Using a fine-tuned model for a different dataset

I have a dataset of different sentences from news articles which I need to classify by their sentiment. For that goal I'm planning to use a fine-tuned model which was fine-tuned on different datasets, for example various comments from forums, reviews, tweets. However, news articles are supposedly quite different from that dataset as they are usually more neutral. I understand that a correct way to approach this issue would be by training a model on my own labeled dataset, however for one reason or another I have to use a fine-tuned model. How bad should I expect the results to be? Is it rather a slight drop in accuracy or results would be completely wrong?