A comparison of statistical methods for meta-analysis

Department of Mathematics and Statistics, Richard Berry Building, The University of Melbourne, Victoria 3010, Australia.
Statistics in Medicine (Impact Factor: 2.04). 03/2001; 20(6):825-40. DOI: 10.1002/sim.650
Source: PubMed

ABSTRACT Meta-analysis may be used to estimate an overall effect across a number of similar studies. A number of statistical techniques are currently used to combine individual study results. The simplest of these is based on a fixed effects model, which assumes the true effect is the same for all studies. A random effects model, however, allows the true effect to vary across studies, with the mean true effect the parameter of interest. We consider three methods currently used for estimation within the framework of a random effects model, and illustrate them by applying each method to a collection of six studies on the effect of aspirin after myocardial infarction. These methods are compared using estimated coverage probabilities of confidence intervals for the overall effect. The techniques considered all generally have coverages below the nominal level, and in particular it is shown that the commonly used DerSimonian and Laird method does not adequately reflect the error associated with parameter estimation, especially when the number of studies is small.

  • [Show abstract] [Hide abstract]
    ABSTRACT: Non-adherence to cardiovascular medications is a problem worldwide, even in Australia, which has a socialized medical system, Medicare. The aim of this systematic review was to evaluate the burden of non-adherence to cardiovascular medications and factors thereof in Australia. Pubmed, Embase, CINAHL, PsycInfo, Cochrane Library databases were searched. Articles were included if they were in English, peer-reviewed and provided empirical data on adherence to cardiovascular medication for an Australian cohort. A meta-analysis of prevalence of medication non-adherence using the double arcsine square root transformed proportion was undertaken. Studies were pooled in homogenous prevalence groups and factors that differed across groups were ascertained. Five studies, including eight datasets and 76,867 subjects were analyzed. Three more or less homogenous prevalence categories were discernable: low [19 %, 95 % confidence interval (CI) 15-24], moderate (26 %, 95 % CI 23-29) and high (43 %, 95 % CI 43-44; this was a single study) prevalence of non-adherence. There were minimal clear patterns across groups in relation to typical factors of non-adherence (patient, condition, healthcare system or socioeconomic factors). Measurements used for non-adherence were similar for six of the eight included datasets, suggesting this did not affect prevalence of non-adherence or inclusion in a prevalence group. Non-adherence to cardiovascular medications is a serious problem in the aging Australian setting with an overall prevalence of between 14 and 43 %. The lack of patterns in the typical factors of non-adherence suggests that another factor, such as patients' beliefs about their conditions and medications, may be playing a stronger role in their non-adherence than clinical or sociodemographic factors. This is an area for further research.
    Drugs & Aging 03/2015; DOI:10.1007/s40266-015-0245-1 · 2.50 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Clostridium difficile infection (CDI) has been extensively described in healthcare settings; however, risk factors associated with community-acquired (CA) CDI remain uncertain. This study aimed to synthesize the current evidence for an association between commonly prescribed medications and comorbidities with CA-CDI. A systematic search was conducted in 5 electronic databases for epidemiologic studies that examined the association between the presence of comorbidities and exposure to medications with the risk of CA-CDI. Pooled odds ratios were estimated using 3 meta-analytic methods. Subgroup analyses by location of studies and by life stages were conducted. Twelve publications (n=56,776 patients) met inclusion criteria. Antimicrobial (odds ratio, 6.18; 95% CI, 3.80-10.04) and corticosteroid (1.81; 1.15-2.84) exposure were associated with increased risk of CA-CDI. Among the comorbidities, inflammatory bowel disease (odds ratio, 3.72; 95% CI, 1.52-9.12), renal failure (2.64; 1.23-5.68), hematologic cancer (1.75; 1.02-5.68), and diabetes mellitus (1.15; 1.05-1.27) were associated with CA-CDI. By location, antimicrobial exposure was associated with a higher risk of CA-CDI in the United States, whereas proton-pump inhibitor exposure was associated with a higher risk in Europe. By life stages, the risk of CA-CDI associated with antimicrobial exposure greatly increased in adults older than 65 years. Antimicrobial exposure was the strongest risk factor associated with CA-CDI. Further studies are required to investigate the risk of CA-CDI associated with medications commonly prescribed in the community. Patients with diarrhea who have inflammatory bowel disease, renal failure, hematologic cancer, or diabetes are appropriate populations for interventional studies of screening. Infect Control Hosp Epidemiol 2014;00(0):1-10.
    Infection Control and Hospital Epidemiology 02/2015; 36(2):132-41. DOI:10.1017/ice.2014.39 · 3.94 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The prevalence of osteoporosis in men with prostate cancer (PCa) on androgen deprivation therapy (ADT) is well documented, with up to 53 % affected by this bone condition. However, there has been less emphasis on the burden of severe bone loss in men with PCa but not undergoing ADT. Therefore, the purpose of this meta-analysis is to compile evidence from the literature on the bone health of hormone-naïve PCa patients and to compare it to the bone health of men with PCa on ADT. Three databases were searched for the relevant literature published from 1990 until January 2014. The pooled prevalence of osteoporosis, low bone mass, and normal bone mass were estimated for this patient group and compared with similar subgroups from a previously published meta-analysis. The prevalence of osteoporosis varies from 4 to 38 % in hormone-naïve PCa patients, and men with more advanced disease have a higher prevalence of osteoporosis. Men with PCa on ADT have poorer bone health than their hormone-naïve counterparts, but the trend toward poorer bone health with metastatic disease remains. In conclusion, it was found that men with PCa experience poor bone health prior to treatment with ADT. These results suggest that all men with PCa should have regular bone health monitoring, whether they commence ADT or not, in order to prevent or indeed minimize the morbidity that accompanies osteoporosis.

Full-text (2 Sources)

Available from
Sep 19, 2014