{
  "abstract": "Introduction The prediction of clinical outcomes in spinal cord stimulation (SCS) for chronic pain is crucial as it enables the optimization of patient selection, enhances therapeutic efficacy, reduces unnecessary interventions, and minimizes costs and risks associated with invasive procedures. Given that 30–40% of patients do not achieve sustained relief after neurostimulator implantation, identifying potential responders beforehand is key to avoiding explantations, patient dissatisfaction, and healthcare system burdens, as emphasized by the American Academy of Pain Medicine in its consensus guidelines. 1 The integration of patient profiles and biomarkers, including gene expression and clinical characteristics, has shown potential to improve response prediction. Gene expression plays an emerging role in personalized medicine for chronic pain, particularly in predicting clinical outcomes and optimizing patient selection for SCS. Gene expression analysis can identify underlying biological processes such as inflammation, neuroimmune modulation, and tissue repair mechanisms, which may influence SCS response.2–4 However, current evidence indicates that while altered genes and biological pathways have been identified in chronic pain patients and SCS responders, no specific gene signatures reliably predict individual clinical responses to SCS to date. Changes in genes like proenkephalin (PENK) and inflammatory cytokines correlate with clinical progression, but their utility as standalone predictive biomarkers remains limited and requires further validation.2 3 The use of gene panels and next-generation sequencing has advanced understanding of pain’s genetic architecture and opened avenues for patient stratification, though routine clinical application is not yet established.5 6Patient clinical profiles, such as age and comorbidities, are also relevant for outcome prediction and personalized SCS candidacy. Older age is associated with higher explantation risk and lower long-term success rates, while obesity increases complication risks and reduces therapy efficacy. Untreated psychiatric comorbidities, particularly depression, are linked to poorer outcomes and are considered relative contraindications.1 Clinical profiles should assess sex, body mass index, pain duration, and neuropathic pain presence, as these influence treatment response. For example, neuropathic pain and female sex correlate with higher response likelihood, whereas advanced age and depression reduce functional improvement odds7 8 While the integration of gene expression and personalized medicine is promising, validated genetic biomarkers for reliable SCS response prediction remain elusive. However, combining clinical, neurophysiological, and genetic data via machine learning models shows potential to enhance prediction and candidate selection, though routine clinical use is not yet established.9–11 Comprehensive clinical profiling and personalized medicine tools are essential to optimize patient selection and maximize SCS benefits for chronic pain.7 8 The ability to predict clinical outcomes in spinal cord stimulation is important to maximize clinical benefit, reduce risks and costs, and move towards precision medicine in the treatment of chronic pain.9 Principles of gene expression and their measurement Fundamental gene expression principles involve DNA transcription to messenger RNA (mRNA) and subsequent translation into functional proteins, which determine cellular phenotype and pathophysiological responses, including chronic pain perception and modulation. In personalized medicine and SCS outcome prediction, precise gene expression quantification helps identify biomarkers, molecular pathways, and response profiles to guide candidate selection and therapeutic optimization.12 13 The main techniques for measuring gene expression include: • Gene expression microarrays: Analyze thousands of known genes simultaneously via labeled mRNA hybridization to specific probes. Useful for cross-group expression comparisons but limited to annotated genes and less sensitive for low-abundance genes.12 14 • RNA Sequencing (RNA-Seq): Uses next-generation sequencing to globally quantify all sample transcripts, including unannotated genes, splice variants, and non-coding RNA. RNA-Seq offers higher sensitivity, dynamic range, and discovery potential than microarrays, making it the preferred method for transcriptomic studies in chronic pain and SCS.14 15 • Proteomic analysis: Directly assesses protein levels and modifications, the functional end-products of gene expression. Though complementary, proteomics is more complex and less routine for clinical prediction but can elucidate action mechanisms and therapeutic targets.16 Gene Expression Applications in Personalized Medicine Gene expression profiles are used in other diseases for prediction and personalized medicine, particularly in oncology. Clinically validated examples include: • Breast cancer: The Oncotype DX test analyzes 21-gene expression in hormone receptor-positive, HER2-negative tumors, stratifying patients into low-, intermediate-, and high-recurrence risk groups. This personalizes chemotherapy decisions, sparing low-risk patients unnecessary toxicity while ensuring high-risk patients receive beneficial therapies. Oncotype DX and similar panels (MammaPrint, PAM50, EndoPredict) are supported by multiple studies and recommended by the Evaluation of Genomic Applications in Practice and Prevention Working Group for early breast cancer.17 18 • Colon cancer: The ColoPrint test (18-gene analysis) stratifies recurrence risk in stage II colon cancer, guiding adjuvant chemotherapy decisions.19 • Acute myeloid leukemia (AML): Gene expression profiles subclassify AML and predict therapy response, though no universally adopted commercial panel exists.20 • Lung cancer: Panels like VeriStrat predict EGFR inhibitor response in advanced non-small cell lung cancer, though use is less widespread than in breast cancer.21 Integrating Clinical and Omics Data Integrating clinical and omics data (genomics, transcriptomics, proteomics, etc.) in predictive models is essential for personalized medicine, as it captures biological complexity and interindividual heterogeneity. Clinical data provide consolidated phenotypic and prognostic information, while omics data reveal underlying molecular mechanisms, functional alterations, and therapeutic targets not detectable with clinical variables alone.22 23 This integration improves clinical outcome prediction accuracy, risk stratification, disease subclassification, and personalized treatment selection, outperforming models based solely on clinical or single-omics data. For example, in breast cancer, combining clinical and multi-omics profiles enhances survival and therapy response prediction.24 25 Multi-omics approaches also identify composite biomarkers and relevant biological pathways, enabling precision medicine where clinical data are insufficient or ambiguous.22 26 Limitations of Clinical-Omics Integration Common limitations include: • Sample heterogeneity: Variability in sample quality, origin, and processing introduces biases and reduces model reproducibility. Differences in data collection, technologies, and studied populations hinder generalizability.27–29 • Overfitting: High omics variable counts versus limited patient samples (‘curse of dimensionality’) can cause models to overfit training data, impairing external cohort performance. Robust validation and variable selection strategies are critical.28 29 • Lack of external validation: Many integrative models lack validation in independent cohorts, limiting their clinical applicability and robustness. The absence of external validation is one of the main reasons why few models are ever implemented in clinical practice.30 Other challenges include missing data, class imbalances (e.g., few events vs. many controls), data harmonization difficulties, and incomplete clinical integration, which may yield less parsimonious models27 31 32 Addressing these requires rigorous study design, quality control, appropriate statistical/machine learning methods, and multicenter validation.28 30 Genomic Expression as a Predictor in Neuromodulation Examples of gene expression profiles for neuromodulation outcome prediction focus on identifying biomarkers and therapeutic targets to guide patient selection and therapy personalization. Transcriptomic studies of dorsal root ganglia and spinal cord have identified gene signatures linked to neuropathic pain and SCS response. For instance, differential expression of genes related to immune signaling, neuronal plasticity, and endogenous opioid modulation has been associated with chronic pain development and maintenance, suggesting predictive potential for neuromodulation33–35 Validated blood-based gene combinations (e.g., SH3BGRL3, TMEM88, CASP9) distinguish neuropathic pain patients from controls and could stratify candidates for advanced therapies like neuromodulation.36 37 Gene expression analysis has also identified potential pharmacological targets, such as the Syk tyrosine kinase pathway, preclinically validated as a pain modulator post-neuromodulation.33 Though not yet standard clinical practice, integrating gene expression with clinical and neurophysiological data represents a promising path for personalized neuromodulation outcome prediction.36 37 In animal neuropathic pain models, SCS modulates inflammation-, glial activation-, and neurotransmission-related genes. Differential programming (e.g., DTMP) reverses microglial and neuronal gene expression toward healthy profiles, correlating with behavioral pain improvement. SCS waveform intensity and type differentially modulate key pain genes (e.g., cFos, GABAbr1, 5HT3ra, TNFα), with gene modulation-clinical improvement correlations observed.38–42 Our research group conducted a genomic association study in humans with persistent spinal pain syndrome type 2 (PSPS 2) to explore SCS response predictors. High-density microarrays analyzed serum gene expression in PSPS 2 patients versus pain-free post-lumbar surgery controls. Multivariate discriminant analysis and gene ontology enrichment identified relevant biological processes. The study found 11 genes significantly downregulated in PSPS 2 patients, linked to heightened inflammation, tissue repair, and proliferative responses. Post-SCS, two additional downregulated genes suggested therapy-modulated inflammatory/restorative processes. However, no genes discriminated SCS responders from non-responders, indicating that while molecular alterations are associated with pain and therapeutic response, current serum gene expression cannot reliably predict individual outcomes.2 Our group also investigated SCS-induced serum proteomic changes in PSPS 2 patients. Mass spectrometry compared protein profiles pre-/post-SCS and between responders/non-responders and controls. Multivariate discriminant and bioinformatic analyses identified relevant biological processes. Results showed PSPS 2 alterations in inflammation-, immunity-, iron metabolism-, and synaptic signaling-related proteins. SCS modulated these processes, reducing inflammatory proteins and increasing neuronal repair markers. Though molecular differences correlated with clinical response, serum protein expression could not reliably predict SCS response, highlighting biomarker potential and limitations for chronic pain personalization.43 Clinical profiles (demographics, comorbidities) showed no significant responder/non-responder differences. Overall, integrating gene expression and clinical data does not yet enable robust SCS response prediction in PSPS 2 but elucidates underlying mechanisms, particularly neuroimmune modulation and tissue repair.2 Clinical application requires further validation.44 Current Challenges and Future Directions Challenges in using gene expression and clinical profiles for neuromodulation outcome prediction include sample heterogeneity, model overfitting, and lack of external validation. Variability in sample processing and phenotypic diversity limit reproducibility. Overfitting arises from high omics variable counts versus small sample sizes, often yielding non-replicable models. Most studies lack robust external validation, restricting clinical applicability.10 11 45–47 Currently, no validated genomic biomarkers reliably predict individual SCS responses, though genes/pathways linked to inflammation, neuronal plasticity, and immune modulation may hold future predictive value.48 49 Integrative models combining clinical, neurophysiological (e.g., intraoperative EEG), and self-report data show promising accuracy but require multicenter validation and standardization before clinical adoption10 11 45 47 Future integration of clinical and omics data via AI/machine learning, composite biomarker development, and large prospective registries will advance personalized neuromodulation, optimizing patient selection and outcomes. However, method standardization, external validation, and multicenter collaboration are essential to overcome current limitations and translate advances into clinical practice10 11 45 47 References Shanthanna H, Eldabe S, Provenzano DA, Bouche B, Buchser E, Chadwick R, et al. Evidence-based consensus guidelines on patient selection and trial stimulation for spinal cord stimulation therapy for chronic non-cancer pain. Reg Anesth Pain Med. 2023;48(6):273–87.Fabregat-Cid G, Cedeno DL, De Andrés J, Harutyunyan A, Monsalve-Dolz V, Mínguez-Martí A, et al. Insights into the pathophysiology and response of persistent spinal pain syndrome type 2 to spinal cord stimulation: a human genome-wide association study. Reg Anesth Pain Med. 2024.De Andrés J, Navarrete-Rueda F, Fabregat G, García-Gutiérrez MS, Monsalve-Dolz V, Harutyunyan A, et al. Differences in gene expression of endogenous opioid peptide precursor, cannabinoid 1 and 2 receptors and interleukin beta in peripheral blood mononuclear cells of patients with refractory failed back surgery syndrome treated with spinal cord stimulation: markers of therapeutic outcomes? Neuromodulation 2021;24(1):49–60.Johnston KJA, Cote AC, Hicks E, Johnson J, Huckins LM. Genetically regulated gene expression in the brain associated with chronic pain: relationships with clinical traits and potential for drug repurposing. Biol Psychiatry. 2024;95(8):745–61.Kringel D, Lötsch J. Knowledge of the genetics of human pain gained over the last decade from next-generation sequencing. Pharmacol Res. 2025;214:107667.Kringel D, Malkusch S, Kalso E, Lötsch J. Computational functional genomics-based ampliseq™ panel for next-generation sequencing of key genes of pain. Int J Mol Sci. 2021;22(2).Bastiaens F, van de Wijgert IH, Bronkhorst EM, van Roosendaal BWP, van Heteren EPZ, Gilligan C, et al. Factors predicting clinically relevant pain relief after spinal cord stimulation for patients with chronic low back and/or leg pain: a systematic review with meta-analysis and meta-regression. Neuromodulation 2024;27(1):70–82.Kapural L, Wu C, Calodney A, Pilitsis J, Bendel M, Petersen E, et al. Demographics and PainDETECT as predictors of 24-month outcomes for 10 kHz SCS in nonsurgical refractory back pain. Pain Physician 2024;27(3):129–39.De Andres J. Neurostimulation in the patient with chronic pain: forecasting the future with data from the present - data-driven analysis or just dreams? Reg Anesth Pain Med. 2024;49(3):155–62.Gopal J, Bao J, Harland T, Pilitsis JG, Paniccioli S, Grey R, et al. Machine learning predicts spinal cord stimulation surgery outcomes and reveals novel neural markers for chronic pain. Sci Rep. 2025;15(1):9279.Hadanny A, Harland T, Khazen O, DiMarzio M, Marchese A, Telkes I, et al. Development of machine learning-based models to predict treatment response to spinal cord stimulation. Neurosurgery 2022;90(5):523–32.Segundo-Val IS, Sanz-Lozano CS. Introduction to the gene expression analysis. Methods Mol Biol. 2016;1434:29–43.Wang D, Kim H, Wang XM, Dionne R. Genomic methods for clinical and translational pain research. Methods Mol Biol. 2012;851:9–46.Perkins JR, Antunes-Martins A, Calvo M, Grist J, Rust W, Schmid R, et al. A comparison of RNA-seq and exon arrays for whole genome transcription profiling of the L5 spinal nerve transection model of neuropathic pain in the rat. Mol Pain. 2014;10:7.Wang D, Karamyshev AL. Next generation sequencing (NGS) application in multiparameter gene expression analysis. Methods Mol Biol. 2020;2102:17–34.Zhang J, Qi L, Sun Y, Chen S, Liu J, Chen J, et al. Integrated bioinformatics analysis of the effects of chronic pain on patients with spinal cord injury. Front Cell Neurosci. 2025;19:1457740.Recommendations from the EGAPP Working Group: does the use of Oncotype DX tumor gene expression profiling to guide treatment decisions improve outcomes in patients with breast cancer? Genet Med. 2016;18(8):770–9.Berton Giachetti PPM, Carnevale Schianca A, Trapani D, Marra A, Toss A, Marchiò C, et al. Current controversies in the use of Oncotype DX in early breast cancer. Cancer Treat Rev. 2025;135:102887.Maak M, Simon I, Nitsche U, Roepman P, Snel M, Glas AM, et al. Independent validation of a prognostic genomic signature (ColoPrint) for patients with stage II colon cancer. Ann Surg. 2013;257(6):1053–8.Bullinger L, Döhner K, Bair E, Fröhling S, Schlenk RF, Tibshirani R, et al. Use of gene-expression profiling to identify prognostic subclasses in adult acute myeloid leukemia. N Engl J Med. 2004;350(16):1605–16.Molina-Pinelo S, Pastor MD, Paz-Ares L. VeriStrat: a prognostic and/or predictive biomarker for advanced lung cancer patients? Expert Rev Respir Med. 2014;8(1):1–4.Olivier M, Asmis R, Hawkins GA, Howard TD, Cox LA. The need for multi-omics biomarker signatures in precision medicine. Int J Mol Sci. 2019;20(19).Karczewski KJ, Snyder MP. Integrative omics for health and disease. Nat Rev Genet. 2018;19(5):299–310.Chakraborty S, Sharma G, Karmakar S, Banerjee S. Multi-OMICS approaches in cancer biology: new era in cancer therapy. Biochim Biophys Acta Mol Basis Dis. 2024;1870(5):167120.Vazquez AI, Veturi Y, Behring M, Shrestha S, Kirst M, Resende MF, et al. Increased proportion of variance explained and prediction accuracy of survival of breast cancer patients with use of whole-genome multiomic profiles. Genetics 2016;203(3):1425–38.Ahmed Z. Practicing precision medicine with intelligently integrative clinical and multi-omics data analysis. Hum Genomics. 2020;14(1):35.López de Maturana E, Alonso L, Alarcón P, Martín-Antoniano IA, Pineda S, Piorno L, et al. Challenges in the integration of omics and non-omics data. Genes (Basel) 2019;10(3).Khan SR, Manialawy Y, Wheeler MB, Cox BJ. Unbiased data analytic strategies to improve biomarker discovery in precision medicine. Drug Discov Today. 2019;24(9):1735–48.Mayer G, Heinze G, Mischak H, Hellemons ME, Heerspink HJ, Bakker SJ, et al. Omics-bioinformatics in the context of clinical data. Methods Mol Biol. 2011;719:479–97.Glaab E, Rauschenberger A, Banzi R, Gerardi C, Garcia P, Demotes J. Biomarker discovery studies for patient stratification using machine learning analysis of omics data: a scoping review. BMJ Open. 2021;11(12):e053674.Volkmann A, De Bin R, Sauerbrei W, Boulesteix AL. A plea for taking all available clinical information into account when assessing the predictive value of omics data. BMC Med Res Methodol. 2019;19(1):162.Zhao J, Feng Q, Wei WQ. Integration of omics and phenotypic data for precision medicine. Methods Mol Biol. 2022;2486:19–35.Ghazisaeidi S, Muley MM, Tu Y, Finn DP, Kolahdouzan M, Pitcher GM, et al. Conserved transcriptional programming across sex and species after peripheral nerve injury predicts treatments for neuropathic pain. Br J Pharmacol. 2023;180(21):2822–36. North RY, Li Y, Ray P, Rhines LD, Tatsui CE, Rao G, et al. Electrophysiological and transcriptomic correlates of neuropathic pain in human dorsal root ganglion neurons. Brain 2019;142(5):1215–26.Pokhilko A, Nash A, Cader MZ. Common transcriptional signatures of neuropathic pain. Pain 2020;161(7):1542–54.Niculescu AB, Le-Niculescu H, Levey DF, Roseberry K, Soe KC, Rogers J, et al. Towards precision medicine for pain: diagnostic biomarkers and repurposed drugs. Mol Psychiatry. 2019;24(4):501–22.Young B, Stephenson J, Islam B, Burke NN, Jennings EM, Finn DP, et al. The identification of human translational biomarkers of neuropathic pain and cross-species validation using an animal model. Mol Neurobiol. 2023;60(3):1179–94.Cedeño DL, Kelley CA, Vallejo R. Effect of stimulation intensity of a differential target multiplexed SCS program in an animal model of neuropathic pain. Pain Pract. 2023;23(6):639–46.Smith WJ, Cedeño DL, Thomas SM, Kelley CA, Vetri F, Vallejo R. Modulation of microglial activation states by spinal cord stimulation in an animal model of neuropathic pain: comparing high rate, low rate, and differential target multiplexed programming. Mol Pain. 2021;17:1744806921999013.Tilley DM, Cedeño DL, Kelley CA, Benyamin R, Vallejo R. Spinal cord stimulation modulates gene expression in the spinal cord of an animal model of peripheral nerve injury. Reg Anesth Pain Med. 2016;41(6):750–6.Vallejo R, Gupta A, Kelley CA, Vallejo A, Rink J, Williams JM, et al. Effects of phase polarity and charge balance spinal cord stimulation on behavior and gene expression in a rat model of neuropathic pain. Neuromodulation 2020;23(1):26–35.Vallejo R, Tilley DM, Cedeño DL, Kelley CA, DeMaegd M, Benyamin R. Genomics of the effect of spinal cord stimulation on an animal model of neuropathic pain. Neuromodulation 2016;19(6):576–86.Fabregat-Cid G, Cedeño DL, Harutyunyan A, Rodríguez-López R, Monsalve-Dolz V, Mínguez-Martí A, et al. Effect of conventional spinal cord stimulation on serum protein profile in patients with persistent spinal pain syndrome: a case-control study. Neuromodulation 2023;26(7):1441–9.de Geus TJ, Franken G, Joosten EAJ. Spinal cord stimulation paradigms and pain relief: a preclinical systematic review on modulation of the central inflammatory response in neuropathic pain. Neuromodulation 2023;26(1):25–34.Harland T, Elliott T, Telkes I, Pilitsis JG. Machine learning in pain neuromodulation. Adv Exp Med Biol. 2024;1462:499–512. Kansal A, Copley S, Duarte RV, Warren FC, Taylor RS, Eldabe S. Systematic review to identify patient-level predictors of treatment response to spinal cord stimulation for neuropathic pain for studies published from 2012 to 2024. Neuromodulation 2025.Mackey S, Aghaeepour N, Gaudilliere B, Kao MC, Kaptan M, Lannon E, et al. Innovations in acute and chronic pain biomarkers: enhancing diagnosis and personalized therapy. Reg Anesth Pain Med. 2025;50(2):110–20.Dib-Hajj SD, Waxman SG. Translational pain research: lessons from genetics and genomics. Sci Transl Med. 2014;6(249):249sr4.Nagel SJ, Hsieh J, Machado AG, Frizon LA, Howard MA, 3rd Gillies GT, et al. Biomarker optimization of spinal cord stimulation therapies. Neuromodulation 2021;24(1):3–12.",
  "authors": [
    {
      "affiliations": [
        "Pain Management Department, Consorcio Hospital General Universitario, Valencia, Spain"
      ],
      "name": "Gustavo Fabregat"
    }
  ],
  "title": "FT21 Validity of genomic expression and patient profiles to predict outcomes",
  "uid": "4a8cdd6e-f2a8-53d6-81f7-1f59fbcf6c0c"
}