Reviewing citations and abstracts identified by searching will be very time consuming, first because of the volume of citations identified and second because the information needed to judge eligibility may not be reported in the title or abstract.
Collect data as for a randomized trial (i.e. details of study, study population, sample size recruited, sample size analysed, etc).
Collect data about what researchers did (NRSMG checklist, or similar).
Collect data about the confounding factors considered.
Collect data about the comparability of groups on confounding factors considered.
Collect data the methods used to control for confounding.
Collect data about multiple effect estimates (both unadjusted and adjusted estimates, if available).