Most available antidepressants do not help children and teenagers with serious mental health problems and some may be unsafe, experts have warned.
A review of clinical trial evidence found that of 14 antidepressant drugs only one, fluoxetine - marketed as Prozac - was better than a "dummy" placebo at relieving the symptoms of young people with major depression.
Another drug, venlafaxine, was associated with an increased risk of suicidal thoughts and suicide attempts.
But the authors stressed that the true effectiveness and safety of antidepressants taken by children and teenagers remained unclear because of the poor design and selective reporting of trials, which were mostly funded by drug companies.
They recommended close monitoring of young people on antidepressants, regardless of what drugs they were prescribed, especially at the start of treatment.
Professor Peng Xie, a member of the team from Chongqing Medical University in China, said: "The balance of risks and benefits of antidepressants for the treatment of major depression does not seem to offer a clear advantage in children and teenagers, with probably only the exception of fluoxetine."
Major depressive disorder affects around 3% of children aged six to 12 and 6% of teenagers aged 13 to 18.
In 2004 the US Food and Drug Administration (FDA) issued a warning against the use of antidepressants in young people up to the age of 24 because of concerns about suicide risk.
Yet the number of young people taking the drugs increased between 2005 and 2012, both in the US and UK, said the study authors writing in The Lancet medical journal.
In the UK, the proportion of children and teenagers aged 19 and under taking antidepressants rose from 0.7% to 1.1%.
The researchers conducted a systematic review of all published and unpublished trials looking at the effects of 14 antidepressants in young people with major depression up to the end of May 2015.
Analysis of 34 trials involving 5,260 participants with an average age of nine to 18 found that only in the case of fluoxetine did benefits outweigh risks in terms of efficacy and tolerability.
Nortriptyline was less effective than seven other drugs and placebo, while imipramine, venlafaxine and duloxetine were the least well tolerated.
Compared with placebo and five other drugs, venlafaxine was linked to an increased risk of suicidal attempts or suicidal thoughts.
Due to a lack of reliable data, the researchers said it was not possible to carry out a comprehensive analysis of "suicidility risk" for all drugs.
Pharmaceutical companies funded 65% of the trials. Ten trials were judged to have shown a high risk of bias while 20 were rated as "moderate". Only in the case of four trials was the risk of a biased outcome considered to be "low".
The overall quality of evidence for primary outcomes was "very low", the researchers concluded.
British lead author Dr Andrea Cipriani, from Oxford University, said: "Without access to individual-level data it is difficult to get accurate effect estimates and we can't be completely confident about the accuracy of the information contained in published and unpublished trials.
"It has been widely argued that there needs to be a transformation of existing scientific culture to one where responsible data sharing should be the norm."