Yes you can use LDA and different variant of LDA to process such a document having 200 words. You can use LDA on abstracts of an article and tweets.

Read more about it on link..

https://towardsdatascience.com/lda-topic-modeling-with-tweets-deff37c0e131

These thresholds are not useful and won't serve to increase confidence in these methods. I suggest thinking instead of validation methods to support that the method works on the data. Besides the number of words, the quality of the text, how was it created, and other factors are also crucial in deciding whether LDA might work on this set of data.

There are some studies about the influence of size of texts over the optimal number of topics of the corpus. While the size diminishes, the optimal number of topics is incremented.

LDA identifies the number of topics, but this result must be validated. Besides, there are techniques to calculate the optimal number of topics

For reference, see this article