Desvenlafaxine (Pristiq) is an antidepressant that has been available since 2008-2009. It is another example, similar to escitalopram, of a new drug being marketed which is simply a chemical "tweak" of another very similar drug. Pristiq is an active metabolite of another common antidepressant, venlafaxine (Effexor). Effexor had been on the market since 1993.
Being new, many studies were done, usually comparing it with placebo, showing that it works. Yet, very few studies were done comparing it with other antidepressants.
Laoutidis and Kioulos (2015) have recently published a review and meta-analysis of desvenlafaxine. http://www.ncbi.nlm.nih.gov/pubmed/26205685
found that while it clearly works better than placebo in short-term
trials, it is significantly inferior to other agents in comparative
studies (i.e. those studies in which desvenlafaxine is compared with a
different antidepressant prospectively).
In a 2014 study by Maity et al., desvenlafaxine and escitalopram were found to be equally effective (actually with a non-statistically-significant" edge favouring escitalopram) for anxious depression. But in this study, it caused more side-effects than escitalopram. http://www.ncbi.nlm.nih.gov/pubmed/25097285
Soares et al (2010) similarly showed no advantage to using Pristiq instead of Cipralex for treating depression in post-menopausal women. Once again, Cipralex had a non-statistically-significant advantage in effectiveness over Pristiq. http://www.ncbi.nlm.nih.gov/pubmed/20539246
(to be clear about what I mean by "non-significant," it is important to know that all statistical findings are probability statements. "Significant" usually refers to a finding which has a less than 5% chance of being due to random variation alone. For many findings, one measure might exceed another, but with a higher than 5% likelihood of the difference being due to chance. It should be considered, though, that from a Bayesian point of view, if you have results which differ, even at a so-called "non-significant" level of confidence, this finding still increases the likelihood somewhat that there is a significant difference. For example, if we toss a coin 10 times, and find that we get 7 heads (instead of the expected 5), we know that there is 17% chance of getting 7 or more out of 10 heads from a fairly balanced coin. Thus this would be "non-significant" with respect to showing that the coin was not fairly balanced. But even so, if one indeed did see 7 heads in 10 tosses, it should increase one's suspicion (in a quantifiable way) that the coin is actually imbalanced. Thus, one should not entirely dismiss "non-significant" results, they should optimally be considered in a large fund of data about an issue, each part of which should reasonably sway our judgment slightly)
In this interesting study by Liebowitz et al (2013)
( http://www.ncbi.nlm.nih.gov/pubmed/23517291 ),
Pristiq was offered at two different doses (10 mg and 50 mg), compared
with placebo, for treating depression. Both doses were superior to
placebo, but were equally effective to each other! Yet, the
"recommended dose" is 50 mg. Pristiq is only available in 50 mg and 100
Pristiq costs about $3.00 per 50 mg pill. A similar drug, Effexor XR, costs $0.75 for a similar dose. Celexa at an equivalent dose costs $0.27, according to Pharmacy Compass (http://www.pharmacycompass.ca/).
So I do not see any reason to recommend Pristiq, except as one of a list of alternatives to try after other options have been tried. There is no reason to expect that it would work better than any other antidepressant, unless a particular person just happens to prefer it (as is sometimes the case). There is evidence to suggest that it has more side effects than alternatives. I do not necessarily think it is a bad drug though: I'm sure that there are some who might try it, and find it very helpful after exploring other options. But based on current evidence it should not be included as a first-line agent.