How to combined positive/negative predictive value in a meta-analysis

[Yang Cheng]

Hi everybody,
We are using Stata 16.0 to conduct a Meta-analysis of diagnostic accuracy studies. The table is

study	year	tp	fp	fn	tn
a1	2004	37	214	1	50
a2	2009	79	511	5	137
a3	2009	22	217	0	75
a4	2012	8	123	0	16
a5	2016	32	269	0	198

Command
midas tp fp fn tn, res(all)

The resulting table is
Parameter Estimate 95% CI

Sensitivity 0.97 [ 0.91, 0.99]

Specificity 0.23 [ 0.16, 0.33]

Positive Likelihood Ratio 1.3 [ 1.1, 1.4]

Negative Likelihood Ratio 0.13 [ 0.03, 0.49]

Diagnostic Odds Ratio 10 [ 2, 41]

But we don't know the commands to get the combined positive/negative predictive value, 95%CI and forest graphs. Could you please tell us how. Thank you for your help!
Kindly, Yang Cheng

[Leonardo Guizzetti]

-metandi- is another user-contributed meta-analysis program that implements a bivariate model for diagnostic accuracy studies.

I think the main reason you will be unlikely to find any such programs to implement such analyses for PPV/NPV are because these values can only be combined with valid population prevalence estimates, and these are probably not available from literature searches. I think then you would need to write your program to do such an analysis if you did have prevalence estimates for each study.