Predictors of response to amine-specific antidepressants

JOURNAL

ELSEVIER

Journal of Affective Disorders 35 (1995) 97-106

OF

AFFECTIVE DISORDERS

Research report

Predictors of response to amine-specific antidepressants R.A. Burns a**, T. Lock a, D.R.L. Edwards b, C.L.E. Katona b, D.A. Harrison ‘, M.M. Robertson b, B. Nairac b, M.T. Abou-Saleh a a Department of Psychiatry. Royal Liverpool Hospital, Liuerpool, UK b Department of Psychiatry. Uniuersity College Medical School, London, Wolfson Building, Middlesex Hospital, London, UK ’ Lilly Industries, Dextra Court. Chapel Hill, Basingstoke, Hampshire, UK Received 29 March 1995; revised 19 April 1995; accepted 28 April 1995

Abstract Discriminant

function

analysis of data from a double-blind

comparative

trial of lofepramine (a noradrenaline-specific involving 183 patients was used to identify predictors of response. Psychic anxiety significantly predicted a positive response to antidepressant medication, whereas psychomotor retardation, observed sadness, subjective lassitude and somatic complaints were significant predictors of nonresponse. Age, gender, endogenicity, duration of illness and number of previous episodes were not predictive of response. Significant differences were found between predictors of response to fluoxetine and lofepramine (P < 0.001 all groups). Predictors of response to lofepramine were similar to overall predictors, i.e., psychic anxiety predicted responders whilst observed sadness, psychomotor retardation, lassitude, inability to feel and somatic complaints predicted nonresponders. In contrast, baseline weight loss predicted response to fluoxetine, whereas anxiety, reduced insight and a tendency to blame others significantly predicted nonresponse. Such findings have practical implications for the management of depressive illness.

reuptake inhibitor) and fluoxetine (a serotonin-specific reuptake inhibitor),

Keywords: Depression;

Antidepressant;

Response;

Predictor

1. Introduction Over the last 3 decades, much research has been carried out on biochemical and clinical predictors of response to antidepressants in depressive illness (Bielski and Friedel, 1976; Joyce and Paykel, 1989)

Corresponding author. Address: St Stephen’s Hospital, Sarsfieldscourt, Glanmire, Co. Cork, Ireland. Fax: (354) (21) 866872. l

0165-0327/95/$09.50 8 1995 Elsevier Science B.V. All rights reserved SSDI 0165-0327(95)00039-9

albeit with somewhat inconclusive results. Clinical studies suggest preferential response to tricyclic antidepressants in nondelusional as opposed to delusional depression (Glassman et al., 1975; Kantor and Glassman, 1977; Nelson et al., 1978; Spiker et al., 1985) and, whilst most authors have suggested that endogenous depression predicts good outcome (Paykel, 1972; R as kr‘n and Crook, 1976), there have been contrasting findings (Simpson et al., 1976; Maier et al., 1988). Psychomotor retardation (Overall

98

R.A. Burns et al. / Journal

of Affectioe Disorders 35 (1995) 97-106

et al., 1966; Paykel, 1972; Downing and Rickels, 1972; Raskin and Crook, 1976) and loss of interest and pleasure (Hollister and Overall, 1965; Raskin and Crook, 1976; Maier et al., 1988) have also been found to be predictors of good response as have absence of personality disorder (Pfohl et al., 19841, neurotic, hypochondriacal or histrionic personality traits (Shawcross and Tyrer, 1985; Hirschfield et al., 1986). Studies have also suggested that older age (Brown et al., 1983) and more frequent recurrences (Raskin et al., 1973; Bielski and Friedel, 1976; Cassano et al., 1983; Keller et al., 1986) predict poor response. There have been contradictory findings for the predictive value of duration of illness (Kupfer and Spiker, 1981; Rush et al., 1983; Keller et al., 1984; Lloyd and Tsuang, 1985), appetite disturbance and sleep disturbance (Kiloh et al., 1972; Raskin and Crook, 1976; Parker et al., 1985; Kocsis, 1990). Few investigators have assessed clinical predictors of response to serotonin-specific reuptake inhibitors (SSRIs) or noradrenaline-specific reuptake inhibitors (NSRIS) despite early claims for ‘serotonin-specific’ and ‘noradrenaline-specific’ depressions with serotonin said to be involved primarily in mood regulation and noradrenaline determining drive vs. retardation (Coppen, 1967; Carlsson et al., 1969; Maas, 1975; Asberg et al., 1976). Claims have been made for preferential response to SSRIs when compared with older less specific antidepressants in terms of overall severity (Dunbar et al., 1991), anxiety (Wakelin et al., 1988; Montgomery et al., 1989) and suicidality (De Wilde et al., 1985; Montgomery et al., 1981; Wakelin et al., 1989). Few studies have been undertaken to examine clinical selectivity of chemically specific antidepressants. These have used relatively small samples and results were inconclusive. Aberg (1981) in a 45-patient study suggested preferential response of somatic anxiety to the SSRI zimelidine when compared with the NSRI desipramine. Muijen et al. (1988) recruited 81 patients and concluded that suicidality was preferentially reduced by the serotonin-specific fluoxetine when compared with the noradrenalinespecific mianserin. Montgomery et al. (1987) found no significant difference between zimelidine and the NSRI maprotiline in terms of antidepressant effect or effects on individual symptoms. Each of these studies, however, examined symptom-specific improve-

ment (i.e., the change of a symptom or symptom group over the study period) and did not investigate subgroups as predictors of overall response. Nystrom and Hallstrom (1985), in a series of 75 outpatients, examined mood, anxiety, retardation and vital symptoms (sleep and appetite) as predictors of response and found no significant difference between zimelidine and maprotiline. However, they found that few prior episodes and few years since the first episode predicted preferential response to maprotiline and that several previous episodes predicted preferential response to zimelidine, thus, adding substance to Van Praag’s theory (Van Praag, 1980) of ‘state-dependent’ vs. ‘trait-dependent’ depression with NA disturbance being related to the depressive state and serotonergic disturbance predisposing towards developing depression. Further investigation may have been deterred by findings of complex yet closely integrated interactions between serotonergic, noradrenergic and other neurones (Chamey et al., 1990) and the suggestion that any specific action may be overshadowed by the overall antidepressant affect of these medications (Montgomery et al., 1987). The purpose of this study was to examine clinical predictors of response with a serotonin-specific and a noradrenaline-specific antidepressant, both individually for each treatment group and overall. We selected fluoxetine, an SSRI, and lofepramine, a tricyclic antidepressant which is itself noradrenalinespecific as is desipramine to which it is partially metabolized (Lancaster and Gonzales, 1989).

2. Methods In a three-centre study with inpatients, outpatients and community-based patients, participants who fulfilled criteria for admission to the study and provided written informed consent were entered in a doubleblind randomized trial of lofepramine or fluoxetine of 6 weeks duration. Participants were aged between 18 and 70 years and fulfilled the DSM III(R) (American Psychiatric Association, 1987) criteria for major depressive disorder with a Hamilton depression rating scale (HDRS; Hamilton, 1960) score of > 17. Recruitment and rating were undertaken by a research psychiatrist at each centre.

R.A. Burns et al./Journal

of Affective Disorders 35 (1995) 97-106

Patients who had taken fluoxetine at any time, lithium or lofepramine in the 3 months immediately preceding entry to the study or during a present episode were excluded as were those requiring ECT or considered to have a serious suicide risk, those with significant medical illness and pregnant women. Use of neuroleptics within the preceding 6 months, monoamine oxidase inhibitors (MAOIS) within the preceding 14 days or any other psychoactive drug apart from short acting benzodiazepines within the preceding 7 days also merited exclusion. On admission to the study baseline ratings on the HDRS, Montgomery-Asberg depression rating scale (MADRS; Montgomery and Asberg, 1979) and the Newcastle diagnostic index (NDI; Camey et al., 1965) were obtained and all patients underwent a physical examination and a haematology and biochemistry screen. The HDRS and MADRS were repeated weekly as well as the clinical global impression of the severity and response (CGI) and the patient global impression of severity and response (PGI). Patients were treated with fluoxetine at a fixed dose of 20 mg daily or lofepramine 70 mg b.d. with the option of adding a further 70 mg after 3 weeks of the trial. Use of placebo tablets ensured that tablet type, numbers and dosage schedule were identical in both fluoxetine and lofepramine treatment groups. Response to medication was defined as a 50% reduction of the HDRS score at baseline and a total score of < 10 on the HDRS to be recorded at the patient’s final visit.

2.1. Statistical methods Our analysis was carried out on an intention to treat basis (ITT) as this mimics clinical practice. Dropouts were, therefore, included in the final analysis. Differences between responders and nonresponders with regard to the HDRS (total, items and subscales), MADRS (total and items), ND1 (total and items), duration of illness, number of previous episodes, gender, age, height and weight were examined using a t test for each drug group and for all patients. Variables selected from this testing were subjected to discriminant function analysis using the

99

SAS procedure STEPDISC for all patients and for the two treatment groups separately. The function was fitted in random samples of 76% of each patient group and tested for goodness of fit in the remaining 33%. CIs for the goodness of fit were determined. Use of the STEPDISC procedure with estimation of goodness of fit reduced the risk of type 1 error. The direction of difference between baseline mean scores on each variable (positive or negative) of nonresponders vs. responders was used to determine whether variables acted as positive or negative predictors of response.

3. Results 183 patients entered the study. The demographic distribution of these patients can be seen in Table 1. Distribution of ND1 scores, numbers of previous episodes and duration of illness is shown in Table 2. There was no significant difference between the treatment groups in terms of demography, previous episodes or duration of illness. Almost half of the sample had symptoms of > 1 year’s duration (n = 89) but few (n = 10) had suffered three or more previous episodes. Endogenicity, as measured by the NDI, followed roughly a normal distribution. 144 patients completed the study. Using the stringent criteria of a HDRS score of < 10 and a 50% reduction of baseline score, 78 patients (42.6%) were classified as treatment responders. There was no significant difference between the treatment groups in response rates (fluoxetine 40%, lofepramine 44%) nor in rates of side effects or drop-outs which are reported elsewhere (Robertson et al., 1994). The number of patients continuing in the study and the number who dropped out at each visit are outlined in Table 3. It is worth noting that all patients in the study had at least one return visit, i.e., at least 1 week of treatment. Statistical analysis (using the t test) of differences between responders and nonresponders overall and within each drug group yielded no significant difference in terms of age, height, weight, duration of episode, gender or overall ND1 score (Table 4). t test analysis revealed significant differences for

R.A. Burns er al. / Journal of Affective Disorders 35 (I 995) 97-l 06

100

15 variables (Table 4) which were then included in the discriminant function analysis (STEPDISC) from which the variables of Table 5 emerged as independent predictors. The process was repeated in several random samples using the same variables to produce an estimate of the goodness of fit. This gave an overall error rate of 31% (95% CI, 23-39%) for all patients, 21% (95% CI, 12-32%) for the lofepramine group and 30% (95% CI, 20-42%) for the fluoxetine group, all of which are significantly lower than the 50% expected by chance (P < 0.001 all groups).

Table 1 Demographic

When responders were compared with nonresponders (combined lofepramine and fluoxetine group data), only the presence of anxiety significantly predicted response to treatment. Apparent sadness, somatic symptoms, psychomotor retardation, lassitude and inability to feel significantly predicted nonresponse. We did not find age, gender, sleep disturbance, duration of illness, number of previous episodes or severity (as measured by total HDRS and MADRS scores) to predict response either overall or with individual antidepressants. Statistically significant predictors of response to lofepramine were iden-

data Lofepramine

Between drug probability

Fluoxetine

n

8

n

%

61 32 93

65.59 34.41 100.00

58 32 90

64.44 35.56 100.00

Sex

Male All

Age Mean SDT n Min Max

Lofepramine

Fhtoxetine

37.26 11.93 93 18.00 66.00

39.89 12.27 90 19.00 68.00

Lofepramine

Fluoxetine

0.87 Between drug probability

0.14

Weight (kg) Mean STD Min Max

Between drug probability

female

male

female

male

57.28 11.82 61 40.60 120.60

75.37 12.25 32 55.80 105.00

61.96 14.67 57 37.00 110.90

75.72 10.17 32 57.50 103.50 0.06 females 0.09 males

Height (cm) Mean STD n Min Max

162.62 7.21 61 145.00 181.00

178.00 7.21 32 169.00 193.00

160.39 6.09 57 145.00 174.00

174.84 8.94 32 152.00 193.00 0.07 females 0.13 males

(P)

R.A. Burns et al./ Journal of Affective Disorders 35 (1995) 97-106 Table 2 Distribution

of NDI, onset, previous episodes

Newcastle diagnostic

index

Lofepramine

0 1 2 3 4 5 6 7 8 9 10 All Number of previous episodes 0 1 2 3 4 5 6 I 10 All Onset of symptoms 1 month 3 months 6 months 12 months > 1 year All

Table 3 Weekly dropouts

n

%

n

%

1 2 8 6 17 12 13 16 9 7 2 93

1.08 2.15 8.60 6.45 18.28 12.90 13.98 17.20 9.68 7.53 2.15 100.00

1 2 7 8 5 19 23 12 10 2 1 90

1.11 2.22 7.78 8.89 5.56 21.11 25.56 13.33 11.11 2.22 1.11 100.00

62 17 9 2 1

66.67 18.28 9.68 2.15 1.08

66 11 8 2

73.33 12.2 8.89 2.22

1 1

1.11 1.11

1 1 93

1.08 1.08 100.00

1 90

1.11 100.00

13 19 17 43 93

1 13.98 20.43 18.28 46.24 100.00

1.08 8 19 15 46 89

1 8.99 21.35 16.85 51.69 100.00

and patients remaining

93 87 83 80 75 68

0.87

0.40 1.12

0.86

loss predicted response to fluoxetine and a tendency to blame others, reduced insight and anxiety predieted nomesponse (Table 5).

in study Fluoxetine

Lofepramine Remaining

1 2 3 4 5 6

Between drug probability

Fluoxetine

tical with those predicting overall response though the order of significance differed. However, differences emerged between the antidepressants as weight

End of week

101

(n)

Dropouts (n)

Remaining

6 4 3 5 7

90 87 81 80 18 76

(n)

Dropouts (n) 3 6 1 2 2

R.A. Burns et al. / Journal of Affectiue Disorders 35 (1995) 97-106

102 Table 4 Probability

of difference

Newcastle Diagnostic Variable Total 2 4 5 6

8 9 10

Age Height Weight Previous episodes Onset Sex ND1 (endo/non) Hamilton total

; 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Subscale Subscale Subscale Subscale MADRS 2 4 5 6 7 8 9 10

1 3 5 6 total

between responders

and non-responders

Index: Lofepramine

Fluoxetine

All patients

0.24 0.94 0.24 0.95 0.11 0.47 0.14 0.19 0.11 0.16 0.39

0.25 0.35 0.41 0.34 0.086 0.74 0.53 0.59 0.42 0.023 0.49

0.10 0.43 0.14 0.48 0.02 0.45 0.53 0.57 0.40 0.67 0.91

0.52 0.51 0.28 0.17 0.90 0.58 0.83 0.28 0.11 0.17 0.30 0.88 0.71 0.24 0.098 0.76 0.96

0.93 0.93 0.14 0.790.24 0.97 0.98 0.660.64 0.060.03 0.29 0.80 0.88 0.53 0.30 0.59 0.65 0.14 0.92 0.64 0.94 0.69 0.13 0.40 0.48 0.14 0.03 0.240.53 0.990.97 0.110.03 0.590.29 0.14 0.03 0.84 0.89 0.8 1 0.26 0.60 0.29 0.33 0.54 0.98

0.56 0.63 0.85

0.32 0.002 0.43 0.79 0.60 0.17 0.91 0.78 0.97 0.15 0.36 0.02 0.006 0.22 0.17 0.81 0.046 0.59 0.01 0.003 0.39 0.12

0.94 0.70

0.07 0.45 0.49 0.56 0.32 0.65 0.15 0.22 0.89 0.17 0.43 0.02 0.13 0.68 0.35 0.05 0.18

0.02 0.0004 0.39 0.39 0.37 0.03 0.21 0.013 0.012 0.95 0.21

RA. Burns er al./Journal

ofAffective Disorders 35 (1995) 97-106

4. Discussion Our sample was larger than most previous predictor studies of predictors of response to antidepressants. As several variables were examined, we made deliberate efforts in our statistical procedures to minimize type 1 error (i.e., false positives occurring by chance). The low response rate is most likely due to our stringent response criteria as recovery rates in many previous antidepressant drug trials were inflated considerably by defining response as a 50% reduction in depressive symptoms (Joyce and Paykel, 1989) and both fluoxetine and lofepramine have been shown to be of similar efficacy to older antidepressants (Lancaster and Gonzales, 1989; Kaspar et al., 1992). In accordance with most of the literature, gender, appetite and sleep disturbance were not found to be predictors of response. Age was found to be nonpredictive though it must be borne in mind that individuals > 70 years of age were excluded. Duration of illness, somewhat surprisingly, did not predict treatment response. A large proportion of our population

Table 5 Discriminate Steu

analysis Variable

103

(48.9%) had illness of > 1 year duration; our findings were, thus, quite robust and should temper widespread clinical pessimism towards illness of longer duration. Previous episodes were not predictive but our findings must be taken in context of the relatively small number (n = 27) who had suffered two or more previous episodes. Neither was endogenicity a predictor of response but it must be noted that our population followed a relatively normal distribution about a mean ND1 of 6 and were, thus, concentrated around the borderline of endogenicity. It has been suggested that treatment response and endogenicity may have a curvilinear relationship with extreme scores at both ends of the ND1 responding poorly (Abou-Saleh and Coppen, 1983). With regard to symptom profile at initial presentation, our results contrast with much of the literature. Our findings for the combined sample did not support those of Paykel (19721, nor of Grunhaus et al. (1988) who found anxiety to predict poorer response to antidepressant medication. However, the finding that anxiety was a positive predictor of response to antidepressants is not without precedent (Robin and

- stepwise selection Partial R2

F statistic

Prob F

NR

R

0.0692 0.0205

13.075 3.662 0.0221 0.0378 3.848 2.224

0.0004 0.0573 3.934 NR 0.05 14 0.1360

NR NR 0.0489

0.0778 0.0560 6.605

9.655 6.783 0.0564 7.170 4.982 0.0120

0.0025 0.0108 5.136 0.0089 0.0283 NR

NR NR 0.0259 NR

0.0598 0.0576 4.225 2.935

5.473 5.200 0.0430 0.0904

0.0216 0.025 1

NR NR R

All patients 1

2 3 4 5 6

M 1 (Apparent sadness) M 12 (Somatic symptoms) ND16 (Depressive psychomotor) M7 (Lassitude) 0.0247 HI0 (Psychic anxiety) M8 (Inability to feel)

Lofepramine group 1 M8 (Inability to feel) 2 H12 (Reduced insight) ND16 (Depressive psychomotor) 3 4 M 1 (Apparent sadness) HlO (Psychic anxiety) 5 6 M7 (Lassitude) 0.0737 Fluoxetine group ND19 (Blames others) 1 H17 (Reduced insight) 2 3 ND14 (Weight loss) 0.0479 0.0342 ND17 (Anxiety) 4 H = HDRS M = MADRS NR = Non-responders R = Responders

4.379 0.0219 0.0130 0.0989 0.0723

NR

NR R

NR

NR R

104

R.A. Burns et al./Joumal

ofAffective

Langley, 1964; Sandifer et al., 1965; Raskin et al., 1970). Furthermore, unlike most previous investigations, we found that psychomotor retardation (and observed sadness) predicted a poorer response to antidepressants. Such a finding, however, also has a precedent as Hordem et al. (1963) found psychomotor retardation to predict poor response to imipramine. The same authors found that loss of interest, which strongly equates to ‘inability to feel’ on the MADRS, also predicted a poor response. In a naturalistic l-year study Parker et al. (1992), found less psychomotor disturbance to predict improved response. Subjective lassitude and somatic complaints as negative predictors of response are consistent with previous investigations (Rickels et al., 1964; Lipowski, 1990). Reasons for the marked discrepancies between our findings and those of other authors are unclear. Many previous studies used smaller populations, different sampling techniques, treatment dosages and response criteria as well as a wide variety of statistical techniques in examining predictors. Sampling procedures may explain some of the differences in that our population was recruited from community primary care clinics, outpatients and inpatients, whereas most previous investigators used hospitalbased populations who probably suffered from more ‘nuclear’ depression although in many cases strict objective criteria for inclusion were not used. In addition, criteria for response to medication were often unclear. Our findings underline the importance of systematic replication of investigations and the standardization of sampling, treatment and response criteria. When comparing lofepramine to fluoxetine, we differed from previous investigators in that we did not examine symptom-specific improvement but rather looked at symptom groups as predictors of treatment response in depression. The finding of anxiety as a positive predictor for response to lofepramine but a negative predictor of response to fluoxetine is at variance with claims for preferential response of anxiety to SSRIs (Wakelin et al., 1988; Montgomery et al., 1989) and consistent with claims for the theory of noradrenergic hyperexcitability in anxiety which have been further supported by positron emission tomographic studies (Reiman et al., 1989). Whilst accepting that the noradrenergic

Disorders 35 (1995) 97-106

theory of anxiety is by no means complete as there is strong pharmacological support for an involvement of serotonin in the pathophysiology of anxiety (Eriksson, 19871, we would nevertheless suggest caution in the use of fluoxetine for depression with anxiety. Pretreatment weight loss as a predictor of response to fluoxetine but not lofepramine is a somewhat surprising finding given that weight loss is a known side effect of SSRIs. However, despite the fact that patients may lose weight whilst taking an SSRI baseline, weight loss has not previously been investigated as a predictor of response to these medications in depression and our results suggest that SSRIs are to be preferred when weight loss has occurred. We found no evidence to substantiate claims for the existence of NA- and SHT-specific depressions with NA disturbance resulting in retardation and loss of drive and retardation and 5HT disturbance primarily affecting mood. Indeed the motor and energy deficits of depressive illness predicted nonresponse to an NA-specific antidepressant but not to a 5HTspecific antidepressant. Therefore, whilst lofepramine may be advantageous for anxious depressives, it is less likely to result in treatment response when used to treat depressives with motor and energy deficits and an SSRI may be preferable. We found no evidence to substantiate Van Praag’s ‘state-dependent’ vs. ‘trait-dependent’ theory as previous episodes were not a predictor of response to either medication. However, in terms of number of previous episodes our population differed considerably from that of Nystrom and Hallstrom’s smaller sample in that 69% had suffered no previous episode and only 14% had > 2 previous episodes, whereas their sample, based solely on outpatients, had 62% with 2 2 previous episodes and only 38% with O-l previous episodes. This probably reflects different patient populations and further research is required in this area. Despite the 30 years availability of antidepressants, we still ‘know disappointingly little about the prediction of response’ (Joyce and Paykel, 1989) to these medications and much less about predictors of response to amine-specific antidepressants. Our results nevertheless have implications for clinical practice and suggest that anxiety, regarded by many as a predictor of nonresponse, may indeed be a positive

R.A. Burns et al./Journal

ofAffectioe

predictor of response whereas variables associated with the motor and energy deficits of major depression predict nonresponse. In terms of a comparison of amine-specific antidepressants, we suggest caution in the use of fluoxetine when anxiety accompanies depression but also note its benefit where weight loss has occurred and its relative benefit with the motor and energy disturbances of depressive illness. In addition, we note the benefits of lofepramine in major depression with anxiety and suggest it should not be the drug of first choice when depression is accompanied by motor and energy deficits. We accept that some of our findings are at variance with much of the literature and that such findings cannot be regarded as conclusive. It is likely that at least some of the variability in the response prediction literature reflects differences in other variables, such as life events, social supports, family dynamics, intelligence and personality. Nevertheless, symptom patterns as predictors of response continue to rightly merit attention and the suggestion that patients with different symptom patterns may show differential responses to SSRI and NSRI antidepressants is of considerable practical importance and deserves further evaluation.

Acknowledgements The research presented in this paper was funded by an educational grant from Lilly Industries. We thank the consultants in Bloomsbury and Islington Health Authority and Liverpool South, Central and East Sectors, G. Stein and his colleagues at Bromley Health Authority and Drs. Ream and Wardale from the Ford Health Centre, Wirral, for allowing us to recruit patients under their care. We are also grateful to J. Turner of Pharmakopius (Services) for undertaking the statistical analysis and to F. Beaumont for her secretarial support.

References Aberg, A. (1981) Controlled cross-over study of a 5HT inhibiting and a NA uptake inhibiting antidepressant. Acta Psychiatr. Stand. (SuppI.) 290, 244-255

Disorders 35 (1995) 97-106

105

Abou-Saleh, M.T. and Coppen, A. (1983) Classification of depression and response to antidepressive therapies. Br. J. Psychiatry 143, 601-603 American Psychiatric Association (1987) Diagnostic and Statistical Manual of Mental Disorders. 3rd Ed. (Rev.). American Psychiatric Association, Washington, DC. Asberg et al. (1976). Bielski R.J. and Friedel R.O. (1976) Prediction of tricyclic antidepressant response. Arch. Gen. Psychiatry, 33, 1479-1489. Brown, R.P., Sweeney, F.A., Kocsis, J.H. and Loutsch, E. (1983) Age as a predictor of treatment response in endogenous depression. J. Clin. Psychophannacol. 3, 176-178. Carlsson, A., Corrodi, H., Fuxe, K. and Hokfelt, T. (1969) Effect of antidepressant drugs on the depletion of intraneuronal brain 5 hydroxytryptamine stores caused by Cmethyl-ethyl-metatyramine. Eur. J. Clin. Pharmacol. 5, 357-366. Camey, M.W.P., Roth, M. and Garside. R.F. (1965) The diagnosis of depressive syndromes and the prediction of ECT response. Br. J. Psychiatry 111, 659-674. Cassano, G.B., Magini, C. and Akiskal, H.S. (1983) Short term subchronic and chronic sequelae of affective disorders. Psychiatr. Clin. N. Am. 6, 55-67. Chamey, D.S., Southwick, S.M., Delgado, P.L. and Krystal, J.H. (1990) Current Status of the Receptor Sensitivity Hypothesis of Antidepressant Action: Implications for the Treatment of Severe Depression. Marcel Dekker, Basel, Switzerland. pp. 13-34. Coppen ( 1967). De Wilde, J., Mertens, C., Fredrigson Overo, K. and Hepfner Peterson, H.E. (1985) Citalopram versus mianserin a controlled double blind trial in depressed patients. Acta Psychiatr. Stand. 72, 89-96. Downing, R.W. and Rickels, K. (1972) Predictors of response to amitriptyline and placebo in three outpatient treatment settings. J. Nerv. Mental Dis. 156, 109. Dunbar, G.C., Cohn, J.B., Fabre, L.F. et al. (1991) A comparison of paroxetine, imipramine and placebo in depressed outpatients. Br. J. Psychiatry, 159, 394-398. Eriksson, E. (1987) Brain neurotransmission in panic disorder. Acta Psychiatr. Stand. 76 (Suppl.), 335, 31-37. Glassman, A.H., Kantor, S.J. and Shostak, M. (1975) Depression, delusions and drug response. Am. J. Psychiatry 132, 176-179. Gnmhaus, L., Harel, Y., Krugler, T., Pande, A.C. and Haskett, R.F. (1988) Major depressive disorder and panic disorder: effects of comorbidity on treatment outcome with antidepressant medications. Clin. Neuropharmacol. 11, 454-461. Hamilton, M. (1960) A rating scale for depression. J. Neurol. Neurosurg. Psychiatry 23, 56-62. Hirschfield et al. (1986). Hollister, L.E. and Overall, J.E. (1965) Reflections on the specificity of action of antidepressants. Psychosomatics 6,361-365. Hordem, A., Hold, N.F., Burt, C.G. et al. (1963) Amitriptyline in depressive states. Br. J. Psychiatry 109, 815-826. Joyce and Paykel (1989). Kantor, S.J. and, Glassman, A.H. (1977) Delusional depressions: natural history and response to treatment. Br. J. Psychiatry 131, 351-360.

106

R.A. Burns et al./Journal

ofAffective Disorders 35 (1995) 97-106

Kaspar, S., Fuger, J. and Moller, H.J. (1992) Comparative efficacy of antidepressants. Drugs 43 (Suppl. 2). 1 l-23. Keller, M.B., Lavori, P.W., Rice, J., Coryell, W. and Hirschfeld, R.M.A. (1986) The persistent risk of chronicity in recurrent episodes of nonbipolar major depressive disorder: a prospective follow-up. Am. J. Psychiatry 143, 24-28. Keller, M.B., Klerman, G.L., Lavori, P.W., Coryell, W., Endicott, J. and Taylor, J. (1984) Long term outcome of episodes of major depression: clinical and public health significance. JAMA 252, 788-792. Kiloh, LG., Andrews, G., Neilson, M. and Bianchi, G. (1972) The relationship of the syndromes called endogenous and neurotic depression. Br. J. Psychiatry 121, 183- 196. Kocsis, J.H. (1990) New issues in the prediction of antidepressants response. Psychophannacol. Bull. 26, 49, 53. Kupfer, D.J. and Spiker, D.G. (1981) Refractory depression: prediction of non-response by clinical indications. J. Clin. Psychiatry 42, 307-312. Lancaster, S.G. and Gonzales, J.P. (1989) Lofepramine, a review of its pharmacodynamic and pharmacokinetic properties and therapeutic efficacy in depressive illness. Drugs 37, 123-140. Lipowski, Z.J. (1990) Somatization and depression. Psychosomatits 31, 13-21. Lloyd, D.W. and Tsuang, M.T. (1985) Duration criteria and long term outcome in affective disorder and schizophrenia. J. Affect. Disord. 9, 35-39. Maas, J.W. (1975) Biogenic amines and depression. Arch. Gen. Psychiatry 32, 1357-1361. Maier, W., Philipp, M., Schlegel, S., Heuser, I., Weidmann, K. and Benkert, 0. (1988) Diagnostic determinants of response to treatment with tricyclic antidepressants: a polydiagnostic approach. Psychiatry Res. 30, 83-93. Montgomery, S.A. and Asberg, M. (1979) A new depression scale designed to be sensitive to change. Br. J. Psychiatry 134, 382-389. Montgomery et al. (1981). Montgomery, S.A., James, D. and Montgomery, D.B. (1987) Pharmacological specificity is not the same as clinical selectivity. In: S.G. Dahl et al. (Eds.), Clinical Pharmacology in Psychiatry. Springer, Berlin, Germany. pp. 179- 183. Montgomery et al. (1989). Muijen, M., Roy, D., Silverstone, T., Mehmer, A. and Christie, M. (1988) A comparative clinical trial of fluoxetine, mianserin and placebo with depressed outpatients. Acta Psychiatr. Stand. 78, 384-390. Nelson et al. (1978). Nystrom, C. and Hallstrom, T. (198.5) Double blind comparison between a serotonin and a noradrenaline reuptake blocker in

the treatment of depressed outpatients. Acta Psychiatr. Stand. 72, 6-15. Overall et al. (1966). Parker et al. (1992). Parker, G., Tennant, C. and Blignault, I. (1985). Predicting improvement in patients with non-endogenous depression. Br. J. Psychiatry 146, 132-139. Paykel, E.S. (1972) Depressive typologies and response to amitriptyline. Br. J. Psychiatry 120, 147-156. Pfohl et al. (1984). Raskin, A., Boothe, H., Schulterbrandt, J.G. et al. (1973) A model for drug use with depressed patients. J. Nerv. Mental Dis. 156, 130-142. Raskin, A. and Crook, T.A. (1976) The endogenous-neurotic distinction as predictor of response to antidepressant drugs. Psychol. Med. 6, 59-70. Raskin, A., Schulterbrandt, J.G., Reatig, N. et al. (1970) Differential response to chlorpromazine, imipramine and placebo. Arch. Gen. Psychiatry 23, 164-173. Reiman, E.M., Raichle, M.E., Robins, E. et al. (1989) Neuroanatomical correlates of a lactate-induced anxiety attack. Arch. Gen. Psychiatry 46, 493-500. Rickels, K., Ward, C.H. and Schut, L. (1964) Different populations, different drug response. Am. J. Med. Sci. 274, 328. Robin, A.A. and Langley, G.E. (1964) A controlled trial of imipramine. Br. J. Psychiatry 110, 419-422. Robertson, M.M., Abou-Saleh, M.T., Harrison, D.A. et al. (1994) Double blind comparative trial of fluoxetine and lofepramine in major depression. J. Psychopharmacol. 115, 261-264. Rush, A.J., Roffwarg, H.P., Giles, D.E., Schlesser, M.A., Fairchild, C. and Tarell, J. (1983) Psychobiological predictors of antidepressant drug response. Pharmacopsychiatrica 16, 192- 194. Sandifer, M.G., Wilson, I.C. and Gambil, J.M. (1965) The influence of case selection and dosage in an antidepressant drug trial. Br. J. Psychiatry 111, 142-148. Shawcross and Tyrer ( 1985). Simpson, G.M., Lee, H.L., Cuche, Z. and Kellner, R. (1976) Two doses of imipramine in hospitalized endogenous and neurotic depressions. Arch. Gen. Psychiatry 33, 1093-l 102. Spiker, D.G., Weiss, J.C., Dealy, R.S., Griffin, S.J., Hanin, I., Neil, J.F., Perel, J.M., Rossi, A.J. and Soloff, P.H. (1985) The pharmacological treatment of delusional depression. Am. J. Psychiatry 142, 430-436. Van Praag, H.M. (1980) Central monoamine metabolism in depressions, 1. Serotonin and related compounds. Compreh. Psychiatry 21, 44-54. Wakelin et al. (1988). Wakelin et al. (1989).