Depression following pegylated interferon-alpha: characteristics and vulnerability

Abstract

Objective:

Interferon-α2 (IFN-α) injections may be capable of triggering depression in some individuals. The first objective was to further characterize this depression, and secondly to examine whether pre-treatment temperament was correlated with subsequent vulnerability to IFN-α.

Methods:

Twenty-three initially euthymic adults undergoing year-long PEG-IFN-α treatment for hepatitis C were evaluated at baseline and then prospectively monitored using both the Structured Clinical Interview for DSM-IV (SCID) and self-report questionnaires.

Results:

A major depressive episode developed within three months in 39%. Principal component analysis of the change in self-report scores after one month of treatment demonstrated three orthogonal factors: (i) a specific increase in depression as manifested in the Beck Depression Inventory (BDI) and Hospital Anxiety and Depression Scale (HADS), (ii) an increase in hostility and anxiety, (iii) and a generalized combination of worse symptoms including somatic symptoms on the Symptom Check List (SCL-90). BDI at one month was predicted by baseline BDI (r=0.76, p=0.004). Hostility at one month was predicted by low baseline agreeableness (r=0.75, p=0.01). Controlling for baseline BDI scores, categorical major depression was predicted by combined high baseline neuroticism and low agreeableness (combined r=0.66, p=0.03).

Conclusion:

These initial results (i) support the depressogenic nature of IFN-α treatment in a subset of vulnerable individuals, (ii) indicate that some individuals are also independently vulnerable to worsened hostility, and (iii) suggest that it may be possible to clinically predict these vulnerabilities in initially euthymic subjects.

Introduction

IFN-α is the principal treatment for hepatitis C (HCV), with about 2.3% of adults in the U.S. being exposed to the virus [1]. However, exogenous interferon-α2 (IFN-α) can potentially trigger major depression (MDD) in a subset of patients [2-4] similar to idiopathic MDD [5-8].

One hypothesized possibility is that endogenous IFN-α may be correspondingly involved in the etiology of primary MDD [9-17]. Examining the variable ‘psychiatric effects triggered by exogenous IFN-α treatment’ may therefore offer a way to prospectively explore variable vulnerability to MDD. However, the nature of the psychiatric syndrome induced by exogenous IFN-α requires further characterization.

In addition, prophylactic therapies are being considered for the prevention of IFN-α-induced depression [18-20]. Although the treatment of pre-existing depression prior to initiating IFN-α is recommended [5, 19, 21-24], the suitability of prophylaxis for euthymic individuals is less certain [25, 26]. Therefore, to characterize the psychiatric symptoms induced by IFN-α and to preliminarily assess potential personality vulnerability factors, we prospectively assessed a set of euthymic individuals before and during IFN-α treatment.

Methods

Participants

Twenty-three euthymic patients (aged 20 to 58; mean = 45; 52% male) initiated combination pegylated interferon-α2 (IFN-α) and oral ribavirin treatment for HCV. To specifically determine IFN-α's effects and minimize concurrent confounding variables, we excluded (a) active mood, anxiety, psychotic, or drug/alcohol use disorders within the past six months; (b) medications that could obscure the relationship between mood and IFN-α such as antidepressants, anticonvulsants, antipsychotics, or alcohol; (c) co-morbid neurological or immunological disorders. Four subjects had hypertension and one had diabetes, but were otherwise without other major co-morbidities. The study was approved by the UPMC Institutional Review Board.

Measures

Prior to initiating IFN-α, participants were evaluated with the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I) by a single interviewer (FEL), and completed the Neuroticism, Extraversion, Openness – Five Factor Inventory (NEO-FFI), the Psychosocial Adjustment to Illness Scale (PAIS), the Barratt Impulsiveness Scale (BIS-11), the Beck Depression Inventory-II (BDI), the Hospital Anxiety and Depression subscales (HD and HA), and the Symptom Checklist-90 (SCL-90) for subscales related to somatic symptoms (SCL-s), obsession-compulsion (SCL-oc), interpersonal sensitivity (SCL-is), depression (SCL-d), anxiety (SCL-a), and hostility (SCL-h). Monthly measures of the BDI, HDS, HAS, and SCL-90 subscales were obtained after initiating IFN-α, and the SCID-I updated monthly.

Participants who developed a major depressive episode (MDE) or concerns about lethality were provided immediate treatment. Psychiatric treatment effectively censored further relevant data regarding the development of interferon-triggered mood symptoms. Consequently, the primary outcome was symptoms after one month of IFN-α, prior to any confounding psychiatric intervention.

Statistics

A. Changes from baseline to month one were assessed with paired t-tests (employing Bonferroni corrections). B. Questionnaire difference scores (month one – baseline) were subjected to principal component analysis (PCA) with varimax rotation, selecting factors with Eigenvalues >1. Appropriateness of PCA was confirmed with a Kaiser-Meyer-Olkin test of sampling adequacy (0.66) and Bartlett's test of sphericity (Χ²₍₃₆₎ = 84.1; P<0.0005). C. BDI and SCL-hostility scores at month one were assessed using univariate regression of pre-treatment variables (employing Bonferroni corrections). Categorical MDD (coded as zero or one), controlling for pre-treatment BDI, was similarly examined. A subsequent multivariate regression for categorical MDD was then implemented, which included any pre-treatment variables that were initially correlated with MDD (p<0.1 not corrected for multiple testing) during univariate testing. All tests used SPSS v13.0. Means +/− standard deviation are reported unless otherwise indicated.

Results

A. Psychiatric symptoms after interferon-α

Categorical MDD developed in 9/23 subjects. Of these, 33% (3/9) had a history of a past major depressive episode and 55% were male; of those who did not develop MDD, 36% (5/14) had a past history and 50% were male; all non-significant differences. Overall, the BDI score increased from 4.6 +/− 4.4 at baseline to 9.7 +/− 6.3 after one month of IFN-α treatment (t₍₂₂₎=5.4; p<0.0005), increasing more in participants who developed categorical MDD (Figure 1).

Figure 1.

IFN-α increases depression symptoms (mean +/− SEM) in subjects developing MDD (squares), and to a much lesser degree in subjects not developing categorical MDD (circles).

The HADS depression subscale (HD) also increased from 2.6 +/− 3.2 to 3.9 +/− 3.6 (t₍₂₁₎=2.9; p=0.04) as did the hostility (SCL-h) score, from 0.2 +/− 0.25 to 0.58 +/− 0.55 (t₍₁₇₎=3.0; p=0.04). There were non-significant trends for the other scores to increase: HA from 4.3 +/− 3.4 to 9.5+/− 2.9; SCL-s from 0.57 +/− 0.46 to 0.79 +/− 0.52; SCL-is from 0.32 +/− 0.44 to 0.4 +/− 0.44; SCL-d from 0.48 +/− 5.2 to 0.62 +/− 0.42; and SCL-a from 0.21 +/− 0.34 to 0.31 +/− 0.30.

B. PCA

After one month of interferon-α treatment, three orthogonal components were identified (Table 1). The first component loaded primarily on items from the SCL-90, including somatic symptoms, interpersonal sensitivity symptoms and non-specific depression symptoms. The second component loaded primarily on the anxiety and hostility components of the SCL-90. The third component principally related to the BDI and HADS.

Table 1.

Results of Principal Component Analysis

	Factor 1	Factor 2	Factor 3
Beck Depression Inventory	.31	.15	.75
HADS depression subscale	.32	−.13	.74
HADS anxiety subscale	−.15	.22	.84
SCL-90 hostility subscale	.34	.87	.21
SCL-90 anxiety subscale	.03	.90	.00
SCL90 somatic subscale	.49	.77	.07
SCL-90 obsessive-compulsive subscale	.79	.49	−.01
SCL-90 interpersonal sensitivity subscale	.80	.10	.28
SCL-90 depression subscale	.87	.26	.17

C. Risk for developing psychiatric symptoms.

As BDI and hostility (SCL-h) loaded onto distinct factors, vulnerability to these two distinct effects was assessed for eight pre-treatment self-reports: neuroticism (mean = 14.7 +/− 5.6), extraversion (28.4 +/− 5.9), openness (25.8 +/− 5.8), agreeableness (33.4+/− 4.8), conscientiousness (32.8 +/− 6.8), total PAIS (18.7 +/− 15.1), BIS (29.0 +/− 13.5), and pre-treatment BDI.

Pre-treatment BDI predicted BDI at one month (r=0.76, p=0.004), which was not predicted by pre-treatment NEO-FFI factors, total PAIS, nor BIS. Hostility (SCL-h) at one month negatively correlated with pre-treatment NEO agreeableness (r=0.75, p=0.01), but not with pre-treatment hostility or the other baseline self-reports. Controlling for pre-treatment BDI, both baseline neuroticism and agreeableness trended toward predicting categorical MDD (uncorrected p=0.095 and p=0.057, respectively) in the initial univariate analyses. When subsequently combined in the multivariate analysis, and continuing to control for pre-treatment BDI, the combination of pre-treatment high neuroticism and low agreeableness predicted MDD with statistical significance (combined r=0.66, p=0.028).

Discussion

Thirty-nine percent of euthymic subjects developed MDD, consistent with a growing literature [3, 7, 27-47], although this literature is complicated by retrospective, non-standardized diagnoses, and/or the present of active baseline mood disorder or antidepressants [23]. Notably, this study was specifically designed to prospectively assess IFN-α's effects while avoiding confounding biases from active neuropsychiatric illness and/or co-prescribed psychoactive medications.

Three orthogonal sets of symptoms were worsened following IFN-α: depression (BDI and HADS), hostility and anxiety (subscales within the SCL-90), and general symptoms on other SCL-90 subscales. This is consistent with a prior observation that IFN-induced depression-specific symptoms are distinct from general somatic and fatigue symptoms [8]. IFN-induced “MDD” may, in fact, potentially reflect a combination of three different phenomena (depression-specific, anxiety/hostility, and general psychiatric/somatic complaints).

The only factor predicting BDI at month one was the baseline BDI, which was consistent with multiple prior reports [23, 48], and which highlights the importance of controlling for baseline BDI in our assessments. Personality styles have been proposed to increase vulnerability to depression, although there have been many methodological problems [49]; for example, depression or subsyndromal depression can contaminate measures of personality such as neuroticism. In our prospective study, the development of a categorical DSM-IV MDE was predicted by a combination of high neuroticism and low agreeableness, when controlling for baseline BDI. The low pre-treatment agreeableness appeared to be related to the development of increased hostility on IFN-α. This supports the hypothesis that personality can independently influence vulnerability, even to a biologic trigger such as IFN-α.

Surprisingly, a past history of MDE did not predict vulnerability to IFN-induced depression. This may reflect several possibilities: difficulties with accurately diagnosing past MDE retrospectively using the SCID, different “types” of MDE have different vulnerabilities, this could have been the first time for some subjects to encounter a truly depressogenic stimulus, we excluded patients with active MDD, and/or vulnerability may change over the life-span.

Important caveats include: (i) Many HCV patients already have co-morbid psychiatric disorders [50-52]. This study was designed to assess IFN-α's influence on euthymic subjects, limiting clinical generalizability to this specific group. (ii) Depression symptoms would likely worsen in the absence of intervention. This was effectively censored by psychiatric treatment, necessary for ethical reasons and necessitating our focus on one month for quantitative data. (iii) A placebo control was not used. However, we estimate that less than 1% developed MDD unrelated to the IFN-α treatment -- given that the incidence of MDD in medical illness is about 2% per year [53-55]. (iv) We did not control for the dose and brand of ribavirin or pegylated IFN-α2. Whether different brands (e.g. PEG-IFN-α2b vs PEG-IFN-α2a vs. non-pegylated) and/or doses can differentially influence these psychiatric side effects will require future controlled trials.

In summary, this study supports the conclusion that even pegylated IFN-α can induce major depression, and it can do so in patients without a history of major depression. Moreover, IFN-α can trigger a distinct condition of worsening hostility. Prophylactic therapy has been considered for euthymic patients to prevent IFN-α-induced depression [20, 35, 56], but this may not be risk free [20, 25, 26]. The results from this study suggest that vulnerability may be clinically predictable, and therefore prophylaxis could be appropriately targeted. This may have implications both for understanding vulnerability to MDD as well as for appropriately targeting prophylactic antidepressant therapy.