Preprocedural superomedial pain localization as a predictor for short-term clinical improvement after cooled radiofrequency ablation in advanced knee osteoarthritis

Abstract

Background

Cooled radiofrequency ablation (CRFA) has shown clinical benefits in patients with advanced knee osteoarthritis (OA); however, predictors of treatment response remain limited. Preprocedural pain localization may offer a simple and clinically relevant predictor for enhanced patient selection for CRFA.

Purpose

To evaluate whether preprocedural superomedial pain localization and overall pain site count are associated with clinical outcomes after CRFA in patients with advanced knee OA.

Methods

This prospective cohort study included 47 patients (54 knees) with Kellgren–Lawrence grade 3 or 4 OA who underwent ultrasound-guided genicular nerve block and standardized CRFA. Pain localization was assessed using an 8-zone knee pain map. The primary outcome was the 3-month change in International Knee Documentation Committee (IKDC) score. Secondary outcomes included Lysholm score, visual analog scale (VAS), and Knee Injury and Osteoarthritis Outcome Score (KOOS) domains. Univariate and multivariate regression analyses were performed.

Results

Superomedial pain, the most common localization (66.7%), was significantly associated with worse baseline IKDC (β = − 8.2, p = 0.02) and KOOS–Pain (β = − 13.9, p = 0.01). However, it was independently associated with greater IKDC improvement after CRFA (β = 9.1, p = 0.02) after adjusting for age, sex, and BMI. Total number of pain sites correlated with worse baseline function but not with treatment response.

Conclusions

Superomedial knee pain was independently associated with superior short-term functional improvement following CRFA. These findings support the utility of pain localization as a prognostic tool for patient selection in advanced knee OA.

Background

Knee osteoarthritis (OA) is a leading cause of disability in aging populations [1], with chronic pain and progressive functional decline substantially impairing mobility and quality of life (QOL) [2]. For patients with advanced OA who are not surgical candidates or do not wish to undergo total knee arthroplasty (TKA), interventional treatments such as genicular nerve radiofrequency ablation (RFA) have emerged as viable alternatives. Among these, cooled RFA (CRFA) offers larger lesion volumes and potentially greater durability than the conventional RFA [2, 3].

Randomized trials and cohort studies have previously demonstrated that CRFA significantly reduces pain and improves function in patients with moderate-to-severe knee OA [4–7], with improvements documented using validated measures, such as the International Knee Documentation Committee (IKDC) [8] and Knee Injury and Osteoarthritis Outcome Score (KOOS) [9]. However, patient responses vary and predicting which patients will benefit the most from CRFA remains a challenge. While diagnostic genicular nerve blocks are widely used to identify candidates for RFA [7, 10, 11], they require procedural expertise and may not fully capture the spatial complexity of OA-related knee pain.

Preprocedural pain localization using a simple, patient-reported body map may be an accessible and informative predictor of CRFA outcomes. Genicular nerves innervate specific sub-regions of the knee, and focal pain in a particular zone may reflect nerve-specific nociceptive inputs amenable to ablation [12–15]. The superomedial region is particularly noteworthy, as it is innervated by the superior medial genicular nerve, which is a common CRFA target [12]. Although pain mapping is routinely used in other musculoskeletal interventions, its predictive value for CRFA outcomes remains understudied [16]. Furthermore, the relationship between the total burden of knee pain (i.e., the number of painful sites) and treatment outcomes remains unclear. It is unclear whether widespread pain reflects central sensitization, which may diminish the response to localized nerve ablation.

This prospective cohort study aimed to evaluate whether pre-procedural knee pain patterns, specifically superomedial pain and the total number of painful sites, are associated with baseline functional status and clinical improvement after CRFA in patients with advanced knee OA. We hypothesized that patients with superomedial pain would have lower baseline function but exhibit greater functional improvement following treatment. We also explored whether the number of painful sites, as a marker of pain diffusion, could predict treatment response to CRFA.

Methods

Study design and setting

This prospective cohort study was conducted at a community‑based orthopedic clinic in Japan. We consecutively enrolled patients with symptomatic knee osteoarthritis who underwent CRFA between June 2023 and October 2024, in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [17]. All procedures (ultrasound‑guided genicular nerve block, CRFA, and subsequent data collection and analysis) were completed by February 2025. All interventions and assessments were performed in full compliance with the Declaration of Helsinki and all applicable institutional, national, and international guidelines and regulations. Written informed consent was obtained from each participant prior to enrolment. Because this research was purely observational and did not involve the prospective assignment of interventions by the investigators, it did not meet the WHO/ICH‑GCP definition of a clinical trial; therefore, no trial registration was required (clinical trial number: not applicable). This study received no specific grant from any public, commercial, or not‑for‑profit funding body.

Human ethics and consent to participate declarations

This study was approved by the Hiroshima University Epidemiology Research Ethics Review Committee (approval number E2023-0254). Written informed consent to participate in the study was obtained from all participants prior to enrollment. All participants provided written informed consent for the publication of aggregated study results; no individual participant is identifiable in this manuscript.

Participants and eligibility criteria

Patients were eligible if they were diagnosed with advanced knee OA (Kellgren–Lawrence grade 3 or 4) and had chronic knee pain lasting ≥ 6 months that was refractory to conservative care, including oral medications, physical therapy, and intra-articular injections. Patients were excluded if they had previously undergone genicular nerve ablation, had inflammatory or neuropathic conditions, or were unable to complete the follow-up due to relocation, occupational limitations, or withdrawal of consent. Patients scheduled to undergo TKA were excluded (Fig. 1). Demographic and clinical data collected included age, sex, height, weight, body mass index (BMI), smoking status, and presence of comorbidities such as type 2 diabetes mellitus (oral or insulin-treated), thyroid disease, collagen vascular disease, complex regional pain syndrome, autoimmune disorders, and psychiatric conditions.

Fig. 1.

Patient Flow Chart

Diagnostic nerve block and CRFA

All patients received a diagnostic ultrasound-guided genicular nerve block targeting the superomedial, inferomedial, and superolateral genicular nerves prior to CRFA. Each site received 1 mL of 1% mepivacaine [18], performed using a SONIMAGE HS1 ultrasound machine (KONICA MINOLTA, Japan) equipped with an L11-3 linear transducer. Patients were eligible for CRFA if they experienced ≥ 50% pain relief for any duration following the block. All subsequent CRFA procedures targeted the same genicular nerve branches using a standardized protocol with bipolar CRFA probes (Coolief™, Avanos) [18]. Electrical stimulation testing confirmed accurate needle positioning before lesioning at 60 °C for 150 s/site. All procedures were performed by two senior orthopedic surgeons (T.K. and N.H.), each with > 10 years of clinical experience.

Pain localization assessment

Prior to CRFA, patients completed a standardized pain mapping form to indicate the presence (1 = yes, 0 = no) or absence of pain in eight anatomic regions: superomedial, inferomedial, superolateral, inferolateral, suprapatellar, infrapatellar, medial popliteal, and lateral popliteal (Fig. 2). This simplified layout was designed to accommodate older patients, who may not accurately identify specific joint lines. Instead, the mapping relies on intuitive surface zones along the femoral and tibial contours. While prior studies on pain mapping have focused predominantly on the patella [16], our eight-zone format provided broader and more reproducible coverage of the knee. The total pain site count (range, 0–8) was calculated as the sum of all positive regions.

Fig. 2.

A standardized pain mapping form

Outcome measures

The primary outcome was the change in the IKDC score from baseline to 3 months (ΔIKDC). The IKDC score is a knee-specific patient-reported outcome measure (PROM) assessing symptoms, daily living functions, and sports activity. It has demonstrated high validity, test–retest reliability, and responsiveness across various knee pathologies, including OA and ligament injuries [19, 20]. The secondary outcomes were changes in the Lysholm score, visual analog scale (VAS) score for pain, and KOOS subdomains: Symptoms, Pain, Activities of Daily Living, Sports, and QOL. The Lysholm score assesses knee function, particularly in individuals with ligament injuries or degenerative joint disease [21, 22]. It evaluates domains, such as limping, support, locking, instability, pain, swelling, stair climbing, and squatting. The VAS is a 100-mm unidimensional scale measuring pain intensity, ranging from “no pain” to “worst imaginable pain [23].” The KOOS is a validated, knee-specific PROM designed to assess both the short- and long-term consequences of knee injury and OA across five functional domains [24, 25]. All PROMs were self-administered under staff supervision.

Statistical analysis

Continuous variables, including all baseline and outcome scores, were summarized as medians with interquartile ranges (IQRs) and rounded to one decimal place. Categorical variables such as sex, pain site involvement, categorized pain site count (1, 2, 3, or ≥ 4), smoking status, and comorbidities were reported as absolute frequencies and percentages.

We first examined the association between each of the eight pain locations and baseline functional scores using univariate linear regression models. Independent regression models were constructed for each site against baseline IKDC, Lysholm, VAS, and KOOS subscores. To assess the influence of the overall pain burden, the total number of pain sites was treated as a continuous predictor in univariable linear regression models for baseline functional scores.

Next, we explored whether pain localization patterns or the total pain site count predicted clinical improvement by assessing associations with score changes from baseline to 3 months (ΔIKDC, ΔLysholm, and ΔVAS). Predictors significantly associated with outcome improvements were entered into multivariate linear regression models adjusting for age, sex, and BMI to account for potential confounders. There were no missing patient-reported outcome measure (PROM) data at baseline or at the 3-month follow-up. All analyses were therefore performed using complete-case data without the need for imputation. All statistical tests were two-sided, and statistical significance was set at a p < 0.05. Analyses were performed using GraphPad Prism (version 10.0; GraphPad Software, San Diego, CA, USA).

Results

A total of 54 knees from 47 patients were analyzed. The median age was 75.0 years (IQR, 69.8–82.3), and 77.8% were female. The median BMI was 25.5 kg/m² (IQR, 22.6–27.8). The most common comorbidity was diabetes (24.1%), followed by thyroid disease (11.1%) and psychiatric disorders (7.4%; Table 1). The number of reported pain sites ranged from 0 to > 4, with a median of 2 (IQR, 1–3). Most patients reported pain in one (33.3%) or two (38.9%) regions, while 18.5% reported pain in > 4 sites.

Table 1.

Demographic and baseline clinical characteristics of study participants

Demographic Data of Patients
Variables	Median (IQR) or n (%)
Age (years)	75.0 (69.8–82.3)
Sex (female)	42/54 (77.8%)
Height (m)	1.55 (1.49–1.58)
Weight (kg)	61.3 (53.3–70.9)
BMI (kg/m²)	25.5 (22.6–27.8)
Current Smoking	3/54(5.6%)
Diabetes (oral)	12/54 (22.2%)
Diabetes (insulin)	1/54 (1.9%)
Thyroid disease	6/54 (11.1%)
Collagen disease	1/54 (1.9%)
CRPS	0%
Autoimmune disease	0%
Psychiatric disorder	4/54 (7.4%)
Superomedial pain	28/54 (51.9%)
Inferomedial pain	41/54 (75.9%)
Superolateral pain	11/54 (20.4%)
Inferolateral pain	7/54 (13.0%)
Suprapatellar pain	7/54 (13.0%)
Infrapatellar pain	10/54 (18.5%)
Medial popliteal pain	5/54 (9.3%)
Lateral popliteal pain	2/54 (3.7%)

Values are presented as median (interquartile range) for continuous variables and number (%) for categorical variables. IQR interquartile range, BMI body mass index, CRPS complex regional pain syndrome

At baseline, the median IKDC, Lysholm, and VAS scores were 31.0 (IQR, 22.7–40.2), 57.0 (IQR, 41.8–72.3), and 65.6 mm (IQR, 47.1–76.0), respectively. The KOOS domain scores indicated impaired QOL and function, with KOOS–QOL at 31.3 (IQR, 23.4–51.5) and KOOS–Sports at 15.0 (IQR, 0–30.0). At 3-month follow-up, all outcome measures improved numerically, including the IKDC (40.2 [31.1–54.0]), Lysholm (70.5 [53.0–82.5]), and VAS (38.5 [18.8–59.4]) scores.

Univariable regression analysis revealed that superomedial pain was significantly associated with worse baseline IKDC (slope=–8.2, p = 0.02) and KOOS–Pain (slope=–13.9, p = 0.01) scores. Superomedial pain was also correlated with lower Lysholm (slope=–11.3, p = 0.04) and KOOS–Sports (slope=–13.9, p = 0.04) scores. These findings were consistent across multiple validated functional domains. Other pain sites did not show any statistically significant associations (Table 2).

Table 2.

Univariable regression analysis of preprocedural pain localization and baseline functional scores

Univariate Analysis
Pain Location	Outcome	Slope	P value	R squared
Superomedial	IKDC (pre)	−8.2	0.02	0.11
	Lysholm (pre)	−11.3	0.04	0.08
	VAS (pre)	0.6	0.96	0.00
	KOOS Symptom (pre)	−3.9	0.41	0.01
	KOOS Pain (pre)	−13.9	0.01	0.12
	KOOS ADL (pre)	−8.6	0.10	0.05
	KOOS Sports (pre)	−13.9	0.04	0.08
	KOOS QOL (pre)	−9.4	0.10	0.05

Linear regression was performed for each pain location and baseline outcome score. IKDC, International Knee Documentation Committee; VASvisual analog scale, KOOSKnee injury and Osteoarthritis Outcome Score, ADLactivities of daily living, QOLquality of life

When treating pain site count as a continuous variable, a higher number of pain sites was significantly associated with lower baseline IKDC (β=–3.3, 95% confidence interval [CI]: − 6.3 to − 0.4, p = 0.03) and KOOS–Pain scores (β=–5.9, 95% CI: − 10.7 to − 1.2, p = 0.02). Similar trends were observed for other functional scores, but these did not reach statistical significance (Table 3).

Table 3.

Univariable regression analysis of pain site count and baseline functional scores

Univariate Analysis
Outcome	Slope	95% CI	P value
IKDC (pre)	−3.3	–6.3 to − 0.4	0.03
Lysholm (pre)	−3.0	–7.9 to 1.9	0.23
VAS (pre)	3.9	–6.3 to 14.1	0.44
KOOS Symptom (pre)	−0.4	–4.5 to 3.8	0.86
KOOS Pain (pre)	−5.9	–10.7 to − 1.2	0.02
KOOS ADL (pre)	−2.6	–7.1 to 2.0	0.26
KOOS Sports (pre)	−4.6	–10.6 to 1.4	0.13
KOOS QOL (pre)	−3.0	–8.0 to 2.1	0.24

Linear regression models were used to evaluate associations between the number of painful sites and each baseline outcome measure

Abbreviations CIconfidence interval, IKDCInternational Knee Documentation Committee, VASvisual analog scale, KOOSKnee injury and Osteoarthritis Outcome Score

We further analyzed whether superomedial pain or total pain site count predicted functional improvement after 3 months. Superomedial pain was significantly associated with greater improvements in IKDC (β = 9.4, p = 0.01), KOOS–QOL (β = 21.2, p = 0.001), KOOS–Sports, and KOOS–Pain domains (Table 4). In contrast, total pain site count was not significantly associated with changes in any outcome measure (Table 5).

Table 4.

Association of superomedial pain with change in functional scores at 3 months

Association of Supermedial Pain with Score Improvement
Score Improvement (3 months-pre)DSFA	Slope	P value	95% CI
∆IKDC	9.4	0.013	2.1–16.8.1.8
∆Lysholm	9.9	0.16	−4.0−23.9
∆VAS	−7.1	0.65	−37.8−23.7
∆KOOS symptoms	7.3	0.16	−3.0−17.5
∆KOOS pain	13.6	0.034	1.1–26.1.1.1
∆DOOS ADL	9.1	0.10	−1.9−20.1
∆KOOS Sports	17.2	0.032	1.5–32.8.5.8
∆KOOS QOL	21.2	0.0012	8.8–33.6.8.6

Univariable linear regression was used to assess whether the presence of superomedial pain predicted changes in outcome scores from baseline to 3 months. IKDC International Knee Documentation Committee, VAS visual analog scale, KOOS Knee injury and Osteoarthritis Outcome Score, ADL activities of daily living, QOL quality of life

Table 5.

Association of pain site count with change in functional scores at 3 months

Association of PAin SIte Count with Score Improvement
Score Improvement (3 months-pre)	Slope	P value	95% CI
∆IKDC	0.7	0.68	−2.7−4.1
∆Lysholm	−4.2	0.18	−10.3-1.9.3.9
∆VAS	−0.1	0.98	−13.3-13.0.3.0
∆KOOS symptoms	−0.5	0.82	−5.2-4.1.2.1
∆KOOS pain	2.3	0.44	−3.5–8.0.5.0
∆DOOS ADL	−0.2	0.95	−5.1-4.8.1.8
∆KOOS Sports	0.5	0.90	−6.8–7.7.8.7
∆KOOS QOL	4.1	0.17	−1.8–10.1.8.1

Linear regression was performed to evaluate the effect of total pain site count on change scores for each outcome. IKDC International Knee Documentation Committee, VAS visual analog scale, KOOS Knee injury and Osteoarthritis Outcome Score, ADL activities of daily living, QOL quality of life

Multivariable linear regression, adjusted for age, sex, and BMI, confirmed that superomedial pain remained independently associated with greater IKDC improvement (β=–9.1, 95% CI: − 16.6 to − 1.5, p = 0.02), indicating a clinically meaningful association. Age, sex, and BMI were not statistically significant predictors (Table 6).

Table 6.

Multivariable linear regression predicting improvement in IKDC score

Multivariable analysis
Variables	Estimate	95% CI	P value
Intercept	2.8	–46.3 to 52.0	0.91
Superomedial Pain (ref: no)	−9.1	–16.6 to − 1.5	0.019
Sex (ref: female)	7.9	–0.9 to 16.7	0.079
Age (years)	0.04	–0.4 to 0.5	0.84
BMI (kg/m²)	0.3	–0.7 to 1.3	0.59

Model includes superomedial pain, sex, age, and BMI as predictors. Superomedial pain was coded as 1 = presence and 0 = absence. Negative β values therefore indicate greater improvement in the presence of superomedial pain. CIconfidence interval, IKDCInternational Knee Documentation Committee, BMIbody mass index

Discussion

This study demonstrated that 4 specific pain localization, particularly in the superomedial region of the knee, is a meaningful predictor of clinical response following CRFA in patients with advanced OA (Kellgren–Lawrence grades 3–4). Patients with superomedial pain presented with significantly lower baseline functional scores (IKDC, Lysholm, KOOS–Pain) and experienced greater improvements at 3 months post-procedure, independent of age, sex, and BMI. These findings support the utility of regional pain mapping in identifying patients who are more likely to benefit from CRFA.

The association between superomedial pain and greater functional improvement may be explained by its anatomical proximity to the target nerves. The superomedial compartment corresponds closely to the distribution of the superior medial genicular nerve, which is a primary target of CRFA [26]. Pain localized in this area may reflect a nociceptive pattern that is more responsive to radiofrequency denervation [27]. This reinforces the clinical value of preprocedural pain topography as a simple, non-invasive tool for prognostic stratification.

While a higher number of reported pain sites was associated with worse baseline functional scores, it did not predict treatment response. This discrepancy suggests that widespread pain may reflect underlying pain sensitization or nociplastic mechanisms, such as central sensitization or pain modulation dysfunction, rather than structural nociception [28]. In these patients, the analgesic benefits of localized nerve ablation may be attenuated. These findings are consistent with the literature linking widespread pain distribution to poor treatment response in other musculoskeletal conditions [29], and highlight the importance of screening for pain modulation phenotypes [30] when selecting candidates for CRFA.

This study had several limitations. The modest sample size and single-center design limit generalizability. We did not formally assess central sensitization, which could have strengthened the interpretation related to pain distribution. Additionally, the use of self-reported pain locations, while pragmatic, may be subject to recall bias or anatomical misclassification. Because of the relatively modest sample size, this study was not specifically powered to detect small effect sizes or subtle associations between pain localization patterns and clinical outcomes. Although statistically significant associations were observed for superomedial pain localization, the possibility of type II error for other pain regions or secondary outcomes cannot be excluded. Therefore, the present findings should be interpreted with caution and considered exploratory, warranting confirmation in larger, multicenter cohorts. Despite these limitations, our study has several strengths. We demonstrated that patient-identified pain location, an easily obtainable clinical variable, can meaningfully predict therapeutic response to CRFA. By restricting the cohort to patients with advanced radiographic OA (Kellgren–Lawrence grades 3–4), disease severity was controlled, enhancing clinical relevance. To the best of our knowledge, this is the first study to show that baseline pain localization can stratify treatment response to genicular nerve ablation in this population.

Conclusions

Preprocedural superomedial pain was associated with poorer baseline function and greater short-term functional improvement after CRFA. Although a higher pain site count correlated with worse baseline function, it did not predict treatment response. These findings suggest that targeted pain localization may enhance patient selection for CRFA and improve clinical outcomes in advanced knee OA.

Abbreviations

OA

Osteoarthritis

QOL

Quality of life

CRFA

Cooled radiofrequency ablation

RFA

Radiofrequency ablation

TKA

Total knee arthroplasty

IKDC

International knee documentation committee

KOOS

Knee injury and osteoarthritis outcome score

VAS

Visual analog scale

BMI

Body mass index

PROM

Patient-reported outcome measure

IQR

Interquartile range

CI

Confidence interval

Authors’ contributions

Naofumi Hashiguchi: Writing, conceptualization, methodology, investigation, data collection, editingAtsuo Nakamae: Methodology, review, editingTsukasa Kanda: Investigation, data collectionKyohei Nakata: Data collectionAkinori Nekomoto: Data collectionKoji Takeda: Data collectionJunya Tsukisaka: Data collectionNobuo Adachi: Review.

Funding statement

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by the Hiroshima University Epidemiology Research Ethics Review Committee (approval number: E2023-0254). The study was conducted in accordance with the Declaration of Helsinki and all applicable institutional guidelines. Written informed consent to participate was obtained from all participants prior to enrollment.

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.