Serum Biomarkers for Chronic Pancreatitis Pain Patterns

Abstract

Objectives:

Chronic pancreatitis (CP) is associated with debilitating refractory pain. Distinct subtypes of CP pain have been previously characterized based on severity (none, mild-moderate, severe) and temporal (none, intermittent, constant) nature of pain, but no mechanism-based tools are available to guide pain management. This exploratory study was designed to determine if potential pain biomarkers could be detected in patient serum and whether they associate with specific pain patterns.

Methods:

Cytokines, chemokines, and peptides associated with nociception and pain were measured in legacy serum samples from CP patients (N=99) enrolled in the North American Pancreatitis Studies. The unsupervised hierarchical cluster analysis was applied to cluster CP patients based on their biomarker profile. Classification and regression tree was used to assess whether these biomarkers can predict pain outcomes.

Results:

The hierarchical cluster analysis revealed a subset of patients with predominantly constant, mild-moderate pain exhibited elevated interleukin-1β (IL-1β), interleukin-6 (IL-6), interleukin-2 (IL-2), tumor necrosis factor alpha (TNFα), and monocyte chemoattractant protein-1 (MCP1) whereas patients with higher interleukin-4 (IL-4), interleukin-8 (IL-8) and calcitonin gene related peptide (CGRP) were more likely to have severe pain. Interestingly, analyses of each individual biomarker revealed that patients with constant pain had reduced circulating TNFα and fractalkine. Patients with severe pain exhibited a significant reduction in TNFα as well as trends towards lower levels of IL-6 and substance P.

Discussion:

The observations from this study indicate that unique pain experiences within the chronic pancreatitis population can be associated with distinct biochemical signatures. These data indicate that further hypothesis-driven analyses combining biochemical measurements and detailed pain phenotyping could be used to develop precision approaches for pain management in patients with chronic pancreatitis.

Introduction

Chronic pancreatitis (CP) is a fibro-inflammatory disease in which ~90% of patients report pain.¹ For non-obstructive cases, there are a variety of medical and psychosocial options available for pain management², but there are no diagnostic tools available in the clinic to guide decision-making. In cases of obstructions (e.g., stones, strictures) or refractory pain, endoscopic or surgical intervention are indicated, however long-lasting pain relief is highly variable and unpredictable.^3–7 Physicians rely on the World Health Organization stepladder approach originally developed for cancer pain management. By moving through a predetermined order of therapeutics, patients may have to try several failed therapies before receiving an effective one and that prolongs suffering. Being able to identify biological mechanisms and prescribe interventions accordingly could bring relief to patients sooner. For patients managed medically, drugs are not always effective and can cause a range of deleterious side effects.⁸ For patients that end up undergoing invasive procedures for pain management (e.g. endoscopy, surgery), only one to two-thirds report significant improvement in pain.^3,5,6 Thus, understanding the biological signaling mechanisms that contribute to different subtypes of CP pain is vital for the identification and development of targeted therapies.

Research groups who have phenotyped their patients’ pain identified distinct pain patterns within the CP population.^9–12 Their focus was on the severity or frequency of painful episodes; the severe and constant pain pattern is an independent predictor of quality of life.^10,13 Animal models of CP have been used to recapitulate this severe ongoing pain.¹⁴ For instance, repeated injection of the cholecystokinin analog cerulein, results in decreased spontaneous activity and mechanical allodynia of the upper abdomen.^15,16 Dibutylin dichloride (DBTC), is a toxic substance excreted by the biliary and pancreatic ducts. Intravenous injection of DBTC induces pancreatitis that is associated with abdominal mechanical allodynia as well as heat hypersensitivity.^17–19 Ductal infusion of trinitrobenzene sulfonic acid (TNBS) also induces pancreatitis and leads to mechanical allodynia, increased immobility, electrically evoked nocifensive behavior and decreased voluntary wheel running.^20–26 Both spontaneous and evoked ‘pain’ behaviors are directly associated with changes in cytokines, chemokines, and peptides involved in nociception. Within the pancreas there is an upregulation of nerve growth factor (NGF); treatment with anti-NGF abrogates both mechanical allodynia and electrically evoked nocifensive behaviors.^20,22,27 Experimental pancreatitis pain is also associated with an upregulation of pro-nociceptive neuropeptides (calcitonin gene related peptide (CGRP) and substance P) and cytokines (tumor necrosis factor alpha (TNFα), interleukin 1-beta (IL-1β), interleukin-6 (IL-6), brain-derived neurotrophic factor (BDNF) and transforming growth factor beta-1 (TGFβ1)) in sensory neurons.^{15,18,20,21,23,25} Moreover, neutralization of TGFβ1, IL-1β or IL-6 is sufficient to attenuate experimental pancreatitis-related pain.^19,24,28

These putative pain biomarkers associated with experimental pancreatitis pain have been validated in clinical studies by analyzing pancreas tissue and pancreatic juice. Specifically, increased expression of CGRP, substance P, and fractalkine within the neuronal fibers innervating pancreas tissue correlates with pain severity and the extent of neuritis.^29–31 The cytokines, IL-8, NGF and BDNF are elevated in the pancreas of CP patients with pain.^29,32,33 Obtaining pancreatic tissue is expensive and invasive; furthermore, procuring extrapancreatic neural tissue is impossible. Therefore, researchers have tried to develop biomarkers with minimally invasive biospecimens such as serum that can easily be tested as part of a regular clinic visit. Several cytokines and chemokines tested in the current study were previously found to be elevated or reduced in serum from CP patients.³⁴ For instance, IL-6, TNFα, fractalkine, monocyte chemoattractant protein 1 (MCP1), TGFβ1, and NGF were elevated in serum from CP patients as compared to healthy controls.^35–39 While these are nociception-related molecules those studies did not distinguish between the absence and presence of pain in their study populations. One recent study, however, reported that IL-1β, IL-8, and MCP1 are significantly higher in serum from CP patients with pain as compared to those with no pain.⁴⁰ While this is an important step towards the identification of CP pain biomarkers, no previous study has examined a large number of potential targets nor have they stratified patients by subtype of pain. The primary goals of this pilot study were to determine if the putative biomarkers identified in pancreatic tissue are also detectable in serum and whether biomarkers for specific pain patterns can be identified.

Material and methods

North American Pancreatitis Studies (NAPS2)

NAPS2 is a series of three cross-sectional studies (original, continuation and validation, ancillary) designed to characterize the complex factors associated with recurrent acute pancreatitis (RAP) and CP.^9,41,42 NAPS2 prospectively enrolled patients with CP from 26 US centers during the years 2000–2014. CP was defined as definitive evidence on imaging studies or histology.⁷ Imaging studies included endoscopic retrograde cholangiopancreatography (ERCP), magnetic resonance cholangiopancreatoraphy (MRCP), computed tomography (CT), or endoscopic ultrasound (EUS). NAPS2 Data Collection: Data were collected in the NAPS2 studies using two comprehensive questionnaires, one completed by patients with the assistance of a clinical research coordinator and the other by the enrolling physician investigator.^9,41,42 The patient questionnaire collected information on demographics, exposure to risk factors, personal and family history, clinical symptoms, hospitalizations and emergency room visits, medication use, and quality of life. The physician questionnaire documented the disease phenotype including etiology and risk factors, exocrine and endocrine insufficiency, findings on imaging studies, treatments tried, and their perceived effectiveness. Study participants also provided a blood sample for research purposes. All patients provided a written informed consent prior to any study procedures. All studies were approved by the institutional review board at each participating institution. Relevant NAPS Data for Current Study: Two specific types of data from the NAPS2 questionnaires were used for the current study, demographic and pain pattern. Based on responses by the patient to questions on the presence of pain in the year preceding enrollment and choosing one of five pre-specified patterns of pain⁹, we selected patients in the following categories in the current study: no pain (N=19), mild/intermittent (N=20), mild/constant (N=20), severe/intermittent (N=20), and severe/constant (N=20). Legacy serum samples from the CP patients was used for the analysis of putative pain biomarker expression. All data and serum samples were de-identified.

Biological Assay

Putative biomarkers (n=15) were chosen based on previous literature (Table 1). Most targets were measured using the Meso Scale Discovery multiplex platform enabling simultaneous measurement of 1–8 analytes. CGRP and substance P were measured individually using ELISA kits (Cayman Chemical). CGRP and substance P were assayed individually using a Thermofisher microplate reader. Concentrations were calculated by SkanIT microplate reader software (Thermofisher). All assays were performed according to the manufacturers’ instructions. All assays included standards to develop a calibration curve and have large detectable ranges (0.1pg/ml-0.1μg/ml). Positive and negative (diluent only) controls were included on each 96 well plate. All samples were run in duplicate. All experimenters performing data collection were blinded and coded samples were randomized across plates.

Table 1.

Intra-class correlation coefficient (ICC) across technical replicates for target expression

Target	ICC
IL-1β	0.923
IL-2	0.917
IL-4	0.624
IL-6	0.937
IL-8	0.998
IL-10	0.498
IFNγ	0.983
TNFα	0.977
MCP1	0.997
TGFβ1	0.929
NGF	0.981
BDNF	0.976
Fractalkine	0.956
Substance P	0.997
CGRP	0.998

Statistical Analysis

Similar to many studies on cytokine markers, the measurement of the cytokine levels suffered from missing values due to detection limits. All markers had technical replicates. The intra-class correlation coefficients (ICC) for those markers, on the logarithm-scale, within their dynamic ranges were used to assess reproducibility. The statistical analyses were done with SAS/STAT version 9.4 and R version 3.4. Exploratory Cluster Analyses: For exploratory analyses, all cytokine markers were transformed into logarithm scale with base 2 and missing values of the cytokine markers were imputed by the lowest value of the observed values within their dynamic ranges minus a small value, 5% of the range. Hierarchical clustering via Euclidean distance was used to cluster the CP patients and the cytokine markers and the corresponding heat map was used to describe how cytokine markers were visually associated with pain outcomes. Classification and Regression Tree (CART) was used to identify patient subgroups that are partitioned according to the cytokine makers with homogeneous pain outcomes. The generated tree was pruned to one with the smallest cross-validation error. Quantitative Analyses: Markers with rare missing values were regarded as quantitative markers. The average of technical replicates was calculated and transformed to the logarithm scale with base 2. The log2 values were compared among pain outcome groups via the analysis of variance. Qualitative Analyses: Markers with a substantial number of missing values were regarded as qualitative. The average of technical replicates was calculated and transformed to the logarithm scale with base 2. Association between qualitative markers and pain outcomes (including both pain severity and pain temporality) were assessed via separate Pearson Chi-square tests. Since this was an exploratory pilot study, unadjusted p-values are presented.

Results

Putative pain pattern biomarkers are reliably detected by MSD and ELISA assays

The demographic information of the cohort included in the current study is outlined in Table 2. Despite the wide detection range of the assays used, several measures were outside the range of detection. This was prevalent in measures of IL-2 (65%), IL-4 (24%), IL-8 (45%), BDNF (21%) and CGRP (19%). Levels being outside the range of detection was not widespread for IL-1β (1%), IL-6 (5%), IL-10 (8%), MCP1 (2%), TGFβ1 (1%) or substance P (2%). There were no missing values for TNFα, IFNγ, NGF and fractalkine. The reliability across technical replicates was high for the majority of targets, with ICC higher than 0.9 with the exception of IL-4 and IL-10 (Table 1).

Table 2.

Patient Characteristics

	No Pain	Intermittent mild/moderate pain	Constant mild/moderate pain	Intermittent severe pain	Constant severe pain	Total
Gender						P=0.76
Male	11 (58)	10 (50)	13 (65)	9 (45)	11 (55)	54 (55)
Age *						P=0.01
	58 (18)	47 (18)	47 (13)	56 (20)	41 (16)	50 (18)
Race						P=0.27
White	18 (95)	17 (85)	15 (75)	19 (95)	18 (90)	87 (88)
Non-White	1 (5)	3 (15)	5 (25)	1 (5)	2 (10)	12 (12)
Drinking						P=0.10
Current	7 (37)	8 (40)	2 (10)	3 (15)	4 (20)	24 (24)
Smoking						P=0.14
None	5 (26)	8 (40)	3 (15)	9 (45)	7 (35)	32 (32)
Past	9 (47)	4 (20)	4 (20)	3 (15)	4 (20)	24 (24)
Current	5 (26)	8 (40)	13 (65)	8 (40)	9 (45)	43 (43)
Etiology						P=0.02
Alcoholic	9 (47)	6 (30)	13 (65)	9 (45)	7 (35)	44 (44)

Notes

^*Mean (standard deviation). Analysis of variance for continuous variable Age. Pearson’s chi-square test for discrete variables.

Exploratory analyses suggest unique signatures distinguish pain features

In order to determine if patients naturally cluster into subgroups based on the expression levels of all of the putative biomarkers, we performed exploratory analyses. An unsupervised hierarchical clustering approach reveal three main clusters (Figure 1). The cluster of patients with high levels MCP1, IL-6, TNFα, IL-1β and IL-2 was comprised of a higher proportion of patients with constant pain. Interestingly, patients with lower levels of these markers but higher IL-8, IL-4 and CGRP levels tended to fall into the severe pain category. To further explore these data and determine ranges of biomarker expression that may segregate patient subgroups, we performed CART analysis, which is a classification system that identifies homogeneous subgroups within a dataset. Patients with severe pain were disproportionately (25/35, 71%) more likely to exhibit moderate IL-1β expression (−0.56 to 5.3), lower TNFα (<2.5) and lower IL-6 (<1.3) compared to only 40% among all participants (Figure 2A). Interestingly, three different cytokine/chemokine profiles were disproportionately (13/16=81%) associated with a constant pain phenotype (Figure 2B). Specifically, patients with lower MCP1(<6.7), higher MCP1 and lower Fractalkine (<11) or lower IL-10 (<−0.96) and lower TNFα (<1) were more likely to report their pain as constant.

Figure 1. Heatmap and hierarchical clustering results.

Unsupervised heatmap of biomarker expression across 99 patients. Patients cluster into 3 distinct clusters base upon pattern of expression of putative biomarkers. Each row is an individual patient (patient number indicated on the right side of the map). Color scale indicates relative level of expression with red being the highest level and white the lowest level. The leftmost column indicates temporality, and the second column indicates severity of the patient. However, pain pattern was not included in the clustering algorithm.

Figure 2. Classification and Regression Tree (CART).

A) A tree showing the probability of patients having no pain, mild/moderate pain, or severe pain predicted by the expression levels of selected markers. Numbers under the leaves refer to the number of observations within that leaf belonging to the class of no pain, mild/moderate pain, and severe pain, respectively. B) A tree showing the probability of patients having intermittent or constant pain predicted by the expression levels of selected markers. Numbers under the leaves refer to the number of observations within that leaf belonging to the class of intermittent and constant pain, respectively. The percentages under the counts refer to the percentage of observations in the leaf. Cutoffs for expression levels of individual markers is given in log2-scale.

Unique decreases in serum biomarkers distinguish severe and constant pain

For the cytokines, chemokines, and peptides defined as quantitative markers, we compared the expression between the different pain patterns, analyzing severity and temporality separately. Patients with severe pain exhibited significantly lower TNFα levels (p=0.04) and trended toward lower IL-6 levels (p=0.07) and substance P levels (p=0.06) compared to no or moderate pain (Table 3). Interestingly, a similar analysis comparing expression between different temporal patterns revealed a distinct biomarker profile. Specifically, there was a trend toward lower Fractalkine levels (p=0.09) and TNFα levels (p=0.08) in patients with constant pain as compared to no or intermittent pain (Table 4). The other quantitative markers were not associated with pain severity nor pain temporality with p-values ranging from 0.16 to 0.94.

Table 3.

Distribution of quantitative markers by pain severity

Markers	No pain	Mild/moderate pain	Severe pain	P-value*
Fractalkine	10.9 (1.2)	10.3 (1.6)	10.4 (1.5)	0.34
IL-1β	1.9 (2.6)	2.0 (3.4)	1.3 (2.0)	0.3
IL-6	2.2 (3.4)	2.6 (4.2)	0.7 (2.9)	0.07
IL-10	0.2 (1.5)	−0.2 (1.9)	0.0 (1.9)	0.93
MCP1	8.7 (1.6)	8.7 (1.8)	8.3 (1.1)	0.21
TNFα	2.6 (2.0)	2.2 (2.5)	1.4 (1.9)	0.04
TGFβ1	14.6 (0.7)	14.7 (0.5)	14.8 (0.5)	0.21
IFNγ	−4.8 (2.5)	−4.5 (2.3)	−4.5 (2.3)	0.65
NGF	−6.2 (1.4)	−6.4 (1.1)	−6.1 (1.4)	0.57
Substance P	5.0 (0.6)	4.9 (0.7)	4.7 (0.5)	0.06

Notes: Data are presented as log2-scale mean (standard deviation).

^*p-value from analysis of variance

Table 4.

Distribution of quantitative markers by pain temporality

Markers	No pain	Intermittent pain	Constant pain	P-value*
Fractalkine	10.9 (1.2)	10.5 (1.8)	10.2 (1.2)	0.09
IL-1β	1.9 (2.6)	1.9 (3)	1.4 (2.6)	0.47
IL-6	2.2 (3.4)	1.5 (3.8)	1.8 (3.7)	0.83
IL-10	0.2 (1.5)	0.1 (2.0)	−0.2 (1.8)	0.43
MCP1	8.7 (1.6)	8.6 (1.6)	8.3 (1.5)	0.26
TNFα	2.6 (2.0)	2.1 (2.4)	1.6 (2.1)	0.08
TGFβ1	14.6 (0.7)	14.7 (0.5)	14.8 (0.5)	0.22
IFNγ	−4.8 (2.5)	−4.2 (2.4)	−4.7 (2.2)	0.94
NGF	−6.2 (1.4)	−6.1 (1.3)	−6.4 (1.3)	0.49
Substance P	5.0 (0.6)	4.8 (0.6)	4.7 (0.5)	0.16

Notes: Data are presented as log2-scale mean(standard deviation).

^*p-value from analysis of variance

Qualitative cytokines to not associate with pain severity or frequency

Several subjects were missing values for measurements of IL-2, IL-4, IL-8 and BDNF. Therefore, we used qualitative analyses to compare these cytokines in each pain category. Patients with severe pain were more likely to have undetectable IL-2 compared to no or mild-moderate pain (Table 5). A similar analysis of each cytokine was performed to compare expression across the different temporal patterns. Patients with constant pain were more likely to have undetectable IL-2 versus no or intermittent pain (Table 6). None of the other qualitative markers were associated with pain severity or pain temporality with p-values of the Pearson chi-square tests varying from 0.25 to 0.96.

Table 5.

Distribution of qualitative markers by pain severity

Markers	Status	No pain	Mild/moderate pain	Severe pain	P-value*
IL- 2	Under LOD	11 (17)	22 (34)	31 (48)	0.09
	Above LOD	8 (23)	18 (51)	9 (26)
IL- 4	Under LOD	4 (17)	8 (33)	12 (50)	0.54
	Above LOD	15 (20)	32 (43)	28 (37)
IL-8	Under LOD	8 (18)	20 (44)	17 (38)	0.76
	Above LOD	11 (20)	20 (37)	23 (43)
BDNF	Under LOD	4 (19)	8 (38)	9 (43)	0.96
	Above LOD	15 (19)	32 (41)	31 (40)
CGRP	Under LOD	4 (20)	11 (55)	5 (25)	0.25
	Above LOD	15 (19)	29 (37)	35 (44)

Notes: Data are presented as log2-scale mean(standard deviation).

^*Pearson chi-square test

Abbreviations: LOD, level of detection.

Table 6.

Distribution of qualitative markers by pain temporality

Markers	Status	No pain	Intermittent pain	Constant pain	P-value*
IL- 2	Under LOD	11 (17)	26 (41)	27 (42)	0.09
	Above LOD	8 (23)	14 (40)	13 (37)
IL- 4	Under LOD	4 (17)	12 (50)	8 (33)	0.54
	Above LOD	15 (20)	28 (37)	32 (43)
IL-8	Under LOD	8 (18)	16 (36)	21 (47)	0.76
	Above LOD	11 (20)	24 (44)	19 (35)
BDNF	Under LOD	4 (19)	6 (29)	11 (52)	0.96
	Above LOD	15 (19)	34 (44)	29 (37)
CGRP	Under LOD	4 (20)	7 (35)	9 (45)	0.25
	Above LOD	15 (19)	33 (42)	31 (39)

Notes: Data are presented as log2-scale mean(standard deviation).

^*Pearson chi-square test

Abbreviations: LOD, level of detection.

Discussion

The majority of CP patients report pain in some form, however, whether specific biological signaling mechanisms can be tied to specific pain patterns remains unknown. In this study, putative pain pattern biomarkers were chosen based on two criteria. First, they have previously been implicated in the regulation of nociception and chronic pain. Second, they have been reported as altered in pancreas tissue/juice or serum from patients with CP. We show here that all of the chosen targets can be successfully and reliably measured in banked serum samples using our platform. Using exploratory cluster analyses, we identified potential patterns of expression that associated more strongly with patients reporting severe pain and distinct patterns that associated more with constant pain. The original prediction was that frequency or severity of pain would correlate with an upregulation of circulating proteins implicated in nociception and inflammation. However, our quantitative analyses verified that patients with severe pain or constant pain had lower levels of TNFα. In addition, these patients exhibited trends toward decreases in cytokines, chemokines and neuropeptides including IL-6, fractalkine, and substance P. While a robust biomarker profile based on pain pattern alone was not discernible from this small pilot study, however, the appearance of several trends suggests this type of profiling could be a key component in the identification of pain type specific signaling mechanisms.

Frequency and Severity are Independent Pain Features

Frequency and severity of pain are the two key features that gastroenterologists consider when thinking about CP pain. However, frequency and severity are independent of one another. In the NAPS2 cohort, only 3.7% of patients report constant severe pain while 4.6% report mild to moderate pain and 44.2% report mild to moderate constant pain with episodes of severe pain⁴². In the sub-cohort used for this study, we tried to maximize the chances of detecting differences by intentionally balancing groups: 50% of patients with constant pain reported it as severe and 50% reported it as mild. Interestingly, our exploratory hierarchical cluster analysis revealed a subset of patients with high levels of IL-2, IL-6, IL-1β, and MCP1. This cluster was predominantly comprised of patients with constant mild-moderate pain. Those with higher expression of CGRP, IL-4 and IL-8, but low expression of IL-2, IL-6, IL-1β, and MCP1 tended to suffer from constant pain. It remains unclear if another disease or life related factor is shared by the patients within each cluster. It is also not clear what distinguishes patients with the same pain pattern in cases where the cluster analysis assigned them to different clusters.

Decreased circulating cytokines and chemokines may associate with worse pain

In a recent study, Robinson and colleagues also used the Meso Scale Discovery platform to measure inflammatory cytokines in CP patients (including those tested in the present study). They found that elevated IL-1β, IL-8, and MCP1 were among the pro-inflammatory cytokines and chemokines associated with diminished quality of life.⁴⁰ Based on these data, we hypothesized that we would see elevated levels of a subgroup of targets that associate specifically with pain and perhaps distinct pain patterns. While we were able to successfully detect all of the putative targets, there was actually a decrease in several markers specifically associated with either continuous or severe pain. While unexpected, this is not unprecedented since a downregulation of serum cytokines has been associated with increased pain in other conditions. Decreased IFNγ, IL-1β, IL-2, and IL-4 predicts increased severity of cancer pain,⁴³ and MCP1 and IL-4 are decreased in patients with chronic low back pain.⁴⁴ Of particular relevance to this study, IL-6 and IL-1β are significantly downregulated in an animal model of CP.⁴⁵ One potential explanation for decreased circulating neuro-immune mediators in patients with severe pain is that the immune cells secreting these proteins have infiltrated and accumulated at the site of injury (i.e. the pancreas) as indicated by multiple assessments of patient specimens.^45–47 This would be supported by distinct biochemical changes in tissue and pancreas juice being inversely correlated with serum levels. A second possibility is that patients with constant or severe pain may no longer be experiencing pain arising from ongoing peripheral stimulation or neurogenic inflammation; it is feasible that there has been a shift toward central nervous system mechanisms. An analysis of the circulating immune cells, radiological (i.e., PET, MRI) and neurophysiology studies could help to validate CP pain biomarkers and further elucidate neuro-immune mechanisms underlying that pain.

Neuropeptides and CP Pain

CGRP and Substance P are neuropeptides implicated in the initiation of inflammatory responses, peripheral nerve sensitization, and pain. In both CP patients and animal models of pancreatitis, substance P and CGRP expression is elevated within the spinal sensory afferents that innervate the pancreas ^{18,20,29,31,48}. CGRP, released from both sensory neurons and endothelial cells drives vasodilation and neurogenic inflammation. Therefore, it is likely that CP pain in patients with elevated serum CGRP may have primarily inflammatory pain. It is also possible that there are systemic sequelae in these patients affecting other organs. Both injury and inflammation drive local increases in production and release of neuropeptides, however, when damage is severe enough that the nerve fibers are gone the local source of substance P is lost. Thus, the specific reduction in substance P could indicate advanced nerve damage and fiber loss. If this is the case, the prediction would be that this subgroup of patients would report more neuropathic pain symptoms. Unfortunately, we do not have this type of data from this cohort, but we know from other CP populations that patients with severe pain have increased amounts of neuritis and nerve damage.⁴⁹ It is also possible that the systemic reduction is reflective of the local increases such that substance P is accumulating at the injury site and being internalized following binding of neurokinin receptors.⁵⁰ Thus, the peptide is being used and broken down in the tissue at a greater rate and not entering into circulation.

Limitations

As this was an exploratory study, the sample size was small. Although we observed some qualitative trends, we had inadequate power to detect significant biologically and clinically relevant effects. Importantly, there are several limiting factors that could modulate levels of circulating nociceptive proteins and impact our results. The biological measurements were reliable and reproducible for the majority of analytes measurements with the exception of IL-2, IL-4 and IL-10. The reduced reliability for regulatory (IL-2) and anti-inflammatory (IL-4, IL-10) cytokines may be related to lower levels of these analytes in this patient population. This is not unexpected as patients with other types of chronic widespread pain have significantly reduced levels of serum IL-4 and IL-10.⁴⁴ Additionally, IL-2 has a short half-life and has anti-hyperalgesic effects.^51,52 Thus, IL-2 measurements could be thought of as a sort of negative control. In line with this perspective, the constant and severe pain groups had the most patients with undetectable IL-2. For subjects suffering from intermittent or episodic pain it is unclear whether their biological samples were collected during a period of pain or no pain. A few ‘no pain’ patients exhibited high expression of some of the nociceptive/inflammatory proteins despite reporting no CP related pain. This could be for a number of reasons including that they have a comorbid inflammatory or painful syndrome unrelated to their pancreatitis. Additionally, if pain is being adequately managed with current therapeutics, patients may report “no pain” despite there being ongoing nociceptive pathology. In the CP pain populations, there was an increased spread of the distribution for several of the targets tested. This suggests that there are definite subgroups within the CP pain population. These pain subgroups may have elevated expression of specific proteins and could benefit from targeted therapies. However, it is clear from the current results that pain pattern is NOT sufficient to identify biologically relevant signatures. Given that pain can arise from multiple etiologies and signaling mechanisms as well as present differently in each individual, the lack of additional clinical, disease-related and psychosocial information is a major limitation of this study. In other words, frequency and severity of pain are not sufficient as stand-alone features to stratify pain subtypes within the CP population.

Conclusion

The identification of pain biomarkers is of great importance. Not only could they reveal novel therapeutic targets but could enable the implementation of directed therapeutic management. The implementation of a personalized medicine approach to pain management could expedite achievement of pain relief by directly targeting specific underlying mechanisms. As new CP cohorts are assembled, it is important to incorporate more comprehensive and mechanistic phenotyping data. The availability of this additional information will improve our ability to interpret results of circulating protein levels. CP pain can involve both visceral and somatic pain as well as a variety of mechanisms including nerve injury, tissue injury, and inflammatory signaling. Whether sensitization of the nervous system is present and how widespread (peripheral versus central) could also impact response to therapy. The addition of sensory testing and assessment of pain experience could point to whether the underlying mechanisms are primarily nociceptive or neuropathic in nature. Further studies combining biological measures and more thorough patient data could lead to the development of a model that predicts response to different analgesic interventions.