Introduction
The 2 currently accepted major classificatory or diagnostic systems for mental disorders—the International Classification of Diseases (ICD), produced by the World Health Organization (WHO), and the Diagnostic and Statistical Manual of Mental Disorders (DSM), produced by the American Psychiatric Association (APA)—shared the aim of rendering order in the definition of mental disorders when they were produced in the 1950s and early 1960s. The impetus for the work of the 2 bodies differed, however, and it is useful to examine what these classifications of illnesses were intended to achieve. For example, the epidemiological work of the WHO in investigating the prevalence of disorders in different countries required the development of a system of defined diagnoses. The APA appears to have taken a practical approach to render assistance to mental health workers by standardizing descriptions of different diagnostic categories.
The early versions of the diagnostic criteria lacked precision, which is perhaps not surprising in view of the comparative lack of development of scientific methodology in the field of psychiatric medicine at the time. For example, the severity of the illness was measured by a count of symptoms regardless of their individual severity. While a higher number of symptoms may generally denote greater severity, it is neither a particularly precise method nor adequate for an illness such as depression, where different presentations produce a different number and pattern of symptoms. These deficiencies were recognized, for example, by Max Hamilton, who at that time created a diagnostic scale that gave different weightings to the symptoms, making some symptoms count more than others. Hamilton’s scale was widely adopted as a very useful measure of severity and change in the severity of depression and became the key scale in prospective placebo-controlled efficacy studies, but the scale’s diagnostic properties were largely forgotten and abandoned.
Development of ICD and DSM
ICD
The early versions of the classifications produced by the WHO ICD committees were marred by imprecise descriptions of disorders, with the somewhat liberal use of terms such as “about” and “usually,” which contributed to a certain lack of focus in the clinical descriptions. This detracted from the usefulness of the classification in providing the precise and careful diagnostic criteria needed to define the presence of depression in the individual patient in different settings. Nevertheless, the classification is able to provide an estimate of approximate numbers in diagnostic categories from the data on hospital admissions, which by treaty obligation are collected by the countries that support the WHO.
Extensive research efforts have contributed to the important improvements in subsequent versions of the ICD. The data gathered emphasize the importance of depression as a very common condition, and have helped to educate users about the different forms and presentations of depressive disorders. However, the ICD-10 Classification of Mental and Behavioural Disorders 1 , 2 still suffers to some extent from lack of precision in the descriptions.
The provision of diagnostic guidelines in ICD-10 relating to the severity of illness in order to identify mild, moderate, and severe depressive episodes was an important development, although the definitions are marred by relying primarily on number of symptoms rather than their individual severity. Moreover, phrases such as “should usually be present” and “is usually distressed” may be general descriptors, but they lead to blurring of boundaries so that the classification is generally considered to be too imprecise for use in research, for example, into efficacy of treatment.
Mixed anxiety and depression
The ICD-10 defines Mixed Anxiety and Depressive Disorder as including those who do not satisfy the criteria for either disorder separately. The description states that these individuals, who have comparatively mild symptoms, are frequently seen in primary care “but many more cases exist among the population which never come to medical or psychiatric attention.” 1 The diagnostic criteria are so imprecise that they include a large section of the normal population rather than being helpful in diagnosing a treatable disorder. In DSM-5, 3 Mixed Anxiety Depression was rejected as a disorder when research revealed that in the test–retest situation, a very large proportion of cases of mixed anxiety and depression were redefined as suffering from a depressive disorder instead. The original Mixed Anxiety and Depressive Disorder was more likely the product of less-than-careful assessment of symptoms combined with vague and imprecise definitions of the condition.
DSM
The first edition of the DSM-I, produced by the APA in 1952, fulfilled the aim to provide a glossary of descriptions of mental disorders in order to standardize the descriptions of different diagnostic categories. My focus here is on depression. The descriptions were influenced by the prevailing psychoanalytic views of the time—that mental disorders represented reactions of the personality to social and psychological factors. While the later DSM-II 4 moved away from markedly psychotherapeutic explanation, substituted terms such as “neurotic” and “reactive” persisted until the DSM-III, 5 in which such concepts relating to postulated origins of the disorder were discarded. A key step was the introduction of a multiaxial classification for (I) major disorders, (II) personality disorders, and (III) physical disorders, with an inherent hierarchy over mental disorders such that, for example, where symptoms similar to those of depression appear as part of a physical disorder, for example, organic mental disorder, this would pre-empt a diagnosis of depression.
This article addresses only those studies that have used the double-blind, randomized, placebo-controlled design, which is regarded by all regulatory authorities as the minimum scientific standard for examining efficacy. The strength of double-blind studies is that neither the patient nor the treating physician should be aware of the allocation of treatment group in order to avoid or reduce the methodological bias introduced by the well-known halo effect of reassurance or the knowledge of being treated with an active or inactive treatment, which affects the response. The substantial biases that arise where double-blind, randomized, placebo-controlled methodology is not adopted are highlighted in the analysis of Bandelow et al.Reference Bandelow, Reitt, Röver, Michaelis, Görlich and Wedekind 6 The presence of a placebo arm allows an estimate of efficacy of a treatment above and beyond the level of response observed during treatment with a nonpharmacological agent, which includes the benefit of the attention and the frequent inquiry about and rating of symptoms that are part of trial procedure. The so-called “placebo response” therefore includes both possible spontaneous improvement and the effects of attention and indirect reassurance, which is part of being in a study. The recorded rates of placebo response has risen over the years and has been shown to increase with the number of scales used and the time spent with the patients by the investigator or the clinical team during the study.
It also appears that in the double-blind, randomized studies that have focused on mild depression, this “placebo response” is indistinguishable from the response observed when patients are treated with antidepressants. This lack of separation has been interpreted as showing that there is no pharmacological benefit of the treatment in the patients with mild depression recruited to these studies.
Claims for the efficacy of treatments such as psychotherapy or exercise in mild depression need to be based on more rigorous scientific methodologies than have generally been the case in order to avoid the bias inherent in knowing the allocation of treatment groups, which currently reduces the confidence in the results of treatment studies. Rigor is also needed in the definition of mild MDD as distinct from moderate Major Depressive Disorder (MDD) with the use of either cut-offs on established severity scales or Clinical Global Impression (CGI) ratings. Surprisingly few nonpharmacological studies have reported subanalyses of mild versus moderate depression using CGI criteria. At the moment in mild depression there is inadequate evidence of efficacy that meets the usual regulatory criteria adopted in this article.
Defining Symptoms
The symptoms selected for defining Mood (Affective) Disorders Depressive episode in ICD-10 or Major Depressive Disorder in DSM are appropriate and are generally consistent with evaluation instruments that were developed on a more scientific basis for the assessment of the presence and severity of symptoms of depression. For example, a considerable overlap has been reported between the core symptoms for diagnosis laid out in the DSM system and the symptoms established during the construction of the Montgomery–Asberg Depression Rating Scale (MADRS)Reference Montgomery 7 , Reference Montgomery and Asberg 8 as being sensitive to change in severity. This is reassuring. The symptoms chosen for both the ICD-10 and DSM definition of depression are in accord with the literature, generally consistent with the concept of core symptoms of depression, and can be defended. These diagnostic criteria avoid the bias seen in some abbreviated versions of the Hamilton Depression Rating Scale (HAM-D) toward favoring a particular class of antidepressants, for example the selective serotonin re-uptake inhibitors (SSRIs). A substantial body of data shows that some treatments have greater efficacy on certain individual symptoms and compromised efficacy on others. For example, the SSRIs are generally considered to be less effective in treating poor appetite and sleep loss, possibly due to their side effects of nausea and nervousness. Similarly, there is also evidence to suggest that they are less effective in treating anhedonia and loss of energy. The full analysis of the placebo-controlled database on the definition of caseness by symptoms in mild depression or the response of symptoms in mild depression is not available, so that it is difficult to define which symptoms are the more important. Both DSM and ICD diagnostic criteria have wisely resisted any move to disregard these symptoms and have continued to include them as core symptoms in defining depression
While both DSM-5 and ICD-10 provide an acceptable characterization of depression, the identification of depressive episodes or major depressive disorder that is appropriate for treatment is less convincing. In DSM-5, as in DSM-IV, at least 5 key symptoms are required to be present for 2 weeks and must cause clinically significant distress or impairment in social or occupational functioning for the diagnosis of major depressive disorder. The introduction of this concept of impairment, interpreted by some as a response to political influence aimed at limiting reimbursement costs, is an important contribution to identifying depressive disorder appropriate for treatment. ICD-10 also requires the presence of a minimum number of symptoms for 2 weeks or more, but is flawed by lack of precision. Words such as “usually” and “about” occur frequently in the descriptions of symptoms, so that it is likely to be unclear whether the individual under consideration satisfies diagnostic criteria. When a phrase such as “about 2 weeks” is used in the criteria, would the condition qualify as depression if it lasts 2 weeks or, say, 10 days or, say, 8 days? This lack of precision detracts from data derived from the ICD-10.
Mild or Absent Disorder
One of the most useful acts a physician can perform for a patient is to distinguish the treatable disorder from the normal variations of symptoms found in normal life. Unfortunately, in some healthcare systems, reimbursement is linked to a diagnosis, with the risk that an incorrect diagnosis may take precedence over reassurance, which may be sufficient instead of mistakenly receiving the label of illness. It is not the individual symptoms selected for diagnosis in ICD or DSM that are problematic but rather that the assessment of their individual severity and the separation of mild depression from the normal population are not adequately addressed in either ICD-10 or DSM-V. Considerable advances have been made in the measurement of the severity of depression with the development of a number of depression severity rating scales, which have been able to establish the validity of individual symptoms and their sensitivity to change; these scales are used extensively to determine whether a treatment is effective. It is a missed opportunity that this body of work on the severity of individual core symptoms appears to be disregarded in both diagnostic systems. Weight is given to the number of often poorly defined symptoms rather than to their clearly defined severity—an unfortunate legacy from an epidemiological approach. The failure to define the symptoms adequately in either the DSM or ICD diagnostic systems contributes to the poor inter-rater reliability in the recognition of mild severity of the individual symptoms. This is one reason that regulatory authorities have established minimum severity and change scores in MDD with improved severity scales that have good inter-rater reliability.
Clear diagnostic criteria are an important starting point for defining those who would be likely to benefit from treatment. In the earlier versions of ICD and DSM, the absence of a minimum severity criterion per symptom brought the risk that even barely noticeable symptoms could count toward a diagnosis—a failure leading unsurprisingly to poor test–retest reliability of diagnoses. The distinction between mild and moderate levels of depressive disorder in DSM-5 appears vague and confused. To make the diagnosis of major depressive disorder, there has to be significant loss of function or distress, yet the specifiers for severity indicate that mild depression must have 5 (or more) distressing but manageable symptoms and minor impairment in functioning. Moderate severity is not defined beyond being greater than mild and less than severe. This somewhat surprising lack of direction seems to leave a gray area for illness that is less than severe, with a lack of clear separation of mild from moderate; this blurring of boundaries reduces the value of DSM-5 as a reliable and useful instrument for the use of clinicians in diagnosing and treating disorders. In ICD-10, mild depression is defined as the presence of at least 2 typical symptoms, but no precise usable definitions are given as to which symptoms are meant or what minimum severity of those individual symptoms would qualify as a diagnostic symptom. Because of the lack of clarity on the minimum level of severity of individual symptoms, there is an unfortunate risk that a substantial proportion of those who may not be ill might be raised to the level of receiving a diagnosis of depression, in other words indirectly, a group of what may be part of the normal population would become medicalized. In the placebo-controlled studies in mild depression, usual care and placebo produced the same response as usual care plus antidepressants.Reference Montgomery 9
In view of the data that show mild depression to be a group where antidepressant treatment may not be appropriate, the lack of focus in the definitions matters. For example, a significant separation of fluoxetine from placebo was reported in DSM-defined MDD patients with moderate depression [HAM-D(17) score 20 or more] but not in a large parallel study (n=372) in patients with scores of 19 or less where there was no difference from placebo at endpoint.Reference Dunlop, Dornseif, Wernicke and Potvin 10 Similar results were reported in meta-analyses of the original raw data from the placebo-controlled studies of moclobemide and imipramine, where the placebo response was seen to increase with the mildness of the depression and reduce with increased severity,Reference Stassen, Angst and Delini-Stula 11 so that no advantage of antidepressant over placebo could be observed in mild depression, and in the meta-analysis of the U.S. Food and Drug Administration (FDA) antidepressant studies which showed no mean difference from placebo in the group with low severity at baseline.Reference Kirsch, Deacon, Huedo-Medina, Scoboria, Moore and Johnson 12 Regulatory authorities now require that only those with moderate or severe depression, confirmed on a severity rating scale rather than relying on a diagnostic scale, should be included in studies to assess efficacy.
Identifying Individuals with Depression Likely to Benefit from Antidepressant Treatment
Neither ICD-10 nor DSM-5 offers an adequate guide to define “caseness” and detect which individuals would benefit from treatment. While there may be differences of opinions on where the severity bar lies, there is general agreement that including a focus on severity can improve specificity for treatment purposes. It is clear that in view of the equivocal results reported, the exclusion of mild depression will improve the assay sensitivity to treatment of a study population. Nevertheless, care is needed to avoid confusion between mild depression and residual or chronic depression seen in those who were moderate or severe in the present episode but who had not responded adequately. This group of nonresponders to treatment is more correctly allocated to the group of treatment-resistant depression requiring treatment until remission is reached. Opinions vary as to the exact definition of moderate depression, with cut-offs of a score of 20 on the HAM-D or 22 on the MADRS proposed on the basis of the data from the placebo-controlled studies. Although the recorded ratings may be affected by external factors, the individual patient who presents with moderate depression defined either by the CGI severity scale or who achieves the level of moderate severity on the recognized severity scales is likely to benefit from treatment. The brief measures for defining severity included in the ICD or DSM systems provide a less reliable guide to the need for treatment.
The standard used in establishing the efficacy of antidepressants in regulatory studies is high: at least 2 positive randomized, double-blind studies are required. In the European Union, long-term, placebo-controlled efficacy also must be established before licensing, and the difference from placebo has to be shown to be clinically relevant. Using these standards, there is no evidence that antidepressants are effective in mild depression, and it is reasonable to conclude that antidepressants are effective only in moderate or severe MDD, which should be the target for treatment.
There are some limitations. The studies on MDD only apply to clear-cut MDD and have excluded major comorbidity with anxiety disorders; OCD adjustment disorders; or bipolar disorder, schizophrenia, or depression that is attributable to another medical condition such as AIDS or depression with myocardial infarction, which are excluded from MDD in the DSM-5 and referred to in ICD-10 as a major depressive-like disorder. MDD excludes depression attributable to neurological disorders such as stroke, dementia, or minimal cognitive dysfunction. For these reasons, the efficacy of antidepressants in moderate and severe MDD established in the large regulatory placebo-controlled database might not necessarily apply to MDD in the presence of comorbidity or to major depressive-like disorders, bearing in mind the suggestions from studies in treatment-resistant depression (TRD) that the presence of comorbidity lowers the response rate and increases the risk of developing TRD.
Three further issues deserve comment: prior duration of the disorder, comorbidity, and resistant depression.
In both ICD-10 and DSM-5, the prior duration of the disorder, defined as 2 weeks or more, is likely to be too short as a basis for a reliable diagnosis. This rather brief duration can be confused with recurrent brief depression, an under-investigated condition lasting a few days, which has been shown to remit spontaneously but then to recur frequently but unpredictably and which has not been shown to benefit from antidepressant treatment. The DSM-5 definition of recurrent brief depression, which required monthly recurrence, is not in accord with the literature. In pivotal placebo-controlled treatment studies of MDD, a longer duration of prior illness of at least 4 weeks is standard. Almost all antidepressants over the last 30 years have been licensed on the basis of efficacy demonstrated in a population of MDD with a minimum prior duration of illness of at least a month.
The results of the placebo-controlled studies on the efficacy of antidepressants in MDD apply only to clear-cut MDD, since they have mostly been conducted in a depressed population without significant comorbidity in order to establish specific efficacy. Most have excluded major comorbidity with anxiety disorders; OCD adjustment disorders; or bipolar disorder, schizophrenia, or depression that is attributable to another medical condition such as AIDS or depression associated with myocardial infarction, which are excluded from MDD in the DSM-5 and referred to in ICD-10 as a major depressive-like disorder. Depression attributable to neurological disorders such as stroke, dementia, or minimal cognitive dysfunction is excluded from the diagnosis of MDD. The efficacy of antidepressants in moderate and severe MDD established in the large regulatory, placebo-controlled database may not apply to MDD in the presence of comorbidity or to major depressive-like disorders. Studies in depression comorbid with anxiety disorders have generally demonstrated response to treatment but with a lower treatment effect. Likewise, comorbidity with physical illness appears to reduce the response of the depression and may increase the rate of resistance to treatment.
Nonresponse to antidepressants always calls for a review of compliance with treatment and of the diagnosis of MDD. For example, bipolar depression is not part of MDD, and evidence from placebo-controlled studies to support the efficacy of antidepressants in these patients is lacking. More appropriate, evidence-based treatment should be selected. Current data do not support the routine raising of the antidepressant dose in those nonresponders with moderate to severe depression to optimize response. Randomized, double-blind studies have compared raising the dose with continuing on the same low dose and found no clinical advantage of dose optimization, and even some disadvantages in the studies with SSRIs.Reference Schweizer, Rynn, Mandos, Demartinis, Garcia-Espana and Rickels 13 – Reference Wernicke, Bosomworth and Ashbrook 15 The evidence to support a dose response relationship within the licensed doses of any antidepressant is very weak, and for some, for example serotonin–norepinephrine reuptake inhibitors (SNRIs), the extra burden of side effects and toxicity outweighs a putative and disputed claim of an added benefit of higher doses. The only exception with convincing data of a significant dose response is vortioxetine.
The issue of treatment resistance is given little attention in either the ICD-10 or DSM-5 despite the growing body of work showing the need to identify this group for whom different treatment strategies may be appropriate. The classification systems of both ICD and DSM will be obliged in the future to accommodate the issue of treatment resistance. The important finding that if there is no early evidence of improvement a later response is significantly less likely stresses the need to adopt alternative strategies for treating nonresponders earlier. A delay of more than 3 or 4 weeks before changing treatment strategy does not address the extra risks of disability, morbidity, mortality, and unnecessary suffering associated with treatment nonresponse. There has been an assumption that nonresponders might benefit from a change of the class of antidepressants, but this is not supported by the data.Reference Ruhé, Huyser, Swinkels and Schene 16 – Reference Souery, Serretti and Calati 18 A single prospective, randomized study of switching nonresponders from citalopram to desipramine or vice versa showed that switching was significantly worse than remaining on the same treatment, which suggests that the effect of discontinuations symptoms and the slow response of the new antidepressant has been underestimated.Reference Souery, Serretti and Calati 19
The currently widely adopted strategy of augmentation treatment with an atypical antipsychotic such as quetiapine or aripiprazole seems to be more useful. Quetiapine, which has the most convincing data, is the only atypical antipsychotic licensed by both the European Medicines Agency (EMA) and the FDA for add-on treatment in MDD. There is particular interest in the fast onset of action (as early as 4 days with quetiapine compared with placebo) and the new findings that add-on glutaminergic compounds such as esketamine may accelerate response in treatment-resistant depression within 24 hours.
Conclusion
If the ICD-10 and DSM-5 are not able to identify which patients will benefit from treatment, what can be done to help clinicians to refine these criteria? ICD-10 is so loosely formulated that a major change would be required to make the criteria for depressive illness fit for treatment purposes. The minimum requirement to raise the minimum severity to moderate depression would help. This could be done using the Clinical Global Impression Scale–Severity Scale (CGI-S)Reference Guy 20 to at least moderate severity. If an individual is rated as having depressive illness on the ICD-10 and additionally is rated as moderately ill on the CGI, then it is reasonable to consider a diagnosis of probable depressive illness that would potentially benefit from treatment.
The symptomatic criteria for MDD on the DSM-5 have hardly changed over the years. This is fortunate, since the only evidence base that we have is that a minimum level of symptoms defined in DSM-III or -IV coupled with a minimum severity criterion equivalent to moderate depression and with a minimum duration of 1 month will identify a population that has a strong evidence base showing that many antidepressants are effective.
The symptoms chosen to define depression in both ICD-10 and DSM-5 can be defended from the large placebo-controlled treatment evidence base. However serious deficiencies in the method of assessing individual symptoms make it difficult to separate symptoms defining depression from the normal mood variations observed in the community. Both ICD-10 and DSM-5 fail to define adequately mild depression, a group in which a failure of active treatment to separate from placebo has been shown in treatment studies. Both diagnostic systems have allowed a large proportion of those who may not be ill enough to qualify as suffering from the defined disorder of depression. This was succinctly captured by Alan Francis, the chair of the DSM-IV committee, in an editorial for the British Journal of Psychiatry: mild major depression is not major, not depression, and not a disorder.Reference Frances and Nardo 21 Both ICD-10 and DSM-5 would have benefitted from a clear symptomatic definition of moderate depression for which there is the strongest evidence base to guide treatment.
Disclosures
Stuart Montgomery has the following disclosures: AstraZeneca, consultant, honoraria; Lundbeck, consultant, honoraria; Pierre Fabre, consultant, honoraria; Richter, consultant, honoraria.
