Association of Protein and Genetic Biomarkers With Response to Lumbar Epidural Steroid Injections in Subjects With Axial Low Back Pain

Abstract

Objective:

The purpose of this observational study was to examine the association of protein and genetic biomarkers with pain and pain-related disability in individuals with axial low back pain undergoing epidural steroid injections.

Design:

Forty-eight adults with axial low back pain undergoing an epidural steroid injection were recruited from an academic medical center. Blood samples were assayed at baseline and follow-up for plasma proteins and functional single-nucleotide polymorphisms associated with pain. Data regarding pain and function were collected at baseline and follow-up. The characteristics of responders (defined as 50% improvement in pain score) and nonresponders were compared, and the association between response and baseline biomarkers was examined.

Results:

Thirty-five percent of subjects were responders to injection. Responders had lower baseline plasma levels of chondroitin sulfate 846 and higher neuropeptide Y and serotonin levels than nonresponders, and baseline neuropeptide Y level correlated with change in disability levels. In addition, subjects with the variant allele for the catechol-O-methyltransferase single-nucleotide polymorphism demonstrated increased odds of responding to the injection.

Conclusions:

These data identify candidates who may have utility for patient selection for spinal procedures and provide support for exploration in prospective studies to assess and validate their predictive ability.

Despite the observation that low back pain represents the leading cause of disability in the United States,¹ outcomes have not improved and cost of care has continued to increase.^2,3 Spinal injections are commonly used to ameliorate axial low back pain associated with degeneration and aging. However, outcomes are variable and difficult to predict, likely related to the difficulty in patient selection. This is especially true for patients with low back pain without radiculopathy. Despite variable outcomes, a review of Medicare claims noted that epidural injections were used as a pain-relieving modality in 36% of patients with axial low back pain.⁴ A recent systematic review of randomized controlled trials revealed level II evidence for the use of epidural injections for axial low back pain without facet or sacroiliac joint pain,⁵ and a randomized double blinded controlled study found greater than 50% relief in 67% of patients with axial low back without radiculopathy after epidural steroid injection (ESI).⁶ Thus, although it appears that a subset of patients may respond to these procedures, the evidence is confounded by indiscriminate use of these procedures in patients not likely to benefit. This results in unnecessary and ineffective utilization of services and associated patient dissatisfaction and complications with poor overall outcomes. Preprocedure clinical and imaging-based predictors have not been clearly identified,⁷ which further complicates the selection of potential patients who might benefit from spinal injections.⁸ Accomplishing improved patient selection and thereby improved outcomes and decreased unnecessary utilization of spine interventions for low back pain requires the identification of novel, objective individualized biomarkers associated with disease activity and pain perception, which serves as the primary motivation of the current study. Molecular biomarkers have garnered much attention recently in other conditions to guide prognosis and treatment and have the potential to serve as predictive tools as well as provide insight into both the etiology of the pain and the mechanisms of therapies. Blood-based protein biomarkers have been explored in low back patients to provide assessment of disease activity and active pain generators. Genetic biomarkers have also generated much interest and have advantages of ease of collection and stability over time. Specifically, the goal of this study was to examine the association of biomarkers, specifically plasma proteins and genetic single-nucleotide polymorphisms (SNPs), with the response to lumbar ESIs (LESIs) in patients with axial low back pain without radiculopathy. Candidate protein and genetic biomarkers were chosen based on previous work demonstrating associations with outcomes in low back pain and other painful conditions.^9–16 A novelty of the current study is that it specifically focused on patients with axial low back pain without radiculopathy, which is a patient population that has been notoriously difficult to treat and phenotype.

METHODS

Participants and Clinical Data Collection

Consecutive patients older than 18 yrs were recruited from academic medical center departments of Physical Medicine and Rehabilitation and Anesthesiology who had already consented for ESI (n = 48) via an interlaminar, transforaminal, or caudal approach, as part of their routine clinical care from July 2012 through May 2014. Importantly, subjects were approached for participation in the study only after being consented for their injection, in an effort to avoid introducing bias into the subject recruitment. All subjects were consented before study participation, and all institutional ethical standards were maintained as approved by the University of Pittsburgh Institutional Review Board (approval 12010334). Formal sample size justification based on statistical power and a fixed 0.05 type I error rate is largely not applicable in this exploratory investigation owing to the greater cost of missed important findings. Nonetheless, assuming an approximately 50% response rate, the sample size was calculated to be able to detect statistical significance with 80% power of a difference corresponding to a Cohen effect size of 0.81 in each biomarker between responders and nonresponders. Subjects were eligible if they had primarily axial low back pain that was more severe than pain in any other part of the body and did not have radiation of pain into the lower limbs, red flag symptoms of serious underlying illness (fever, weight loss, bowel or bladder changes, or neurologic decline), recent oral steroids, uncontrolled psychiatric illness, or systemic inflammatory conditions. Sex, age, body mass index, race, education, employment status, smoking status, previous treatments and exercise, treatment expectation, and medications were collected at the time of enrollment, before the epidural injections (baseline). To remain pragmatic in the study’s approach, recruitment was not limited based on duration of low back pain. Clinical data collected both at baseline and at clinical follow up, approximately 2 wks postinjection, included numeric pain rating scale (NPRS), Oswestry Disability Index (ODI), Roland Morris Disability, McGill Pain Questionnaire (MPQ), generalized anxiety disorder, Patient Health Questionnaire (PHQ)-9, 10 m-walking speed, Fear Avoidance Beliefs Questionnaire, Catastrophizing Scale, and Modified Cumulative Illness Rating Scale. Patients were also queried regarding exercise frequency and expectation of treatment response. The responders to injection were defined as those who had 50% reduction in NPRS at clinical follow-up; otherwise, subjects were categorized as nonresponders. The 50% reduction was chosen as an aggressive threshold to define responders and is consistent with previous work using 30%–50% improvement in pain score as indications of successful clinical treatments.^17,18

LESI Procedures

ESIs were performed by fellowship trained and/or subspecialty board-certified physiatrists or pain anesthesiologists under the guidance of fluoroscopy with the use of 2–3 ml of Omnipaque 180 (GE Health care, Marlborough, MA) or approximately 2 ml of Isovue-m 300 (BIPSO GmbH, Singen, Germany). For the transforaminal ESI (TFESI), typical injections contained 1–1.5 ml of 40 mg/ml Depo-Medrol (methylprednisolone acetate) (Pfizer Inc, New York, NY) and 2–3.5 ml lidocaine (1% preservative free) (Fresenius Kabi LLC, Lake Zurich, IL). For the interlaminar ESI and caudal ESIs, 2 ml of 40 mg/ml Depo-Medrol and 8 ml preservative-free normal saline (USP Hospira Inc, Lake Forest, IL) were used. In the current study, all injections were single-level injections, and 25 were TFESI, 18 were interlaminar ESI, and 5 were caudal approaches. For TFESI and interlaminar ESI, most injections were L5/S1, with the exception of one performed at L3/L4.

Biosample Collection

Blood samples were collected at baseline and clinical follow-up. Samples were immediately placed on ice and the plasma was obtained within 2 hrs by centrifugation, aliquoted, and stored at −80°C until the protein biomarker analyses. A separate sample of whole blood was stored at 4°C and sent to a core facility (Genomics Research Core, University of Pittsburgh) for DNA extraction from the buffy coat. DNA was purified using the Gentra Puregene Blood kit (Qiagen) extraction protocol and reagents.

Biomarker Analyses

Plasma biomarkers were assayed by enzyme-linked immunosorbent assay. Kits were purchased from the following sources: neuropeptide Y (NPY), RANTES (Regulated on activation, normal T-cell-expressed and secreted), and E-selectin from Ray Biotech (Norcross, GA); serotonin from Enzo Life Sciences Inc (Farmingdale, NY); CS846 (aggrecan chondroitin sulfate 846) from IBEX Pharmaceuticals Inc (Montreal, Quebec, Canada); and CTX-II (C-terminal telopeptide of type II collagen) from Immunodiagnosticsystems, IDS Inc (Gaiersburgh, MD). All samples were assayed in duplicate and averaged following the manufactures’ instructions and reported as concentration (ng/ml).

SNP and Haplotypes

Functional SNPs were selected from the literature for NPY (rs16147),¹⁶ COMT (rs4680, rs4633, rs4818, rs16559, rs6269), GCH1 (rs752688, rs3783641, rs4411417, rs998259), and AVPR1 (rs10877969). All SNPs were genotyped by TaqMan allelic discrimination with the ABI Prism 7000 Sequence Detection System, Taqman genotyping assays, and SDS software v1.2.3 (Thermo Fisher Scientific Inc, Waltham, MA). Genotypes were double called by two individuals blinded to phenotypes and discrepancies addressed by review of the raw data and, if necessary, regenotyped. The distribution of haplotypes was constructed for five SNPs, namely, rs6269, rs4633, rs4818, and rs4680, plus rs165599 (SNP numbers cited from National Center for Biotechnology Information databases) in the COMT gene, as defined previously.^19,20 Haplotypes and haplotype frequencies were calculated using Haploview software (version 4.2, The Broad Institute).

Statistical Analysis

Appropriate descriptive statistics were used to summarize participant baseline characteristics and clinical measures for responders and nonresponders. There were no statistical differences regarding the type of injection between the responders and the nonresponders (P = 0.41). Moreover, when TFESI and interlaminar ESI were compared, there were no significant differences in the percentage change in pain after the injections (P = 0.45). Therefore, all types of injections were grouped for the remainder of the analysis. When the two groups, that is, responders and nonresponders, were compared for continuous variables, depending on the distribution of the data, independent-samples t or Wilcoxon rank sum were used; for noncontinuous variables, χ² or Fisher’s exact tests were used. Pearson correlation coefficients were also used to examine the association between pain, ODI improvements, and serum biomarkers. Finally, logistic regression models were used to examine the association between the baseline biomarkers and the response, with a view toward predicting the likelihood of benefit of LESI using response status as the dependent variable and individual (or combinations of) various baseline biomarkers as independent variables. SAS version 9.3 (SAS Institute, Inc, Cary, NC) and SPSS 21.0 (SPSS, Chicago, IL) were used for these statistical analyses. In addition, MedCalc statistical software was used to compute odds ratio (OR) for the various genetic SNPs and COMT haplotype in comparing responders vs. nonresponders and the 95% confidence interval was used. P < 0.05 defined as statistically significantly difference.

RESULTS

Baseline Characteristics of Clinical Data Between Responders and Nonresponders

At clinical follow-up after LESI, 17 subjects reported 50% or greater reduction in the pain score, whereas the remaining 31 subjects did not. The choice of definition of responder is supported by the consistency of this cutoff with this study’s data for the other clinical measures, including a clinically significant reduction of 2 points or more in NPRS¹⁸ and change in pain-related function, including Roland Morris Disability change of greater than or equal to 5 points and ODI change of greater than or equal to 10 points^18,21 (Fig. 1A–C). In addition, there was also a statistically and clinically significant increase in the normal gait speed in the responder group (Table 1), consistent with improvement in the group categorized as responders. There were no significant differences in the clinical measurements at baseline between the groups. Four subjects reported pain less than 3 mos and the remainder of subjects reported pain for greater than 3 mos. Subjects with pain less than 3 mos had no difference in baseline clinical characteristics or response rate compared with subjects with pain greater than 3 mos.

FIGURE 1.

Significant differences between responders and nonresponders in (A) NPRS changes, (B) ODI (% disability), and (C) Roland Morris Disability.

TABLE 1.

Summary of baseline clinical measurements and their changes from baseline at the follow up between the LESI responders (>50% pain relief) vs. nonresponders

Measure	Responders (n = 17)	Nonresponders (n = 31)	P
FABQ work subscale
Baseline	12 (2.4)	18 (2.4)	0.11
Change	−1.3 (1.9)	0.9 (0.9)	0.2
FABQ physical activity subscale
Baseline	17.8 (0.8)	18.2 (1.0)	0.700
Change	−2.8 (1.4)	−0.7 (0.7)	0.1
GAD
Baseline	4.2 (1.3)	7.4 (1.3)	0.11
Change	−0.8 (0.8)	−1.48 (0.6)	0.52
CS
Baseline	6.9 (1.6)	11.1 (1.7)	0.090
Change	−1.6 (1.2)	−0.9 (1.1)	0.69
Gait speed
Baseline	1.7 (0.09)	1.6 (0.12)	0.56
Change	0.92 (0.15)	0.54 (0.08)	0.044a
MCIRS
Baseline	19.5 (0.8)	20.0 (0.8)	0.7
Change	0	0	1

Numbers represent mean (SE). P values were obtained using independent-samples t test.

FABQ indicates Fear Avoidance Beliefs Questionnaire; GAD, generalized anxiety disorder; CS, Catastrophizing Scale; gait speed, 10-m walking speed (meter/second); MCIRS, Modified Cumulative Illness Rating Scale.

^aStatistically significant change.

There were no significant differences in the baseline demographic variables and other clinical characteristics between responders and nonresponders regarding age or sex, body mass index, race, education, employment status, smoking status (Global Adult Tobacco Survey), and nonsteroidal anti-inflammatory drug medications (Table 2). In addition, there were no differences in the baseline symptom severity as measured by NPRS (mean ± SD, 6.1 ± 1.8 vs. 6.1 ± 2.2; P = 0.9) or ODI (40.6 ± 12.9 vs. 35.7 ± 11.6; P = 0.2) between responders and nonresponders. However, the responders were found to have significantly lower PHQ and MPQ scores than nonresponders (Table 3), representing lower depression and pain experience. In addition, before injection, the responders also reported exercising more frequently than the patients who did not respond. Finally, when asked whether they were expecting the pain relief from LESI, among the 48 recruited LESI patients, no subject answered “no”; 28 subjects answered “yes,” whereas another 19 subjects answered “unsure.” There were significantly less “unsure” answers in the responder group compared with that in the nonresponder group (Table 3).

TABLE 2.

Summary of baseline clinical characteristics by LESI responders (>50% pain relief) vs. nonresponders that showed no differences between the two groups

Baseline Measure	Responders (n = 17)	Nonresponders (n = 31)	P
Age, mean (SE)	52 (3)	49 (5)	0.47
BMI, mean (SE)	29.5 (1.9)	30.0 (1.5)	0.800
GATS total, mean (SE)	8.6 (2.8)	10.6 (2.0)	0.06
	No. Patients (%)	No. Patients (%)
Sex			0.24
Male	8 (47)	14 (45)
Female	9 (53)	17 (55)
Race			0.33
Nonwhite	2 (12)	2 (6.5)
White	15 (88)	29 (93.5)
Education			0.4
High school	0 (0)	2 (6.5)
>High school	17 (100)	29 (93.5)
Employed			0.2
No	9 (53)	19 (61.3)
Yes	8 (47)	12 (38.7)
Previous physical therapy			0.24
No	14 (82.4)	23 (74.2)
Yes	3 (17.6)	8 (25.8)
NSAIDs			0.6
No	14 (82.4)	26 (83.9)
Yes	3 (17.6)	5 (16.1)
Opioids			0.73
No	13 (76.5)	25 (80.6)
Yes	4 (23.5)	6 (19.4)
Neuropathic pain medications			0.25
No	12 (70.6)	27 (87.1)
Yes	5 (29.4)	4 (12.9)
Muscle relaxants			0.64
No	16 (94)	27 (87.1)
Yes	1 (6)	4 (12.9)

P values were obtained using independent-samples t test or Fisher’s exact test.

High school indicates some high school and high school graduate (or equivalent); >high school, some college or university, associate/bachelor’s degree, master’s degree, and doctorate/professional degree. NSAIDs reported included aspirin and ibuprofen. Opioids reported included hydrocodone-acetaminophen and oxymorphone. Neuropathic pain medications reported included amitriptyline, duloxetine, and gabapentin. Muscle relaxants reported included methocarbamol, cyclobenzaprine, carisoprodol, and baclofen.

BMI indicates body mass index; GATS, Global Adult Tobacco Survey; NSAID, nonsteroidal anti-inflammatory drug.

TABLE 3.

Summary of baseline clinical characteristics by LESI responders (>50% pain relief) vs. nonresponders that showed significant differences between the two groups of patients

Baseline Measure	Responders (n = 17)	Nonresponders (n = 31)	P
PHQ total, mean (SE)	6 (1.2)	12.5 (1.6)	0.007
MPQ sensory, mean (SE)	1.09 (0.08)	1.40 (0.12)	0.045
MPQ affective, mean (SE)	0.63 (0.11)	1.19 (0.16)	0.016
MPQ total, mean (SE)	0.96 (0.08)	1.35 (0.12)	0.03
	No. Patients (%)	No. Patients (%)
Past exercise			0.012
<3 days/wk	7 (41)	23 (74)
3–5 days/wk	9 (53)	4 (13)
>5 days/wk	1 (6)	4 (13)
Expect relief			0.03
Yes	14 (82)	14 (47)
No	0 (0)	0 (0)
Unsure	3 (18)	16 (53)

P values were obtained using independent-samples t test or Fisher’s exact test.

Baseline Characteristics of Biomarkers Related to the Responses to LESI

The responders had lower baseline plasma levels of the matrix biomarker CS846, higher NPY, and higher serotonin compared with the nonresponders (Fig. 2A–C).

FIGURE 2.

Significant differences between responders and nonresponders in (A) baseline levels of CS846, (B) baseline levels of NPY, and (C) baseline levels of serotonin. Significant differences levels are indicated in the figure.

Baseline NPY levels, although not correlated with the baseline ODI or pain score, were significantly correlated with the ODI change from that before to that after injection and demonstrated a trend with pain score changes (Fig. 3A and B). Upon evaluation of combinations of biomarkers, NPY and serotonin together demonstrated a trend toward an association with reduction in pain score (P = 0.0734), and NPY was significantly associated with pain reduction (P = 0.0417) independent of serotonin levels.

FIGURE 3.

Pearson correlations (p and ρ values) between baseline NPY levels and changes in (A) ODI and (B) NPRS. A significant correlation between baseline NPY levels and changes in ODI was observed, whereas there was a trend toward significant correlation between baseline NPY levels and changes in pain NPRS.

Characteristics of Pain-Related SNPs Between Responders and Nonresponders

Subjects with the variant allele for the SNP COMT rs4680 demonstrated increased odds of responding to the injection (OR, 18; P = 0.05). Subjects with the variant allele for the SNP COMT rs4633 showed a trend toward having increased odds of responding to an injection (OR, 6.7; P = 0.07) (Fig. 4A and B). Subjects with the variant allele for the SNP AVPR1 rs10877969 showed a trend toward having decreased odds of responding to a LESI (OR, 0.1; P = 0.08) (Fig. 4C).

FIGURE 4.

Percentage of responders and nonresponders to LESI with each genotype. (A) COMT rs4680: non-A-allele (GG) carriers (n = 9) and A-allele (AG/AA) carriers (n = 35), which are grouped together for statistical comparison with the non-A-allele (GG) homozygotes. Subjects with the variant allele A had increased odds of responding to the injection (OR, 18; P = 0.05). (B) COMT rs4633: non-A-allele (GG) carriers (n = 8) and A-allele (AG/AA) carriers (n = 36), which are grouped together for statistical comparison with the non-A-allele (GG) homozygotes. Subjects with the variant allele A showed a trend toward having increased odds of responding to the injection (OR, 6.7; P = 0.07). (C) AVPR1 rs10877969: non-G-allele (AA) carriers (n = 34) and G-allele (AG/GG) carriers (n = 9), which are grouped together for statistical comparison to the non-G-allele (AA) homozygotes. Subjects with the variant allele G showed a trend toward having decreased odds of responding to a LESI (OR, 0.1; P = 0.08). (D) COMT haplotype blocks for the five SNPs investigated using Haploview v4.2. Haplotype blocks indicate regions of a chromosome that are tightly linked and inherited together. This figure demonstrates that four of the SNPs investigated are in one haplotype block (rs6269, rs4633, rs4818, and rs4680), in which the COMT haplotype of ACCG occurred in higher-than-expected frequency in the nonresponder group (P = 0.04).

Association of COMT Haplotypes With ESI Response

Because multiple SNPs were measured for COMT, the distribution of haplotypes was constructed for five SNPs in the COMT gene and assessed for the association with response to injection.¹⁹ Linkage disequilibrium, which measures the nonrandom associations between the two alleles based on expectations relative to allele frequencies in the patient population in this study, is shown in Figure 4D. Consistent with the previous report,²² the COMT SNPs included one haplotype block and one loosely associated SNP. The correlation coefficient (r²) between rs6269, rs4633, rs4818, and rs4680 was 0.568 and showed high correlations. The COMT haplotype of ACCG occurred in higher-than-expected frequency in the nonresponder group (χ² = 4.246, P = 0.04; Fig. 4) and showed a trend toward increased odds of nonresponse to injection compared with all other haplotypes (OR, 6.46; P = 0.08).

DISCUSSION

Current guidelines have shown good evidence for the management of lumbar disc herniation or radiculitis with interlaminar or TFESIs, but evidence is only fair or limited for axial or discogenic lumbar pain without disc herniation or radiculitis.⁶ Our study was consistent with these findings, as only 17 of 48 patients (35%) had a 50% reduction in their pain score at the 2-wk follow-up after a LESI. This underscores the importance of identifying these subjects before an injection, to facilitate consideration of injections only for those patients most likely to respond. The results from the current study demonstrate the association of clinical data, plasma proteins, and genetic SNPs with response to LESI in patients with axial low back pain without radiculopathy and may represent candidates for use as predictive tools. There are few previous studies that have examined biomarkers as predictors of response to LESI. Previous work has shown limited value of clinical characteristics alone to predict response to injection, with only health-related quality of life as measured by the EQ-5D showing association with response, but this was only for leg, not low back, symptoms.²³ In addition, radiographic predictors have demonstrated limited utility, with only contrast flow pattern during the injection significantly associated with outcome and no useful preprocedure predictors.⁷ A previous exploratory study evaluating 48 blood based inflammatory biomarkers demonstrated associations of biochemical mediators with response to ESI.²⁴ However, this study included a small number of patients (n = 16) with various diagnoses, that is, disc herniation, degenerative disc disease, and spinal stenosis. In addition, a previous study using epidural lavage demonstrated associations of a fibronectin fragment with response to injection, but the utility of this assay is limited by the need for an invasive approach to collect the sample for measurement.²⁵ The current study identifies candidate biomarkers for use before a planned procedure.

Clinical Biomarkers

For subjects in the study, clinical data collected showed that there were no differences between the responders and the nonresponders for baseline demographics or NPRS pain scores between the two groups before LESI. However, there were several baseline clinical measures that were significantly different. Interestingly, patients who were responders were less likely to be unsure regarding expecting relief. This is consistent with previous results that demonstrate the importance that treatment expectancy plays in response to injections²⁶ and suggests a potential expectation bias that could not be statistically addressed given the sample size of the current study. However, this raises the question regarding a possible placebo effect. In fact, it has been shown in patients with lumbar spinal stenosis that there was no benefit of LESI over lidocaine injection alone.²⁷ In fact, the COMT SNP rs4680 A-allele-carriers, which were more prevalent in subjects who were responders in our study, has previously been associated with a higher placebo response.²⁸ This underscores the importance of patient expectation and patient education when considering LESI, which should be explored clinically to determine appropriate use of these procedures. In addition, the potential presence of biologic differences associated with placebo response represents an important avenue for future exploration in the treatment of pain.

The PHQ, which has been shown to be a valid measure for patient depression, was also lower in responders, which may be reflective of the role that mental and emotional health can have in low back pain. Similarly, nonresponders had a statistically significant higher score in all domains of the MPQ, which has been shown to be a valid measure for patient low back pain. This may represent an important indicator of the affective aspect of low back pain that may not respond to LESI. It is interesting to note that responders differed in PHQ and MPQ, which represent depression and the experience or emotional response to pain, but did not differ in their pain score or reported level of disability. This, coupled with our findings of a treatment expectation effect and differences in previous exercise between responders and nonresponders, may suggest an important role in the patients’ resilience or self-efficacy and the affective component of the pain experience in influencing treatment outcomes. This may represent an important educational target for patient treatment, and this topic warrants further exploration in future studies. In fact, recent data demonstrating noninferiority of injections without steroid has raised question regarding the utility of steroid in the observed treatment effect,²⁷ underscoring the importance of addressing important contributors to pain beyond inflammatory mediators.

Protein Biomarkers

Blood-based biomarkers have shown great potential to be used in the diagnosis, prognosis, and treatment validation in musculoskeletal conditions. Specific to this study, the authors chose to assess three categories of serum molecular biomarkers, including tissue metabolites, inflammatory cytokines or chemokines, and pain-related biomarkers. Lower levels of CS846, an epitope present in newly synthetized aggrecan molecules, correlated with response to LESI. CS846 is increased in the joints and synovial fluid of patients with rheumatoid arthritis or osteoarthritis¹² and may reflect pathologic increased turnover of matrix. However, it cannot be determined using peripheral samples the tissue source of this epitope. Distinct from matrix turnover markers, pain-related biomarkers may provide information relevant to the subjects’ overall pain experience, which may be more relevant to predicting response to a procedure which targets pain, such as an ESI.

NPY has been found in the annulus fibrosus of human intervertebral discs removed for back pain,²⁹ and circulating NPY levels show a strong correlation with pain and pain-related function in older adults with axial low back pain.⁹ In the current study, NPY levels differed between responders and nonresponders and demonstrated an association with changes in pain and pain-related disability. Previous studies have reported that lower NPY levels predict higher emotion-induced activation of the amygdala, as well as diminished resiliency as assessed by pain/stress-induced activation.¹⁶ Thus, the observation in the current study that responders had higher NPY levels may be related to a greater capacity and resilience to experience pain improvement after LESI. Even though the exact role that NPY plays in low back pain remains unclear, the results from our study implicate NPYas being a particularly useful serum protein biomarker to predict improved pain response and pain-related function to LESI. Lower levels of serotonin were also seen in nonresponders. Serotonin has been well known to be involved with pain perception and analgesia as well as depression³⁰ and is thought to play an active role in the neurochemical basis of chronic pain and depression.³¹ In our study, as the nonresponders had lower basal levels of serotonin, and a higher PHQ depression scores than responders, depression may represent one of the important factors influencing the responses to ESI.

Genetic Biomarkers

Genetic factors have been implicated to contribute to the variability in the experience of low back pain. Recent evidence has shown certain genetic SNPs were predictive of recovery in low back pain.³² In addition, the genes measured in this study have been previously associated with human pain.

COMT encodes for the enzyme catechol-O-methyl transferase, which is responsible for metabolizing catecholamine neurotransmitters that are involved in adrenergic, noradrenergic, and dopaminergic signaling, all of which have important roles in the pain perception pathway. Our results showed for the single genetic SNP COMT rs4680, which locates in the coding area for the gene enzyme product, subjects with the variant allele (Val/G substituted with Met/A) had increased odds of responding to LESI. In addition, for the COMT SNP rs4633, subjects with the variant allele showed a trend toward having increased odds of responding to a LESI. Previous studies have shown that the COMT variant allele may contribute to an increased functionality of the COMTenzyme that has a protective role in clinical pain syndromes,³³ which is consistent with our finding of the increased odds of response in patients with the variant allele.

It is generally accepted that the COMT haplotype including rs 4680 and rs 4633 and related diplotype are better indicators of the variations in sensitivity to experimental pain.^19,22 ACCG COMT haplotype and related diplotype have been associated with lower enzyme activity and high pain sensitivity to experimental pain and higher pain rating after shoulder surgery.³⁴ Consistent with the literature, our analysis showed that the COMT haplotype of ACCG occurred in a higher-than-expected frequency in the LESI nonresponder group and showed a trend of increased odds of nonresponse to LESI compared with all other haplotypes. This may have resulted from the higher pain sensitivity in those patients. Similarly, subjects with the variant allele for the SNP AVPR1 rs10877969 showed a trend toward having decreased odds of responding to a LESI. The AVPR1 gene encodes for the vasopressin-1A receptor, and the SNP in the promoter region of AVPR1 helps to modulate the chemical/inflammatory pain sensitivity¹⁵ and impact depressive symptoms and pain catastrophizing in patients with other musculoskeletal conditions.³⁴ Therefore, the differential response to LESI in subjects with and without variant allele in this particular SNP may be related to modulatory and nonnociceptive factors in the experience of pain.

Overall, the results of current study demonstrate the potential utility of using clinical data, serum proteins, and genetic SNPs as biomarkers to help predict LESI response in subjects with axial low back pain without radiculopathy. As this is a complex disease process, having several different categories of biomarkers (clinical, protein, and genetic) is advantageous to use in combination to create a multibiomarker panel, which should be explored in future work to develop robust clinical predictive capacity. Furthermore, such a multibiomarker panel may lead to individualized approaches to patient care and more cost-effective treatment protocols, as well as minimize risk and maximize benefit in patients with axial low back pain without radiculopathy.

There are several limitations in this study that should be noted. First, the study contained a relatively small sample size, particularly for genetic assessments. Consequently, subjects had similar ethnicities, with most being white, which limits the generalizability of the study, and certain SNP variants might be disproportionately represented. Future studies including a larger and more ethnically diverse sample size will be necessary to evaluate the ability of these and other genetic SNPs to predict LESI treatment response. In addition, our study only assessed for response at one follow-up time-point and as such did not evaluate synergy with other treatments or long-term improvement, and the outcome assessment relied on patient report to categorize patients as responders or nonresponders. However, in the current study, responders to the injection were deemed as having a 50% reduction in their pain score as this has been consistently validated in the literature for clinical relevance.^17,18,35 Our data support this stratification, as the responders also had a clinically meaningful decrease in their NPRS pain score and disability rating as reflected by their Roland Morris Disability and ODI scores after LESI,¹⁸ as well as improved gait speed. Thus, as LESI is intended for use as a pain-relieving procedure, these represent relevant and important outcomes. Finally, based on study design, the data only suggests an association and does not imply a mechanistic link between the biomarkers and LESI response, and these biomarkers are not readily available for clinical use. However, these data provide multiple potential targets to be evaluated in future studies evaluating their predictive utility.

In conclusion, this study demonstrates that select clinical characteristics, plasma proteins, and genetic SNPs demonstrate associations with response to LESI in subjects with axial low back pain without radiculopathy. Improving upon current treatment effects will require improved ability to phenotype patients and correctly and efficiently select appropriately matched treatment, and identification of relevant biomarkers is the first step in the development of robust prediction models to individualize treatments based on clinical and biologic phenotypes, thereby exposing patients only to risks associated with a treatment with a higher likelihood of efficacy, and also decrease the costs of care by eliminating procedures with low likelihood of benefit. As the costs of these biologic tests continue to decrease and clinical data become more readily accessible, such an approach represents a potential innovative solution to this complex problem.

What Is Known

• Axial low back pain represents a significant cause of functional limitation. Despite conflicting evidence, epidural steroid injections demonstrate benefit in a subset of patients. However, difficulty in identifying patients likely to respond to epidural injections limits the clinical utility of this treatment.

What Is New

• Through a pragmatic approach, associations were identified between patient response to epidural steroid injections and clinical, plasma protein, and genetic biomarkers. Biomarkers were identified that may provide mechanistic targets for the development of novel treatments, and associations with patients’ experience and pain-related biomarkers related to resilience provide an avenue for individualizing care.