A systematic review of the literature was undertaken comparing proton pump inhibitors with H2-receptor blockers for the prevention of gastro-intestinal bleeding in ICU patients.

a) Name the type of graph illustrated in the above figure. *(10% marks)*

b) What does it show? *(25% marks)*

c) What are the benefits of this type of analysis? *(25% marks)*

d) What are the disadvantages of this analysis? *(40% marks)*

## College Answer

a)

Forest plot

b)

Combining the trials together, PPI use results in an odds ratio of 0.35 or reduction in the risk of bleeding compared to H2RA. Alternatively, PPI use results in 65% reduction (1- 0.35) in bleeding.

c)

Combines small studies with limited power, increasing the number and thus the ability to pick up a positive effect. Small studies with low power (due to small effect, small numbers) run the risk of a Type II error.

d)

Individual studies might have different patient populations (with different risk of bleeding) or different definitions of outcome.

Individual studies might have been conducted with different degrees of rigour (blinding, etc.)

There is publication bias to positive studies so that negative studies are not reported. Need full disclosure how the studies were selected, their scientific grading, subgroup analyses and assessment of heterogeneity.

## Discussion

I have no idea whether the college actually used this exact image, but certainly the paper was correctly identified by LITFL. My hat is off to Chris Nickson, who managed to track down the exact PPI vs H2A study which had this exact forest plot and OR / RRR. It was indeed the Alhazzani study from 2013.

So:

**a) and b) **are reviewed in greater detail in the chapter on forest and box-and-whisker plots. In short:

- This is a forest plot.
- The horizontal lines – confidence intervals of the OR
- The position of the square – point estimate of the OR
- The size of the square – the weight of the study
- The vertical line: OR of 1 (no association)
- If the CI of the summed results crosses the vertical line, the treatment is no more effective than control.
- This study shows that PPIs are better than H2As in reducing the risk of bleeding.

**c) and d)**

**Advantages of meta-analysis**

- A more objective appraisal of evidence
- Reduces the probability of false negative results
- May explain heterogeneity between the results of different studies
- Avoids Simpson’s paradox, in which a consistent effect in constituent trials is reversed when results are simply pooled.

**Disadvantages of meta-analysis**

- Frustrated by heterogeneity of population samples and methodologies
- Selection of studies may be biased
- Negative studies are rarely published, and thus may not be included
- The meta-analysis uses summary data rather than individual data

### References

Alhazzani, Waleed, et al. "Proton pump inhibitors versus histamine 2 receptor antagonists for stress ulcer prophylaxis in critically ill patients: a systematic review and meta-analysis*." *Critical care medicine* 41.3 (2013): 693-705.

Methodological Expectations of Cochrane Intervention Reviews

Schriger, David L., et al. "Forest plots in reports of systematic reviews: a cross-sectional study reviewing current practice." *International journal of epidemiology*39.2 (2010): 421-429.

Lewis, Steff, and Mike Clarke. "Forest plots: trying to see the wood and the trees." *Bmj* 322.7300 (2001): 1479-1480.

Anzures‐Cabrera, Judith, and Julian Higgins. "Graphical displays for meta‐analysis: An overview with suggestions for practice." *Research Synthesis Methods* 1.1 (2010): 66-80.

Cochrane: "Considerations and recommendations for

figures in Cochrane reviews: graphs of statistical data" 4 December 2003 (updated 27 February 2008)

Reade, Michael C., et al. "Bench-to-bedside review: Avoiding pitfalls in critical care meta-analysis–funnel plots, risk estimates, types of heterogeneity, baseline risk and the ecologic fallacy." *Critical Care* 12.4 (2008): 220.

DerSimonian, Rebecca, and Nan Laird. "Meta-analysis in clinical trials."*Controlled clinical trials* 7.3 (1986): 177-188.

Biggerstaff, B. J., and R. L. Tweedie. "Incorporating variability in estimates of heterogeneity in the random effects model in meta-analysis." *Statistics in medicine* 16.7 (1997): 753-768.

The Cochrane Handbook: 9.5.4 "Incorporating heterogeneity into random-effects model"