This was a multicentre, prospective and observational study carried out in outpatient psychiatric clinics between October 2007 and January 2009. The study was approved by the local ethics committee of the Hospital Clínico de San Carlos (Madrid) and was conducted according to the Helsinki declaration for research in the human being. Patients were requested to give written informed consent before taking part in the study. Patients were evaluated at Baseline, 3 and 6 months visits. Functional and clinical outcome measures were completed in all three visits and included the following instruments: Hamilton Anxiety Rating Scale, Montgomery-Asberg Depression Rating Scale (baseline and 6 month visit only), Clinical Global Impression-Severity of Illness Scale, Sleep Scale from Medical Outcomes Study (MOS-sleep), World Health Organization Disability Assessment Schedule II (baseline and 6 month visit only), and the quality of life questionnaire, EQ-5D (all described in detail below). At baseline, socio-demographic data, current therapy and psychiatric and medical illnesses information were collected. For the 3 and 6 months visits, psychiatrists collected data related to patient’s follow-up and current treatments. Participation in the study did not modify the usual clinical management of participating physicians. Psychiatrists participating in the study made a confirmatory diagnosis of GAD in 6 consecutive patients, 3 according to DSM-IV criteria and 3 according to broad criteria. Although patients could have symptoms and receive treatment before entering the study, they hadn’t been diagnosed before by the psychiatrist.

Eligible patients were men and women over 18 years of age that were diagnosed with GAD by a trained psychiatrist at the baseline visit. Exclusion criteria included previous GAD diagnosis, inability or difficulty to understand patient-reported-outcomes questionnaires written in Spanish. Diagnosis was carried out by a trained psychiatrist, familiar with the study procedures, and who worked in an outpatient setting.

A stratified multistage probabilistic sample without replacement was drawn. Sampling frame included all public and private healthcare settings within the 17 regions of Spain. First stage consisted on a selection of the outpatient psychiatry clinics (mental health centres) in each healthcare region. The number of clinics selected in each region was proportional to the region’s population, being the probability of choosing each clinic relative to the population in the area covered by the clinic. In the second stage, one psychiatrist per clinic was chosen randomly within those with previous experience in clinical and epidemiological research in Psychiatry, and with at least 5 year experience in mental health diseases diagnosis. If a selected psychiatrist refused to participate, he or she was replaced by another from the same clinic (also selected randomly). Finally, patients were enrolled at the third stage by a systematic sampling strategy. Participant psychiatrists were asked to invite in a consecutive manner those patients from the daily list of appointments that fulfilled the inclusion criteria. Psychiatrists were encouraged to maintain the above mentioned proportion of 3 subjects diagnosed following DMS-IV criteria and 3 subjects diagnosed following the broad criteria.

Sample size was calculated taking into account study’s main variable: evolution of anxiety symptoms according to the total HAM-A scores for each patient group. A sample of 3400 evaluable patients was estimated assuming a 2-tailed 95% confidence interval for the total HAM-A scores lower than 0.25 points from the observed mean. Based on the results of a previous study, a standard deviation lower than 8 was assumed [4].

GAD was diagnosed following the DSM-IV criteria (DSM-IV criteria group) in one group [1]. The other group (broad criteria group) was diagnosed according to broader criteria consisting on shortening of anxiety duration from “at least 6 months” to “at least 1 month (but less than 6 months)”, including excessive or non-excessive worry and reducing the associated symptoms from a minimum of 3 to a minimum of 2. As described previously, patients were diagnosed by a trained psychiatrist.

Hamilton Anxiety Rating Scale (HAM-A): The validated Spanish version of the HAM-A scale was used to evaluate the evolution of anxiety symptoms during all 3 visits of the study [19]. The patient was interviewed about the anxiety symptoms experienced the previous week. The HAM-A scales includes 14 items, each of them rates from 0 (absence) to 4 (severe). The total score comes from the addition of all 14 items with a maximum of 56. Scores higher than 24 were considered as severe anxiety; between 16 and 24, moderate; between 10 and 15, mild; and below 9, no anxiety. A psychic anxiety sub domain (addition of items 1-6 and 14) and a somatic anxiety sub domain (addition of items 7-13) were also calculated, with a maximum value of 28 in each subscale.

Montgomery-Asberg Depression Rating Scale (MADRS): The validated Spanish version of the MADRS scale was used to measure the intensity of depression symptoms [19]. This observer-rated depression scale consists on 10 items regarding different depressive symptoms, each item being scored from 0 to 6 depending on the severity of symptoms; the total score results from adding the score of each item, obtaining a minimum of 0 (no symptoms) and a maximum of 60 (very severe).

Clinical Global Impression-Severity of Illness Scale (CGI-S): Participant psychiatrists evaluated patient’s global severity with the CGI-S scale, which rates illness severity from 0 to 7, where 0 was considered as “not evaluated”, 1 as “no disease”, 2 “borderline ill”, 3 “mildly ill”, 4 “moderately ill”, 5 “markedly ill”, 6 “severely ill”, and 7 “extremely ill” [20].

Sleep Scale from Medical Outcomes Study (MOS-SLEEP): MOS-SLEEP scale evaluates the impact or interference with sleep of any disease or treatment. This 12-item scale is made up of 7 subscales and 2 overall index scores. The 7 subscales are sleep disturbance (4 items), snoring (1 item), awaken short of breath or with headache (1 item), quantity of sleep (1 item), optimal sleep (1 item), sleep adequacy (2 items), and somnolence (3 items). This scale also has a sleep problems index, rated from 0 (no interference) to 100 (maximum interference), and a sleep problems subscale. Each item is rated independently with more interference scoring higher, except for the sleep adequacy and quantity and optimal sleep subscales [21, 22].

World Health Organization Disability Assessment Schedule II (WHO-DAS II): The WHO-DAS II is a psychometric instrument that measures day to day functioning and disability in the following six activity domains: understanding and communicating, moving and getting around, self care, getting along with people, work activities, and participation in society. The 12-item WHO-DAS II scale is rated from 1 (none or nothing) to 5 (extreme or inability to perform a given task). The total calculated score ranges from 0 to 100, with higher scores indicating higher degrees of disability, for the total scale as well as for each subscale [23].

EQ-5D: This is a standard self-reported quality of life questionnaire developed in several European countries and validated for Spain [24]. This questionnaire consists of two parts, the first one aims to evaluate patient’s health profile in five dimensions (mobility, self-care, daily living activities, pain, and anxiety/depression). The patient rates his/her current status as no problems, some/moderate problems, and severe/extreme problems for each dimension. Results are then converted into 1 of the 243 different EQ-5D health state descriptions which are used to calculate a unique index or tariff, between 1 (as healthy as possible) and 0 (usually equivalent to death). This index is used to calculate the time-trade off tariff which allows calculating quality-adjusted-life-year (QALY) gain (further explanation below). The second part consists on a visual analogue scale (VAS or thermometer) in which the patient has to rate his health status between 0 (equivalent to death) and 100 (the best possible healthy status).

For statistical analysis, last observation carried forward approach was used for those patients for whom data from the 6 months visit were missing. If the only missing data was that from the 3 months visit, this was estimated proportionally to the change observed at the 6 months visit, given that the missing data at 3 month visit were completely at random. Patients were considered to respond to treatment when a higher than 50% reduction was observed for the total HAM-A scale score at the study final visit when compared to that of the baseline visit.

Descriptive statistics were prepared for the continuous variables in the study, including the assessment of central position and dispersion (two-tailed 95% confidence interval). The Kolmogorov-Smirnov test was applied to check adjustment of data to a Gaussian distribution. For categorical variables absolute and relative frequencies were calculated. For comparisons, Student t tests and chi-square test were used for continuous and categorical variables, respectively, to test homogeneity of baseline data between groups. Analysis of Covariance (ANCOVA) or binary logistic regression models were fitted comparing DSM-IV criteria versus broad criteria groups in most variables adjusting by baseline scoring, sex, age, body mass index, educational level, marital status and percentage of patients with comorbid major depression or depressive disorder. Analysis of variance (ANOVA) was applied for quantitative measures and Chi square (McNemar’s test) for qualitative measures in case of paired comparisons.

The clinical relevance of observed differences in baseline characteristics was ascertained by means of the effect size that was calculated using the difference of means for each group, divided by the combined standard deviation of the variables [25].

QALYs gain after the 6-months follow-up was calculated from patients’ responses to the EQ-5D as described previously using tariff calculated with this instrument. We used Spanish values for tariffs calculation [24]. Gained QALYs were estimated by using the trapezoidal approach, with baseline and 6- month visit values used as reference values [26].

All statistical tests were two-tailed, and α error of < 0.05 was accepted as statistically significant. Data were analyzed using SAS version 8.2 statistical software.

A total of 3,549 patients were finally recruited by 618 psychiatrists in Spanish psychiatric clinics for this observational study. Four hundred and fifty three patients (12.8%) were excluded because they did not meet all inclusion criteria. Remaining patients were diagnosed of GAD according to DSM-IV criteria (1,815 patients) or to broad criteria (1,281 patients) as described on the Methods section. A total of 529 (17.0%) patients drop-out from study, with no significant differences between study groups (DSM-IV criteria: 313 patients, broad criteria: 216 patients; p=0.899) (Fig. 1). However, main reasons for drop-outs were a somewhat different between groups with 52.7% lost to follow-up in DSM-IV group, compared with 67.1% in broad criteria group (p=0.033), and unknown reasons in 22.0% of withdrawals in DSM-IV group compared with 13.6% in broad criteria group (p=0.023). Socio-demographic characteristics are shown in Table 1. The mean age was 45.5 years for the DSM-IV group and slightly lower for the broad criteria group (42.9 years, p<0.001 vs. DSM-IV group). Regarding GAD, all features studied were slightly, but significantly higher in the standard DSM-IV criteria group. Thus, the mean number of anxiety symptoms was 4.7 for the DSM-IV group compared with 4.3 for the broad criteria group (p<0.001). Significant differences were also observed in the percentage of patients presenting with each individual GAD symptom. All HAM-A scores indicated severe anxiety on both patient groups, being the scores in the DSM-IV group significantly higher (Table 1). Likewise, statistical differences were observed at baseline in the CGI and MADRS scales between the two groups. However, the magnitude of the effect was low for all quantitative variables at baseline (Effect size < 0,50 ) so the statistically significant differences observed were of small or negligible clinical relevance.

Fig. (1). Step-by-step cohort selection according to inclusion criteria.

Most patients in both groups (87.9% for DSM-IV; 82.0% for broad criteria) were following a pharmacological treatment in the six months previous to the baseline visit. As described in Fig. (2), benzodiazepines were the most commonly used drugs followed by antidepressants and antiepileptics. At the end of the study, there was an increase in the number of patients under pharmacological treatment (97% in both groups, p<0.001 versus baseline in all cases). The use of antiepileptic drugs increased by 3.2 and 3.8 fold in the DSM-IV and broad criteria groups, respectively, by the end-of-trial (p<0.001 in all cases). The percentage of antiepileptic drugs users in DSM-IV group at baseline and at 6 months was significantly higher, albeit only slightly (p=0.002 in both cases, Fig. 2). Both groups showed a significant reduction in the percentage of benzodiazepines users after 6 months of follow-up (p<0.001), being this reduction even higher in the broad criteria group (p=0.011). Also a slight but statistically significant increase was observed in the percentage of anti-depressive users in both groups (p<0.001, see Fig. 2).

Fig. (2). Drugs used for GAD treatment at basal and 6 month visits. A) DSM-IV criteria group. B) Broaden criteria group. Data presented as percentage of patients in each group. Differences not significant when not indicated. All changes, except for other, were statistically significant when compared to the baseline visit (p<0.001 in all cases). # p< 0.05, † p<0.01 and ‡ p<0.001 when comparing DSM-IV and broad criteria groups in both visits.

At the study entry, the percentage of subjects in DSM-IV and broad criteria groups that had at least one psychiatric comorbidity was 51.4% and 43.7%, respectively, (p<0.001). Table 2 describes the most frequent psychiatric comorbidities and medical disorders in both study groups. However, neither the mean number of comorbid psychiatric disorders

nor the medical illnesses were different between both criteria groups, except for comorbid major depression and other depressive disorders which were slightly more frequent, but statistically significant in DSM-IV group (p<0.01 and p=0.05, respectively, Table 2). Regarding psychiatric disorders, most of them were present around 5 months before the diagnosis of GAD, without significant differences between groups in disease evolution. Major depression and panic disorders were the most frequent comorbid psychiatric disorders in both study groups. For medical illnesses, chronic pain was present in more than half of patients in both groups, although a higher percentage was seen in DSM-IV group; 64.3% versus 52.0%, p<0.001 (Table 2).

Anxiety severity was assessed with HAM-A and CGI-S scales at 3 and 6 month visits. Fig. (3) shows the evolution of HAM-A total and sub-scores over study time in the DSM-IV and broad criteria groups. All HAM-A scores improved after 3 and 6 months in both criteria groups. Moreover, the percentage of patients without symptoms of anxiety, showing HAM-A<9, increased up to 22.3% and 31.4% at the 3 month visit for the DSM-IV and broad criteria groups (p<0.001 vs. basal visit in both groups, and p=0.006 between groups), and up to 49% and 58% at the 6 month visit (p<0.001 vs. basal visit in both groups, and p=0.261 between groups), indicating similar results for both groups at the end of the study. Also, patients responding to anxiety treatment increased when considering the responder criteria (>50% reduction of baseline scores); 30.7% and 38.2% for the DMS-IV and broad criteria, respectively, at the 3 months visit (p= 0.004); increasing up to 59.7% in the DMS-IV and 67.7% at the 6 month visit (p=0.103).

Fig. (3). Reduction on HAM-A scale and subscale scores at 3 and 6 months visit for both DSM-IV and broad criteria groups. A) Psychiatric anxiety sub-scores, B) Somatic anxiety sub-scores and C) Total scores. Data presented as change in scale score. * p< 0.001 vs. basal visit, # p< 0.05 DSM-IV vs. broad criteria scores.

With respect to the observed results for the CGI-S scores (Fig. 4), differences between criteria groups were statistically significant at the study time points. Regarding each GAD symptom, Fig. (5) shows the reduction in the percentage of patients presenting with each anxiety symptom. Individually, significant improvements were observed for each anxiety symptom in both study groups. However, higher significant reductions were observed in DSM-IV group for the following symptoms: fatigue, irritability and sleep disturbance (p<0.05 in all cases, Fig. 5).

Fig. (4). Score evolution for the CGI-S scale scores by study group. All changes were statistically significant when compared to the basal visit (p<0.001). *p< 0.001 and # p< 0.05 when comparing DSM-IV and broad criteria groups.

Fig. (5). Reduction versus baseline in the percentage of patients presenting with each individual anxiety symptom at the 6 months visit by study group. All changes were statistically significant when compared within group to the baseline visit (p<0.001). # p< 0.05 when comparing DSM-IV and broad criteria groups.

In this study, depression symptoms were evaluated using the MADRS. When compared to scores at baseline visit (Table 1), patients in both criteria groups improved their MADRS scores with a mean reduction of 12.1 and 12.5 points for the DMS-IV and broad criteria groups, respectively (p=0.264 DSM-IV vs. broad criteria).

MOS-Sleep scale scores revealed an overall improvement in sleep patterns over study time in both criteria group. As described in Table 3, all sleep domains in the MOS-Sleep scale changed significantly after 6 months when compared to basal visit scores. Differences between the DSM-IV and broad criteria groups were observed for the total score (p= 0.018), sleep disturbance domain (p=0.012), sleep quantity (p=0.016), sleep adequacy (p=0.031), and sleep problems (p=0.019) scores, always favouring the broad criteria group.

Health-related quality of life was measured by the EQ-5D questionnaire. As seen in Table 3, both the health status (visual analogue scale) and the frequency of severe problems in each domain had significantly improved at the 6 month visit in both study groups (p<0.001 in all cases). Also, comparisons between both criteria groups in quality-of-life gain by domain were significant in favour of the broad criteria group in health status, daily living activities and mood status. Mean QALYs gain at the 3 month and 6 month visits were 0.034 and 0.103 years, respectively, for the DSM-IV criteria group and 0.034 (3 month) and 0.101 (6 month) for the broad criteria patients, without statistical significance between groups (Table 3). At the 6 month visit, the percentage of patients reporting an improvement in their health status as compared with that of one year earlier increased by 13.8% for the DSM-IV criteria group, and 16.6% for the broad criteria group (differences not significant).

For measuring the disability degree in both groups of patients, the WHO-DAS scale was used. As shown in Table 4, there was a significant improvement in all 7 domains of the WHO-DAS II scale after the 6 month follow-up in both criteria groups. The improvements were not significantly different in the DSM-IV group when compared to the new criteria group except for household activity domain, in which the improvement in the DSM-IV group was significantly better when compared to the broad criteria group. Nevertheless, a trend toward statistical significance was observed in the total scoring of the disability scale favouring broad criteria group.