Is there any reference or guideline regarding the minimum number of studies for conducting a meta-analysis?
There's actually no minimal (or maximal) number needed for a pooled analysis (e.g. meta-analysis). I've conducted systematic reviews including one trial to over 250 trials. As long as you can analyze data from two separate sources then it is a meta-analysis. If you have a limited number of studies contributing to the meta-analysis then some analyses are not advisable (e.g. publication bias tests, meta-regressions, subgroup/ sensitivity analyses, etc.) due to low power.
Hope this helps.