Measuring Inconsistency in Meta-Analyses

MRC Biostatistics Unit, Institute of Public Health, Cambridge CB2 2SR.
BMJ (online) (Impact Factor: 17.45). 10/2003; 327(7414):557-60. DOI: 10.1136/bmj.327.7414.557
Source: PubMed
Download full-text

Full-text

Available from: Jonathan J Deeks, Feb 03, 2015
  • Source
    • "The adjusted ORs or RRs, and 95 % CIs for each determinant of home or nursing home death compared to hospital death were extracted from the studies. Heterogeneity was measured using the I 2 statistic.[11]Meta-analyses were performed when more than one study was available for a given determinant and in the absence of considerable heterogeneity using the inverse variance method and a random effects model. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background: Most Canadians die in hospital, and yet, many express a preference to die at home. Place of death is the result of the interaction among sociodemographic, illness- and healthcare-related factors. Although home death is sometimes considered a potential indicator of end-of-life/palliative care quality, some determinants of place of death are more modifiable than others. The objective of this systematic review was to evaluate the determinants of home and nursing home death in adult patients diagnosed with an advanced, life-limiting illness. Methods: A systematic literature search was performed for studies in English published from January 1, 2004 to September 24, 2013 that evaluated the determinants of home or nursing home death compared to hospital death in adult patients with an advanced, life-limiting condition. The adjusted odds ratios, relative risks, and 95 % confidence intervals of each determinant were extracted from the studies. Meta-analyses were performed if appropriate. The quality of individual studies was assessed using the Newcastle-Ottawa scale and the body of evidence was assessed according to the GRADE Working Group criteria. Results: Of the 5,900 citations identified, 26 retrospective cohort studies were eligible. The risk of bias in the studies identified was considered low. Factors associated with an increased likelihood of home versus hospital death included multidisciplinary home palliative care, preference for home death, cancer as opposed to other diagnoses, early referral to palliative care, not living alone, having a caregiver, and the caregiver's coping skills. Conclusions: Knowledge about the determinants of place of death can be used to inform care planning between healthcare providers, patients and family members regarding the feasibility of dying in the preferred location and may help explain the incongruence between preferred and actual place of death.
    Preview · Article · Dec 2016 · BMC Palliative Care
  • Source
    • "Since advanced usage of the command can be challenging, users should utilize these outputs to ensure they are modelling the desired structure, and re-specify if necessary. Regarding setting heterogeneity levels and their interpretation, there are numerous practical guides that can prove helpful (Fletcher 2007; Higgins, Thompson, Deeks, and Altman 2003). Although we present information on data missingness mechanisms in the help file, interested users can find more details in an excellent overview provided by Horton and Kleinman (2007). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Simulations are a practical and reliable approach to power calculations, especially for multi-level mixed effects models where the analytic solutions can be very complex. In addition, power calculations are model-specific and multi-level mixed effects models are defined by a plethora of parameters. In other words, model variations in this context are numerous and so are the tailored algebraic calculations. This article describes ipdpower in Stata, a new simulations-based command that calculates power for mixed effects two-level data structures. Although the command was developed having individual patient data meta-analyses and primary care databases analyses in mind, where patients are nested within studies and general practices respectively, the methods apply to any two-level structure.
    Full-text · Article · Feb 2016
  • Source
    • "We combined our studies using the DerSimonian and Laird random effects model, because this method partially accounts for variability within and between studies[16]. We calculated the I 2 statistic to assess heterogeneity among studies, and classified values less than 50% as minimal , 50–75% as moderate, and >75% as substantial[14,17]. To assess clinical heterogeneity based on characteristics of study population and interventions , we performed subgroup analyses of: (i) patients given different VDRAs; (ii) patients with different baseline eGFRs (<60 mL/min/1.73 "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background: Vitamin D receptor activators (VDRAs) can protect against mineral bone disease, but they are reported to elevate serum creatinine (SCr) and may also reduce glomerular filtration rate (GFR). Methods: We conducted a systematic review and meta-analysis of randomized clinical trials (RCTs) to evaluate the effect of VDRAs on kidney function and adverse events. MEDLINE, EMBASE, the Cochrane Controlled Trials Register were searched for RCTs that evaluate vitamin D receptor activators (alfacalcidol, calcitriol, doxercalciferol, falecalcitriol, maxacalcitol and paricalcitol) up to March 2015. Results: We included 31 studies, all of which were performed between 1976 and 2015, which enrolled 2621 patients. Patients receiving VDRAs had lower eGFR (weighted mean difference WMD -1.29 mL/min /1.73 m2, 95% CI -2.42 to -0.17) and elevated serum creatinine (WMD 7.03 μmol/L, 95% CI 0.61 to 13.46) in sensitivity analysis excluding studies with dropout rate more than 30%. Subgroup analysis of the 5 studies that not use SCr-based measures did not indicated lower GFR in the VDRAs group(WMD -0.97 mL/min/1.73 m2, 95% CI -4.85 to 2.92). Compared with control groups, there was no difference in all-cause mortality (relative risk RR 1.41, 95% CI 0.58 to 3.80), cardiovascular disease (RR 0.84, 95% CI 0.42 to 1.71), and severe adverse events (RR 1.15, 95% CI 0.75 to 1.77) for the VDRAs groups. Episodes of hypercalcemia (RR 3.29, 95% CI 2.02 to 5.38) were more common in the VDRAs group than in the control group. Conclusions: Administration of VDRAs increased serum creatinine levels. Subgroup analysis of studies that did not use SCr-based measures did not indicate a lower GFR in the VDRA group. Future studies with non-SCr-based measures are needed to assess whether the mild elevations of serum creatinine are of clinical significance.
    Full-text · Article · Jan 2016 · PLoS ONE
Show more

Questions & Answers about this publication