Major depression during interferon-alpha treatment: vulnerability and prevention.

Major Depressive Disorder (MDD) during interferon-alpha (IFN-alpha) treatment can occur within a few months of therapy, and shares many homologies with other forms of MDD. Most patients are resilient to the side effect of interferon-induced depression (IFN-MDD), but 15% to 40% are vulnerable. Several studies have employed antidepressants to prevent the incidence of an IFN-MDD episode, and the results suggest that prophylactic antidepressants may be specifically useful in those with pre-existing subthreshold depressive symptoms and/or a history of prior MDD episodes. Several other potential markers of vulnerability for IFN-MDD have been implicated in assessments of nondepressed patients before they start IFN-alpha. These include poor sleep quality, premorbid elevations in inflammatory cytokines, genetic polymorphisms in the serotonin system, personality, and social support. The interplay of these factors strongly predicts who is at risk for IFN-MDD, and indicates several potentially modifiable targets for the personalized prevention of IFN-MDD.

Major Depressive Disorder (MDD) is common, costly,[1-3] and notably heterogeneous. Unfortunately, the accurate prediction and subsequent prevention of MDD episodes (MDEs) has been challenging. There is evidence that MDEs are variously associated with elevated psychosocial stress, the postpartum period, hypothyroidism, circadian changes, cerebrovascular disease, administration of inflammatory cytokines such as interferon-α (IFN-α), etc. Therefore, one approach for preventing a MDE could be to avoid stressful circumstances, pregnancy, cerebrovascular disease, and/or IFN-α therapy. However, this is often impractical. Thankfully, most people who are exposed to these various“triggers” do not develop MDD.

Identifying modifiable markers of risk in specificallyvulnerable people, and then mitigating these before MDD occurs, could be a better approach for preventing MDD. However, identifying causal risk factors that pre-exist in nondepressed people requires prospective studies, and the incidence of an MDE over 1 year is less than 2%.[4-6] The necessarily large epidemiologic studies have successfully identified predictive risk markers such as gender, age, cohort, family history, marital status, socioeconomic status, and stressful life events[6,7] - but each of these is difficult or impossible to mitigate. Another strategy is needed for prospectively assessing nondepressed people for modifiable risk factors, and a related strategy is needed for examining whether specifically alleviating these vulnerabilities prevents MDE.

MDD during IFN-α therapy

One approach for delineating modifiable risk factors is to examine homogeneous groups of people who are definitively known to soon be exposed to a specific MDD-evoking situation. Towards this end, patients receiving IFN-α may be ideal candidates for examining MDD vulnerability.[8-12] MDD during IFN-α treatment (IFN-MDD) typically develops within the first 2 or 3 months of administration,[13-17] and occurs in about 15% to 40% of patients.[18] Thus, prospectively assessing IFN-MDD onset is feasible - and consequently it may be possible to determine predictive modifiable vulnerabilities in the 15% to 40% who subsequently develop IFN-MDD.

Of course, it is conceivable that IFN-MDD is unique, and that other forms of MDD are completely distinct from it. However, several lines of evidence indicate that IFN-MDD may successfully inform us about MDD in general. First, a variety of studies have found a robust relationship between IFN-α and MDE, including those demonstrating a dose-response relationship,[19,20] studies with control groups,[16,19-21] and prospective documentations of worsening depression during IFN-a treatment with a return to baseline mood after discontinuation.[23-25] Thus, IFN-MDD is a replicable finding in prospective studies. Second, IFNMDD has phenomenological resemblance to MDD diagnosed in other situations.[21-24] That is, IFN-MDD is not simply fatigue and malaise but - similarly to MDD - involves anhedonia, depressed mood, irritability, anxiety, social withdrawal, poor concentration, altered sleep, personality- changes, and suicidal ideation (Table I). Third, MDD and IFN-MDD may share similar pathophysiologic mechanisms, as indicated by various independent lines of investigation:

Many inflammatory cytokines are elevated during MDD.[51-53]

Psychosocial stress can increase the levels of inflammatory cytokines.[54,55]

IFN-α and other cytokines can affect central monoaminergic systems plausibly involved in MDD.[56-63]

Peripheral cytokines and IFN-α have access to the CNS through a variety of routes in addition to being synthesized in the brain.[64-66]

Endogenous IFN-α mRNA can be induced in the cortex, hippocampus, and hypothalamus, with correlated changes in behavior in animal models of depression.[64-67]

Systemic administration of IFN-α and other cytokines can affect amotivation and anhedonia behaviors in rodent models of depression.[68-75]

Once IFN-MDD is diagnosed, it responds to treatments that are effective for idiopathic MDD, ranging from selective serotonin reuptake inhibitors (SSRIs) and tricyclic antidepressants to electroconvulsive therapy,[15,76-85] with about 79% to 85% of patients responding to antidepressants.[86,87]

IFN-α administration can influence frontal lobe and anterior cingulate function,[88,89] dopaminergic activity,[60] and serotonergic function,[90-93] - all of which may contribute to the development of depression in a manner homologous to other types of MDD.

Of further public health significance, the use of IFN-a is not rare. Almost 2% of the U.S. currently has chronic hepatitis C (HCV).[94] IFN-a is the only FDA-approved treatment for HCV,[95,96] whereas about 170 million people worldwide have been infected with HCV.[97] Supporting IFN-α's widespread use, untreated chronic HCV can lead to cirrhosis, hepatocellular cancer, and liver failure, resulting in about 10 000 deaths per year in the US,[98] a rate which exceeds that from acquired immunodeficiency syndrome.[99] Unfortunately, IFN-MDD can potentially result in suicide,[38] dose reduction with risk for viral relapse,[16] discontinuation of treatment,[100,101] and lower quality of life.[102,103] Therefore, the two rationales for preventing IFN-MDD are that (i) this is a common and disabling syndrome; and (ii) it may be and ideal strategy for informing us about ways to prevent MDD in general.

Prevention studies of IFN-MDD

A few prophylactic trials using selective serotonin reuptake inhibitors (SSRIs) have transpired. These prevention studies initiated SSRIs in patients who were not currently experiencing any MDE prior to beginning the IFN-a therapy (Table II)[80,83,85,104-107]. The first randomized placebo-controlled trial (RCT) was done in patients with metastatic melanoma, using very high doses of intravenous IFN-a. This initial study found strong evidence for prevention of IFN-MDD, with only 2/18 par oxe tinetreated patients (11%) developing IFN-MDD, as compared with 45% of the placebo-treated group.[80] Similarly, in three open-label trials of prophylactic SSRIs given to nondepressed HCV patients, only 3/32 patients (9%) developed IFN-MDD, despite all 32 patients having a prior history of affective disorder. These open-label studies are thus consistent with this RCT study, supporting the conclusion that preventative treatment with SSRIs may be useful.

However, two small RCT studies have now been completed in patients with HCV (Table II). Neither study found IFN-MDD prevention.[85,106] Prophylactic SSRIs may therefore not be universally effective. Despite these two negative findings, one of these studies did report that 24/29 patients in the placebo group developed elevated depression symptoms compared with 10/23 in the paroxetine group.[106] Additionally, further exploratory analyses indicated that prevention may have been most successful for those subjects who already had high pretreatment baseline levels of depressive symptoms.[106] This would be an example of “indicated prevention” whereby treating “subthreshold” depression symptoms may prevent subsequent worsening to full categorical MDD.[108-111] It has been well-replicated that higher levels of pretreatment depression symptoms are associated with the development of IFN-MDD,[18,112-115] and these subthreshold symptoms may be an appropriate target for using preventive SSRIs. Another open possibility is that prophylactic SSRIs specifically prevented IFN-MDD in those with past histories of MDD in remission. This type of prevention would be consistent with the use of antidepressants to prevent recurrence of remitted MDD.[116-119]

To explore this latter possibility, we prospectively followed 31 patients who were not depressed at the onset of IFN-α therapy (as determined using a Structured Clinical Interview of DSM-IV Axis I diagnoses). All of these patients had no MDEs within 6 months prior to starting IFN-α, but they did have a history of past MDD. Ten of these patients were stably taking SSRIs. Only 20% (2/10) of the patients on SSRIs developed IFN-MDD, while 47.6% (10/21) not on antidepressants did. These results are numerically similar to the RCTs reviewed above. This very limited analysis suggests a more targeted use of SSRIs to prevent recurrence, limiting prophylactic SSRI to those patients who are known to have past MDD histories.

However, all of these studies have been very limited in size, and therefore power. Assessing all of the six published prevention studies and our open-label data combined - in a very exploratory type of meta-analysis - 15/97 (15%) patients receiving SSRIs prior to starting IFN-α developed IFN-MDD, compared with 36/99 (36%). This is a significant difference, χ2=8.2;P<0.001. However, limiting the meta-analysis to the three RCTs, 10/55 (18%) subjects randomized to pretreatment paroxetine developed IFN-MDD while 21/68 (31%) randomized to placebo did. The trend is numerically similar to the larger meta-analysis, but does not have the power to be significant in a chi-square test (χ2=1.98).

At this point, only tentative conclusions are possible: (i) Prophylactic SSRIs may plausibly cut in half the incidence of IFN-MDD. To conclusively determine this, however, will require a larger-size trial than those performed to date; (ii) SSRIs may specifically benefit subjects with either pre-existing depressive symptoms (ie, subthreshold depression) and/or a history of prior MDD. This is consistent either with studies of “indicated prevention” in which patients with subthreshold depression are prevented from worsening to full categorical MDD by about 30%,[108-110] or with studies preventing recurrence of MDD.[116-119] A more targeted prevention RCT would be valuable to examine these two possibilities; (iii) Even if SSRIs are found to be effective prophylactics for some people, about 15% to 20% of patients still developed IFN-MDD even when prescribed SSRIs, there fore antidepressants may not be universally effective. Other targets and approaches for prevention are needed; (iv) Most importantly, about half of the patients with a history of MDD remain resilient even during IFN-α treatment. Identifying the source of this resilience for potential replication in other patients would be beneficial.

Modifiable risk factors for IFN-MDD

The goal for this work is preventative treatments that can be targeted towards specifically mitigating those mechanisms underlying vulnerability. Poor sleep quality prior to IFN-α treatment may be one such risk factor.[121,122] Patients with scores greater than 10 on the Pittsburgh Sleep Quality Index, a validated self-report assessment of sleep quality,[123] were ten times more like to subsequently develop IFN-MDD than patients sleeping better than this.[122] This large effect size was evident even when controlling for other depression symptoms. It is also consistent with large epidemiological studies wherein insomnia predicted the subsequent development of MDD over follow-up intervals of 1 to 35 years.[124-127] As many treatments for sleep exist, this may be a potentially modifiable risk factor for preventing IFN-MDD. This has previously been suggested for MDD,[128] but may now be readily testable in patients about to be treated with IFN-α.

There is also evidence that increased age may be another risk factor for IFN-MDD,[129] although this is certainly not a consistent finding.[130,131] Despite the fact that age itself is not modifiable, this could indicate the presence of agerelated modifiable risk factors. Related to this, elevated levels of inflammatory cytokines, such as interleukin-6 (IL-6), prior to IFN-a therapy have been associated with subsequent IFN-MDD.[132,133] Additionally, a polymorphism in IL-6 that has been associated with increased IL-6 levels is predictive of IFN-MDD.134 In the subset of people with increased IL-6 during IFN-α administration, the IL6 levels temporally predicted next month's depression symptoms.[133] This is consistent with cross-sectional studies in which elevated IL-6 levels are associated with MDD.[54,132,135-140] Thus, increased IL-6 may be another plausibly modifiable target for preventive intervention in depressed individuals. Interestingly, IL-6 increases with age but can be modified by diet[141] and/or exercise.[142-143]

Potential premorbid risk factors for IFN-MDD that may be modifiable through psychosocial interventions could include social isolation[144] and neuroticism.[115,145] However, when controlling for other premorbid risk factors, the effect size for these is fairly small.[146] Another risk factor may be a hyperactive stress response in the hypothalamic-pituitaryadrenal (HPA) axis.[147] Given the common association between abnormalities in the HPA axis and MDD,[148-150] this may also be a potentially useful predictive marker. Interestingly, HPA axis responsiveness can be therapeutically modifiable by antidepressants.[154] It is therefore plausible that patients with overactive HPA responses may be the subjects who benefit most from antidepressant prophylaxis. Consistent with this, stress-reactivity did correlate with depressive symptoms prior to IFN-α therapy[147] - and thus elevated stress-reactivity may be a potential predictor of the need for ”indicated“ SSRI prevention.

Genetic polymorphisms within the serotonergic system have also been associated with vulnerability to IFN-MDD.[134,146,155] Two studies have replicated the finding that a short allele in the serotonin transporter robustly increases risk for IFN-MDD.[134,146]Vulnerability to tryptophan depletion has also been associated with polymorphisms in the 5-HT reuptake transporter.[59] Because IFN-MDD has been associated with lowered tryptophan levels during treatment,[57,91,93]-[156] this suggests that differences in serotonergic tone may leave some people vulnerable to IFN-MDD. It is also plausible that these are the same subjects who may benefit from SSRI prophylaxis, a possibility that requires testing.

Interestingly, gender has not been a consistent predictor of IFN-MDD,[37,157,158] which suggests that IFN-MDD maybe partially distinct from some forms of MDD that are unique to females. Also, as long as patients remain abstinent, a past history- of drug and alcohol abuse is not predictive of increased risk.[15,37,131] This suggests that risks for drug and alcohol abuse are distinct from risk for IFN-MDD. One critical implication is that a past history of drug use, in remission, is not a contraindication to prescribing IFN-α. Nonetheless, several leads are now suggested by these various predictive risk factors, several of which may be amenable to modification. The IFN-MDD paradigm has now been used in several studies to examine whether SSRIs can prevent depression. It may now be useful to determine whether other preventive treatments are effective.

Other populations at selective risk for MDD

In summary, encouraging results indicate that: (i) specific patients may be at elevated risk for IFN-MDD; (ii) this vulnerability may be identifiable prior to IFN-α treatment; (iii) some sources of this vulnerability (such as poor sleep) may be modifiable; and (iv) therefore personalized prevention is testable and could become a reality. Because of the high incidence of IFN-MDD in the first few months of treatment, and the ability to recruit nondepressed patients prior to IFN-α treatment, examining these possibilities appears to be practical and feasible in this population. Several studies with prophylactic SSRIs have already occurred. Furthermore, because of the homologies between IFN-MDD and MDD in general, any lessons learned from IFN-MDD may be translatable to other types of MDD. As examples, MDD occurs at higher rates in populations with multiple chronic illnesses,[159] during bereavement,[160] in caregivers of demented patients,[161] in stroke survivors,[162-164] in postpartum mothers,[110,165,166] and there is preliminary evidence that MDD incidence could potentially be reduced in these settings.[161,167] Similar to IFN-MDD, most people in these settings are resilient to developing MDD, with only a subset who are vulnerable.[168]

Conclusion

It remains an intriguing possibility that modifiable risk factors identified for IFN-MDD may also be modifiable risk factors in these other settings. Thus, targeting the appropriate prevention to the appropriate patient may be possible, and this may soon lead to the personalized prevention of MDD.

Table I.

Comparison of Major Depressive Disorder (MDD) and interferon-a depressive disorder (IFN-MDD) during interferon-a treatment.

Signs and symptoms	MDD	IFN-MDD
Anhedonia	Yes	Yes[24-29]
Depressed mood	Yes	Yes[26-30]
Sleep problems	Yes	Yes [26,27,29-32]
Amotivation	Yes	Yes[27]
Decreased appetite	Yes	Yes [27,22-30]
Concentration changes	Yes	Yes [27-29,33-36]
Tearfulness	Yes	Yes[27]
Social withdrawal	Yes	Yes[27]
Guilt	Yes	Yes [26]
Interpersonal sensitivity	Yes	Yes[37]
Suicidal ideation	Yes	Yes[27,29,38]
Associated symptoms
Irritability	Yes	Yes[26,27,29,30,33]
Increased neuroticism	Yes[40,41]	Not known
Care-seeking behaviors	Yes[42]	Seen in illness behavior[43]
Dependency/acting out
regression/somatization	Yes[44,45]	Senn in illness behavior[45-50]

Table II.

Studies examining prevention of IFN-MDD using antidepressants. Three randomized placebo-controlled trials (RCT), and four open-label studies examining the prevention of major depressive disorder (MDD), diagnosed using criteria from the Diagnostic and Statistical Manual of Mental Disorders-IV (DMS-IV), the Hamilton Depression rating scale (HAM-D), or the Montgomery-Asberg Depression Rating Scale (MADRS).

SSRI	Trial type	(N)	Baseline characteristics	Diagnosis	Comments
Paroxetine[80]	RCT	18 vs 20	Melanoma patients; average HAM-D>5	DSM-IV	Prevented INF-MDD. 2/18 vs 9/20
Paroxetine[85]	RCT	14 vs 19	Average HAM-D<3	DSM-IV	Did not prevent INF-MDD overall. 5/14 vs 6/19
Paroxetine[106]	RCT	23 vs 29	Median MADRS = 3	DSM-IV or MADRS<15	Did not prevent INF-MDD overall 3/23 vs 6/29 Benefit for patients with baseline MADRS>3
Citalopram[104]	Open label	10 vs 0	MDD history in remission	HAM-D =17	1/10 had recurrence of INF-MDD
Paroxetine or Citalopram [120]	Open label	8 vs 0	History of previous INF-MDD (Comparaison with prior IFN-α trial)	HADS>8	0/8 had recurrence of IFN-MDD Small average increase in HADS scores
Citalopram[107]	Open label	14 vs 11	Average MADRS>10; History of affective disorder	DSM-IV	2/14 developed INF-MDD vs 7/11 in the comparison group
Various	Open label	10 vs 21	History of any DSM-IV affective disorder	DSM-IV	2/10 developed INF-MDD vs 10/21 in the comparison group