Treatment Efficacy for Non-Cardiovascular Chest Pain: A Systematic Review and Meta-Analysis

Background Non-cardiovascular chest pain (NCCP) leads to impaired quality of life and is associated with a high disease burden. Upon ruling out cardiovascular disease, only vague recommendations exist for further treatment. Objectives To summarize treatment efficacy for patients presenting with NCCP. Methods Systematic review and meta-analysis. In July 2013, Medline, Web of Knowledge, Embase, EBSCOhost, Cochrane Reviews and Trials, and Scopus were searched. Hand and bibliography searches were also conducted. Randomized controlled trials (RCTs) evaluating non-surgical treatments in patients with NCCP were included. Exclusion criteria were poor study quality and small sample size (<10 patients per group). Results Thirty eligible RCT’s were included. Most studies assessed PPI efficacy for gastroesophageal reflux disorders (GERD, n = 10). Two RCTs included musculoskeletal chest pain, seven psychotropic drugs, and eleven various psychological interventions. Study quality was high in five RCTs and acceptable in 25. PPI treatment in patients with GERD (5 RCTs, 192 patients) was more effective than placebo [pooled OR 11.7 (95% CI 5.5 to 25.0, heterogeneity I2 = 6.1%)]. The pooled OR in GERD negative patients (4 RCTs, 156 patients) was 0.8 (95% CI 0.2 to 2.8, heterogeneity I2 = 50.4%). In musculoskeletal NCCP (2 RCTs, 229 patients) manual therapy was more effective than usual care but not than home exercise [pooled mean difference 0.5 (95% CI −0.3 to 1.3, heterogeneity I2 = 46.2%)]. The findings for cognitive behavioral treatment, serotonin reuptake inhibitors, tricyclic antidepressants were mixed. Most evidence was available for cognitive behavioral treatment interventions. Limitations Only a small number of studies were available. Conclusions Timely diagnostic evaluation and treatment of the disease underlying NCCP is important. For patients with suspected GERD, high-dose treatment with PPI is effective. Only limited evidence was available for most prevalent diseases manifesting with chest pain. In patients with idiopathic NCCP, treatments based on cognitive behavioral principles might be considered.


Background
In the United States, 6 million patients present to emergency departments with chest pain each year, at an annual cost of $8 billion [1,2]. Sixty to ninety percent of the patients that present to emergency departments with chest pain have no underlying cardiovascular disease [3][4][5][6]. The proportion of patients with cardiovascular disease is higher in specialized units (cardiology emergency departments, CCU, ICU) [7] and lower in the primary care setting [6,[8][9][10]. After serious illnesses have been ruled out, physicians often assume that patients with non-cardiovascular chest pain (NCCP) have an excellent prognosis [11,12]. However, patients with NCCP have a high disease burden; many patients that seek care for NCCP complain of persisting symptoms in a 4year follow-up [13]. Furthermore, patients with non-cardiac chest pain experience an impaired quality of life and greater number of medical visits compared with patients with cardiac pain [14].
In patients with chest pain, the diagnostic work-up focuses primarily on cardiovascular disease and is often performed by cardiologists. Upon ruling out cardiovascular disease, only vague recommendations exist for further treatment, delaying appropriate treatment and causing uncertainty for patients [15]. A recent systematic synthesis of diagnostic tests [16] showed that patients with gastroesophageal reflux disorder (GERD) can be identified by their response to proton pump inhibitor (PPI) treatment, and certain clinical findings can guide clinicians to the most appropriate treatments (e.g., pain increase with movement or decrease on medication were associated with musculoskeletal chest pain). However, limited data are available regarding the efficacy of treatments for patients with NCCP.
The present systematic review aimed to summarize the current evidence about the efficacy of different treatments based on randomized controlled trials (RCTs) for patients that seek care for NCCP.

Literature search and study selection
This search, conducted in July 2013, followed the PRISMA statement [17]. We searched six databases: Medline (OvidSP), including In-Process & Other Non-Indexed Citations, Daily and OLDMEDLINE; Web of Knowledge, including Biosis and Web of Science; Embase (OvidSP); EBSCOhost, including CINAHL and PsycINFO; Cochrane Reviews and Trials; and Scopus. We used the following search terms as medical subject headings (MeSH terms) and other subject headings: 'thoracic pain', 'chest pain', 'non-cardiac chest pain', 'atypical chest pain', 'musculoskeletal chest pain', 'esophageal chest pain', and 'thoracic spine pain'. The findings were limited to studies published in the last 20 years. We applied no limits regarding study setting or language. Table S1 depicts two detailed search strategies.
To ensure search completeness, one reviewer (BJ) conducted a thorough search of the bibliographies of all included studies. Potential eligible references were also included in the full text review.

Eligibility criteria
Eligible studies were randomized controlled trials (RCTs) published in the last 20 years. Inclusion criteria were studies reporting on patients aged $18 years seeking care for NCCP. NCCP was defined as chest pain after cardiac or other vascular disease (e.g., cardiovascular disease, aortic dissection, pulmonary embolism) had been ruled out. Studies with less than 10 patients per group were excluded.

Study selection, data extraction, and synthesis
Two reviewers (MW and BJ) independently screened 5372 references by title and abstract. Both reviewers independently reviewed the full text of 62 studies that met the eligibility criteria. Disagreements were discussed and resolved by consensus or third party arbitration (JS). Researchers with specific language proficiencies reviewed non-English language references. When the same study was included in several publications without change in treatment, outcome, or follow-up, the most recent publication was chosen and missing information was added from previous publications.
All information regarding the treatment and control groups, treatment duration, follow-up duration, and patient population was extracted and grouped according to the disease investigated.

Quality assessment
Study quality was assessed using the Scottish Intercollegiate Guidelines Network (SIGN) methodology checklist for RCTs [18]. Overall bias risk and study quality were rated according to the SIGN recommendations. The ratings included high quality (++; the majority of criteria met; little or no risk of bias; results unlikely to be changed by further research.), acceptable quality (+; most criteria met; some flaws in the study with an associated risk of bias; conclusions may change in the light of further studies), and low quality (0; either most criteria not met, or significant flaws relating to key aspects of study design; conclusions likely to change in the light of further studies).
It was not possible to include all studies in the meta-analysis because data was missing for some outcomes. It may have happened that the studies originally considered several different outcome measures, but only reported the measures that provided significant results. Copas et al. [19] refer to this as outcome reporting bias, which is defined as outcome reporting driven by the significance and/or direction of the effect size. All studies that were not included in the meta-analyses were assessed for a potential outcome reporting bias using the 9-item outcome reporting in trails (ORBIT) tool [20]. Risk of bias was rated from low (outcome of interest was not measured) to high [trial report states that outcome was analyzed, but only reports that the result was not significant (typically stating p.0.05)].

Outcome
The outcome of primary interest was chest pain, including chest pain frequency and intensity.
We also assessed psychological outcome measures. In particular, we aimed to assess the efficacy of treatment interventions on anxiety, depression, and panic disorders. All measures were extracted, and validity of the outcome measure used was assessed.

Statistical analysis
Descriptive statistics were used to summarize findings across all groups of diagnostic studies. These included number of patients, mean patient age, and gender distribution.
In order to summarize findings across studies, different pain score scales were re-scaled to a 0-to 10-point scale where necessary. In addition, the frequency scores were homogenized to present results on a monthly basis. To present counted pain events in the treatment and control arms as odds ratios, we used number of events and number of patients in both groups, with a cut-off of .50% improvement. To present changes from baseline to followup in the treatment and control arms as mean differences, we used mean change, standard deviation of change, and number of patients in both groups. If the necessary information was not directly available from the original publication, we derived these quantities following instructions described in the Cochrane Handbook [21]. We assumed a random effects model to obtain a pooled estimate of the effect if more than one trial was available in a subgroup. A restricted maximum-likelihood estimator was used to quantify the amount of heterogeneity.
Risk of bias was assessed using a funnel plot. Funnel plot asymmetry was assessed with the regression test proposed by Egger [22].

Study characteristics
Treatment efficacy for NCCP due to gastroesophageal reflux disease (GERD) Only similar RCTs (n = 7, 771 patients) were included in the meta-analysis (Figure 2). A PPI was administered in most RCTs twice daily for 1 to 8 weeks. In two RCTs, GERD-positive and GERD-negative patients were not distinguished [30,32]; in both of these studies, the efficacy of treatment was more effective than placebo. The pooled odds ratio for a reduction in chest pain of $ 50% was 4.2 (95% CI 2.7 to 6.7, heterogeneity I 2 = 26.6%).
While the heterogeneity among GERD-positive patients was low and three of five RCTs emanated from the same center [26,27,29], there was some heterogeneity among the GERDnegative patients. Regarding GERD-negative patients, two RCTs found a trend toward more chest pain in the treatment group [26,27] while two RCTs found a trend toward less chest pain in the PPI group [29,31]. Funnel plots for the three groups (GERDpositive and GERD-negative mixed, GERD-positive alone, and GERD-negative alone) are depicted in Figure S1. There was no evidence for funnel plot asymmetry in the mixed group (GERDpositive and GERD-negative, p = 0.27) or in the GERD-negative group (p = 0.68); however, some evidence of funnel plot asymmetry was observed in the GERD-positive group (p = 0.04), as assessed using Egger's regression test [22]. We refrained from statistical adjustment for outcome reporting bias as proposed by Copas et al. [19] because only one PPI treatment study was not included in our meta-analysis [28]. Dore et al. did not analyze the outcome of interest in our study and therefore the study was associated with a low risk of outcome reporting bias (Table S3)  week follow-up [33]. Detailed results for all studies are summarized in Table S4.

Treatment efficacy regarding the intensity and frequency of non-cardiac chest pain
Treating patients with NCCP without a specific diagnosis of psychiatric disease with the selective serotonin reuptake inhibitor sertraline was more effective after 2 months [54], but not after 8 months of follow-up for chest pain intensity (Figure 3) [52]. Although use of the tricyclic antidepressant imipramine exhibited a trend towards less chest pain intensity, the effect was not statistically significant [49]. Findings were mixed regarding treatments based on the principles of cognitive behavioral therapy (CBT). One study showed that CBT (4 to 12 60-min sessions) more effectively reduced chest pain intensity at 1 year than usual care at the cardiology department [48]. Another study found that cognitive skill training (5 bi-weekly and 6 30-min monthly sessions), alone or supplemented with sertraline, was no more effective than placebo or sertraline alone in reducing chest pain intensity at 8 months of follow-up [52]. A brief CBT intervention was not more effective in reducing chest pain intensity (Figure 3), but more effectively reduced chest pain frequency (Figure 3) at 3 months of follow-up [39]. Two smaller RCTs found no [44] or only a trend towards [41] less frequent chest pain after CBT interventions (Figure 3). In a small pilot study Johrei, a process of transmission of healing energy, was compared to a waiting list [56]. While the treatment group experienced a significant improvement in symptom intensity these findings need further validation. The heterogeneity of the study settings and treatments used prevented us from conducting a meta-analysis.
One study of patients with musculoskeletal NCCP found that manipulation reduced pain more effectively than acupuncture and usual care (Figure 3) [35]. Another study found that chiropractic treatment and home exercise were similarly effective [36]. The pooled mean difference for the manipulation therapy was 0.5 (95% CI 20.3 to 1.3, heterogeneity I 2 = 46.2%).

Efficacy of treatment regarding anxiety and depression
Only similar studies were included in meta-analyses. CBT more effectively reduced anxiety scores in all studies (Figure 3). The pooled mean difference for CBT in two similar studies [39,43] was 4.4 (95% CI 0.5 to 8.4, heterogeneity I 2 = 32.7%).
CBT treatment reduced depression scores more effectively than usual general practitioner (GP) care after 1 year [41,47]. The pooled mean difference for CBT was 3.1 (95% CI 0.5 to 5.7, heterogeneity I 2 = 52.6%). A study with four arms compared CBT or no treatment and sertraline or placebo [52]. While CBT with and without sertraline was more effective at reducing anxiety scores, they were no more effective at reducing depression scores than placebo. At baseline, depression scores were lower than anxiety scores in this study population (depression 8.9-10.2, range 0-63; anxiety 34-41, range 20-80).

Main findings
The systematic analysis of 30 RCTs involving patients with NCCP demonstrated that PPI treatment was effective in patients with GERD. In NCCP patients without GERD, PPI treatment was no more effective than placebo. Treatment based on cognitive behavioral principles reduced chest pain frequency compared with 'general practitioner treatment' in three RCTs, while one study found no effect. Most studies that compared the efficacy of serotonin antagonists or tricyclic agents with placebo demonstrated no difference or only a trend towards less pain intensity in the treatment arms. Manipulative treatment interventions produced conflicting results for patients with musculoskeletal chest pain, and acupuncture was no more effective than usual care. For most prevalent diseases that manifested with chest pain, only a few studies were available.

Results in light of the existing literature
To our knowledge, this is the first comprehensive systematic review and meta-analysis to summarize the current evidence on treatment efficacy based on RCTs for various diseases presenting in patients with NCCP. Recently new therapies for NCCP of gastrointestinal origin were discussed [57]. Treatment interventions including nitrates, Phosphodiesterase-5 inhibitors, anticholinergics, calcium channel blockers, and endoscopic injection of botulinum toxin may be effective in a subset of patients with gastrointestinal diseases. Most of these interventions have been studied in non-randomized trials or case series [58]. This systematic review and meta-analysis confirms that limited RCTs are available for many interventions and highlights the need for further studies. Non-randomized trials tend to overestimate treatment effects [59]. Further, this comprehensive overview addresses the need for interdisciplinary evaluation and care in patients with NCCP and summarizes the evidence for treatment interventions in underlying diseases oftentimes not considered. A recent systematic review found evidence that NCCP patients have similar levels of psychological morbidity than patients with cardiac chest pain and higher levels than healthy controls [60]. While gastroesophageal diseases are common these findings indicate that other diseases might not be diagnosed.
Only limited evidence was available for most prevalent diseases that manifest with chest pain. Only two RCTs investigated treatment efficacy of manual therapy in patients with musculoskeletal chest pain [35,36]. The efficacy of psychotropic drugs on chest pain intensity, anxiety and depression scores were in line with a recently published meta-analysis that analyzed the efficacy of CBT compared with pharmacotherapy in adults with major depressive disorder (21 RCTs, 2027 patients) or anxiety disorder (21 RCTs, 1266 patients) [61]. The authors found CBT to be equally effective as pharmacotherapy in patients with depression whereas CBT was somewhat more effective than pharmacotherapy in anxiety disorders [61]. It has been shown that patients discharged from the emergency department with the diagnosis of NCCP had elevated anxiety levels compared to healthy individuals [62]. Anxiety disorder might be an underlying disease for subjects with chest symptoms to seek evaluation in emergency departments. Interestingly, CBT was more effective in patients with panic disorders [61]. This information may be relevant for further management of patients with unexplained chest pain. No study was identified that investigated panic disorders in patients with NCCP. However, in patients that present with NCCP to the emergency department, panic disorders are often not diagnosed [63,64].
A recent meta-analysis using a hierarchical Bayesian model demonstrated the diagnostic value of the response to high-dose PPI treatment in patients with GERD [posterior mean sensitivity 0.89 and specificity of 0.88 [16]]. Together with the current findings in patients with NCCP, in which GERD is suspected, PPI treatment should be initiated early and PPI treatment response should be evaluated after 2 weeks [16]. While the findings are in line with previous published meta-analyses on PPI treatment studies [65,66], this is the first meta-analysis to assess study quality and the risk for outcome reporting bias. In comparison to Kahrilas et al. [65,66] and Cremonini et al. [65,66], one additional study was included [30]. Further, Cremonini et al. [65,66] included open-label studies and did not distinguish between GERD-positive and GERD-negative patients. We also used odds ratios as the effect measure for pooling because of its favorable mathematical properties over the relative risk. This includes that the odds ratio is unbounded regardless of the underlying event rate [67]. Our study expands on a relevant aspect in the clinical setting where the PPI response is often used for the diagnosis of GERD. We showed that the effect in a mixed patient sample is smaller compared to patients where GERD was diagnosed by a reference test (e.g. 24hour pH monitoring). Therefore, a lack of response to PPI treatment after 2 weeks should lead to discontinuation of PPI treatment, while a response indicates underlying GERD for which PPI treatment is effective.

Strengths and limitations
This review comprehensively evaluates the currently available studies. The search was inclusive; no language restrictions were applied, a thorough bibliographic search was conducted to identify all relevant studies, and rigorous methodology was applied. The extraction process was performed in accordance with current guidelines and supported by an experienced statistician. Potential factors influencing treatment efficacy were identified by a multidisciplinary team (an internist, general practitioner, statistician, and methodologist).
The main limitation of this systematic review and meta-analysis was the limited number of RCTs. Many interventions used in clinical practice in patients with NCCP were not assessed in RCTs. Of the included studies, many were only of moderate methodological quality. Furthermore, NCCP is a collective term with potentially different underlying diseases and therefore might present differently. Treatment efficacy in one population in which the prevalence for one disease is high might be entirely different for another population [68]. In addition, the heterogeneity of the outcome measures used and follow-up durations reported prevented us from including most studies in our meta-analysis.
The results of these studies should be interpreted on an individual study level within the context of the study population. We have tried to balance this by providing a thorough description of the studies inclusion and exclusion criteria and the study settings. This information will allow readers to judge to whom various study results apply.

Research implications
Additional research should compare diagnostic indicators (e.g., pain increase with movement or decrease on medication were associated with musculoskeletal chest pain) [16] plus a corresponding treatment intervention to usual care alone in defined patient populations (e.g., emergency departments, primary care). Future research should also aim to contribute to our knowledge about diagnostic processes and treatment decisions for patients with NCCP. Although most patients with chest pain consult primary care physicians [69], few studies are performed in this setting. Additional research is needed to strengthen the evidence in a primary care setting. Screening questionnaires for panic and anxiety disorders could be used to identify patients that need further specialized assessment and would respond well to treatments based on cognitive behavioral principles. No such study was found in the current analysis.

Implication for practice
Patients with NCCP incur high healthcare costs owing to extensive and often invasive diagnostic testing, as well as the effect of NCCP on quality of life. Early identification of underlying diseases is essential to avoid delayed treatment and chronicity of complaints. Symptoms and clinical findings may provide important information to guide treatment of an underlying illness [16]. In patients with typical GERD symptoms, twice-daily high-dose PPI treatment is the most efficient diagnostic and therapeutic approach. GERD is very likely if a positive treatment response occurs after 1 week, and is unlikely if there is no response after 4 weeks of PPI treatment [16]. In patients that do not respond to PPI, PPI treatment should be stopped if an endoscopy reveals no pathological findings.
Panic and anxiety disorders are often missed in clinical practice [70]. For patients with anxiety, treatments based on cognitive behavioral principles might be more effective than pharmacologic treatment. To date, evidence for the efficacy of serotonin antagonists or tricyclic agents in patients with NCCP is weak.

Conclusion
Timely diagnostic evaluation and treatment of the underlying disease is important for patients with NCCP. The current systematic review and meta-analysis showed a lack of RCTs for many diseases presenting with NCCP or treatment interventions proposed in the literature. Only limited evidence was available for prevalent diseases that manifest with chest pain. In addition, many treatment interventions that have been shown to be effective in non-randomized trials have not been studied in RCTs and might overestimate treatment efficacy. In patients suspected to have GERD high-dose treatment with a PPI is effective. In otherwise unexplained NCCP treatments based on cognitive behavioral principles might be considered. There is a need for further high quality studies addressing the gaps highlighted in this review.