Figure 2 of this paper by Khan et al (2012) shows to my eyes at least that all we have is placebo - since every intervention works better than doing nothing but no better than each other - I say 'to my eye' because others have argued against this point - but whatever the conclusion the fact remains almost every meta-analysis that has been done in recent years shows that none of the treatments we currently have are effective for a majority of patients (even when shown to be marginally more effective than placebo)

http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.0041778

The American Psychiatric Association (APA) and the British Association of Psychopharmacology (BAP) has recently published an updated edition of their guidelines for the treatment of major depression, both dated 1993. Both groups used the nosology of the Diagnostic and Statistical Manual of Mental Disorders, 4th revision (DSM-IV), which is easy for a comparative reading of the two texts, and have restricted the content of the elaborate unipolar major depressive disorder. Furthermore, with regard to pharmacotherapy, both groups have used as a reference the same meta-analysis and both treatment guidelines are presented as a support for the doctor and the specialist and not as a set of therapeutic protocols to be applied in a manner slavish according to the diagnosis. However, it is reasonable to expect from these elaborate a clear address on using the most appropriate of the various active ingredients in different clinical conditions.

Making a decision about which antidepressant is best for a person can be complex. The decision is made in consultation with a doctor, after careful assessment and consideration. People can help the doctor's assessment by providing as much information as possible about themselves and their medical history. Important factors include the person's age, symptoms, other medications and, if female, whether they are pregnant or breastfeeding.

There are many different types of antidepressant medication which have been shown to work, but their effectiveness differs from person to person. Antidepressants take at least two weeks before they start to help, and it may also take some time for the doctor to find the most suitable medication and dosage.

http://www.beyondblue.org.au/the-facts/depression/treatments-for-depression/medical-treatments-for-depression

Try checking pubmed for papers from Irving Kirsch. He's somewhat of a luminary on this topic. 

Thank you for the pointer Helge - the Kirsch paper linked below is very interesting because it raises a further issue about the influence of initial severity - and that is whilst a decrease of a couple of points on a rating scale might be statistically significant, the question has to be, is this of clinical significance for the patient?

http://www.plosmedicine.org/article/info%3Adoi%2F10.1371%2Fjournal.pmed.0050045

Hello

Depression and anxiety is normally distributed therefore to measure the intensity of these disorders we should use a different mathematical approach. 

It is hard to promote own works but I am doing disclosure now , 

Conference Paper

A Fuzzy-Semiotic Framework for Modeling Imprecision in the Assessment of Depression.

Conference: Proceedings of the Joint 2009 International Fuzzy Systems Association World Congress and 2009 European Society of Fuzzy Logic and Technology Conference, Lisbon, Portugal, July 20-24, 2009

Without any doubt researches agreement made in DSM IV and V is misunderstanding what you observed in depression. (Impaired global functioning means nothing )

What you are observing in reality is:

According to Seligman, depressed people have learned to be helpless. In other

words, depressed this is what affects frontal lobes and planning. (It is true information instead of global functioning)

These markers stay with us after actions, therefore severely depressed patient is avoiding any action due to fears about the outcome. The outcome is affected by another cognitive component as anxiety and of course influences on attention.

Depressed patients are more realistic sometimes with their assessment (of course not these who are psychotic) but there is problem with their actions and interactions. We can say that there is established the wrong balance what needs to be crashed in severe cases or changed in mild/moderate.cases 

All medical interventions described in http://www.plosmedicine.org/article/info%3Adoi%2F10.1371%2Fjournal.pmed.0050045 were done mainly to register their medication with the short period of observation and with huge errors when you are thinking that depressive symptoms are black or white , 0 or 1 or ordinal data type.

As I see there are 2 problems with treatment of depression. One, depression as a symptom (or to use an archaic phrase, sadness) is different from clinical depression. Secondly, the diagnosis is clinical depression, and response to treatment, both are based on many biological factors such as 5HTT gene, neural circuitry in amygdala, and lesser recognised factors such as BDNF. Thus, all patients with depression are not same just because they exhibit the same symptoms. But these tests have not been incorporated in clinical practice or in research settings, contrary to the data present on their utility. For example, those with long form of 5HTT gene are more responsive to 

SSRI medications as compared to those with the shorter form.

All clinical trials, and clinical practice for that matter, is still dictated by verbally reported symptoms, and no biological tests to underpin a subtype of depression. This is the major shortcoming for all disorders in psychiatry. Hence, it is no surprise that the response rates are so low. 

Having said that, the response rates for medications in other disorders are not so great either. Leucht et al compared effect sizes from general medicine to that in psychiatric disorders - "For that purpose a total of 94 meta-analyses have been reviewed, of which 48 cover general medical medications in 20 medical diseases, and 16 meta-analyses cover psychopharmacologic compounds for 8 psychiatric disorders. Mean differences and standardized mean differences, as well as absolute and relative risk reductions, were calculated and the observational time of the respective trials was shown. While there were some medical drugs with impressive high effect sizes, overall it can be concluded that psychiatric drugs generally were not less effective than most other medical drugs" ( Leucht S, Hierl S, Kissling W, Dold M, Davis JM: Putting the efficacy of psychiatric and general medicine medication into perspective: review of meta-analyses.

Br J Psychiatry 2012, 200:97-106).

Maybe as psychiatrists we are no better off than other branches of medicine where medications are used. 

I would estrongly recommend the reading of the article by Hendrie et al (J Psychophamacol 2013, 27(5):407; http://jop.sagepub.com/content/27/5/407.long) to have a clear vision of the state-of-the-art of the pharmacological treatment of depression and related experimental approach. Really interesting!

30% efficiency means by comparing to placebo.

I my 35 y. experience, antidepressants are roughly equally effective. Delay before effect ls between 2 and 30 days, usually around 10. For a given patient, effects is nil under a given dose, then increases (if efficient) till a higher dose, the decreases, usually with appearance of intolerance signs.

Changing molecule if inefficient after a given time, according to the treated problem, adding an other complementary one if effect is good but not perfect at optimal dosage, leads to approximately 80% of perfect or excellent results.

Solution lies in therapeutic methodology.

bmaroy.free.fr/depressionF&E

The 30% is not in addition to placebo unfortunately but the absolute effect.

Since placebo produces a 30-40% effect it is reasonable to assume that this is also a placebo effect

Fig 2 of the Khan et al (2012) paper I linked up earlier shows in graphic detail that no treatment (behavioural or drug) is any better than placebo.

It is clear however that doing something/anything is better than doing nothing (see waiting list response)

http://img.photobucket.com/albums/v462/chendrie/Khan%20fig%202.png

The large scale  NIMH Sequenced Treatment Alternatives to Relieve Depression (STAR*D) Study confirms this low 30% remission rate with antidepressants, with another 10% deriving some benefit at Level 1 - and again this 40% effect is within the range expected with placebo treatment

http://www.nimh.nih.gov/funding/clinical-trials-for-researchers/practical/stard/allmedicationlevels.shtml

STAR-D went on to treat those that didn't respond at Level 1 with different treatments and repeated this a total of 4 times.

Incorporating all 4 levels the claim is that more than 70% of patients were successfully treated.

This figure is somewhat inflated because they do not include the significant number of patients that dropped out of this program and the total successfully treated comes down to less than 60% when these are taken into account

The most parsimonious explanation to my mind is that the antidepressants are no more effective than placebo and that therefore any effect they do have can be attributed to a placebo effect.

Importantly however, the STAR-D program also shows that there are individual differences as to which placebo each person will respond to and so there are potential benefits to trying different treatments if earlier ones didn't work.

It would be useful to have some properly controlled studies to demonstrate this point though (i.e. those incorporating an additional placebo treatment for each medication/therapy that is tried)

Unfortunately, phenomenology of depression and methodology of assessment the intensity above mentioned process changed significantly recently.

Depression is normally distributed and only psychotic depression is an example from this rule, therefore new mathematical model needs to be used "fuzzy logic" because symptoms are fuzzy too.

Other factors play important role in this process but the art is in comparison.

By the definition important is extremum  of your function which can give you information what type of severity is your depressive episode and you should compare only episodes with the same severity.

The outcome of typical trials is a simple manipulation for the purpose and there are a lot of commercial biases

If fact, most studies use only one molecule at a fixed dosage.

Before stating that the patient fails to respond, it's of paramount importance to precise the protocol used.

I have developed an algorithm, based on progressive increase of dosage, changing the molecule if inefficient and associating molecules if the result fails to be perfect.

Lack of efficiency lies around 10%.