A random‐effects regression model for meta‐analysis

Berkey, Catherine S.; Hoaglin, David C.; Mosteller, Frederick; Colditz, Graham A.

doi:10.1002/sim.4780140406

Cited by 747 publications

(629 citation statements)

References 8 publications

Supporting

Mentioning

624

Contrasting

Unclassified

Order By: Relevance

“…Results from all studies were then pooled using random-effects models. (17) To assess heterogeneity, we calculated the Q statistic, a measure of the statistical significance of heterogeneity, and the I 2 index, a measure of the extent of heterogeneity. (18) To assess potential publication bias, we used Begg's and Egger's tests and Begg's funnel plot (19,20) ; no evidence of bias was seen.…”

Section: Methodsmentioning

confidence: 99%

Milk intake and risk of hip fracture in men and women: A meta-analysis of prospective cohort studies

Bischoff‐Ferrari

Dawson‐Hughes

Baron

et al. 2010

Journal of Bone and Mineral Research

124

View full text Add to dashboard Cite

Milk contains calcium, phosphorus, and protein and is fortified with vitamin D in the United States. All these ingredients may improve bone health. However, the potential benefit of milk on hip fracture prevention is not well established. The objective of this study was to assess the association of milk intake with risk of hip fracture based on a meta-analysis of cohort studies in middle-aged or older men and women. Data sources for this study were English and non-English publications via Medline (Ovid, PubMed) and EMBASE search up to June 2010, experts in the field, and reference lists. The idea was to compare prospective cohort studies on the same scale so that we could calculate the relative risk (RR) of hip fracture per glass of milk intake daily (approximately 300 mg calcium per glass of milk). Pooled analyses were based on random effects models. The data were extracted by two independent observers. The results show that in women (6 studies, 195,102 women, 3574 hip fractures), there was no overall association between total milk intake and hip fracture risk (pooled RR per glass of milk per day ¼ 0.99; 95% confidence interval [CI] 0.96-1.02; Q-test p ¼ .37). In men (3 studies, 75,149 men, 195 hip fractures), the pooled RR per daily glass of milk was 0.91 (95% CI 0.81-1.01). Our conclusion is that in our meta-analysis of cohort studies, there was no overall association between milk intake and hip fracture risk in women but that more data are needed in men. ß

show abstract

Section: Methodsmentioning

confidence: 99%

Milk intake and risk of hip fracture in men and women: A meta-analysis of prospective cohort studies

Bischoff‐Ferrari

Dawson‐Hughes

Baron

et al. 2010

Journal of Bone and Mineral Research

124

View full text Add to dashboard Cite

show abstract

“…Therefore, differences in one‐stage and two‐stage meta‐analysis results may be a consequence of different estimation methods. For example, it is well known that the ML method tends to underestimate the between‐study heterogeneity 69, 70, 71, 72, 73, 74 and that REML is preferred in comparison. However, for non‐continuous outcomes, one‐stage models typically use ML estimation as it is usually the only option.…”

Section: Key Reasons Why Meta‐analysis Results May Differ For the Onementioning

confidence: 99%

Meta-analysis using individual participant data: one-stage and two-stage approaches, and why they may differ

2016

View full text Add to dashboard Cite

Meta‐analysis using individual participant data (IPD) obtains and synthesises the raw, participant‐level data from a set of relevant studies. The IPD approach is becoming an increasingly popular tool as an alternative to traditional aggregate data meta‐analysis, especially as it avoids reliance on published results and provides an opportunity to investigate individual‐level interactions, such as treatment‐effect modifiers. There are two statistical approaches for conducting an IPD meta‐analysis: one‐stage and two‐stage. The one‐stage approach analyses the IPD from all studies simultaneously, for example, in a hierarchical regression model with random effects. The two‐stage approach derives aggregate data (such as effect estimates) in each study separately and then combines these in a traditional meta‐analysis model. There have been numerous comparisons of the one‐stage and two‐stage approaches via theoretical consideration, simulation and empirical examples, yet there remains confusion regarding when each approach should be adopted, and indeed why they may differ.In this tutorial paper, we outline the key statistical methods for one‐stage and two‐stage IPD meta‐analyses, and provide 10 key reasons why they may produce different summary results. We explain that most differences arise because of different modelling assumptions, rather than the choice of one‐stage or two‐stage itself. We illustrate the concepts with recently published IPD meta‐analyses, summarise key statistical software and provide recommendations for future IPD meta‐analyses. © 2016 The Authors. Statistics in Medicine published by John Wiley & Sons Ltd.

show abstract

“…Between-study variation was modelled as a random effect, and heterogeneity over studies was assessed by the significance of the between-study variance (Berkey et al, 1995;Takkouche et al, 1999). The within-study variance was taken to be the estimated variance of the log relative risks for each study, giving more precise estimates greater weights in the summary measure (Greenland, 1998).…”

Section: Discussionmentioning

confidence: 99%

“…The regression was weighted by the inverse variance of the log relative risk for each category. The correlation between categories was estimated using a previous method (Greenland and Longnecker, 1992).We used a mixed effects weighted regression model to combine estimates from BMI categories from the individual studies.Between-study variation was modelled as a random effect, and heterogeneity over studies was assessed by the significance of the between-study variance (Berkey et al, 1995;Takkouche et al, 1999). The within-study variance was taken to be the estimated variance of the log relative risks for each study, giving more precise estimates greater weights in the summary measure (Greenland, 1998).…”

mentioning

confidence: 99%

Obesity and renal cell cancer – a quantitative review

et al. 2001

View full text Add to dashboard Cite

Although many epidemiological studies have shown that obesity increases the risk of renal cell cancer among women, the evidence for men is considered weaker (Wolk et al, 1996;McLaughlin and Lipworth, 2000).The incidence of renal cell cancer, the predominant type of kidney cancer, has been increasing both in the US and in most Western countries (Black et al, 1997;Liu et al, 1997). Today, renal cell cancer accounts for about 2% of cancers in the US (Landis et al, 1999), as well as worldwide (Parkin, 1998), with 30 000 cases occurring in the US in 1999 (Landis et al, 1999). The incidence varies more than 10-fold over the world. The highest rates are found in North America and Europe and the lowest in Asia (Parkin, 1998). Renal cell cancer occurs about twice as often among men, as among women.Obesity has also been increasing throughout the world. More than half of the adult US population is considered to have an excess weight (body mass index, BMI ≥ 25.0 kg m -2 ) and nearly one quarter are clinically obese (BMI ≥ 30.0 kg m -2 ) (Flegal et al, 1998). The increasing prevalence of obesity might therefore, at least partly, explain the increasing incidence of renal cell cancer.The purpose of this review was to evaluate the existing evidence that obesity increases the risk of renal cell cancer among both men and women. We conducted a quantitative summary analysis to estimate the magnitude of the association taking into account sex and different study characteristics, such as study design and size. We also investigated the possible effect modification between obesity and these factors. MATERIALS AND METHODS Literature reviewWe identified studies investigating the relation between obesity and kidney cancer, available in MEDLINE 1966 through 1998. We used the MeSH-terms 'kidney neoplasms' plus 'obesity', 'body weight' or 'body mass index' and selected original epidemiological studies. Additional studies were identified from systematical examinations of the list of references in the identified articles and previous reviews (Wolk et al, 1996;McLaughlin and Lipworth, 2000). Inclusion criteriaIn adults, cancer of the kidney is classified as either cancer of the parenchyma (renal cell) or the renal pelvis. In many descriptive and some analytic studies tumours of the renal parenchyma and the renal pelvis are combined and the term kidney cancer is used. Since renal cell cancer is the predominant type, responsible for more than 80% of all adult kidney neoplasms (Devesa et al, 1990), studies unable to disentangle the 2 entities were included in our review.In total, we identified 30 studies on obesity and renal cell cancer risk published between 1966 and 1998 for our review. Each studybase was eligible only once. When multiple reports were available for the same study-base we chose the one analysing incident renal cell cancer as outcome, and/or defining obesity in terms of body mass index (BMI, kg m -2 ) and in more detail and/or with the greater number of cases. We excluded 6 studies (Whittemore et al, 1985;Lindblad et al, 1994;Mel...

show abstract

A random‐effects regression model for meta‐analysis

Cited by 747 publications

References 8 publications

Milk intake and risk of hip fracture in men and women: A meta-analysis of prospective cohort studies

Milk intake and risk of hip fracture in men and women: A meta-analysis of prospective cohort studies

Meta-analysis using individual participant data: one-stage and two-stage approaches, and why they may differ

Obesity and renal cell cancer – a quantitative review

Contact Info

Product

Resources

About