In my opinion the article has serious methodological flaws. The main problem is "confounding by indication", which is characteristic of observational studies, especially if they are retrospective. 

By comparing the cohort of medicated and non-medicated depressed patients, and by not controlling for the severity of the cases (an unforgivable error), the results of the comparison cannot be attributed to the treatment, but to the characteristics of the patient, which is what determines whether they are treated or not. 

In other words, the lack of differences between the two groups cannot be attributed to the antidepressants, as the treated patients were most likely more severe than the untreated ones. 

It is like comparing patients with covid-19 treated with corticosteroids and without: mortality is higher in patients treated with corticosteroids, but corticosteroids are given to the most severe patients precisely to save their lives.

I believe that there is still a long way to go for depression treatments to solve all the problems associated with depression, but today's treatments are saving lives and helping millions of people to regain their previous life, work and family. They can be improved and should be combined with psychotherapy, but this study does not provide valid or reliable information to help improve patient care.