There is a definition of the bandwidth for the filter and amplifiers such that it is limited by -3 dB points. At these points the power drops to half its pass band value. But purposely you can choose any other level according to the allowed fluctuations of the signal power in the pass band. It seems that you have large ripples in the pass band especially at low drain currents if i could read your curve order correctly.
I think if you allow 6 dBs variations in the passband you will be able to define an approximate bandwidth easier.
It depend in you applications.
The other solution is to reduce the ripple in the pass band to get more flat pass band response.
thanks for your points. I considered 5 dBs variations in the passband to approximate the bandwidth. But I think you also agree with that idea to allow 6 dBs variations.
In my opinion, considering the trends of the frequency respons of the filters, you could also find the frequency (fc) for which the ratio between the integral of the squared magnite up to fc and that total turns out to be equal to- 3 dB (half power).