Hamilton depression rating scale and montgomery–asberg depression rating scale in depressed and bipolar I patients: psychometric properties in a Brazilian sample

The general term depression can be applied to a wide range of states, and is defined by symptoms that can be present in a number of different clinical or psychiatric conditions, associated with the use of psychoactive drugs, or even manifest under normal conditions (such as grief or sadness) [1]. By contrast, patients with major depressive disorder present with a group of consistent symptoms that occur throughout most of the day, almost every day, for at least 2 weeks [2].

As is the case for certain other psychiatric disorders, specific biological markers are absent for mood disorders, which means that diagnosis tends to be based on the presence of a syndrome, i.e., a group of relatively stable or recurrent signs and symptoms. Furthermore, some symptoms can be confounded with physical disorders, such as sleep disturbances and changes in weight or libido; these complicating factors facilitate underdiagnosis and confusion with bipolar disorder [3-7].

Consequently, the use of clinical rating scales is required to improve diagnosis quality, to reduce bias caused by physical symptoms, to assess prognosis during treatment, and to evaluate outcomes [8,9]. A large number of scales for evaluating depressive conditions are available, but the Hamilton Depression Rating Scale (HAM-D) [8] and the Montgomery–Asberg Depression Rating Scale (MADRS) [10] are the most common rating scales in use, and are commonly applied to establish clinical criteria for distinguishing levels of severity and for measuring evolution of and recovery from a depressive episode [1].

Despite the differences in content, form of scaling and number of items, most studies that compare HAM-D and MADRS have indicated that the scales are correlated. Some studies reveal correlation coefficients between 0.65 and 0.94 [9,11-13], which suggests that the two instruments may actually measure slightly different aspects of depression. For this reason, we will describe the instruments in sequence.

The HAM-D was developed for administration with patients previously diagnosed with affective disorder, as a measure to quantify the results obtained during the clinical interview, but it has largely been used to assess the efficacy of antidepressive treatment [8,14]. Considered the “gold standard” of depression measures, the HAM-D serves as a point of reference for more recently developed scales. The HAM-D is a hetero-rated scale: it requires a trained rater with sufficient knowledge of the instrument and the symptoms of the depressive syndrome. In order to ensure fidelity, a standardized interview guide was created as part of the training [11,15].

Since first published by Hamilton [8], the scale has been largely applied, and many studies of it have been performed. Bagby et al. [11] used the MEDLINE database to access and review studies that used the 17-item version and that were published between the instrument’s original release in 1979 and 2003. Results indicated good internal, interrater, and retest reliability estimates for the overall scale, but weak interrater and retest coefficients at the item level. As expected, evidence for good convergent, discriminant, and predictive validity was also found. Other studies have found coefficients ranging from 0.83 to 0.94. Interrater reliability of the scale proved consistent, exceeding 0.85 [16]. Nonetheless, the authors highlight some deficits of the scale, mainly due to Hamilton’s conception of depression, which was created more than 40 years ago.

In Brazil, there are few studies about the psychometric proprieties of the HAM-D. The first study with the HAM-D-17 was conducted in the 1980s. In 1987, Dratcu, Ribeiro and Calil [17] assessed severity of patients’ depressive levels with the HAM-D and compared these results with those of the MADRS. Results indicated a moderate correlation between the HAM-D and the MADRS, and low sensitivity of the HAM-D-17 for assessing severity of symptoms. Carvalho [18] designed a translation study, using the HAM-D with 63 bilingual undergraduate students; and Fleck et al. [19] examined the factorial structure of the HAM-D in 130 depressed inpatients from France and Brazil. The authors found differences in the distribution of the factor solution, such that the Brazilian solution indicated four factors, one more than in the French version, as well as differences in the way anxiety items were distributed. No existing reliability studies with Brazilian samples were found for the HAM-D or the MADRS.

The MADRS items are designated to remedy the psychometric limitations of the HAM-D and to measure change during treatment. The scale’s main advantage is its focus on core symptoms to the exclusion of somatic and psychomotor items. The MADRS has been translated into more than 24 languages. Reliability analyses have confirmed the ability to discriminate changes during treatment; interrater reliability coefficients have ranged from 0.89 when with a psychiatrist and a general practitioner and applied on a Swedish sample that lives on North America. Results indicated a reliability of 0.95 when applied by American researches, 0.97 Swedish researchers, 0.97 with a Swedish and an English applicator, and 0.93 with an american psychiatrist with a nurse [10,20]. Zimmerman et al. [21] conducted a literature review of the MADRS to verify the cutoff for defining remission, using MEDLINE data from 1966 to 2002, resulting in ten studies. According to them, ≥4 was the best cutoff score to delineate significant symptoms of depression; whereas the majority of the literature defines an optimal cut off < 9. Zimmerman et al. [21] recommend caution in interpreting their data, however, because not all studies used the same inclusion criteria, and not all patients were being assessed for major depressive disorder (MDD). The previously mentioned work by Dratcu, Ribeiro and Calil [17] is the only Brazilian study on the MADRS.

Both scales have ample psychometric evidence that indicates they are valid measures of symptom outcome in major depression. Validity studies are primary conditions to a test, considering that ensures the degree to which an empirical evidence is supported [22,23].

Validity is often treated as a unitary concept, but the process of validation should be an ongoing process of accumulating various kinds of evidence [23]. Where translation of an instrument is concerned, the first form of evidence relates to content. A literal translation does not ensure that the test measures the same constructs as the original instrument did. Results obtained by Fleck et al. [19] comparing French and Brazilian samples demonstrated the importance of using appropriate statistical analysis, such as factor analysis, to examine and avoid translation threats. Variations in depression scores with the HAM-D have been reported in other studies, indicating the importance of considering cultural differences when validating scales [24,25]. Other important and common evidence is related to other variables, that address the degree that measure are consistent with the construct, as in the study by Dractu et al. [17]; however, this study was performed more than 20 years ago, underscoring the dearth of Brazilian studies despite the scales’ being the most commonly used worldwide.

As was mentioned earlier, no reliability study with a Brazilian sample was found. Reliability is an important component of verifying whether a test is able to measure the proposed construct; i.e., the scale should yield similar answers about the same or similar constructs [26]. In the case of the HAM-D and the MADRS, such reliability is particularly important for ensuring efficacy of clinical drug trials. Within this context, the present study aims to explore the psychometric properties (validity and reliability) of the HAM-D-17 and the MADRS. Secondarily, it aims to compare responses of bipolar I and unipolar patients.