The effectiveness of intradiscal biologic treatments for discogenic low back pain: a systematic review

BACKGROUND CONTEXT

There are limited treatments for discogenic low back pain. Intradiscal injections of biologic agents such as platelet-rich plasma (PRP) or stem cells (SC) are theorized to have regenerative properties and have gained increasing interest as a possible treatment, but the evidence supporting their use in clinical practice is not yet well-defined.

PURPOSE

Determine the effectiveness of intradiscal biologics for treating discogenic low back pain.

STUDY DESIGN

PRISMA-compliant systematic review.

PATIENT SAMPLE

Patients with discogenic low back pain confirmed by provocation discography or clinical and imaging findings consistent with discogenic pain.

OUTCOME MEASURES

The primary outcome was the proportion of individuals with ≥50% pain relief after intradiscal biologic injection at 6 months. Secondary outcomes included ≥2-point pain score reduction on NRS; patient satisfaction; functional improvement; decreased use of other health care, including analgesics and surgery; and structural disc changes on MRI.

METHODS

Comprehensive literature search performed in 2018 and updated in 2020. Interventions included were biologic therapies including mesenchymal stem cells, platelet rich plasma, microfragmented fat, amniotic membrane-based injectates, and autologous conditioned serum. Any other treatment (sham or active) was considered for comparative studies. Studies were independently reviewed.

RESULTS

The literature search yielded 3,063 results, 37 studies were identified for full-text review, and 12 met established inclusion criteria for review. The quality of evidence on effectiveness of intradiscal biologics was very low. A single randomized controlled trial evaluating platelet-rich plasma reported positive outcomes but had significant methodological flaws. A single trial that evaluated mesenchymal stem cells was negative. Success rates for platelet-rich plasma injectate in aggregate were 54.8% (95% Confidence Interval: 40%-70%). For mesenchymal stem cells, the aggregate success rate at six months was 53.5% (95% Confidence Interval: 38.6%-68.4%), though using worst-case analysis this decreased to 40.7% (95% Confidence Interval: 28.1%-53.2%). Similarly, ≥30% functional improvement was achieved in 74.3% (95% Confidence Interval: 59.8%-88.7%) at six months but using worst-case analysis, this decreased to 44.1% (95% Confidence Interval: 28.1%-53.2%).

CONCLUSION

Limited observational data support the use of intradiscal biologic agents for the treatment of discogenic low back pain. According to the Grades of Recommendation, Assessment, Development and Evaluation System, the evidence supporting use of intradiscal mesenchymal stem cells and platelet-rich plasma is very low quality.

Association of Spinal Cord Stimulator Implantation With Persistent Opioid Use in Patients With Postlaminectomy Syndrome

IMPORTANCE

The results of studies evaluating spinal cord stimulation (SCS) for postlaminectomy syndrome (PLS) have yielded mixed results. This has led to an increased emphasis on objective outcome measures such as opioid prescribing.

OBJECTIVE

To determine the association between SCS and long-term opioid therapy (LOT) for PLS.

DESIGN, SETTING, AND PARTICIPANTS

In this cohort study, adults with PLS were identified using the TriNetx Diamond Network and separated based on whether they underwent SCS. Patients were stratified according to baseline opioid use (opioid-naive or receiving LOT) and subsequent opioid therapy over the 12-month period ranging from 3 to 15 months post-SCS implantation or post-PLS index date. Statistical analysis was performed from June to December 2021.

EXPOSURE

SCS.

MAIN OUTCOMES AND MEASURES

The main outcome was cessation of opioid use among patients receiving LOT or abstinence from opioids among opioid-naive patients. Opioid-naive patients were defined as those receiving at most 2 opioid prescriptions per year, and patients on LOT were those receiving at least 6 opioid prescriptions per year.

RESULTS

Among 552 937 eligible patients treated between December 2015 and May 2021, 26 179 with PLS received an SCS implant. The median (IQR) patient age was 60 (51-69) years; 305 802 patients (55.3%) were female. Among those reporting racial identify (37.0% [204 758 patients]), 9.3% (18 971 patients) were African American, 0.3% (648 patients) were Asian, and 90.4% (185 139 patients) were White. Compared with those who did not receive an SCS, individuals who received an SCS were more likely to be using opioids preimplantation (mean [SD] prescriptions: 4.3 [8.5] vs 4.1 [9.3]; P < .001) but less likely to be using opioids after SCS implantation (mean [SD] prescriptions: 3.8 [8.2] vs 4.0 [9.4]; P = .006). In the 12-month study period, similar proportions in the SCS and no-SCS groups receiving baseline LOT remained on LOT (70.3% [n = 74 585] vs 69.2% [n = 3882], respectively; P = .10). In opioid-naive patients, SCS was associated with a small decreased likelihood of patients subsequently receiving LOT (7.6% vs 7.0%; difference, -0.6% [95% CI, -1.0% to -0.2%]; P = .003). In multivariable analysis, SCS was associated with an increased likelihood of not being on opioids in both opioid-naive (adjusted odds ratio [OR], 0.90 [95% CI, 0.85-0.96]; P < .001) and LOT patients (adjusted OR, 0.93 [95% CI, 0.88-0.99]; P = .02). White patients were significantly more likely to be diagnosed with PLS (ie, underwent surgery) (90.4% vs 85.2%; difference, 5.2% [95% CI, 5.1%-5.4%]; P < .001) and receive an SCS (93.7% vs 90.3%; difference, 3.4% [95% CI, 2.9% to 4.0%]; P < .001) than patients of other racial identities.

CONCLUSIONS AND RELEVANCE

These findings suggest that under real-life conditions, SCS was associated with small, clinically questionable associations with opioid discontinuation and not starting opioids in the context of PLS.

Permanent Paraplegia as a Complication of Injection of Contrast Media at L2-L3 Vertebral Level

BACKGROUND

The lumbar intrathecal (subarachnoid) space is accessed for both therapeutic and diagnostic purposes. Occasionally, the needle may unintentionally enter the intrathecal space during lumbar interlaminar epidural steroid injections (LESIs)-one of the most commonly performed medical procedures in the United States. Ordinarily, this merely constitutes a minor complication or even a desired placement (in the case of some diagnostic procedures). However, some patients have a rare condition wherein the spinal cord terminates below the L2 vertebral level (tethered cord). In such cases, injections administered at the lumbar level may potentially result in spinal cord damage and irreversible paraplegia if the physician performing the intervention does not recognize the intramedullary position of the needle.

OBJECTIVE

The aim of this study is to describe and analyze an unintentional L2-L3 injection of contrast medium into a tethered spinal cord. Many physicians may consider lumbar injections "safe" because the spinal cord usually terminates at or above the L2 vertebral level. However, complacency stemming from this false impression of safety contributes to nonadherence to practice guidelines, which may lead to catastrophic neurological complications. Presented here is the first published occurrence of paraplegia that resulted from contrast medium injection into the spinal cord during a myelography study performed below the L2 vertebral level.

CONCLUSIONS

Disregard of the procedural guidelines by the physicians performing an elective diagnostic intervention may cause devastating neurological complications. The described casualty occurred because of failure to review previous imaging studies, injection of the contrast medium despite unsuccessful attempts to aspirate cerebrospinal fluid, and an unwillingness to terminate the procedure immediately when the patient reported an unusual sensation in both of his lower extremities. Consequently, we suggest that not only for cervical and thoracic but also for lumbar interlaminar ESIs, previous imaging studies should be reviewed before the injection.

Guideline summary review: an evidence-based clinical guideline for the diagnosis and treatment of low back pain

BACKGROUND CONTEXT

The North American Spine Society's (NASS) Evidence Based Clinical Guideline for the Diagnosis and Treatment of Low Back Pain features evidence-based recommendations for diagnosing and treating adult patients with nonspecific low back pain. The guideline is intended to reflect contemporary treatment concepts for nonspecific low back pain as reflected in the highest quality clinical literature available on this subject as of February 2016.

PURPOSE

The purpose of the guideline is to provide an evidence-based educational tool to assist spine specialists when making clinical decisions for adult patients with nonspecific low back pain. This article provides a brief summary of the evidence-based guideline recommendations for diagnosing and treating patients with this condition.

STUDY DESIGN

This is a guideline summary review.

METHODS

This guideline is the product of the Low Back Pain Work Group of NASS' Evidence-Based Clinical Guideline Development Committee. The methods used to develop this guideline are detailed in the complete guideline and technical report available on the NASS website. In brief, a multidisciplinary work group of spine care specialists convened to identify clinical questions to address in the guideline. The literature search strategy was developed in consultation with medical librarians. Upon completion of the systematic literature search, evidence relevant to the clinical questions posed in the guideline was reviewed. Work group members utilized NASS evidentiary table templates to summarize study conclusions, identify study strengths and weaknesses, and assign levels of evidence. Work group members participated in webcasts and in-person recommendation meetings to update and formulate evidence-based recommendations and incorporate expert opinion when necessary. The draft guideline was submitted to an internal and external peer review process and ultimately approved by the NASS Board of Directors.

RESULTS

Eighty-two clinical questions were addressed, and the answers are summarized in this article. The respective recommendations were graded according to the levels of evidence of the supporting literature.

CONCLUSIONS

The evidence-based clinical guideline has been created using techniques of evidence-based medicine and best available evidence to aid practitioners in the diagnosis and treatment of adult patients with nonspecific low back pain. The entire guideline document, including the evidentiary tables, literature search parameters, literature attrition flowchart, suggestions for future research, and all of the references, is available electronically on the NASS website at https://www.spine.org/ResearchClinicalCare/QualityImprovement/ClinicalGuidelines.aspx.

Guidelines for Composing and Assessing a Paper on the Treatment of Pain: A Practical Application of Evidence-Based Medicine Principles to the Mint Randomized Clinical Trials

Objective

To perform a thorough assessment of the recently published Mint Trials in order to illustrate how to read and analyze a study critically, according to principles of evidence-based medicine.

Design

Narrative review.

Method

We have applied the recently published guidelines for composing and assessing studies on the treatment of pain to a recently published article describing a large study that claimed its "findings do not support the use of radiofrequency denervation to treat chronic low back pain." These guidelines describe the critical components of a high-quality manuscript that allows communication of all relevant information from authors to readers.

Results

Application of evidence-based medicine principles to the publication describing the Mint Trials reveals significant issues with the methodology and conclusions drawn by the authors. A thorough assessment demonstrates issues with inclusion/exclusion criteria, diagnostic block protocols, radiofrequency neurotomy technique, co-interventions, outcome measurement, power analysis, study sample characteristics, data analysis, and loss to follow-up. A failure to definitively establish a diagnosis, combined with use of an inadequate technique for radiofrequency neurotomy and numerous other methodological flaws, leaves the reader unable to draw meaningful conclusions from the study data.

Conclusions

Critical analysis, rooted in principles of evidence-based medicine, must be employed by writers and readers alike in order to encourage transparency and ensure that appropriate conclusions are drawn from study data.

Fluoroscopically Guided vs Landmark-Guided Sacroiliac Joint Injections: A Randomized Controlled Study

OBJECTIVES

To determine the prevalence of intra- and extra-articular sacroiliac joint (SIJ) pain, which injection is more beneficial, and whether fluoroscopy improves outcomes.

PATIENTS AND METHODS

This patient- and evaluator-blinded comparative effectiveness study randomized 125 participants with SIJ pain from April 30, 2014, through December 12, 2017, to receive fluoroscopically guided injections into the joint capsule (group 1) or "blind" injections to the point of maximum tenderness using sham radiographs (group 2). The primary outcome was average pain on a 0 to 10 scale 1 month after injection. A positive outcome was defined as at least a 2-point decrease in average pain score coupled with positive (>3) satisfaction on a Likert scale from 1 to 5.

RESULTS

For the primary outcome, no significant differences were observed between groups (mean ± SD change from baseline, -2.3±2.4 points in group 1 vs -1.7±2.3 points in group 2; 95% CI, -0.33 to 1.36 points for adjusted difference; P=.23), nor was there a difference in the proportions of positive blocks (61% vs 62%) or 1-month categorical outcome (48% vs 40% in groups 1 and 2, respectively; P=.33). At 3 months, the mean ± SD reductions in average pain (-1.8±2.1 vs -0.9 ± 2.0 points; 95% CI, 0.11 to 1.58 points for adjusted difference; P=.02) and worst pain (-2.2±2.5 vs -1.4±2.0 points; 95% CI, 0.01 to 1.66 points for adjusted difference; P=.049) were greater in group 1 than 2, with other outcome differences falling shy of statistical significance.

CONCLUSION

Although fluoroscopically guided injections provide greater intermediate-term benefit in some patients, these differences are modest and accompanied by large cost differences.

TRIAL REGISTRATION

clinicaltrials.gov Identifier: NCT02096653.

The Effectiveness and Risks of Fluoroscopically Guided Lumbar Interlaminar Epidural Steroid Injections: A Systematic Review with Comprehensive Analysis of the Published Data

Objective

To determine the effectiveness and risks of fluoroscopically guided lumbar interlaminar epidural steroid injections.

Design

Systematic review of the literature with comprehensive analysis of the published data.

Interventions

Three reviewers with formal training in evidence-based medicine searched the literature on fluoroscopically guided lumbar interlaminar epidural steroid injections. A larger team consisting of five reviewers independently assessed the methodology of studies found and appraised the quality of the evidence presented.

Outcome Measures

The primary outcome assessed was pain relief. Other outcomes such as functional improvement, reduction in surgery rate, decreased use of opioids/medications, and complications were noted, if reported. The evidence on each outcome was appraised in accordance with the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) system of evaluating evidence.

Results

The search yielded 71 primary publications addressing fluoroscopically guided lumbar interlaminar epidural steroid injections. There were no explanatory studies and all pragmatic studies identified were of low quality, yielding evidence comparable to observational studies.

Conclusions

The body of evidence regarding effectiveness of fluoroscopically guided interlaminar epidural steroid injection is of low quality according to GRADE. Studies suggest a lack of effectiveness of fluoroscopically guided lumbar interlaminar epidural steroid injections in treating primarily axial pain regardless of etiology. Most studies on radicular pain due to lumbar disc herniation and stenosis do, however, report statistically significant short-term improvement in pain.

The Effectiveness and Risks of Non-Image-Guided Lumbar Interlaminar Epidural Steroid Injections: A Systematic Review with Comprehensive Analysis of the Published Data

OBJECTIVE

To determine the effectiveness and risks of non-image-guided lumbar interlaminar epidural steroid injections.

DESIGN

Systematic review.

INTERVENTIONS

Three reviewers with formal training and certification in evidence-based medicine searched the literature on non-image-guided lumbar interlaminar epidural steroid injections. A larger team of seven reviewers independently assessed the methodology of studies found and appraised the quality of the evidence presented.

OUTCOME MEASURES

The primary outcome assessed was pain relief. Other outcomes such as functional improvement, reduction in surgery rate, decreased use of opioids, and complications were noted, if reported. The evidence was appraised in accordance with the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) system of evaluating evidence.

RESULTS

The searches yielded 92 primary publications addressing non-image-guided lumbar interlaminar epidural steroid injections. The evidence supporting the effectiveness of these injections for pain relief and functional improvement in patients with lumbar radicular pain due to disc herniation or neurogenic claudication secondary to lumbar spinal stenosis is limited. This procedure may provide short-term benefit in the first 3-6 weeks. The small number of case reports on significant risks suggests these injections are relatively safe. In accordance with GRADE, the quality of evidence is very low.

CONCLUSIONS

In patients with lumbar radicular pain secondary to disc herniation or neurogenic claudication due to spinal stenosis, non-image-guided lumbar interlaminar epidural steroid injections appear to have clinical effectiveness limited to short-term pain relief. Therefore, in a contemporary medical practice, these procedures should be restricted to the rare settings where fluoroscopy is not available.

Epidural steroid injections compared with gabapentin for lumbosacral radicular pain: multicenter randomized double blind comparative efficacy study

OBJECTIVE

To evaluate whether an epidural steroid injection or gabapentin is a better treatment for lumbosacral radiculopathy.

DESIGN

A multicenter randomized study conducted between 2011 and 2014. Computer generated randomization was stratified by site. Patients and evaluating physicians were blinded to treatment outcomes.

SETTINGS

Eight military, Veterans Administration, and civilian hospitals.

PARTICIPANTS

145 people with lumbosacral radicular pain secondary to herniated disc or spinal stenosis for less than four years in duration and in whom leg pain is as severe or more severe than back pain.

INTERVENTIONS

Participants received either epidural steroid injection plus placebo pills or sham injection plus gabapentin.

MAIN OUTCOME MEASURES

Average leg pain one and three months after the injection on a 0-10 numerical rating scale. A positive outcome was defined as a ≥ 2 point decrease in leg pain coupled with a positive global perceived effect. All patients had one month follow-up visits; patients whose condition improved remained blinded for their three month visit.

RESULTS

There were no significant differences for the primary outcome measure at one month (mean pain score 3.3 (SD 2.6) and mean change from baseline -2.2 (SD 2.4) in epidural steroid injection group versus 3.7 (SD 2.6) and -1.7 (SD 2.6) in gabapentin group; adjusted difference 0.4, 95% confidence interval -0.3 to 1.2; P=0.25) and three months (mean pain score 3.4 (SD 2.7) and mean change from baseline -2.0 (SD 2.6) versus 3.7 (SD 2.8) and -1.6 (SD 2.7), respectively; adjusted difference 0.3, -0.5 to 1.2; P=0.43). Among secondary outcomes, one month after treatment those who received epidural steroid injection had greater reductions in worst leg pain (-3.0, SD 2.8) than those treated with gabapentin (-2.0, SD 2.9; P=0.04) and were more likely to experience a positive successful outcome (66% v 46%; number needed to treat=5.0, 95% confidence interval 2.8 to 27.0; P=0.02). At three months, there were no significant differences between treatments.

CONCLUSIONS

Although epidural steroid injection might provide greater benefit than gabapentin for some outcome measures, the differences are modest and are transient for most people.Trial registration ClinicalTrials.gov Identifier: NCT01495923.

Epidural Steroid Injections are Safe and Effective: Multisociety Letter in Support of the Safety and Effectiveness of Epidural Steroid Injections

BACKGROUND

In April 2014, the Food and Drug Administration (FDA) issued a Drug Safety Communication requesting that corticosteroid labeling include warnings that injection of corticosteroids into the epidural space of the spine may result in rare but serious adverse events, including loss of vision, stroke, paralysis, and death.

RESULTS

The International Spine Intervention Society spearheaded a collaboration of more than a dozen other medical societies in submitting the letter below to the FDA on November 7, 2014. We are publishing the letter to ensure that the readership of Pain Medicine is aware of the multisociety support for the safety and effectiveness of these procedures. A special note of thanks to all of the societies who signed on in support of the message.

Epidural steroid injections, conservative treatment, or combination treatment for cervical radicular pain: a multicenter, randomized, comparative-effectiveness study

BACKGROUND

Cervical radicular pain is a major cause of disability. No studies have been published comparing different types of nonsurgical therapy.

METHODS

A comparative-effectiveness study was performed in 169 patients with cervical radicular pain less than 4 yr in duration. Participants received nortriptyline and/or gabapentin plus physical therapies, up to three cervical epidural steroid injections (ESI) or combination treatment over 6 months. The primary outcome measure was average arm pain on a 0 to 10 scale at 1 month.

RESULTS

One-month arm pain scores were 3.5 (95% CI, 2.8 to 4.2) in the combination group, 4.2 (CI, 2.8 to 4.2) in ESI patients, and 4.3 (CI, 2.8 to 4.2) in individuals treated conservatively (P = 0.26). Combination group patients experienced a mean reduction of -3.1 (95% CI, -3.8 to -2.3) in average arm pain at 1 month versus -1.8 (CI, -2.5 to -1.2) in the conservative group and -2.0 (CI, -2.7 to -1.3) in ESI patients (P = 0.035). For neck pain, a mean reduction of -2.2 (95% CI, -3.0 to -1.5) was noted in combination patients versus -1.2 (CI, -1.9 to -0.5) in conservative group patients and -1.1 (CI, -1.8 to -0.4) in those who received ESI; P = 0.064). Three-month posttreatment, 56.9% of patients treated with combination therapy experienced a positive outcome versus 26.8% in the conservative group and 36.7% in ESI patients (P = 0.006).

CONCLUSIONS

For the primary outcome measure, no significant differences were found between treatments, although combination therapy provided better improvement than stand-alone treatment on some measures. Whereas these results suggest an interdisciplinary approach to neck pain may improve outcomes, confirmatory studies are needed.

Incidence of neuropathic pain after cooled radiofrequency ablation of sacral lateral branch nerves

OBJECTIVE

To determine the incidence of neuropathic pain after cooled radiofrequency ablation (RFA) of the sacral lateral branches for the treatment of chronic posterior sacroiliac joint complex pain.

DESIGN

Retrospective chart review of all patients with chronic posterior sacroiliac joint complex pain who underwent cooled RFA of the sacral lateral branches in our practice between July 2011 and February 2014.

SETTING

Single academic pain practice at a tertiary care medical center.

SUBJECTS

Thirty-six patients with chronic posterior sacroiliac joint complex pain.

METHODS

All charts were reviewed to determine the procedure date, unilateral or bilateral, number of levels treated, and number of individual lesions. Side effects were assessed for their presence or absence, character, intensity, duration, and whether treatment was initiated or symptoms resolved spontaneously.

RESULTS

Forty-eight separate procedures were performed, with a total of 193 levels and 430 lesions. Three patients had transient postprocedure neuropathic pain yielding a 0.7% (95% confidence interval [CI]± 0.4%) rate of this complication per lesion. This proportion increases to 6.2% (95% CI ± 3.5%) per procedure and to 9.4% (95% CI ± 5.2%) per patient.

CONCLUSION

The incidence of postprocedural neuropathic pain after cooled RFA for posterior sacroiliac joint complex denervation is low and in a similar range to that in the lumbar spine. We consider this procedure safe to be utilized by pain medicine practitioners.

Percutaneous mechanical disc decompression using Dekompressor device: an appraisal of the current literature

OBJECTIVES

The objective of this study was to determine if the available literature answers the following questions: does percutaneous disc decompression using Dekompressor device relieve radicular pain caused by a small disc herniation? Is the duration of response stable and clinically worthwhile? Is relief of pain corroborated by improvements in physical and social function? Does relief of pain result in reduction in the use of other health care? Is there a risk of serious side effects or complications? DESIGN/SETTINGS: The study was designed as a narrative review and description of the available evidence, drawn from the databases of PubMed, EMBASE, and the Cochrane Library. Innovatively, the concept of the "context of the patient" was introduced in the assessment. It required the assessors to consider the alternatives the patients and their treating practitioners faced.

RESULTS

The literature search identified three nonrandomized clinical trials, and a single case series. All studies were reasonably rigorous in reporting relief of pain and the use of analgesics. Evidence with respect to physical functioning was scarce. Although investigators reported on the relief of pain, they lacked rigor when reporting associated outcome measures such as use of other health care and physical functioning.

CONCLUSIONS

Unfortunately, the context of a patient with persistent radicular pain caused by a small disc herniation is the lack of good alternatives to Dekompressor procedure. The moral question is whether Dekompressor is any less valid an option than perpetual opioids or discectomy. This question would be much easier to answer if the literature on Dekompressor was more rigorous and more compelling in its evidence.

Combination therapy for neuropathic pain: a review of current evidence

Neuropathic pain is a debilitating chronic condition that remains very difficult to treat. Recently, a number of clinical studies have compared the effectiveness of combination drug therapy with monotherapy for neuropathic pain treatment. In this article, we summarize up-to-date clinical studies of combination therapy for the treatment of both cancer- and non-cancer-related neuropathic pain. Despite a relatively small number of clinical studies on this topic, several positive indications have emerged. First, clinical studies using gabapentin (five positive trials) and pregabalin (five positive trials and one negative trial) in combination with an opioid, cyclo-oxygenase-2 inhibitor or antidepressant have shown positive responses greater than the respective monotherapies for pain related to diabetic neuropathy and postherpetic neuropathy. Second, high-concentration (8%) topical capsaicin and a 5% lidocaine patch seem to be effective add-on therapies (a modality of combination therapy) for various neuropathic pain conditions. Third, combination therapy for cancer-related neuropathic pain has yielded only limited success based on a number of small-scale clinical studies. While there are benefits of using combination therapy for neuropathic pain treatment, including better pain relief and reduced adverse effects, more clinical studies are required in order to (i) make head-to-head comparisons between combination and single-drug therapies, (ii) identify symptom-specific combination therapies for distinctive clinical neuropathic pain conditions, (iii) explore combination therapies that include non-drug modalities such as physical therapy, psychological coping and biofeedback to facilitate functional restoration and (iv) develop new and objective evaluation tools for clinical outcome assessment.

Price Prediction in a Trading Agent Competition

The 2002 Trading Agent Competition (TAC) presented a challenging market game in the domain of travel shopping. One of the pivotal issues in this domain is uncertainty about hotel prices, which have a significant influence on the relative cost of alternative trip schedules. Thus, virtually all participants employ some method for predicting hotel prices. We survey approaches employed in the tournament, finding that agents apply an interesting diversity of techniques, taking into account differing sources of evidence bearing on prices. Based on data provided by entrants on their agents' actual predictions in the TAC-02 finals and semifinals, we analyze the relative efficacy of these approaches. The results show that taking into account game-specific information about flight prices is a major distinguishing factor. Machine learning methods effectively induce the relationship between flight and hotel prices from game data, and a purely analytical approach based on competitive equilibrium analysis achieves equal accuracy with no historical data. Employing a new measure of prediction quality, we relate absolute accuracy to bottom-line performance in the game.