openMetaAnalysis: collaborative and continuous

News 08/07/2021: registration data required in a new column 4. Enter No if not registered or NA if either unknown resistration or you do not intend a subgroup by registration. Note, this only addresses whether registration done and not whether done timely and without changes in primary outcome or other attributes that could bias the results. The rationale for this is re-analysis of the role of tramadol for chronic back pain (PMID 29204613). In this example, the data is further weakened by the outcome in the Lee study not matching the outcome in the trial registration.

Enter inputs

To perform a meta-analysis of interventions:

• Revise or replace the example data below as needed or
• Upload a csv file formatted like these examples with dichotomous or continous outcomes.
  1st row is column names? TRUE FALSE

Required columns: 1) study name, 2) year, 3) PMID, 4) trial registration number, 5) events or mean (SD) in intervention group, 6) total in intervention group, 7) events or mean (SD) in control group, 8) total in control group.
Optional columns for up to five cofactors
Separate columns with commas. Click here to add commas.

The initial data shown below is fictitious and is instructional.

Wells, 2012, 12345678, NA, 36, 100, 50, 100, high-dose, 30, Ford, 2013, 12345678, NA, 50, 100, 50, 100, low-dose, 35, Freud, 2014, 12345678, NA, 38, 100, 50, 100, high-dose, 40, Pavlov, 2015, 12345678, NCT12345678, 52, 100, 50, 100, low-dose, 45, Malthus, 2016, 12345678, NCT12345678, 40, 100, 50, 100, high-dose, 50,

Meta-analysis without subgroup Meta-analysis with subgroup by trial registration Meta-analysis with subgroup using cofactor 1 Meta-analysis with subgroup using cofactor 2 Meta-analysis with subgroup using cofactor 3 Meta-analysis with subgroup using cofactor 4 Meta-analysis with subgroup using cofactor 5 Meta-analysis with subgroup using cofactor 6 Meta-analysis with subgroup using cofactor 7 Meta-analysis with subgroup using cofactor 8 Meta-analysis with subgroup using cofactor 9 Meta-analysis with subgroup using cofactor 10 Meta-regression Funnel plot

Odds ratio Relative risk Mean difference Standardized mean difference Ratio of means

Meta-analysis settings

Sort order: Cofactor 1 (ascending) Cofactor 2 (ascending) Cofactor 3 (ascending) Cofactor 4 (ascending) Cofactor 5 (ascending) Cofactor 6 (ascending) Cofactor 7 (ascending) Cofactor 8 (ascending) Cofactor 9 (ascending) Cofactor 10 (ascending) Baseline value in control group Study name Weight (descending) Year (ascending)

 

Left label:

Analysis method:

Knapp-Hartung adjustment should be clicked unless heterogeneity is insig per PMID: 24727843. However, unclick this if yields wacky confidence intervals when less than 5 studies available.

Random effects with Knapp-Hartung adjustment Random effects Fixed effects (ok if heterogeneity is insignificant)

 

Right label:

Meta-regression settings

No labels for points Below points Left of points Above points Right of points

Select independent variable Study size Year Control rate Treatment rate Cofactor 1 Cofactor 2 Cofactor 3 Cofactor 4 Cofactor 5 Cofactor 6 Cofactor 7 Cofactor 8 Cofactor 9 Cofactor 10

 

Topic (for title - optional): None theme KU

Update Plot (below)

Example analyses

Teaching example 1: Soma (fictitious)

Fictitious example based on Brave New World

Soma dataset

Before making the plot

• Do you believe Soma will show a benefit when the studies are pooled?

Now, look at the plot

• What does I² indicate? (help)

Soma dataset, subgrouped by dose

Before making the plot

• Do you believe Soma will show a benefit when the studies are pooled?

Now, look at the plot

• What does I² indicate? (help)

Teaching example 2: back pain treatment with tramadol

From PMID 29204613. Note that the trial by Lee was prospectively registered, the primary outcome in the publication was not prospectively registered.

Tramadol dataset

Tramadol dataset, subgrouped

Meta-analysis of binary outcomes

Example: vitamin D for the prevention of osteoporotic fractures.

Include all studies; no subgroups When pooled, is there a significant difference? What does i-squared suggest?

Subgroup What does subgrouping reveal?

Meta-analysis of continuous outcomes

Example: hypertonic saline to reduce length of stay in hospitalized children with bronchiolitis

Include all studies; no subgroups When pooled, is there a significant difference? What does i-squared suggest?

About standardized mean differences

The standardized mean difference (SMD) is:

In the interpretation of SMD:

• 0.2 represents a small effect
• 0.5 a moderate effect
• > 0.8 a large effect

The 'standardization' allows pooling of different types of outcomes.

From: Cochrane Handbook. 12.6 Interpreting results from continuous outcomes (including standardized mean differences)

Meta-regression

Example: Early goal-directed therapy for reducing mortality from septic shock.

Meta-analysis of reduction in mortality

Meta-regression of reduction in mortality by baseline lactate level

Meta-regression of reduction in mortality by control event rate What is the relationship between reduction in mortality and the baseline serum lactate? How about with the morality in the control group?