What to add to nothing? Use and avoidance of continuity corrections in meta‐analysis of sparse data

Sweeting, Michael; Sutton, Alex J.; Lambert, Paul C.

doi:10.1002/sim.1761

Cited by 1,436 publications

(1,143 citation statements)

References 39 publications

(48 reference statements)

Supporting

Mentioning

1,109

Contrasting

Unclassified

Order By: Relevance

“…It thereby provides a more exact likelihood specification from which to make inferences 11, 58. The one‐stage approach may additionally, but perhaps more critically, be different from the two‐stage approach if there are zero event counts for binary outcomes because the two stage method typically requires a continuity correction (such as +0.5 to cells in the available two‐by‐two tables 59, 60, 61), whereas this is not necessary with the one‐stage logistic regression model 11. The issue that effect estimates are not normally distributed is possibly less of a concern than the issue of zero cells and subsequent continuity corrections, as the latter actually influence the magnitude of effect estimates and their estimated variances, which may introduce bias that then feeds in to the second stage.…”

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

confidence: 99%

“…As described in Reason I, stage 2 in a two‐stage approach is problematic when outcomes are rare and studies are small and may even be problematic when some studies have unbalanced sample size in the treatment and control group numbers 10, 59, 60. To address this, alternative two‐stage approaches have been proposed, which use a different weighting scheme to the inverse variance method, such as the Peto method 65 and the Mantel–Haenszel method 66, which also avoid the use of continuity corrections when there are zero cells.…”

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

confidence: 99%

See 1 more Smart Citation

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

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

confidence: 99%

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

show abstract

“…For all outcomes in which ORs were used as the estimates, rates of event in patient‐years were used rather than number of events alone because the length of follow‐up of each trial varied. If there were no outcome events in one of the treatment groups, we applied the treatment arm continuity correction (the reciprocal value of the opposite treatment group size) 19. To test the robustness of the findings for cardiovascular outcomes, we performed subgroup analysis confined to RCTs.…”

Section: Methodsmentioning

confidence: 99%

Cardiovascular Safety, Long‐Term Noncardiovascular Safety, and Efficacy of Sodium–Glucose Cotransporter 2 Inhibitors in Patients With Type 2 Diabetes Mellitus: A Systemic Review and Meta‐Analysis With Trial Sequential Analysis

Zhang

Zhu

Chen

et al. 2018

JAHA

104

View full text Add to dashboard Cite

BackgroundThe cardiovascular and long‐term noncardiovascular safety and efficacy of SGLT2 (sodium–glucose cotransporter 2) inhibitors have not been well documented.Methods and ResultsFor cardiovascular outcomes, we performed a meta‐analysis with trial sequential analysis of randomized controlled trials and adjusted observational studies, each with a minimum of 26 weeks and 2000 patient‐years of follow‐up. For long‐term noncardiovascular safety and efficacy outcome analyses, we included only randomized controlled trials with at least 2 years and 1000 patient‐years of follow‐up. Five studies with 351 476 patients were included in cardiovascular outcomes analysis. Meta‐analyses showed that SGLT2 inhibitors significantly reduced the risks of major adverse cardiac events (hazard ratio [HR]: 0.80; 95% confidence interval [CI], 0.69–0.92; P=0.002), all‐cause mortality (HR: 0.67; 95% CI, 0.54–0.84; P<0.001), cardiovascular mortality (HR: 0.77; 95% CI, 0.60–0.98; P=0.03), nonfatal myocardial infarction (HR: 0.86; 95% CI, 0.76–0.98; P=0.02), hospitalization for heart failure (HR: 0.62; 95% CI, 0.55–0.69; P<0.001), and progression of albuminuria (HR: 0.68; 95% CI, 0.58–0.81; P<0.001). No significant difference in nonfatal stroke was found. Analyses limited to randomized controlled trials showed similar findings. Trial sequential analysis provided firm evidence of a 20% reduction in major adverse cardiac events, all‐cause mortality, and hospitalization for heart failure with SGLT2 inhibitors, but evidence remains inconclusive for cardiovascular mortality. Nine randomized controlled trials contributed to long‐term noncardiovascular and efficacy analyses. SGLT2 inhibitors reduced incidence of hypoglycemia and acute kidney injury but increased the risks of urinary tract and genital infections.Conclusions SGLT2 inhibitors showed remarkable cardiovascular‐ and renal‐protective effects and good long‐term noncardiovascular safety with sustained efficacy.

show abstract

“…We used meta-analysis of RCTs for evaluating drug safety based on all available trials. 46 We analyzed sparse adverse effects data with various statistical methods 43,[47][48][49][50][51] for robustness by comparing statistical significance and magnitude of the harms. In cases of multi-arm trials, we created a single pair-wise comparison.…”

Section: Methodsmentioning

confidence: 99%

Preventive Pharmacologic Treatments for Episodic Migraine in Adults

et al. 2013

View full text Add to dashboard Cite

ELIGIBILITY CRITERIA: English-language randomized controlled trials (RCTs) of preventive drugs compared to placebo or active treatments examining rates of ≥50 % reduction in monthly migraine frequency or improvement in quality of life. STUDY APPRAISAL AND SYNTHESIS METHODS:We assessed risk of bias and strength of evidence and conducted random effects meta-analyses of absolute risk differences and Bayesian network meta-analysis. RESULTS: Of 5,244 retrieved references, 215 publications of RCTs provided mostly low-strength evidence because of the risk of bias and imprecision. RCTs examined 59 drugs from 14 drug classes. All approved drugs, including topiramate (9 RCTs), divalproex (3 RCTs), timolol (3 RCTs), and propranolol (4 RCTs); offlabel beta blockers metoprolol (4 RCTs), atenolol (1 RCT), nadolol (1 RCT), and acebutolol (1 RCT); angiotensin-converting enzyme inhibitors captopril (1 RCT) and lisinopril (1 RCT); and angiotensin II receptor blocker candesartan (1 RCT), outperformed placebo in reducing monthly migraine frequency by ≥50 % in 200-400 patients per 1,000 treated. Adverse effects leading to treatment discontinuation (68 RCTs) were greater with topiramate, off-label antiepileptics, and antidepressants than with placebo. Limited direct evidence as well as frequentist and exploratory network Bayesian meta-analysis showed no statistically significant differences in benefits between approved drugs. Off-label angiotensin-inhibiting drugs and beta-blockers were most effective and tolerable for episodic migraine prevention. LIMITATIONS:We did not quantify reporting bias or contact principal investigators regarding unpublished trials. CONCLUSIONS: Approved drugs prevented episodic migraine frequency by ≥50 % with no statistically significant difference between them. Exploratory network meta-analysis suggested that off-label angiotensin-inhibiting drugs and beta-blockers had favorable benefit-to-harm ratios. Evidence is lacking for longterm effects of drug treatments (i.e., trials of more than 3 months duration), especially for quality of life.KEY WORDS: migraine; evidence based medicine; adverse drug effects.

show abstract

What to add to nothing? Use and avoidance of continuity corrections in meta‐analysis of sparse data

Abstract: Many routinely used summary methods provide widely ranging estimates when applied to sparse data with high imbalance between the size of the studies' arms. A sensitivity analysis using several methods and continuity correction factors is advocated for routine practice.

Cited by 1,436 publications

References 39 publications

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

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

Cardiovascular Safety, Long‐Term Noncardiovascular Safety, and Efficacy of Sodium–Glucose Cotransporter 2 Inhibitors in Patients With Type 2 Diabetes Mellitus: A Systemic Review and Meta‐Analysis With Trial Sequential Analysis

Preventive Pharmacologic Treatments for Episodic Migraine in Adults

Contact Info

Product

Resources

About