This uncommissioned systematic review was subject to full peer-review.
Systematic review: the treatment of noncardiac chest pain with antidepressants
Article first published online: 13 JAN 2012
© 2012 Blackwell Publishing Ltd
Alimentary Pharmacology & Therapeutics
Volume 35, Issue 5, pages 493–500, March 2012
How to Cite
Nguyen, T. M. T. and Eslick, G. D. (2012), Systematic review: the treatment of noncardiac chest pain with antidepressants. Alimentary Pharmacology & Therapeutics, 35: 493–500. doi: 10.1111/j.1365-2036.2011.04978.x
- Issue published online: 3 FEB 2012
- Article first published online: 13 JAN 2012
- Manuscript Accepted: 16 DEC 2011
- Manuscript Revised: 12 DEC 2011
- Manuscript Received: 20 OCT 2011
- Manuscript Revised: 13 OCT 2011
Noncardiac chest pain (NCCP) is a common condition, affecting approximately 25% of the general population. The cause of NCCP can be classified as gastro-oesophageal reflux disease (GERD)-related NCCP, where antireflux therapy is the main treatment modality or alternatively as non-GERD-related NCCP, where pain modulators, including antidepressants, are utilised.
To provide a systematic review evaluating the evidence for the use of antidepressants in the treatment of non-GERD-related NCCP.
A computerised literature and manual search was conducted to identify relevant randomised, placebo-controlled studies, published in any language for the evaluation of the effectiveness of antidepressant as a therapeutic intervention for NCCP.
Six randomised placebo-controlled trials of antidepressant treatment for NCCP were identified. The medications included were selective serotonin reuptake inhibitors [paroxetine (n = 2), sertraline (n = 1)], tricyclic antidepressant [impramine (n = 1)], serotonin–norepinephrine reuptake inhibitor [venlafaxine (n = 1)] and a triazolopyridine [trazodone (n = 1)]. The percentage reduction in chest pain was statistically significant with venlafaxine (50% vs. 10%; P < 0.001), sertraline (63% vs. 15%; P = 0.02) and imipramine (52% vs. 1%; P = 0.03). The improvement in chest pain symptoms was independent of improvement in depression scores. Clinical global improvement also significantly improved in patients on venlafaxine, sertraline, paroxetine and trazodone. The percentage of patients in treatment groups reporting adverse effects were relatively high compared with those in placebo groups, although majority were statistically insignificant or significance was not reported. Nonetheless, adverse events were the reported reason for discontinuation of trials in 53% of patients from the antidepressant groups compared with 29% from the placebo group.
There is modest evidence for the benefit of antidepressants in reducing NCCP and improving patients’ general health. However, there is significant heterogeneity amongst the studies with several study limitations to warrant more rigorous trials and to assess the usefulness of low-dose antidepressants in non-GERD NCCP.
Noncardiac chest pain (NCCP) is defined as recurrent episodes of substernal chest pain or discomfort of noncardiac origin. It is a common problem affecting approximately 25% of the population, with a trend to decrease with increasing age. NCCP is a heterogeneous disorder, with oesophageal causes accounting for up to 40–60% of NCCP. It includes oesophageal hypersensitivity, oesophageal motor dysfunction and gastro-oesophageal reflux disease (GERD). Non-oesophageal causes include musculoskeletal (10–20%), psychological (20–60%) and other miscellaneous causes such as pulmonary disease and breast conditions.
Most NCCP patients experience impairment of their daily activities, emotional distress, and as many as 36% have significant impaired quality of life with marked increase in the utilisation of healthcare resources.[2, 5, 6] Their prognosis is relatively poor, and patients will continue to have chest pain for many years after their initial evaluation.
Between 17% and 43% of patients with NCCP have psychiatric co-morbidities, with high prevalence in panic disorder (24–70%), anxiety (33–50%) and major depression (11–22%). It was for this reason that psychoactive agents were initially trialed for the treatment of NCCP. However, as GERD is the most common cause of NCCP, estimated to have a role in as many as 60% of all cases, proton pump inhibitors (PPI) has become the first-line therapy for NCCP, with majority of patients responding to this treatment.
Many patients with non-GERD NCCP who do not respond to proton pump inhibitors have evidence of visceral hypersensitivity, defined as an increased perception to visceral stimulus that is enhanced independent of the intensity of the stimulus.[6, 9] Low-dose antidepressants have been shown to have beneficial effects in other disorders with known visceral hyperalgesia. These include functional disorders such as, functional dyspepsia, irritable bowel syndrome, and chronic pain disorders such as diabetic neuropathy, postherpetic neuralgia, headache and fibromyalgia.[10, 11] Although, the mechanism of action remains unclear, it was postulated that antidepressants may also be effective for NCCP patients with visceral hyperalgesia.
The aim of this study was to determine the efficacy and safety of antidepressant medications in the treatment of non-GERD NCCP. We reviewed the literature for randomised placebo-controlled trials for this systematic review.
We followed the PRISMA guidelines in performing our systematic review. We conducted a systematic search through MEDLINE (from 1950), PubMed (from 1946) and EMBASE (from 1949) to identify relevant articles. The search used the terms antidepressants, functional chest pain, unexplained chest pain, noncardiac chest pain, heartburn, oesophageal disorders, which were searched as text word and as exploded medical subject headings where possible. The reference lists of relevant articles were also searched for appropriate studies. No language restrictions were used in either the search or study selection. A search for unpublished literature was not performed.
We excluded studies that met the following exclusion criteria: (i) Use of healthy volunteers with no prior history of NCCP as participants; (ii) No record of pain severity/frequency/intensity recorded by patients; (iii) Interventions such as balloon distention were used to provoke chest pain for assessment; (iv) Majority of patients having severe coexisting psychiatric illness; and (v) Patients currently prescribed regular cardiac medications for their chest pain.
The data extraction used a standardised data extraction form, collecting information on the publication year, study design, number of patients, number of controls, total sample size, population type, country, mean age, presence of crossover effects in the crossover study, percentage of patient and placebo dropouts, class and dose of the antidepressants, percentage of case and placebo adverse events, percentage reduction in chest pain symptoms, clinical improvement assessment data and depression scores. Quality of the studies was not assessed, and authors were not contacted for missing data.
The literature search identified 437 citations of which 11 randomised clinical trials were identified as potentially meeting the inclusion criteria, and the remaining 426 citations were excluded. Of those excluded, 54 were review articles, 177 had wrong titles and abstracts, 135 were of unrelated gastrointestinal disorders, four had unsuitable study designs, four were experimental reports, four were editorials or letters and 48 were duplicate citations.
From the aforementioned 11 randomised clinical trials, another five articles were subsequently excluded. One was excluded as majority of patients had severe coexisting psychiatric illness, whilst patients in another had possible cardiac causes of chest pain and were on cardiac medications. One article was excluded, as pain records reported by patients were not utilised as a measure of outcome and another as interventions such as balloon distention were used to provoke chest pain. Finally, an article was excluded as participants were healthy volunteers with no prior history of NCCP (Figure 1).
Ultimately, six RCTs were included in this systematic review, producing a total of 251 participants[7, 11-15] and included studies on two selective serotonin reuptake inhibitor (SSRI), paroxetine[14, 15] and sertraline, a tricyclic antidepressant (TCA), imipramine, a serotonin norepinephrine reuptake inhibitor (SNRI), venlafaxine and a triazolopyridine, trazodone. Individual details of these RCTs are summarised in Table 1.
|Authors||Year||Country||Randomised & double blinded||ITT||Total n||Patients||Placebo||% Female||Mean age||% Patient Drop out||% Placebo Drop out||Class||Drug||Dose||% Patient A/E||% Placebo A/E|
|Doraiswamy, et al.||2006||US||Yes||Unreported||50||27||23||42%||53.3||19% (5)||9% (2)||SSRI||Paroxetine||5–50 mg||74% (20)||65% (15)|
|Spinhoven, et al.||2010||the Netherlands||Yes||Yes||46||23||23||48%||57.4||30% (7)||17% (4)||SSRI||Paroxetine||10–40 mg||Unreported||Not reported|
|Varia, et al.||2000||US||Yes||Yes||30||15||15||Unreported||Unreported||7% (1)||27% (4)||SSRI||Sertraline||50–200 mg||33% (5)||Not reported|
|Cannon, et al.||1994||US||Yes||Yes||40||20||20||73%||49.5||0%||0%||TCA||Imipramine||1st week 25 mg then 50 mg||75% (15)||55% (11)|
|Lee, et al.||2010||Korea||Yes||Yes||50 (cross-over)||25||25||14%||23.5||8% (2)||20% (5)||SNRI||Venlafaxine||75 mg||52% (13)||12% (3)|
|Clouse, et al.||1987||US||Yes||No||35||18||17||72%||47.7||17% (3)||18% (3)||Triazolopyridine||Trazodone||100–150 mg||47% (7)||21% (3)|
Limitations of the six RCTs included, amongst other things, a failure to state the method of randomisation,[7, 11-15] small sample sizes[7, 11-15] and large drop-out rates.[7, 11, 13-15] Given the large drop-out rates, a majority of the studies notably failed to discuss whether the baseline characteristics of participants who withdrew were statistically significantly different to the remaining participants[7, 11, 13, 15], and two of the studies did not use or failed to mention whether analysis was by intention-to-treat.[7, 15] In addition, only one of the six studies followed up participants monitoring the long-term effectiveness and the tolerability profile of the studied antidepressant, whereas another did not monitor changes in depression level.
The outcomes measured differed considerably between the six studies and varyingly included chest pain frequency, intensity and/or severity. To achieve consistency for reporting purposes, the percentage reduction of any combination of these outcomes was calculated (Figure 2). The percentage reduction in chest pain severity with sertraline (P = 0.02), frequency with imipramine (P = 0.03) and intensity (severity × frequency) with venlafaxine (P < 0.001) were all statistically significant. Interestingly, although Cannon et al. also measured and demonstrated a significantly lower intensity of chest pain during the administration of imipramine, when compared with the placebo group, this was not significant. Similarly, the percentage reduction in pain score between paroxetine and placebo in the Doraiswamy et al. study and the percentage reduction in pain index (derived from chest pain duration and intensity) in the Spinhoven et al. study were both statistically insignificant. The Clouse et al. study was not included in Figure 2 as sufficient data could not be extrapolated from the graphs provided. However, the study showed no significant difference between the placebo and trazodone group in chest pain frequency and intensity.
Clinical improvement scores were also collated in five of the six studies. The studies examined different clinical improvement areas such as chest pain, functional impairment, anxiety scale, body pain, general health, emotions and global improvement. With the exception of the paroxetine study by Doraiswamy et al., clinical improvement scores were obtained from participants, allowing them to provide feedback at the end of the trial on their overall progress during the study period. Conversely, in the Doraiswamy et al. study, physicians rated the clinical global impression of improvement for the patients and, it should be noted that this was the only statistically significant finding from the study – Clinical Global Severity ratings, and Short Form 36 Health Survey (SF-36) were also measured, but were not statistically significant.
Figure 3 displays the statistically significant areas of improvement, with the exception of data from the imipramine study, which despite showing significant improvement in all patients, failed to show significant difference between the treatment and placebo groups. In the venlafaxine study, health-related quality of life was assessed by SF-36, showing statistically significant improvement in the domains of pain, emotional state and general health perception in the venlafaxine group compared with the placebo group, but no significant difference in the other domains of SF-36. In the sertraline study, SF-36 subscales were all insignificant with the exception of the SF-36 subscale for general health.
Table 2 provides a hierarchy of antidepressants of choice for chest pain reduction and global health improvement. Venlafaxine and sertraline are the two antidepressants of choice to achieve both significant chest pain reduction and global health improvement. Paroxetine was ranked last in the global health improvement despite achieving a statistically significant result in clinical global impression of improvement, as this was determined by the clinicians and was the only statistically significant result in the study.
|Pain reduction||Global health improvement|
|1. Venlafaxine||1. Venlafaxine|
|2. Sertraline||2. Sertraline|
|3. Imipramine||3. Trazodone|
|4. Trazodone||4. Imipramine|
|5. Paroxetine||5. Paroxetine|
The percentage of patients in treatment groups reporting adverse effects were relatively high compared with those in placebo groups,[7, 11, 12, 15] although majority were not statistically significant or significance was not reported. The Spinhoven et al. paroxetine study was not included as there was no discussion of adverse events. Discussed below are the adverse effects documented and whether it was the reason for discontinuation of five of the six studies:
- The incidence of adverse effects with imipramine was 75% compared with 55% for the placebo group, and there was no mention of whether this was significant. Imipramine, however, produced a significant prolongation of the corrected QT (QTc) interval (range of prolongation, −0.01 to 0.04 s; P = 0.02), but there were no signs of proarrhythmic effect even following an average follow-up of 21 months (range 9–33). No patients discontinued therapy in this study due to adverse effects of imipramine.
- The frequency of adverse effects of venlafaxine was reported to be more significant compared with placebo (52% vs. 12% P = 0.005), with reports of sleep disturbance, nausea, dizziness, loss of appetite, fatigue and constipation. The authors did not consider the adverse effects serious enough to necessitate drug withdrawal, and stated that the occurrence of adverse events may be reduced by further dose adjustments. One patient discontinued treatment because of nonlife-threatening adverse effects (sleep disturbance and loss of appetite).
- Thirty-three per cent of participants receiving trazodone reported CNS side effects (dizziness, drowsiness or fatigue) compared with 21% of patients in the placebo group. However, all were rated as ‘mild’, and any differences between the trazodone and placebo group were insignificant. Two participants did not complete the study because of the presumed adverse effects of trazodone (increased symptoms after each capsule in one and sedation in the other) and one participant from the placebo group because of possible side effects.
- The sertraline study reported mild adverse effects in 33% of participants receiving sertraline. Adverse effects included restlessnesss, nausea, decreased libido and delayed ejaculation. The incidence in placebo group was not reported, and there was no comment on whether the incidences were significantly different between the two groups. Whilst five patients withdrew early from the study due to noncompliance or scheduling conflicts, there was no mention of adverse effects being a reason for withdrawal.
- The paroxetine study reported a high incidence of headache (74% vs. 65%) and somnolence (37% vs. 26%) in both the treatment and placebo group. Other adverse events included insomnia, dry mouth, nausea, constipation, joint/muscle pain, pharyngitis, rhinitis and abnormal ejaculation. Again, it was not mentioned whether the difference in incidence of adverse effects between the paroxetine and placebo groups was significant. One patient in this study withdrew because of an allergic reaction (tongue and facial swelling) to paroxetine.
When comparing the overall percentage of patients who discontinued the studies as a result of adverse effects across all five studies, the percentage was much greater in patients who were randomised to the antidepressant groups (53% vs. 29%). The sertraline study was not included in this calculation as the article did not provide the reasons for why participants discontinued the trial.
The paroxetine, sertraline and venlafaxine studies used the BDI to measure depression scores and showed no differences in the depression score from baseline to the end of treatment. The imipramine study that utilised the NIMH Global Scale-Depression and the trazodone study that utilised an unspecified depression scale also did not show any changes to the depression scale during the course of the study.
This systematic review of published RCTs suggests that antidepressants may reduce the symptoms of noncardiac chest pain (NCCP). The percentage reduction of chest pain scores was significant in the sertraline (severity), imipramine (frequency but not pain intensity) and venlafaxine (severity and frequency) studies (Figure 2). However, in both the paroxetine studies,[14, 15] placebo groups showed greater reductions in pain scores and duration/intensity of chest pain than the paroxetine groups. Paroxetine was also found to be less effective than cognitive-behavioural therapy (CBT) in reducing chest pain scores in the Spinhoven et al. study. When commenting on this latter study, Palsson referred to low treatment acceptability amongst patients randomised to the paroxetine, suggesting that the patients’ response to paroxetine may have been impaired by their perception that their therapy was not as beneficial as those randomised to CBT. This is of importance, as this article also reported that approximately 80% of potential NCCP patients declined participation in the study because of the possibility of being randomised to receiving an antidepressant.[5, 14]
Overall, adverse effects from the antidepressants were considered mild and did not raise concerns amongst clinicians. Nonetheless, adverse events were the reported reason for discontinuation of trials in 53% of patients from the antidepressant groups compared with 29% from the placebo group. This implies that, although not severe enough to warrant clinician concern, the adverse effects may have caused patient distress. To this extent, although the adverse events were reported as mild and transient with low-dose antidepressants, the studies need to statistically assess the impact of these adverse effects on the individual. Equally, there is a need to comment on whether differences in the incidences of adverse events between antidepressant and placebo group are statistically significant to better understand their risk and benefit profiles.
The improvement of general health was consistently significant in four of the five studies measuring clinical improvements.[7, 11, 13, 15] In two of these studies, global health improvement was reported as significant despite having an insignificant reduction in chest pain.[7, 15] As such, it may be possible for antidepressants to offer an overall positive effect on patients’ general health and well-being irrespective of whether it can significantly reduce chest pain symptoms. This was demonstrated in the trazodone study, where trazodone reduced the perception of symptoms, as measured by ratings of oesophageal symptom distress, more than what was measured in the objective rating of total oesophageal symptoms.
The high prevalence of psychiatric illness amongst patients with NCCP also raises the question of whether the psychological effect of antidepressants contributes to the reduction in NCCP. In this regard, five of the six studies in this analysis assessed depression scores and all reported no changes in those scores over the duration of the trials.[7, 11-13, 15] This supports the notion of antidepressants as pain modulators for NCCP with a mechanism of action independent of their psychological effects. The imipramine study provides further support by demonstrating that a diagnosis of a psychiatric illness is not a positive predictor of response. That is, patients with psychiatric disorders at baseline were not more likely to have a greater reduction in frequency in chest pain than those without a psychiatric disorder, and psychiatric improvement was evident over time irrespective of the treatment received.
Several other studies also attempted to demonstrate positive predictors of outcome. The sertraline study showed no correlation between treatment and whether, amongst other things, the patient has previously received an angiogram. There was also no difference in the SF-36 subscales and BDI results between patients with a negative angiogram and those with a negative stress test alone. Similarly, the beneficial effect of imipramine was independent of results obtained from extensive cardiac, oesophageal and psychiatric testing. No positive predictors of outcome could also be established amongst the following variables in the trazodone study: age, sex, duration of oesophageal symptoms, presence of chest pain, contraction abnormalities, psychiatric illness, affective disorder, anxiety state and treatment received. However, the inability to detect positive predictors of response to antidepressant treatment may again be a consequence of the small sample size of the individual studies.
There were several limitations to the studies discussed in this review. One of the more critical limitations was the small sample sizes across all six studies, with two likely contributing factors. First, there was a reported unwillingness amongst patients to be randomised to an antidepressant for the treatment of their NCCP,[14, 15] and an unwillingness that is also likely to be shared by most patients with NCCP. This may stem from a combination of the perceived stigma surrounding the administration of antidepressants and a lack of understanding of how antidepressants may help in the relief of chest pain. Given the difficulties that this would pose in patient recruitment, it would therefore be important for future studies to provide clearer evidence of the benefits vs. the risks for this therapy. Such evidence would not only better guide clinicians’ understanding but also the patients’ understanding of this treatment. The second factor contributing to the small sample size is the presence of very selective exclusion/inclusion criteria. This limits both the recruitment of patients and the clinical application of the results to NCCP patients who may have medical backgrounds different from those found in the studies. Perhaps with a less rigid selection criterion in future studies, a larger sample size as well as a more reliable analysis of positive predictors of response could be obtained.
In addition, whilst a majority of the studies reported high drop-out rates, only one commented on whether the baseline characteristics of those who discontinued therapy were significantly different from the participants who completed the trials. This is information vital to the analysis, particularly given the small sample size of the trials. Likewise, a majority of the studies did not have patient follow-up, which is essential in assessing the patients’ compliance to the medications, incidence of adverse events and long-term effects or satisfaction with the treatment.
Currently, there is no clear definition of the role of antidepressants in the treatment of NCCP. Some authors have suggested a trial of nitrates, calcium channel blocker or prokinetic medications before trialing antidepressants as an empirical therapy for visceral hyperalgesia. Others have suggested theophylline as first-line treatment for patients with hypersensitivity, with low-dose antidepressants only used if patients are unable to tolerate or have contraindications to theophylline. One particular paper suggested that it is reasonable to use low-dose antidepressants in patients with non-GERD related nonspastic oesophageal motility disorder, and in patients with spastic oesophageal disorders, and a trial with calcium channel blockers or low-dose antidepressants as a visceral analgesic was recommended. Notwithstanding this lack of a clear role and despite the total body of published trials in this area being relatively small, the results based on our review of six randomised controlled trials is promising. There is modest evidence for the benefit of antidepressants in reducing NCCP and improving patients’ general health. Nonetheless, more rigorous trials are needed to address the limitations of the current trials to provide stronger evidence for the use of antidepressants in relieving non-GERD NCCP.
Declaration of personal and funding interests: None.
- 8Diagnostic and therapeutic use of proton pump inhibitors in non-cardiac chest pain: a meta-analysis. Am J Gastroenterol 2005; 100: 1226–32., , , et al.Direct Link:
- 15A randomized controlled trial of paroxetine for noncardiac chest pain. Psychopharmacol Bull 2006; 39: 15–24., , , et al.
- 16Diagnosis and management of esophageal chest pain. Gastroenterol Hepatol 2011; 7: 50–2..