This is a systematic review of a collection of 11 (short duration) clinical trials (2 RCTs and 9 against active comparator)  on a therapeutic substance (citalopram), as submitted to the regulatory authorities at the time of registration (1988). 9 of these (short duration) trials were unpublished at that time.
It is an interesting historic approach, and relevant to appraise the current use of the drug, and the Defined Daily Dose, as fixed by the WHO (Collaborating Centre on Drugs Statistics Methodology).
In the discussion, two later reviews of placebo-controlled trials are discussed (One in 1994, and one in 2000, including all the trials of the previous review). If my interpretation is correct the full evidence base is 12 RCTs, of which only 6 were published.
The article focuses on the original set, and argues that no evidence was given for the effectiveness of doses below 45 mg. On the other hand, through Drug Utilisation Research in the Nordic Prescription Database (2004-2007) it was established that most patients use this drug at a dosis below 45 mg per day. In the discussion the data from the later reviews of RCTs are analysed, with the conclusion that there is no evidence for efficacy below a dosing of 45 mg/pd.

The point made is that the evidence base for the use of low dose citalopram (below 45 mg) was small at the time of registration, with studies of short duration, poor quality, and mainly unpublished. Hence, the registration for citalopram in these low doses in 1988 was not apppropriate.
This did not fundamentally change later on. Yet, the DDD was set at 20 mg, at the time of registration, and the
majority of the use of the drug between 2004 and 2007 is in the low dose range.
MAJOR REMARK
The analysis of the authors is interesting, but the conclusions are not outspoken on the consequences of the findings:
1) should the DDD have been changed ?
2) Is citalopram (and other antidepressants) used in too low dosages, in too short duration, as a placebo, to patients who did not need an antiperssant in the first place, or to patients who should have been treated with higher doses or other molecules.
3) In case higher doses of citalopram will be used, how will the risk benifit profile be affected.
4) Should there be a regulatory review after this article, and should this regulatory review not have been done, also without this article ?
Is the Cochrane Collaboration review cycle working properly ?
MINOR REMARKS
What is the exact cut off point (less then 45 mg vs. 45 mg and more) ? Now the text says : less then 45 vs more then 45 mg)
The discussion of the RCTs before and after registration is in the discussion and can not be considered as a full systematic review, which should be conducted as a Cochrane review.
The theory of inhibition of d-citolopram on s-citalopram is brought up late in the discussion (on escitalopram as an alternative) and is not well documented and not clear.
It is not clear whether the two placebo controlled RCTs of the pre-registration analysis were both incluced in the 2000 review.
Please mention that review and a protocol for a Cochrane Review has been submitted recently, for comparisons of (s) Citalopram against other antidepressants (by the same group of authors referred to in the lancet
comparions of 2009).
A critical remark that first, a systematic review of the placebo controlled trials is necessary, would be at order.
Escitalopram versus other antidepressive agents for depression Andrea Cipriani, Claudio Santilli, Toshi A Furukawa, Alessandra Signoretti, Atsuo Nakagawa, Hugh McGuire, Rachel Churchill, Corrado Barbui October 2009
Citalopram versus other anti-depressive agents for depression Giuseppe Imperadore, Andrea Cipriani, Alessandra Signoretti, Toshi A Furukawa, Norio Watanabe, Rachel Churchill, Hugh McGuire, Corrado Barbui, Marianna Purgato
July 2009

2.

-Escitalopram is not studied, and should not be mentioned in the conclusion

-The authors state in the results that “Studies with citalopram with an average dose above 45 mg per day clearly show that the antidepressant effect of citalopram is similar to other antidepressant drugs (Illustration 2).” This interpretation of Illustration 2 is imprecise. The mentioned meta-analysis is inconclusive.

-In the discussion, the authors refer to a study by Montgomery et.al. In the second paragraph, it is stated that in this study “only 40 mg was effective”. This statement corresponds to the results in that study. In the fifth paragraph, however, the authors claim that this study “showing no effect of 20 mg citalopram”. In the study the phrase is; “no statistically significant difference between the citalopram 20 mg group and placebo”. This is not in correspondence with each other, because “absence of evidence is not evidence of absence” (Ref: Altman BMJ 1995).

5.

-The article raises some very interesting issues, but it lacks focus. The authors should ask themselves what is really their research question. E.g. The last two sentences of the background should be a separate paragraph, and the beginning of the discussion should be better in correspondence with this.

-Parts of the Methods section seems to be results.

6.

-In the meta-regression, the authors assume a linear relationship. It would be interesting to see whether other assumptions lead to different results.

-In the sixth paragraph in the discussion, three other trials comparing citalopram and placebo are mentioned. It would be interesting to see the impact of these studies on the meta-analysis.

7.

-The last sentence of the fifth paragraph in the discussion should be deleted. If an average effect is considered statistically significant, its size is not relevant by itself. A similar comment is written in the sixth paragraph; “a difference of 1.5 points must be regarded as not clinically useful”. This is also an average and hence, the comment is unnecessary and inappropriate.

-The introduction of d-citalopram and s-citalopram in the discussion lacks some introduction to these terms or should be deleted.

Misprints:

-A minus is missing in a confidence interval in the abstract. According to illustration 2, this should have been (-0.87 to -0.21)

-At the end of the second paragraph in the methods section, it says “depression sale”. I guess this should be “depression scale”

-In the methods section, it is stated that they performed 2 meta-analyses, but in Illustration 2, there seems to be 3, and according to the results, there were another 3 meta-analyses with ITT data

-In the third paragraph under results, the authors report a confidence interval of (-1.2 to 0.21) around an estimate of -0.68. This confidence interval is not symmetric around the estimate in contrary to the other confidence intervals. This should be checked.

-The email address to the last author lacks an “e”.

I have been involved in numerous systematic reviews and performed several meta-analyses, some of these have been in fielads related to this article

The aim and research question could be clrearer.

Several years of experience with systematic reviews and meta analyses of efficacy studies within the mental health field.

Just some minor corrections:

Page 2  left column, line 5 from below  and

Page 4  left column, line 18 from below

     P=0.65 (See Table 1)

 Page 3 left column, line 20 from below "depression scale"

Page 3 right column, line 3 from below  "... of citalopram at endpoint..."

Page 5 right column, line 3 from below  "."

I have published several papers in this area.