There are many types of Meta-analysis, did you make it?

Author | Zhang Yaotian

There are many types of Meta-analysis

The common ones are as follows

1. Conventional Meta-analysis

This type of Meta-analysis is based on the merging of effect sizes of randomized controlled trials, non-randomized controlled trials, cohort studies, and case-control studies. This type of Meta-analysis is the most mature method and has the largest number of publications.

Example : Effect of sodium-glucose co-transporter 2 (SGLT-2) inhibitors on cardiovascular outcomes in patients with type 2 diabetes.

(From: Wu JH, et al. Lancet Diabetes Endocrinol. 2016;4:411-9.)

2. Individual Data Meta-analysis

Individual data meta-analysis has been referred to as the gold standard for systematic reviews. Instead of performing Meta-analysis using summarized data from published findings, it obtains raw data from the authors of the original study for each subject and performs Meta-analysis on these data. However, this type of article is not for the average researcher and is suitable for discipline leaders to lead the operation.

Example : The relationship between BMI and all-cause mortality.

(From: Global BMI Mortality Collaboration. Lancet. 2016;388:776-86.)

3. Meta-analysis of single-group rates

Meta-analysis also allows for the combination of single-group rates. The outcome indicators for this type of Meta-analysis are mostly incidence, prevalence, morbidity, mortality, detection, awareness, infection, etc. The original studies are mostly cross-sectional. For Meta-analyses of single group rates, the difficulty lies in controlling for heterogeneity. Subgroup analysis and Meta-regression analysis are important methods to deal with heterogeneity.

Example : What proportion of children with disabilities have experienced sexual violence?

(From: Jones L, et al. Lancet 2012;380:899-907.)

4. Diagnostic test Meta-analysis

Evaluating the diagnostic value of a certain measure for a disease, mainly evaluating the sensitivity, specificity, and the area under the ROC curve.

For example: the diagnostic value of calcitoninogen and C-reactive protein for early abdominal infection in patients undergoing elective colorectal surgery.

(From: Cousin F, et al. Ann Surg. 2016;264:252-6.)

5. Cumulative Meta-analysis

Cumulative Meta-analysis is the sequential addition of individual included studies, in a certain order (such as time of publication, sample size, and quality score of the study, etc.), to a multiple Meta-analysis. Meta-analysis is performed once every time a new study is included, which reflects the dynamic trend of the study results and assesses the impact of individual studies on the composite results.

Example : Cardiovascular risk of rofecoxib.

(From: Jüni P, et al. Lancet. 2004;364:2021-9.)

6. Sequential Meta-analysis

Sequential Meta-analysis is similar to Cumulative Meta-analysis, except that it is considered as an interim analysis at the time each new study is included. analyses.) Sequential Meta-analysis overcomes the shortcomings of traditional Meta-analysis, especially cumulative Meta-analysis, and maximizes the control of Type I error (alpha).

Example: Cancer risk of antihypertensive drugs.

(From: Bangalore S, et al. Lancet Oncol 2011; 12: 65-82.)

7. Dose-response meta-analysis

Dose-response meta-analysis improves the statistical power of the meta-analysis by combining multiple original studies of dose-response relationships.

Example: Does high blood glucose increase the risk of pancreatic cancer?

(From: Liao WC, et al. BMJ. 2015;349:g7371.)

8. Reticulated Meta-analysis

If there are a range of drugs available to treat a particular disease, but there are no or very few comparisons of a particular number of drugs with each other, indirect comparisons are needed in this case. Reticulated Meta-analysis focuses on indirect comparisons, where all interventions in the same body of evidence are evaluated and ranked together at the same time.

Example : Effectiveness and safety of hypoglycemic drugs.

(From: Palmer SC, et al. JAMA. 2016;316:313-24.)

9. Other types of Meta-analysis

① Meta-analysis of P-value alone

When the included studies do not give an effect value but only P-value, and need to be merged, it can be considered that the P-value alone can be considered. However, Meta-analysis of P-values alone was not considered. However, Meta-analysis of P-values alone has many shortcomings and is not recommended by most people.

② Prospective Meta-analysis

Prospective Meta-analysis refers to a kind of Meta-analysis that systematically searches, evaluates, and develops inclusion and exclusion criteria before the results of the included studies are available.

③ Other types of Meta-analysis

Such as Meta-analysis of adverse effects, Meta-analysis of cost-effectiveness/utility/benefit, Meta-analysis of patient-reported outcomes, Meta-analysis of genome-wide association studies, and pooled analysis of Meta-analysis.

10. Meta-regression

Meta-analysis may be needed to evaluate the magnitude and source of heterogeneity between studies when heterogeneity exists between original studies and cannot be explained by subgroup analysis.

Example: Effect of blood pressure lowering on cardiovascular morbidity and mortality.

(From: Ettehad D, et al. Lancet. 2016;387:957-67.)

Article adapted from

Zeng XT, et al. Meta-analysis series one: types of meta-analysis. Chinese Journal of Evidence-Based Cardiovascular Medicine. 2012;4:3-5.