Abstract
Symptoms that occur during amphetamine abstinence have been conventionally thought of as mildly physiological compensations to be expected after a protracted period of wakefulness, starvation and continuous activity [1]. However, many patients may have severe withdrawal symptoms and need some intensive treatments, including hospitalisation [2, 2,3]. Because of the wide-ranging severity of amphetamine withdrawal, a reliable and valid measure is needed for making comparisons of cases and data meaningful.
The available measures
Amphetamines and cocaine produce similar behavioural effects since both substances principally increase synaptic catecholamine levels [4]. As there is no measure for assessing amphetamine withdrawal symptoms, the application of a cocaine withdrawal measure for rating amphetamine withdrawal severity might be acceptable. However, such a measure with high reliability and validity cannot be found. While two measures were used, but without testing their reliabilities and validities [5, 5,6], another visual analogue scale could not give any elaborated results [7]. Although Voris et al. [8] have presented a reliable measure comprising four cocaine withdrawal symptoms, their measure does not cover several prevalent withdrawal symptoms, such as, loss of interest or pleasure, slowing of movement, increased appetite, and craving for sleep.
Aim of the study
The investigators proposed to create a short, reliable and valid questionnaire for the evaluation of amphetamine withdrawal symptoms, which we shall call the Amphetamine Withdrawal Questionnaire (AWQ). In addition, a factor analysis to examine its components was conducted.
Method
Construction of the questionnaire
Based on the DSM-IV diagnostic criteria [9] and a comprehensive review [4], 11 items of amphetamine withdrawal symptoms were included in the AWQ. These items were: (i) drug craving; (ii) dysphoric mood; (iii) loss of interest or pleasure; (iv) anxiety; (v) slowing in movement; (vi) agitation; (vii) decreased energy or fatigue; (viii) increased appetite or eating too much; (ix) vivid or unpleasant dreams; (x) craving for sleep or sleeping too much; and (xi) insomnia. Each item was scored on a 0–4 scale (choices: ‘Not at all’, ‘Very little’, ‘A little’, ‘Quite a lot’ and ‘Very much'). The evaluation of withdrawal severity was mainly based on the patient's condition during the 7 days before assessment. To use with the patients, the AWQ was literally translated into Thai.
Subjects
After the agreement on content and face validities could be reached, a field trial for assessing the AWQ reliability, validity and factor structure was conducted in both outpatients and inpatients with amphetamine withdrawal. The inclusion criteria were: (i) aged 15–65 years; (ii) fulfilling the DSM-IV diagnostic criteria for amphetamine withdrawal; (iii) positive urine test for amphetamine; (iv) negative urine test for opiates; (v) no psychotic symptoms; and (vi) no serious physical illness.
There were two groups of patients in this study. The first group of 30 patients were requested to fill in the AWQ three times. The first set of questionnaire was completed after 24-120 h of amphetamine abstinence (day 0), before they received any treatments. The second and third responses were done on day 7 and day 8, respectively. On day 7 and day 8, they were given a set of psychosocial treatments, including therapeutic milieu, group and individual therapy and psycho-education. Apart from occasional use of lorazepam in patients with moderate to severe anxiety or insomnia, no other psychotropic drugs were administered to these 30 patients. To minimise the variation of clinical features which might affect the test-retest reliability, the doses of lorazepam used on day 7 and day 8 were the same. The second group were 72 patients with 24–120 h of amphetamine abstinence who were requested to fill in the AWQ only once before they received any treatments.
Reliability test
The internal consistency was assessed by analysing the first 30 patients’ data obtained after 24–120 h of amphetamine abstinence. The data obtained on day 7 and on day 8 were analysed for assessing the test-retest reliability. One day interval between test and retest was applied in this study since the clinical features of amphetamine withdrawal may change over time. However, to minimise the ability to recall the response on day 7, the investigators applied an equivalent, but new format of the AWQ for the assessment on day 8. The new format of AWQ was created by reversing both item and severity sequences of the original AWQ applied on day 7. Cronbach's alpha and Spearman rank order correlation coefficients were computed for assessing the internal consistency and the test-retest reliability, respectively. For each item, a p-value less than 0.05 indicated a significant correlation of test-retest reliability.
Validity test
To evaluate the AWQ criterion validity, the general condition of amphetamine withdrawal was also assessed by using the other two accepted measures. Together with the AWQ scoring, the first 30 patients with 24-120 h of amphetamine abstinence were requested to placed an ‘X’ on the visual analogue scale of a 20 cm line that best approximated the ‘feels sick/feels well’. This method is the subscale of a measure presented by Voris et al. [8]. In addition, a trained nurse also rated the severity of amphetamine withdrawal by using the Clinical Global Impression (CGI) scale [10]. Spearman rank order correlation coefficients were calculated for testing the AWQ criterion validity.
Factor analysis
All 102 patients’ data obtained after 24-120 h of amphetamine abstinence were included in the factor analysis. The AWQ items were subjected to principal component analysis for identifying the distinct factors. Eigen value-one test was applied to keep or discard factors. Finally, varimax rotation was performed to elicit the factor components. In addition, a Cronbach's alpha was computed to evaluate the internal consistency of each factor.
Results
Subjects for reliability and validity tests
A total of 28 men and two women participated in the study of AWQ reliability and validity. The means of age, amphetamine-use duration and amphetamine-abstinence period were 20.3 (SD = 6.6) years, 22.9 (SD = 12.2) months and 51.4 (SD = 22.4) hours, respectively. On day 7 and day 8, six of 30 patients were given lorazepam with a mean dose of 0.67 (SD = 0.26) mg day−1.
Mean and 95% confidence interval (CI) and Spearman rank order correlation coefficient (p-value) of each item
Subjects for factor analysis
A total of 99 men and three women participated in the factor analytic study. The means of age, amphetamine-use duration and amphetamine-abstinence period were 20.2 (SD = 5.0) years, 25.5 (SD = 15.1) months and 46.3 (SD = 20.9) hours, respectively.
Mean score (95% confidence interval) of each item
Table 1 presents the mean score (95% confidence interval [CI]) of each item and the mean total score (95% CI) of the 10-item AWQ. These scores were calculated by analysing the first 30 patients’ data obtained after 24-120 h of amphetamine abstinence. It can be seen that the mean insomnia score (95% CI) was very low. Taken together with the fact that 21 and seven patients rated this item as 0 and 1, respectively, we omitted this item from any subsequent analyses. After discarding the insomnia item, the AWQ final version comprised 10 items.
Reliability test
The Cronbach's alpha for AWQ was 0.77. Regarding the test-retest reliability, the Spearman correlation coefficients of all items, except the ‘decreased energy’ item, were between 0.59 and 0.89 (Table 1). The test-retest correlation coefficient of the AWQ total score was 0.79. All item scores and the total score on day 7 were significantly correlated (p < 0.05) with those on day 8.
Varimax rotated factor structure of the Amphetamine Withdrawal Questionnaire
Validity test
The AWQ total score obtained after 24-120 h of amphetamine abstinence was significantly and positively correlated with the CGI score (r = 0.62, p = 0.00). In the correlation with the ‘feels sick/feels well’ visual analogue scale, the AWQ total score and the score of visual analogue scale were significantly and negatively correlated (r = − 0.50, p = 0.01).
Factor analysis
Varimax rotation of the principal component analysis yielded a three-factor model of the AWQ (Table 2). The first factor, which accounted for 39% of the variance and should be called ‘hyperarousal factor’, comprised drug craving, agitation and vivid or unpleasant dreams. The second factor, which accounted for 15.8% of the variance and should be called ‘reversed vegetative factor’, comprised decreased energy, increased appetite and craving for sleep. The last factor, which accounted for 11.1% of the variance and should be called ‘anxiety factor’, comprised loss of interest or pleasure, anxiety and slowing of movement. The Cronbach's alphas of these first, second and third factors were 0.67, 0.68 and 0.72, respectively. For the ‘dysphoric mood’ item, this symptom moderately loaded in both factor 1 (0.52) and factor 2 (0.50).
Discussion
This study demonstrates that the AWQ is a short, reliable and valid measure for the evaluation of amphetamine withdrawal symptoms. The factor analysis also indicates that AWQ can be divided into hyperarousal, reversed vegetative and anxiety subscales.
Insomnia, which is included in the DSM-IV diagnostic criteria for amphetamine withdrawal, appears to be a low prevalence symptom. Incorporating such a low prevalence item into a questionnaire may result in measurement errors, such as, random responding, lying [11], so that this item was discarded from the original AWQ. The final version of AWQ therefore consists of only 10 items.
Although a Cronbach's alpha value can be raised by adding more questionnaire items [12], according to the original concepts, it was preferred to keep the number of AWQ items as small as possible. After discarding the insomnia item, the 10-item AWQ internal consistency with the Cronbach's alpha of 0.77 should be satisfactory. Regarding the test-retest reliability, the Spearman correlation coefficients of AWQ total score and all item scores, except one, are considered acceptable. The low correlation coefficient (r = 0.39) of ‘decreased energy’ item may be owing to the fact that this symptom is particularly susceptible to variation between individuals and over time.
For criterion validity, the AWQ total score significantly correlates with the scores of CGI and ‘feels sick/feels well’ visual analogue scales. Both correlation coefficients (r) of 0.62 and −0.50 indicate that the data obtained by means of the AWQ and the other two accepted measures are not different [11]. However, the correlation coefficients being less than 0.70 also suggest that the AWQ and the other two measures do not tap the very same thing. These differences may be attributable to the more elaborate measurement of AWQ.
The factor analysis results of AWQ seem to help understanding more about the components of amphetamine withdrawal symptoms. Since amphetamines affect several systems of the brain, the three components of withdrawal symptoms may be caused by the different pathophysiological processes that have not been known.
It is of interest to note that the ‘dysphoric mood’ item loaded highly in both hyperarousal and reversed vegetative factors. Although this phenomenon infers that a dysphoric mood may be a common symptom of both factors, as a preliminary factor study, it was decided to exclude this symptom from any factors. Further study with a larger number of patients may be able to clarify the place of the ‘dysphoric mood’ symptom.
Some limitations should be considered in this study. First, the number of patients participating in the factor analysis is quite small. As some experts suggest to include at least five subjects per variable [13], the 10:1 subjects-variable ratio in this factor study should be sufficient. However, the sample size of 102 is just marginally larger than the accepted minimum figure of 100 [13]. Second, the symptoms in this study were specific to young men with amphetamine withdrawal. Third, the three-factor model of AWQ was developed from a factor analysis of data obtained shortly after abstinence (within 24-120 h). Further studies with a larger number of patients, other groups of patients and a different duration of amphetamine abstinence should be conducted to confirm the results of this factor analysis.
In conclusion, the AWQ is a short, reliable and valid measure for the evaluation of amphetamine withdrawal symptoms. The factor analysis suggests that the AWQ may be divided into hyperarousal, reversed vegetative and anxiety subscales. More studies with a larger number of patients, other groups of patients and a different duration of amphetamine abstinence should be conducted.
Footnotes
Acknowledgements
We wish to thank Ms Montira Viriya for her coordination and Associate Professor Keith Godfrey for his language assistance.