I would suggest a sensitivity test, so that you run the meta-analyses both with the percentage score included and with it included. If its inclusion impacts the final outcome (and there is no other reason for this than the the effect size measure), then I would check 1) if the effect size measures are comparable (it is not clear from your question which effect size measures you will use), and 2) consider removing the percentage change score.

If you have the percentage change score and the baseline, then just use (percentage change score) * baseline divided by 100

Dear Jiaqi, you can use the percent changes if you are using the Standardized Mean Difference (SMD) method - this is commonly done. With the SMD method, it does not matter whether you have the percent changes or the original units. If you are using the Mean Difference (MD) method, then I agree with Adrian.

you can try to contact the author the get the mean (SD) score including both baseline or post-interventions, if you have only the change score of percentages, you can calculate the mean score referring to the handbook of Cochrane handbook of systematic review. in an easy way, it's excluding this article from your meta-analysis

Turn percentages into absolute numbers and you will have dichotomous data that you can use in the meta-analysis

I have the same problem as you, some colleagues think it is not appropriate to use SMD to merge both percentages and values with units, and I was wondering if you have found a suitable method recently. In addition, if the original value is multiplied by the percentage directly, does it make sense to derive the SD.