Could you provide a link to the study? I’d like to take a look at it. Perhaps we could discuss it in a separate thread.
Having no intuition or experience in this particular area, I’d be concerned with automatically granting this most recent study higher credibility than the prior studies (even acknowledging your valid criticisms of methodological issues related to definition of these medical conditions).
It would seem there are severe problems with the measurement/data generation process in critical care clinical trials. This calls into question the sampling models for the experiments, and the trustworthiness of the reported statistics. Despite that, there is still reason to believe something can be learned, even if it is only to design better experiments.
Following the guidelines discussed in the following thread on meta-analysis, I would focus on identifying sources of heterogeneity using meta-regression if data permitted it.