III score and predict the outcome for different doses through a single analysis.

Multivariate meta-analysis of multiple outcomes: characteristics and predictors of borrowing of strength from Cochrane reviews

OBJECTIVES

Multivariate meta-analysis allows the joint synthesis of multiple outcomes accounting for their correlation. This enables borrowing of strength (BoS) across outcomes, which may lead to greater efficiency and even different conclusions compared to separate univariate meta-analyses. However, multivariate meta-analysis is complex to apply, so guidance is needed to flag (in advance of analysis) when the approach is most useful.

STUDY DESIGN AND SETTING

We use 43 Cochrane intervention reviews to empirically investigate the characteristics of meta-analysis datasets that are associated with a larger BoS statistic (from 0 to 100%) when applying a bivariate meta-analysis of binary outcomes.

RESULTS

Four characteristics were identified as strongly associated with BoS: the total number of studies, the number of studies with the outcome of interest, the percentage of studies missing the outcome of interest, and the largest absolute within-study correlation. Using these characteristics, we then develop a model for predicting BoS in a new dataset, which is shown to have good performance (an adjusted R² of 50%). Applied examples are used to illustrate the use of the BoS prediction model.

CONCLUSIONS

Cochrane reviewers mainly use univariate meta-analysis methods, but the identified characteristics associated with BoS and our subsequent prediction model for BoS help to flag when a multivariate meta-analysis may also be beneficial in Cochrane reviews with multiple binary outcomes. Extension to non-Cochrane reviews and other outcome types is still required.

Effects of high-frequency temperature variabilities on the morbidity of chronic obstructive pulmonary disease: Evidence in 21 cities of Guangdong, South China

BACKGROUND

While temperature changes have been confirmed as one of the contributory factors affecting human health, the association between high-frequency temperature variability (HFTV, i.e., temperature variation at short time scales such as 1, 2, and 5 days) and the hospitalization of chronic obstructive pulmonary disease (COPD) was rarely reported.

OBJECTIVES

To evaluate the associations between high-frequency temperature variabilities (i.e., at 1, 2, and 5-day scales) and daily COPD hospitalization.

METHODS

We collected daily records of COPD hospitalization and meteorological variables from 2013 to 2017 in 21 cities of Guangdong Province, South China. A quasi-Poisson regression with a distributed lag nonlinear model was first employed to quantify the effects of two HFTV measures, i.e., the day-to-day (DTD) temperature change and the intraday-interday temperature variability (IITV), on COPD morbidity for each city. Second, we used multivariate meta-analysis to pool the city-specific estimates, and stratified analyses were performed by age and sex to identify vulnerable groups. Then, the meta-regression with city-level characteristics was employed to detect the potential sources of the differences among 21 cities.

RESULTS

A monotonic increasing curve of the overall exposure-response association was observed, suggesting that positive HFTV (i.e., increased DTD and IITV) will significantly increase the risk of COPD admission. Negative DTD was associated with reduced COPD morbidity while positive DTD elevated the COPD risk. An interquartile range (IQR) increase in DTD was associated with a 24% (95% CI: 12-38%) increase in COPD admissions. An IQR increase in IITV_0-1 was associated with 18% (95% CI: 7-27%) increase in COPD admissions. Males and people aged 0-64 years appeared to be more vulnerable to the DTD effect than others. Potential sources of the disparity among different cities include urbanization level, sex structure, industry structure, gross domestic product (GDP), health care services, and air quality.

CONCLUSIONS

The increases of DTD and IITV have significant adverse impacts on COPD hospitalization. As climate change intensifies, precautions need to be taken to mitigate the impacts of high-frequency temperature changes.

Multivariate meta-analysis of critical care meta-analyses: a meta-epidemiological study

Background

Meta-analyses typically consider multiple outcomes and report univariate effect sizes considered as independent. Multivariate meta-analysis (MVMA) incorporates outcome correlation and synthesises direct evidence and related outcome estimates within a single analysis. In a series of meta-analyses from the critically ill literature, the current study contrasts multiple univariate effect estimates and their precision with those derived from MVMA.

Methods

A previous meta-epidemiological study was used to identify meta-analyses with either one or two secondary outcomes providing sufficient detail to structure bivariate or tri-variate MVMA, with mortality as primary outcome. Analysis was performed using a random effects model for both odds ratio (OR) and risk ratio (RR); borrowing of strength (BoS) between multivariate outcome estimates was reported. Estimate comparisons, β coefficients, standard errors (SE) and confidence interval (CI) width, univariate versus multivariate, were performed using Lin’s concordance correlation coefficient (CCC).

Results

In bivariate meta-analyses, for OR (n = 49) and RR (n = 48), there was substantial concordance (≥ 0.69) between estimates; but this was less so for tri-variate meta-analyses for both OR (n = 25; ≥ 0.38) and RR (≥ -0.10; n = 22). A variable change in the multivariate precision of primary mortality outcome estimates compared with univariate was present for both bivariate and tri-variate meta-analyses and for metrics. For second outcomes, precision tended to decrease and CI width increase for bivariate meta-analyses, but was variable in the tri-variate. For third outcomes, precision increased and CI width decreased. In bivariate meta-analyses, OR coefficient significance reversal, univariate versus MVMA, occurred once for mortality and 6 cases for second outcomes. RR coefficient significance reversal occurred in 4 cases; 2 were discordant with OR. For tri-variate OR meta-analyses reversal of coefficient estimate significance occurred in two cases for mortality, nine cases for second and 7 cases for third outcomes. In RR meta-analyses significance reversals occurred for mortality in 2 cases, 6 cases for second and 3 cases for third; there were 7 discordances with OR. BoS was greater in trivariate MVMAs compared with bivariate and for OR versus RR.

Conclusions

MVMA would appear to be the preferred solution to multiple univariate analyses; parameter significance changes may occur. Analytic metric appears to be a determinant.

Umbrella review and multivariate meta-analysis of diagnostic test accuracy studies on hybrid non-invasive imaging for coronary artery disease

BACKGROUND

The diagnosis of coronary artery disease (CAD) remains challenging. It is uncertain whether hybrid imaging can improve diagnostic accuracy for CAD.

METHODS

This is a systematic review and multivariate meta-analysis. We searched PubMed and The Cochrane Library for recent (≥ 2010) systematic reviews of diagnostic test accuracy studies on non-invasive imaging for CAD. Study-level data were extracted from them, and pooled with pairwise and multivariate meta-analytic methods, using invasive coronary angiography (ICA) or invasive fractional flow reserve (FFR) as reference standards, focusing on sensitivity and specificity.

RESULTS

Details from 661 original studies (71,823 patients) were pooled. Pairwise meta-analysis using ICA as reference showed that anatomic imaging was associated with the best diagnostic accuracy (sensitivity = 0.95 [95% confidence interval 0.94-0.96], specificity = 0.83 [0.81-0.85]), whereas using FFR as reference identified hybrid imaging as the best test (sensitivity = 0.87 [0.83-0.90], specificity = 0.82 [0.76-0.87]). Multivariate meta-analysis confirmed the superiority of anatomic imaging using ICA as reference (sensitivity = 0.96, specificity = 0.83), and hybrid imaging using FFR as reference (sensitivity = 0.88 [0.86-0.91], specificity = 0.82 [0.77-0.87]).

CONCLUSIONS

Non-invasive hybrid imaging tests appear superior to anatomic or functional only tests to diagnose ischemia-provoking coronary lesions, whereas anatomic imaging is best to diagnose and/or rule out angiographically significant CAD.

SYSTEMATIC REVIEW REGISTRATION

PROSPERO Registry Number CRD42018088528.

Combining tumour response and progression free survival as surrogate endpoints for overall survival in advanced colorectal cancer

BACKGROUND

Progression free survival (PFS) and tumour response (TR) have been investigated as surrogate endpoints for overall survival (OS) in advanced colorectal cancer (aCRC), however their validity has been shown to be suboptimal. In recent years, meta-analytic methods allowing for use of multiple surrogate endpoints jointly have been proposed. Our aim was to assess if PFS and TR used jointly as surrogate endpoints to OS improve their predictive value.

METHODS

Data were obtained from a systematic review of randomised controlled trials investigating effectiveness of pharmacological therapies in aCRC, including systemic chemotherapies, anti-epidermal growth factor receptor therapies and anti-angiogenic agents. Multivariate meta-analysis was used to model the association patterns between treatment effects on the surrogate endpoints (TR, PFS) and the final outcome (OS).

RESULTS

Analysis of 33 trials reporting treatment effects on all three outcomes showed reasonably strong association between treatment effects on PFS and OS, however the association parameters were obtained with a large uncertainty. A weak surrogate relationship was noted between the treatment effects on TR and OS. Modelling the two surrogate endpoints, TR and PFS, jointly as predictors of treatment effect on OS gave no marked improvement to surrogate association patterns. Modest improvement in the precision of the predicted treatment effects on the final outcome was noted in studies investigating anti-angiogenic therapy, however it was likely due to chance.

CONCLUSION

The joint use of two surrogate endpoints did not lead to marked improvement in the association between treatment effects on surrogate and final endpoints in advanced colorectal cancer.

The effect of temperature on childhood hand, foot and mouth disease in Guangdong Province, China, 2010-2013: a multicity study

Background

Hand, foot and mouth disease (HFMD) is a serious infectious disease, which has become a public health problem. Previous studies have shown that temperature may influence the incidence of HFMD, but most only focus on single city and the results are highly heterogeneous. Therefore, a multicity study was conducted to explore the association between temperature and HFMD in different cities and search for modifiers that influence the heterogeneity.

Methods

We collected daily cases of childhood HFMD (aged 0–5 years) and meteorological factors of 21 cities in Guangdong Province in the period of 2010–2013. Distributed lag non-linear model (DLNM) with quasi-Poisson was adopted to quantify the effects of temperature on HFMD in 21 cities. Then the effects of each city were pooled by multivariate meta-analysis to obtain the heterogeneity among 21 cities. Potential city-level factors were included in meta-regression to explore effect modifiers.

Results

A total of 1,048,574 childhood cases were included in this study. There was a great correlation between daily childhood HFMD cases and temperature in each city, which was non-linear and lagged. High heterogeneity was showed in the associations between temperature and HFMD in 21 cities. The pooled temperature-HFMD association was peaking at the 79th percentile of temperature with relative risk (RR) of 2.474(95% CI: 2.065–2.965) as compared to the median temperature. Latitude was the main modifier for reducing the heterogeneity to 69.28% revealed by meta-analysis.

Conclusions

There was a strong non-linear and lagged correlation between temperature and HFMD. Latitude was strongly associated with the relationship between temperature and HFMD. Meanwhile, it had an effect on modifying the relationship. These findings can conducive to local governments developing corresponding preventive measures.

A Guide to Conducting a Meta-Analysis with Non-Independent Effect Sizes

Conventional meta-analytic procedures assume that effect sizes are independent. When effect sizes are not independent, conclusions based on these conventional procedures can be misleading or even wrong. Traditional approaches, such as averaging the effect sizes and selecting one effect size per study, are usually used to avoid the dependence of the effect sizes. These ad-hoc approaches, however, may lead to missed opportunities to utilize all available data to address the relevant research questions. Both multivariate meta-analysis and three-level meta-analysis have been proposed to handle non-independent effect sizes. This paper gives a brief introduction to these new techniques for applied researchers. The first objective is to highlight the benefits of using these methods to address non-independent effect sizes. The second objective is to illustrate how to apply these techniques with real data in R and Mplus. Researchers may modify the sample R and Mplus code to fit their data.

A tutorial on individual participant data meta-analysis using Bayesian multilevel modeling to estimate alcohol intervention effects across heterogeneous studies

This paper provides a tutorial companion for the methodological approach implemented in Huh et al. (2015) that overcame two major challenges for individual participant data (IPD) meta-analysis. Specifically, we show how to validly combine data from heterogeneous studies with varying numbers of treatment arms, and how to analyze highly-skewed count outcomes with many zeroes (e.g., alcohol and substance use outcomes) to estimate overall effect sizes. These issues have important implications for the feasibility, applicability, and interpretation of IPD meta-analysis but have received little attention thus far in the applied research literature. We present a Bayesian multilevel modeling approach for combining multi-arm trials (i.e., those with two or more treatment groups) in a distribution-appropriate IPD analysis. Illustrative data come from Project INTEGRATE, an IPD meta-analysis study of brief motivational interventions to reduce excessive alcohol use and related harm among college students. Our approach preserves the original random allocation within studies, combines within-study estimates across all studies, overcomes between-study heterogeneity in trial design (i.e., number of treatment arms) and/or study-level missing data, and derives two related treatment outcomes in a multivariate IPD meta-analysis. This methodological approach is a favorable alternative to collapsing or excluding intervention groups within multi-arm trials, making it possible to directly compare multiple treatment arms in a one-step IPD meta-analysis. To facilitate application of the method, we provide annotated computer code in R along with the example data used in this tutorial.

Study characteristics influence the efficacy of substance abuse treatments: A meta-analysis of medications for alcohol use disorder

BACKGROUND

Understanding study characteristics' influence on treatment efficacy could improve methodologies and interpretation of findings. We examine study characteristics as predictors of outcomes in clinical trials of medications for alcohol problems.

METHODS

We obtained data on 23 trials of naltrexone and 22 trials of acamprosate from Cochrane reviews. We extracted data for 14 study characteristics and 3 dependent variables (odds ratio; percent abstinent in placebo; medication conditions). We used general linear models to determine which study characteristics explained the variability among outcomes after controlling for medication characteristics.

RESULTS

Study characteristics accounted for 45% of the variance in odds ratio when trials of different medications were combined, 19% among acamprosate, and 48% among naltrexone trials above and beyond medication characteristics. When medications were combined, greater odds ratios were predicted by having more dropouts in the placebo than medication conditions, an earlier publication year, and not receiving industry funding. Whether this was due to effects on placebo or medication conditions was unclear. Among acamprosate trials, smaller sample sizes predicted greater odds ratios, which appeared to be due to more abstinence in medication conditions. Among naltrexone trials, greater odds ratios were predicted by having more dropouts in the placebo than medication conditions and a greater probability of randomizing participants to treatment. This appeared to be due to less abstinence in placebo conditions.

CONCLUSION

Study characteristics influence the assessment of treatment efficacy beyond medication characteristics in alcohol treatment trials. Future studies are needed to determine which study characteristics reliably influence efficacy to help investigators design and help clinicians interpret trials.

The role of secondary outcomes in multivariate meta-analysis

Univariate meta‐analysis concerns a single outcome of interest measured across a number of independent studies. However, many research studies will have also measured secondary outcomes. Multivariate meta‐analysis allows us to take these secondary outcomes into account and can also include studies where the primary outcome is missing. We define the efficiency E as the variance of the overall estimate from a multivariate meta‐analysis relative to the variance of the overall estimate from a univariate meta‐analysis. The extra information gained from a multivariate meta‐analysis of n studies is then similar to the extra information gained if a univariate meta‐analysis of the primary effect had a further n(1−E)/E studies. The variance contribution of a study's secondary outcomes (its borrowing of strength) can be thought of as a contrast between the variance matrix of the outcomes in that study and the set of variance matrices of all the studies in the meta‐analysis. In the bivariate case this is given a simple graphical interpretation as the borrowing‐of‐strength plot. We discuss how these findings can also be used in the context of random‐effects meta‐analysis. Our discussion is motivated by a published meta‐analysis of 10 antihypertension clinical trials.

Mental disease-related emergency admissions attributable to hot temperatures

OBJECTIVE

The association between high temperature and mental disease has been the focus of several studies worldwide. However, no studies have focused on the mental disease burden attributable to hot temperature. Here, we aim to quantify the risk attributed to hot temperatures based on the exposure-lag-response relationship between temperature and mental diseases.

METHOD

From data on daily temperature and emergency admissions (EA) for mental diseases collected from 6 major cities (Seoul, Incheon, Daejeon, Daegu, Busan, and Gwangju in South Korea) over a period of 11years (2003-2013), we estimated temperature-disease associations using a distributed lag non-linear model, and we pooled the data by city through multivariate meta-analysis. Cumulative relative risk and attributable risks were calculated for extreme hot temperatures, defined as the 99th percentile relative to the 50th percentile of temperatures.

RESULTS

The strongest association between mental disease and high temperature was seen within a period of 0-4days of high temperature exposure. Our results reveal that 14.6% of EA for mental disease were due to extreme hot temperatures, and the elderly were more susceptible (19.1%). Specific mental diseases, including anxiety, dementia, schizophrenia, and depression, also showed significant risk attributed to hot temperatures. Of all EA for anxiety, 31.6% were attributed to extremely hot temperatures.

CONCLUSIONS

High temperature was responsible for an attributable risk for mental disease, and the burden was higher in the elderly. This finding has important implications for designing appropriate public health policies to minimize the impact of high temperature on mental health.

Short-term effects of air quality and thermal stress on non-accidental morbidity-a multivariate meta-analysis comparing indices to single measures

Air quality and thermal stress lead to increased morbidity and mortality. Studies on morbidity and the combined impact of air pollution and thermal stress are still rare. To analyse the correlations between air quality, thermal stress and morbidity, we used a two-stage meta-analysis approach, consisting of a Poisson regression model combined with distributed lag non-linear models (DLNMs) and a meta-analysis investigating whether latitude or the number of inhabitants significantly influence the correlations. We used air pollution, meteorological and hospital admission data from 28 administrative districts along a north-south gradient in western Germany from 2001 to 2011. We compared the performance of the single measure particulate matter (PM10) and air temperature to air quality indices (MPI and CAQI) and the biometeorological index UTCI. Based on the Akaike information criterion (AIC), it can be shown that using air quality indices instead of single measures increases the model strength. However, using the UTCI in the model does not give additional information compared to mean air temperature. Interaction between the 3-day average of air quality (max PM10, max CAQI and max MPI) and meteorology (mean air temperature and mean UTCI) did not improve the models. Using the mean air temperature, we found immediate effects of heat stress (RR 1.0013, 95% CI: 0.9983-1.0043) and by 3 days delayed effects of cold stress (RR: 1.0184, 95% CI: 1.0117-1.0252). The results for air quality differ between both air quality indices and PM10. CAQI and MPI show a delayed impact on morbidity with a maximum RR after 2 days (MPI 1.0058, 95% CI: 1.0013-1.0102; CAQI 1.0068, 95% CI: 1.0030-1.0107). Latitude was identified as a significant meta-variable, whereas the number of inhabitants was not significant in the model.

Using structural equation modeling for network meta-analysis

Background

Network meta-analysis overcomes the limitations of traditional pair-wise meta-analysis by incorporating all available evidence into a general statistical framework for simultaneous comparisons of several treatments. Currently, network meta-analyses are undertaken either within the Bayesian hierarchical linear models or frequentist generalized linear mixed models. Structural equation modeling (SEM) is a statistical method originally developed for modeling causal relations among observed and latent variables. As random effect is explicitly modeled as a latent variable in SEM, it is very flexible for analysts to specify complex random effect structure and to make linear and nonlinear constraints on parameters. The aim of this article is to show how to undertake a network meta-analysis within the statistical framework of SEM.

Methods

We used an example dataset to demonstrate the standard fixed and random effect network meta-analysis models can be easily implemented in SEM. It contains results of 26 studies that directly compared three treatment groups A, B and C for prevention of first bleeding in patients with liver cirrhosis. We also showed that a new approach to network meta-analysis based on the technique of unrestricted weighted least squares (UWLS) method can also be undertaken using SEM.

Results

For both the fixed and random effect network meta-analysis, SEM yielded similar coefficients and confidence intervals to those reported in the previous literature. The point estimates of two UWLS models were identical to those in the fixed effect model but the confidence intervals were greater. This is consistent with results from the traditional pairwise meta-analyses. Comparing to UWLS model with common variance adjusted factor, UWLS model with unique variance adjusted factor has greater confidence intervals when the heterogeneity was larger in the pairwise comparison. The UWLS model with unique variance adjusted factor reflects the difference in heterogeneity within each comparison.

Conclusion

SEM provides a very flexible framework for univariate and multivariate meta-analysis, and its potential as a powerful tool for advanced meta-analysis is still to be explored.

Electronic supplementary material

The online version of this article (doi:10.1186/s12874-017-0390-9) contains supplementary material, which is available to authorized users.

Spatiotemporal variations of extreme low temperature for emergency transport: a nationwide observational study

Although recent studies have investigated the effect of extreme heat on emergency transport, few have investigated the spatiotemporal variations of extreme low temperature for emergency transport on a national scale. Data pertaining to emergency ambulance transport and weather variation in the 47 prefectures of Japan between 2007 and 2010 were obtained. Nonlinear and delayed relationships between temperature and morbidity were assessed using a two-stage analysis. First, a Poisson regression analysis allowing for overdispersion in a distributed lag nonlinear model was used to estimate the prefecture-specific effects of temperature on morbidity. Second, a multivariate meta-analysis was applied to pool estimates on a national level. Of 15,868,086 emergency transports over the study period, 5,375,621 emergency transports were reported during the winter months (November through February). The overall cumulative relative risk (RR) at the first percentile vs. the minimum morbidity percentile was 1.24 (95 % CI = 1.15-1.34) for all causes, 1.50 (95 % CI = 1.30-1.74) for cardiovascular diseases, and 1.59 (95 % CI = 1.33-1.89) for respiratory diseases. There were differences in the temporal variations between extreme low temperature and respiratory disease morbidity. Spatial variation between prefectures was observed for all causes (Cochran Q test, p < 0.001; I ² = 34.0 %) and respiratory diseases (Cochran Q test, p = 0.026; I ² = 18.2 %); however, there was no significant spatial heterogeneity for cardiovascular diseases (Cochran Q test, p = 0.413; I ² = 2.0 %). Our findings indicated that there were differences in the spatiotemporal variations of extreme low temperatures for emergency transport during winter in Japan. Our findings highlight the importance of further investigating to identify social and environmental factors, which can be responsible for spatial heterogeneity between prefectures.

Multivariate meta-analysis with an increasing number of parameters

Meta-analysis can average estimates of multiple parameters, such as a treatment's effect on multiple outcomes, across studies. Univariate meta-analysis (UVMA) considers each parameter individually, while multivariate meta-analysis (MVMA) considers the parameters jointly and accounts for the correlation between their estimates. The performance of MVMA and UVMA has been extensively compared in scenarios with two parameters. Our objective is to compare the performance of MVMA and UVMA as the number of parameters, p, increases. Specifically, we show that (i) for fixed-effect (FE) meta-analysis, the benefit from using MVMA can substantially increase as p increases; (ii) for random effects (RE) meta-analysis, the benefit from MVMA can increase as p increases, but the potential improvement is modest in the presence of high between-study variability and the actual improvement is further reduced by the need to estimate an increasingly large between study covariance matrix; and (iii) when there is little to no between-study variability, the loss of efficiency due to choosing RE MVMA over FE MVMA increases as p increases. We demonstrate these three features through theory, simulation, and a meta-analysis of risk factors for non-Hodgkin lymphoma.

Univariate and bivariate likelihood-based meta-analysis methods performed comparably when marginal sensitivity and specificity were the targets of inference

OBJECTIVES

To compare statistical methods for meta-analysis of sensitivity and specificity of medical tests (e.g., diagnostic or screening tests).

STUDY DESIGN AND SETTING

We constructed a database of PubMed-indexed meta-analyses of test performance from which 2 × 2 tables for each included study could be extracted. We reanalyzed the data using univariate and bivariate random effects models fit with inverse variance and maximum likelihood methods. Analyses were performed using both normal and binomial likelihoods to describe within-study variability. The bivariate model using the binomial likelihood was also fit using a fully Bayesian approach.

RESULTS

We use two worked examples-thoracic computerized tomography to detect aortic injury and rapid prescreening of Papanicolaou smears to detect cytological abnormalities-to highlight that different meta-analysis approaches can produce different results. We also present results from reanalysis of 308 meta-analyses of sensitivity and specificity. Models using the normal approximation produced sensitivity and specificity estimates closer to 50% and smaller standard errors compared to models using the binomial likelihood; absolute differences of 5% or greater were observed in 12% and 5% of meta-analyses for sensitivity and specificity, respectively. Results from univariate and bivariate random effects models were similar, regardless of estimation method. Maximum likelihood and Bayesian methods produced almost identical summary estimates under the bivariate model; however, Bayesian analyses indicated greater uncertainty around those estimates. Bivariate models produced imprecise estimates of the between-study correlation of sensitivity and specificity. Differences between methods were larger with increasing proportion of studies that were small or required a continuity correction.

CONCLUSION

The binomial likelihood should be used to model within-study variability. Univariate and bivariate models give similar estimates of the marginal distributions for sensitivity and specificity. Bayesian methods fully quantify uncertainty and their ability to incorporate external evidence may be useful for imprecisely estimated parameters.

Multivariate meta-analysis of individual participant data helped externally validate the performance and implementation of a prediction model

OBJECTIVES

Our aim was to improve meta-analysis methods for summarizing a prediction model's performance when individual participant data are available from multiple studies for external validation.

STUDY DESIGN AND SETTING

We suggest multivariate meta-analysis for jointly synthesizing calibration and discrimination performance, while accounting for their correlation. The approach estimates a prediction model's average performance, the heterogeneity in performance across populations, and the probability of "good" performance in new populations. This allows different implementation strategies (e.g., recalibration) to be compared. Application is made to a diagnostic model for deep vein thrombosis (DVT) and a prognostic model for breast cancer mortality.

RESULTS

In both examples, multivariate meta-analysis reveals that calibration performance is excellent on average but highly heterogeneous across populations unless the model's intercept (baseline hazard) is recalibrated. For the cancer model, the probability of "good" performance (defined by C statistic ≥0.7 and calibration slope between 0.9 and 1.1) in a new population was 0.67 with recalibration but 0.22 without recalibration. For the DVT model, even with recalibration, there was only a 0.03 probability of "good" performance.

CONCLUSION

Multivariate meta-analysis can be used to externally validate a prediction model's calibration and discrimination performance across multiple populations and to evaluate different implementation strategies.

The temperature-mortality relationship in China: An analysis from 66 Chinese communities

BACKGROUND

Previous studies examining temperature-mortality associations in China focused on a single city or a small number of cities. A multi-city study covering different climatic zones is necessary to better understand regional differences in temperature risk on mortality in China.

METHODS

Sixty-six communities from 7 regions across China were included in this study. We first used a Distributed Lag Non-linear Model (DLNM) to estimate community-specific effects of temperature on non-accidental mortality during 2006-2011. A multivariate meta-analysis was then applied to pool the estimates of community-specific effects.

RESULTS

A U-shaped curve was observed between temperature and mortality at the national level in China, indicating both low and high temperatures were associated with increased mortality risk. The overall threshold was at about the 75th percentile of the pooled temperature distribution. The relative risk was 1.61 (95% CI: 1.48-1.74) for extremely cold temperature (1st percentile of temperature), and 1.21 (95% CI: 1.10-1.34) for extreme hot temperature (99th percentile of temperature) at lag0-21 days. The temperature-mortality relationship is different for different regions. Compared with north China, south China had a higher minimum mortality temperature (MMT), and there was a larger cold effect in the more southern parts of China and a more pronounced hot effect in more northern parts.

CONCLUSIONS

Both cold and hot temperatures increase mortality risk in China, and the relationship varies geographically. Our findings suggest that public health policies for climate change adaptation should be tailored to the local climate conditions.

Multivariate meta-analysis helps examine the impact of outcome reporting bias in Cochrane rheumatoid arthritis reviews

OBJECTIVES

Outcome reporting bias (ORB) is a threat to validity of systematic reviews. Multivariate meta-analysis (MVMA) can potentially reduce the impact of ORB when outcomes are correlated. The aim of this study was to assess ORB in Cochrane systematic reviews of rheumatoid arthritis and to demonstrate how MVMA may examine its impact.

STUDY DESIGN AND SETTING

Reviews were assessed for ORB in relation to eight outcomes for rheumatoid arthritis using a nine-point classification system. Impact of ORB was assessed by comparing estimates from univariate meta-analysis and MVMA models.

RESULTS

ORB assessment was applied in 21 included reviews, and all contained missing data on at least one of the eight outcomes. ORB was highly suspected in 247 (22%) of the 1,118 evaluable outcomes from 155 assessable trials. MVMA and univariate results sometimes differed importantly. The maximum change in treatment effect estimate between MVMA and univariate meta-analysis approach was found to be 176% for one of the outcome considered.

CONCLUSION

ORB has the potential to affect the conclusions in meta-analyses. This could be avoided if trialists reported on all measured outcomes in full. If missing outcome data are unobtainable, MVMA is useful to examine the impact of missing outcomes and ORB on conclusions.

Joint synthesis of multiple correlated outcomes in networks of interventions

Multiple outcomes multivariate meta-analysis (MOMA) is gaining in popularity as a tool for jointly synthesizing evidence coming from studies that report effect estimates for multiple correlated outcomes. Models for MOMA are available for the case of the pairwise meta-analysis of two treatments for multiple outcomes. Network meta-analysis (NMA) can be used for handling studies that compare more than two treatments; however, there is currently little guidance on how to perform an MOMA for the case of a network of interventions with multiple outcomes. The aim of this paper is to address this issue by proposing two models for synthesizing evidence from multi-arm studies reporting on multiple correlated outcomes for networks of competing treatments. Our models can handle continuous, binary, time-to-event or mixed outcomes, with or without availability of within-study correlations. They are set in a Bayesian framework to allow flexibility in fitting and assigning prior distributions to the parameters of interest while fully accounting for parameter uncertainty. As an illustrative example, we use a network of interventions for acute mania, which contains multi-arm studies reporting on two correlated binary outcomes: response rate and dropout rate. Both multiple-outcomes NMA models produce narrower confidence intervals compared with independent, univariate network meta-analyses for each outcome and have an impact on the relative ranking of the treatments.