Comparison of efficacy and safety between individualized and empiric dose regimen of amitriptyline in the treatment of major depressive episode


Dr Dragan R. Milovanovic, Center for Clinical and Experimental Pharmacology, Zmaj Jovina 30, PO Box 179, 34000 Kragujevac, Yugoslavia. Email: piki@ptt.yu


Abstract  The most efficient method for amitriptyline dose individualization has not been established as yet. For this purpose the authors developed and clinically assessed the modified Bayesian method supported by original basic computer program. Twenty-one male and 39 female subjects (32–65 years old), with major depressive disorder (International Classification of Diseases, 10th revision), were randomly assigned and single-blinded to take individualized (experimental group, n = 30) or empiric (control group, n = 30) doses of amitriptyline for 8 weeks. Both treatments were effective. However, the mean daily doses (112.25 ± 29.85 vs 124.50 ± 32.25 mg/day) and plasma concentrations of amitriptyline plus nortriptyline (145.43–161.95 vs 157.63–197.84 ng/mL) were lower in the experimental group (P < 0.05). Total Hamilton Rating Scale for Depression scores at baseline, 14th, 28th, 42nd and 56th day were significantly lower in experimental (mean ± SD: 26.73 ± 3.92, 18.73 ± 4.01, 11.76 ± 4.43, 9.73 ± 3.89, 8.60 ± 3.72) than in control patients (27.56 ± 4.28, 20.23 ± 4.23, 14.56 ± 3.96, 11.56 ± 4.06, 10.70 ± 4.30). Clinical Global Impression Scale severity of illness scores in the experimental (5.76 ± 0.62, 4.90 ± 0.84, 3.53 ± 1.30, 2.53 ± 1.30, 2.10 ± 1.32) and the control group (5.96 ± 0.80, 5.03 ± 0.80, 4.33 ± 0.92, 3.26 ± 1.20, 2.83 ± 1.41), as well as global improvement and therapeutic effect scores, also favored individualized regimen. The adverse effects were less frequent in the experimental group. It is concluded that the modified Bayesian method is more effective and safe than empiric treatment.


Different opinions exist about strength of correlation between plasma concentrations of amitriptyline and therapeutic effect in major depressive disorder. However, it is generally accepted that the correlation exists, and a therapeutic window has been established, from 60 to 220 ng/mL. If one wishes to individualize the amitriptyline dose regimen, the available options are far from ideal.1

The most frequently used is the Bayesian method of dose individualization using nomograms, specific for a particular patient population. Only one measurement of plasma concentration of the drug after its single dose is performed,2,3 but if the appropriate nomograms do not exist (as in the case for the Serbian population) this method cannot be used.

The most precise is the multiple-point method based on estimation of the clearance of the drug for each patient. However, this method is time-consuming and requires multiple measurements of plasma drug concentrations, significant technical support and experienced medical staff. Consequently, it was used less frequently in clinical practice.

Finally, there is the single-dose method where clearance of the drug is estimated from the single plasma concentration measured after administration of the test dose. The method is simple to apply but suffers from severe limitations: the variability is too high; low plasma concentrations are associated with significant errors; and pharmacokinetic parameters are not individually calculated.

In the present study we have developed and clinically assessed the modified Bayesian method for the individualization of amitriptyline dose. The results show that the method is suitable for a clinical setting when appropriate nomograms do not exist and minimal technical requirements are satisfied.


A total of 60 adult patients (21 men and 49 women, between 32 and 65 years old) with major depressive disorder (according to International Classification of Diseases, 10th revision; ICD-10) were admitted to the Psychiatric Clinic during the year 1997 and randomly (using random table, in blocks of 10) divided in two groups. The patients in group A were taking individualized doses of amitriptyline (experimental group) and the patients in group B were taking amitriptyline in usual (empiric) dose regimen (control group). The doses of amitriptyline were taken orally in both groups. The patient did not know under which dose regimen they were being treated, but the rater did (a single-blind method). The treatment course was 8 weeks long, in a psychiatric clinical setting. The plasma steady-state concentration of both amitriptyline and its active metabolite nortriptyline were measured every 2 weeks during the treatment course in all patients, by the fluorescent immunoassay (TDX system; Abbott Diagnostika, Wiesbaden, Germany). The blood samples for measurement of amitriptyline and nortriptyline concentrations were taken from the cubital vein, in the beginning of the last third of the dose interval. Indications and dose ranges of amitriptyline used in the present study have been already approved for clinical practice in Serbia. Nevertheless, the written informed consent was obtained from participants before enrolment.

In group A, doses of amitriptyline in the beginning of the treatment course were calculated by means of an original computer program written in basic computer language. The program is protected by patent and any interested reader can obtain it directly from the authors. The program calculates doses on the basis of therapeutic steady-state concentration of 80 ng/mL, as well as on the basis of a patient's sex, weight, age, creatinine plasma concentration, albumin plasma concentration, and volume of the liquid in the ‘third space’ (if present). In the initial version of the program we used average pharmacokinetic parameters of amitriptyline plus nortriptyline from the general population, published in Kaplan and Sadock's textbook of psychiatry1 assuming linear elimination kinetics of amitriptyline.4 However, in a pilot study we adjusted clearance of amitriptyline  for  the  Serbian  population  and  a  final  version of the program was established.5 In short, it was revealed that in our patients, clearance of amitriptyline was a bit less than in the patients of developed (Western) countries. In the present study the measurement of steady-state plasma concentration was performed on the 14th, 28th and 42nd days. If the measured concentration was outside the therapeutic range, the correction of the dose was done by using an additional part of the computer program devoted to the correction of the dose.

The patients from both groups were evaluated clinically by Hamilton Depression Rating Scale (HAM-D; with 21 items) and Clinical Global Impression Scale (CGI).6 The type and frequency of adverse effects were also recorded. Assessments were done at baseline and on the 14th, 28th, 42nd, and 56th days from the beginning of the treatment course. The statistical differences between and within the experimental and control group were tested using the Mann–Whitney U-test and Wilcoxon's test, respectively. The differences in demographic factors and adverse effects frequency between the groups were tested using Student's t-test and χ2 test, respectively. In general, the level of significance was set at P < 0.05.



In general, the demographic characteristics of the patients were similar in the experimental and control groups (Table 1).

Table 1.  Patient data
n (%)
n (%)
  1. NS, not significant.

Gender  0.0011.0 (NS)
 Male11 (37)10 (33)  
 Female19 (63)20 (67)  
 Total30 (100)30 (100)  
Age (years)  1.2891.00 (NS)
 31–40 6 (20) 5 (16)  
 41–5013 (43)11 (37)  
 51–60 7 (23)11 (37)  
 61–65 4 (14) 3 (10)  
 Mean (± SD)48.1 ± 8.749.5 ± 8.6  
Education  0.6440.73 (NS)
 Elementary13 (43)16 (53)  
 High15 (50)12 (40)  
 Academic 2 (7) 2 (7)  
Marital status  0.9880.61 (NS)
 Married22 (73)20 (67)  
 Single 5 (17) 8 (27)  
 Divorced 3 (10) 2 (6)  
Location  0.6030.96 (NS)
 Rural18 (60)14 (47)  
 Urban12 (40)16 (53)  

Drug doses and plasma concentrations

Throughout the treatment course the mean doses of amitriptyline were lower in the experimental than in the control group. Statistically significant difference between the experimental and control groups was met at the start (U = 298.50, P < 0.05), augmented at the 14th day (U = 272.5, P < 0.05) and remained to the end of the study. Plasma concentrations of amitriptyline plus nortriptyline were also lower in patients in the experimental group. Statistical differences were found on the 14th day (U = 291.5, P < 0.05), augmented at the 28th day (U = 259.0, P < 0.01) and at the same level to the 42nd day. Within the groups, both the drug doses and the plasma drug concentrations decreased at subsequent visits, being the lowest at the end of the study (Table 2).

Table 2.  Drug daily doses and steady-state plasma concentrations of amitriptyline plus nortriptyline during the treatment course
Mean ± SD
Control group
Mean ± SD
  • *

    P < 0.01;

  • **

    P < 0.05.

Dose (mg)
 Baseline128.33 ± 14.28137.50 ± 19.42
 Day 14133.33 ± 17.77**148.33 ± 21.70*
 Day 28121.25 ± 23.01*128.33 ± 21.50*
 Day 42104.58 ± 31.05*113.33 ± 31.30*
 Day 56 88.75 ± 34.16* 95.00 ± 35.59*
 Mean (range: 0–56)112.25 ± 29.85
124.50 ± 32.25
Plasma concentrations (ng/mL)
 Day 14150.91 ± 38.82178.95 ± 42.49
 Day 28161.95 ± 27.18*197.84 ± 49.56*
 Day 42145.43 ± 23.08*157.63 ± 45.32*

Hamilton Depression Rating Scale scores

According to the HAM-D scores, both amitriptyline treatments were effective (Table 3). The first signs of improvement were observed after 2 weeks of therapy and full therapeutic effect was achieved at the end of the study. However, the total HAM-D scores were significantly lower in the experimental group, beginning at day 28 (U = 404.00, P < 0.05) until day 56 (U = 317.00, P < 0.05).

Table 3. Hamilton Depression Rating Scale scores
HAM-D itemsStart
Mean ± SD
Day 14
Mean ± SD
Day 28
Mean ± SD
Day 42
Mean ± SD
Day 56
Mean ± SD
  • HAM-D, Hamilton Depression Rating Scale.

  • *

    P < 0.01.

Total score
 Experimental26.73 ± 3.9218.73 ± 4.01*11.76 ± 4.43* 9.73 ± 3.89* 8.60 ± 3.72*
 Control27.56 ± 4.2820.23 ± 4.23*14.56 ± 3.96*11.56 ± 4.06*10.70 ± 4.30*

Clinical Global Impression Scale scores

On severity of illness item, scores were statistically significantly lower in the experimental group beginning at day 28 (U = 309.50.00, P < 0.05), to day 56 (U = 315.50, P < 0.05). Within both groups the difference was detected at day 14 and lasted to the end. Global improvement scores had a significant difference between the groups but only at the third visit (U = 309.50.00, P < 0.05). Finally, therapeutic effect scores in the experimental group were the lowest at day 56 (U = 334.00, P < 0.05). For the two last items, the difference within the groups was achieved on days 28 and 42 but no further improvement was noted (Table 4).

Table 4.  Clinical Global Impression Scale scores
CGI itemStart
Mean ± SD
Day 14
Mean ± SD
Day 28
Mean ± SD
Day 42
Mean ± SD
Day 56
Mean ± SD
  • CGI, Clinical Global Impression Scale.

  • *

    P < 0.01.

Severity of illness
 Experimental5.76 ± 0.624.90 ± 0.84*3.53 ± 1.30*2.53 ± 1.30*2.10 ± 1.32*
 Control5.96 ± 0.805.03 ± 0.80*4.33 ± 0.92*3.26 ± 1.20*2.83 ± 1.41*
Global improvement
 Experimental2.70 ± 0.592.06 ± 0.86*1.63 ± 0.76*1.50 ± 0.73
 Control2.73 ± 0.582.36 ± 0.49*1.86 ± 0.68*1.76 ± 0.77
Therapeutic effect
 Experimental3.30 ± 0.704.10 ± 0.84*4.43 ± 0.77*4.63 ± 0.66
 Control3.13 ± 0.773.80 ± 0.55*4.16 ± 0.69*4.30 ± 0.47

Adverse effects

The adverse effects observed in both groups were of type A (i.e. pharmacological side-effects). Their frequency was significantly higher in the control group on the 14th and 28th days of the study (Table 5).

Table 5.  No. adverse effects observed
Adverse effectDays
  • E, experimental group; C, control group.

  • *

    Significant difference.

Dry mouth 8121013 6 7 6 7
Nausea 2 3 3 3 1 2 1 6
Constipation 2 2 2 3 4 5 4 4
Tachycardia 2 3 1 2 1 1  
Blurred vision  2 1 2    
Urinary retention  1  2    
Sweating 1 2 1 3    
Fatigue 2 5 3 5 1 3 1 2
Tremor 1 1 1 5    
Confusion    1    
Postural hypotension 3 5 1 2    
Arrhythmia    1    
χ2 ( d.f. = 1)P = 0.0384*P = 0.0135*P = 0.3586P = 0.5562


Whether plasma steady-state concentrations of amitriptyline and nortriptyline are correlated with clinical effect remains a controversial issue.7 Most clinicians determine doses of amitriptyline empirically but with the cost of significant incidence of adverse effects.8 In contrast, a number of studies defined the therapeutic range of plasma steady-state concentration as 60–220 ng/mL.1,8 However, the most efficient method for individualization has not been established,9 presumably because of the well-known broad individual variability in the elimination of amitriptyline in humans.10

In the present study we compared modified Bayesian method of dose individualization with empiric treatment. Taking into account the above-mentioned limitations of three usual methods (Bayesian, multiple-point, single-dose) we assumed that the modified Bayesian method would be reasonably accurate within current clinical settings in Serbia. We had to use average pharmacokinetic parameters for general population from relevant pharmacological literature and to calculate individualized doses assuming linear elimination kinetics of amitriptyline and its metabolite, nortriptyline.11 In order to make this calculation practical, we wrote an original computer program for rapid processing. However, due to overestimated values for pharmacokinetic parameters, the plasma concentrations of the drug at the beginning of the treatment course were too high and almost similar to non-individualized doses. At this point we made a correction to the starting doses that resulted in significantly lower therapeutic doses and plasma drug concentrations in the experimental group.

The therapeutic effects of the treatment were assessed by two tools widely used in clinical trials concerning antidepressants, HAM-D and CGI, which showed that both treatments were effective. Clinical improvement was observed after 2 weeks of the treatment and extended throughout the whole course. These findings are in agreement with the well-known therapeutic efficacy of amitriptyline12 and time course of the clinical action of antidepressants.13 However, the individualization of the amitriptyline doses was more effective, as shown by significantly lower HAM-D and CGI total scores.

Plasma drug concentrations in the majority of our patients (in all except two) were within the proposed therapeutic window. Some of the older studies suggested that there were complex relationships between plasma antidepressant concentrations and drug effects.14,15 A recent meta-analysis of clinical studies that investigated the relationship between serum amitriptyline concentration and the effect has shown a biphasic pattern: the weak positive correlation was noted for the concentrations within the therapeutic window, while the concentrations outside the window did not correlate with the effect at all.16 However, from a practical point of view it seems rational to target the lower part of the therapeutic range of serum amitriptyline concentrations: the present study has shown that the concentrations in the lower part of therapeutic range were at least as effective as concentrations in the upper part, while the adverse effects were significantly less frequent.

It seems that the optimal amitriptyline dose regimen is a necessary prerequisite for optimal therapeutic outcome in the treatment of major depressive disorder, in the majority of cases. The doses should be sufficient to produce clinical benefit but not too high to produce unacceptable adverse effects.17 Because the adverse effects of tricyclic antidepressants are dose dependent,18 the modified Bayesian method might also result in a better safety profile. Clinicians treating major depression have sometimes had difficulty in achieving full clinical improvement. One of the reasonable alternatives in that situation should be a selection of the proper method of dose individualization of antidepressants.

In conclusion, we have demonstrated that the modified Bayesian method used in the present study was more effective than empiric treatment, with lower doses and serum drug concentrations. The frequency of adverse effects was decreased also. If larger clinical trials, using more patients as well as other antidepressants, confirm our findings, the modified Bayesian method could be widely recommended as an easy-to-use tool in everyday psychiatric practice dealing with major depressive disorder.