About half the time, just using forward and backward searching will find every article in the field (Robinson et al., 2014). This study in the BMJ suggests that database searches usually only find about 25% of the final sample in a systematic review; forward and backward searching often finds more than 50%.
In general, it's best to download these citation searches, remove duplicates, and screen them in Abstrackr or Covidence so they're tracked. But, most reviewers won't ask. Therefore, sometimes you can just do it as you go: the person reviewing reference lists can screen for "possibly relevant" while reviewing the list. If they find anything, download and add for full-text screening.