Bone marrow lesion coexisted with subchondral bone attrition and/or subchondral bone cyst is associated with knee pain in knee osteoarthritis regardless of cartilage lesion: the Bunkyo health study

Abstract

Background

Subchondral bone changes, which include bone marrow lesion (BML), subchondral bone attrition (SBA) and subchondral bone cyst (SBC) by magnetic resonance imaging (MRI) analysis, are reportedly implicated for knee pain in knee osteoarthritis (OA). However, the relationship between these subchondral bone changes and OA knee pain and the effect of articular cartilage lesion on the pain remain elusive.

Methods

Elderly subjects (1,145 subjects, 72.9 years old on average) in the Bunkyo Health Study, 71.5% of whom had knee OA with ≥ Kellgren-Lawrence grade 2, were enrolled. Knee pain in daily life over the past few days was measured using Visual Analogue Scale (VAS) in the Japanese Knee Osteoarthritis Measure. The subjects without or with pain were defined if they indicated a pain VAS score of 0 or a pain VAS score of ≥ 1. The association was examined between knee pain and MRI-detected OA structural changes which were determined according to the Whole Organ Magnetic Resonance Imaging Score.

Results

While 62.2% of the subjects were free from knee pain, 37.8% of the subjects had knee pain. Knee pain was not related with cartilage lesion without subchondral bone changes (odds ratio [OR]: 1.10 [95% confidence interval [CI]: 0.83–1.46]) or BML alone (OR: 1.32 [95% CI: 0.95–1.83]). However, knee pain was significantly associated with BML coexistent with SBA (OR: 2.22 [95% CI: 1.25–3.97]), SBC (OR: 1.79 [95% CI: 1.28–2.51]), or both SBA and SBC (OR: 2.18 [95% CI: 1.35–3.53]). Similar positive relationships between knee pain and coexisted subchondral bone changes were obtained regardless of the presence or absence of cartilage lesion present above the BML region. When BML was not coexistent with either SBA or SBC regardless of cartilage lesion above the subchondral bone changes, BML was not associated with knee pain (OR: 1.26 [95% CI: 0.90–1.77]) or (OR: 2.16 [95% CI: 0.89–5.23]).

Conclusions

BML coexistent with SBA and/or SBC, but not BML without the coexistence, was associated with knee pain in the elderly with knee OA regardless of the presence or absence of cartilage lesion.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13075-025-03644-2.

Background

Knee osteoarthritis (OA) is a prevalent joint disease with chronic pain in adults [1], and is a leading cause of functional impairment, such as diminished activities of daily living and decline of walking speed that is one of the health indicators and a predictor of lifespan in elderlies [2, 3, 4, 5]. Research on OA is advancing toward the goal of early diagnosis and early treatment of this disease. A genome-wide association study meta-analyses including our large-scale cohort data revealed > 900 independent associations with plenty of associations which have not been reported or with several biological processes such as circadian clock, glial-cell-related processes and pathways with an established role in OA (transforming growth factor β, fibroblast growth factor, Wnt, and bone morphogenetic protein, retinoic acid signaling, and extracellular matrix organization) [6]. However, at present, surgical treatment such as knee joint replacement is occasionally required in advanced cases of knee OA, and knee pain is the major reason for hospital visits in knee OA patients and is still an important target for treatment because of the lack of any disease-modifying drugs [3, 7, 8, 9]. Therefore, early intervention for knee pain will be extremely important, and understanding the pathophysiology of knee pain is even more important.

Although early-stage OA knee joint shows no or negligible abnormalities on radiographs, magnetic resonance imaging (MRI) is an excellent tool for the early detection of structural abnormalities of knee joint [10, 11, 12]. Accumulated lines of evidence obtained by MRI examination have disclosed that knee pain in knee OA patients is closely associated with concentration of mechanical load on local area in the joint, bone marrow changes and synovitis [13, 14, 15, 16]. Eventually, subchondral bone changes are suggested as an initial change in knee OA and a risk factor for OA progression [17].

Subchondral bone changes include bone marrow lesion (BML), subchondral bone attrition (SBA) and subchondral bone cyst (SBC), all of which may be engendered by accelerated concentration of mechanical load to a local area [13]. Among the subchondral bone changes, many previous studies have shown the association between BML and knee pain [15, 18, 19, 20]. However, in an observational study on the subjects without radiographic OA changes, prevalence of SBA and SBC as well as BML is reported to be significantly higher in a group with knee pain compared to that without knee pain [10]. Several region-specific longitudinal studies on the relationships among these subchondral bone abnormalities in knee OA have indicated that SBCs tend to longitudinally develop in areas where BMLs are present [21], and SBA is highly associated with BMLs in the same knee subregion [22]. Thus, these data suggest the possibility that BML, SBA and SBC may be generated by their interactions each other [13] and that combined lesions of the subchondral bone changes may be more closely related to knee pain. While most previous studies on the associations between knee pain and subchondral bone changes focused on only BML [23, 24, 25], some have suggested that BML alone may not be sufficient to cause pain but may require the presence of other subchondral bone changes. Actually, BMLs are more strongly associated with knee pain when they are accompanied by subchondral bone defects, i.e. SBA, or when the subchondral bone plate is exposed due to loss of articular cartilage [19, 26, 27]. However, relationship between the coexistence of BML with SBA and/or SBC and knee pain remains elusive. In addition, it remains unknown whether cartilage lesion present above the subchondral bone changes is related to association between subchondral bone changes and knee pain.

In this study, we investigated associations between knee pain and subchondral bone changes including BML, SBA, and SBC and also examined whether these associations are affected by the presence or absence of cartilage lesion in elderly populations using the data from the large-scale cohort, the Bunkyo Health Study (BHS) cohort [28, 29, 30, 31, 32].

Methods

Subjects of BHS cohort

The study protocol of BHS has been approved by the ethics committee of Juntendo University from November in 2015 (No. 2015078, 2016138, 2016131, and 2017121). The BHS has been developed as a prospective cohort study over 10 years to identify risk factors for needing long-term care for elderly [28, 29, 30, 31, 32]. Elderly subjects between 65 and 84 years of age living in Bunkyo-ku, an urban area in Tokyo, Japan were recruited, and a total of 1,630 subjects were participated from November 2015 to September 2018 [28]. The study is conducted in accordance with the principles outlined in the Declaration of Helsinki. At the orientation meeting, participants were told that they have the right to withdraw from the trial at any time and, then, written informed consent was obtained from all the participants. Collected data of the subjects are coded with non-identifying numbers and stored securely in password-protected files. Accessibility to the files of the collected data of BHS was limited to the principal investigators.

Inclusion and exclusion criteria of the present study

This study was a cross-sectional study using the baseline data of the BHS. While recruited 1,630 subjects were agreed to participate in the BHS at baseline, the elderlies who had implanted pacemakers or defibrillators or had diabetes with insulin therapy were excluded from the BHS cohort. Thus, 1,191 subjects were included in the BHS at baseline [28]. Then, they underwent radiography of both knee joints and full-limb in the standing position and MRI of the knee joints. As the 46 subjects showed lateral-type knee OA with tibiofemoral angle (TFA) < 173°, they were excluded and thus the remaining 1,145 subjects were included in the present study [6, 31, 32].

The radiographic evaluation of knee OA

All subjects underwent knee radiographs by the weight-bearing antero-posterior radiographs of the femorotibial joint for both knees by the bilateral standing extended view and by the 45° of flexion view (Rosenberg view) [33] at one institution. The radiographic severity of knee OA was evaluated by the Kellgren-Lawrence (K/L) classification [34] by the bilateral standing extended view with slight modification. Specifically, KL grade 2 was defined as the presence of “at least 2 definite osteophytes” or “one definite osteophyte plus definite joint space narrowing” according to our recent study [6]. By using this definition, we evaluated the presence or absence of radiographic knee OA, which was defined by having a K/L grading score ≥ 2. The 45° of flexion view (Rosenberg view) was used to evaluate joint space narrowing for the cases, in which the evaluation of joint space width was difficult due to unclear joint line by the bilateral standing extended view. TFA was evaluated by the angle between the mechanical axes of the femur and tibia based on the full-limb standing radiographs [31, 32]. All radiographs were taken by experienced technicians. Three experienced senior orthopedic surgeons (HK, KY, and KK) independently and blindly evaluated K/L grade and TFA. Intra-observer reliability of them were 0.894 [95% confidence interval [CI]: 0.853–0.924] (HK), 0.863 [95% CI: 0.775–0.919] (KY), and 0.929 [95% CI: 0.914–0.941] (KK), while inter-observer-reliability of them were 0.825 [95% CI: 0.784–0.859] (HK and KY), 0.914 [95% CI: 0.879–0.939] (HK and KK), and 0.841 [95% CI: 0.804–0.871] (KY and KK) in the present study. When the readers disagreed on the presence of radiographic OA, the radiograph readings were adjudicated by a panel of three readers. The left knee was selected for analysis of radiography in the subjects without any symptoms, but in the subjects who had symptoms such as knee pain, stiffness, and/or restricted range of motion, the affected knee was selected for the analysis.

MRI evaluation

The 0.3-Tesla MRI system (Hitachi AIRIS Vento; Hitachi Medical Corporation, Tokyo, Japan) with a knee coil was used to analyze knee joints of the subjects [31, 32]. The structural changes of the knee joints were examined by sagittal and coronal proton density-weighted fat-suppressed (PDFS) spin-echo sequences (in-plane resolution = 0.39 × 0.39 mm²; TR/TE = 1500/21.2 msec; section thickness = 3 mm). The MRI data were evaluated in the whole knee joint, and OA structural changes were scored according to the Whole Organ Magnetic Resonance Imaging Score (WORMS) [35]. Absence and presence of osteophyte, meniscus lesion, BML, SBA, SBC, synovitis, and ligament were WORMS grade 0 and WORMS grade ≥ 1, respectively, whereas absence and presence of cartilage lesion were WORMS grade ≤ 1 and WORMS grade ≥ 2, respectively [10]. Cartilage lesion and subchondral bone changes (BML, SBA, and SBC) were evaluated in 15 regions for cartilage lesion, BML, and SBC or in 14 regions for SBA. The prevalence of structural changes was assessed in the medial and lateral compartments and whole knee joints according to WORMS [35].

Subchondral bone changes are occasionally associated with cartilage lesion [17], although prevalence rate of the coexistence remains unknown. Therefore, we examined subchondral bone changes in the connection with coexistence of cartilage lesion in each subregion. By focusing on BML and cartilage lesion, MRI lesions of the subjects could be classified into 8 different patterns: BML without either SBA or SBC (Fig. 1A and E), BML coexistent with SBA alone (Fig. 1B and F), BML coexistent with SBC alone (Fig. 1C and G), and BML coexistent with both SBA and SBC (Fig. 1D and H) in the absence or presence of cartilage lesion above the BML region. Cartilage lesion, subchondral bone changes and the lesions coexisted each other were analyzed in fat-suppressed T2-weighted MRI of knee joints in 14 regions.

Fig. 1.

Schematic presentation of the eight different MRI patterns of subchondral bone changes and their representative MRI images of the knee joints. Cartilage lesion, bone marrow lesion (BML), subchondral bone attrition (SBA), and subchondral bone cyst (SBC), which were observed in either femur or tibia, were analyzed in sagittal fat-suppressed T2-weighted MRI images of the knee joints of the subjects. Based on the data, MRI lesions of the subjects were classified into 8 different patterns, which are depicted in the tibia of each schema. A - D, BML without either SBA or SBC (A), BML coexistent with SBA alone (B), BML coexistent with SBC alone (C), and BML coexistent with both SBA and SBC (D) in the absence of cartilage lesion above the BML region. E - H, BML without either SBA or SBC (E), BML coexistent with SBA alone (F), BML coexistent with SBC alone (G), and BML coexistent with both SBA and SBC (H) in the presence of cartilage lesion above the BML region. Boxed areas in each panel are shown at higher magnification to the right. Arrowheads, BML; red arrows, SBA; asterisks, SBC; yellow arrows, cartilage lesion

The left knee was selected for the MRI analysis in the subjects without any symptoms. However, in the subjects who had symptoms such as knee pain, stiffness, and/or restricted range of motion, the affected knee was selected for the analysis.

Measurement of knee pain severity

Knee pain in daily life over the past few days was measured using Visual Analogue Scale (VAS) in the Japanese Knee Osteoarthritis Measure (JKOM) (JKOM I, ranging from 0 to 100) [36]. VAS scores were obtained according to the instruction of JKOM I as follows: “The following questions will ask you about the degree of knee pain you experience. Your replies will range from the far left side or “no pain at all (0)” to the far right or “the most severe pain you’ve ever had (100)”. Mark an X indicating where you think the level of pain you experienced during the last few days fell.” In the scale, a score of 0 was defined as “no pain,” and a score of 100 was defined as “pain as bad as it could be,” with standardized instructions. The subjects were defined as those without pain if they indicated a pain VAS score of 0 and those with pain if they indicated a pain VAS score of ≥ 1.

Statistical analysis

The associations between knee pain and MRI-detected OA structural alterations including cartilage lesion, osteophyte, BML, SBA, SBC, meniscus lesion, synovitis, and ligament lesion were analyzed by univariate logistic regression analysis and multivariate logistic regression analysis. The p value < 0.05 was considered to be significant. A SPSS Statistics 25.0 software program (IBM, Armonk, NY) was used for all statistical analyses in the present study.

Results

Characteristics of the subjects

Among 1,145 subjects, 683 subjects (55.7%) were female and the remaining 507 subjects (44.3%) were male (Table 1) [31]. Age and body mass index (BMI) of the subjects was 72.9 years of age on average and 22.8 kg/m² on average, respectively. While approximately 30% (28.5%, 326 among 1145) of the subjects showed no radiographic knee OA, mora than 70% (71.5%, 819 among 1145) of them showed radiographic knee OA (≥ K/L grade 2) (Table 1). The TFA of the subjects was 177.6° on average. The 432 subjects (37.8%) had knee pain and remaining 713 subjects (62.2%) were free from knee pain. Pain VAS of the subjects were 8.6 ± 16.2.

Table 1.

Baseline characteristics of the subjects

Characteristics	Data
Number of subjects	1,145
Gender, n (%)	Female, 638 (55.7%); Male, 507 (44.3%)
Age, years	72.9 (5.4) (65–84)
BMI, kg/cm2	22.8 (3.0) (15.4–36.6)
Radiographic findings
K/L grades 0/1,2,3,4 (n)	326 (28.5%), 674 (58.9%), 102 (8.9%), 43 (3.7%)
TFA, degree	177.6 (2.5) (173–191)
Pain (+/-)	37.8% (432) / 62.2% (713)
Pain VAS (0-100)	8.6 (16.2) (0–98)

Data are expressed as mean value (standard deviation, if not indicated). BMI, body mass index; K/L grade, Kellgren-Lawrence grade; TFA, tibiofemoral angle; VAS, visual analog scale

Prevalence of MRI-detected OA structural changes

When structural changes of knee OA were evaluated by MRI according to the WORMS, 35.4%, 96.7%, 76.2%, 33.3%, 95.7%, 66.6%, and 60.6% of the subjects had cartilage lesion, osteophyte, meniscus lesion, BML, SBA, SBC, and synovitis, respectively (Supplemental Table 1). The ligament lesion was seen in only 8.8% of the subjects. Among the subchondral bone lesions, SBA was the most prevalently observed (Supplemental Table 1). Importantly, all the MRI lesions were more frequently observed in the medial femorotibial joint as compared to the lateral femorotibial joint of the subjects (Supplemental Table 1).

BML without either SBA or SBC regardless of the presence or absence of cartilage lesion (Fig. 1A and E) was seen in 16.2% of the subjects (Supplemental Table 2). While 5.9% and 17.9% of the subjects had BML coexistent with SBA (Fig. 1B and F) and BML coexistent with SBC (Fig. 1C and G) regardless of the presence or absence of cartilage lesion, respectively, 7.0% of the subjects had BML coexistent with both SBA and SBC (Fig. 1D and H) regardless of the presence or absence of cartilage lesion (Supplemental Table 2). When BML alone and BML coexistent with SBA and/or SBC were analyzed individually by the absence or presence of cartilage lesion, the prevalence of BML without either SBA or SBC and BML coexistent with SBA and/or SBC were higher in the absence of cartilage lesion than in the presence of cartilage lesion (Supplemental Table 2). These results show that BML without cartilage lesion above the same BML region is observed more prevalently than BML with cartilage lesion.

Association between knee pain and MRI-detected OA structural changes in the knee joint

We evaluated the associations between pain and OA structural changes in whole knee joint by following the previous study [37]. We found that knee pain is significantly associated with cartilage lesion, BML, and SBC (Table 2). However, there were no correlations between knee pain and osteophyte, meniscus lesion, synovitis, or ligament injury (Table 2).

Table 2.

Association between knee pain and MRI-detected OA structural changes in the whole knee joint of the subjects

		Prevalence, n (%)				Adjusted OR (95%CI)
		Pain: − n = 713		Pain: + n = 432
Cartilage	A	495	(66.9)	245	(33.1)	1.00	(ref.)
	P	218	(53.8)	187	(46.2)	1.41	(1.09–1.84)
OP	A	22	(57.9)	16	(42.1)	1.00	(ref.)
	P	691	(62.4)	416	(37.6)	0.81	(0.41–1.58)
Meniscus	A	188	(69.1)	84	(30.9)	1.00	(ref.)
	P	525	(60.1)	348	(39.9)	1.30	(0.96–1.76)
BML	A	518	(68.0)	244	(32.0)	1.00	(ref.)
	P	195	(50.9)	188	(49.1)	1.73	(1.33–2.24)
SBA	A	24	(49.0)	25	(51.0)	1.00	(ref.)
	P	689	(62.9)	407	(37.1)	0.56	(0.31–1.02)
SBC	A	279	(72.8)	104	(27.2)	1.00	(ref.)
	P	434	(57.0)	328	(43.0)	1.73	(1.33–2.24)
Synovitis	A	298	(66.1)	153	(33.9)	1.00	(ref.)
	P	415	(59.8)	279	(40.2)	1.27	(0.98–1.63)
Ligament	A	663	(63.5)	381	(36.5)	1.00	(ref.)
	P	50	(49.5)	51	(50.5)	1.44	(0.94–2.21)

In a logistic regression analysis, the dependent variable is pain [without pain: VAS = 0, with: VAS ≥ 1], and the independent variable is each lesion [without subchondral bone lesions: score 0, with: score 1–3; without cartilage lesions: score ≤ 1, with: score ≥ 2; without synovitis: score 0, with: score 1–3] adjusted for age, gender, and BMI. OP, osteophyte; BML, bone marrow lesion; SBA, subchondral bone attrition; SBC, subchondral bone cyst; A, absent; P, present; ref., reference

Then, we examined more detailed associations of knee pain with combined cartilage lesion and subchondral bone changes. Cartilage lesion without coexisting with subchondral bone changes showed no association with knee pain (odds ratio [OR]: 1.10 [95% CI: 0.83–1.46]) (Table 3). Importantly, BML without other subchondral bone changes (Fig. 1A and E) showed no significant association with knee pain (OR: 1.32 [95% CI: 0.95–1.83]) (Table 3). On the other hand, significant associations between knee pain and subchondral bone changes were obtained in BML coexistent with SBA (Fig. 1B and F) (OR: 2.22 [95%: CI 1.25–3.97]), SBC (OR: 1.79 [95% CI: 1.28–2.51]) (Fig. 1C and G), or both SBA and SBC (Fig. 1D and H) (OR: 2.18 [95% CI: 1.35–3.53]) (Table 3). Similar positive relationships between knee pain and BML were obtained regardless of the presence or absence of cartilage lesion which is present above the BML region (Table 3). When BML was not coexistent with either SBA or SBC in the absence or presence of cartilage lesion above the subchondral bone changes (Fig. 1A and E), BML was not associated with knee pain (OR: 1.26 [95% CI: 0.90–1.77]) or (OR: 2.16 [95% CI: 0.89–5.23]) (Table 3).

Table 3.

Association between knee pain and BML with or without coexistent subchondral changes in the absence of cartilage lesion above the same BML region in the whole knee joint of the subjects

Subchondral bone changes			CL		Prevalence, n (%)				Adjusted OR (95%CI)
BML	SBA	SBC			Pain: − n = 713		Pain: + n = 432
-	-	-	+	A	546	(63.7)	311	(36.3)	1.00	(ref.)
				P	167	(58.0)	121	(42.0)	1.10	(0.83–1.46)
+	-	-	+/-	A	612	(63.8)	348	(36.3)	1.00	(ref.)
				P	101	(54.6)	84	(45.4)	1.32	(0.95–1.83)
+	+	-	+/-	A	688	(63.6)	393	(36.4)	1.00	(ref.)
				P	25	(39.1)	39	(60.9)	2.22	(1.25–3.97)
+	-	+	+/-	A	622	(66.2)	318	(33.8)	1.00	(ref.)
				P	91	(44.4)	114	(55.6)	1.79	(1.28–2.51)
+	+	+	+/-	A	682	(64.0)	383	(36.0)	1.00	(ref.)
				P	31	(38.8)	49	(61.3)	2.18	(1.35–3.53)
+	-	-	-	A	616	(63.4)	355	(36.6)	1.00	(ref.)
				P	97	(55.7)	77	(44.3)	1.26	(0.90–1.77)
+	+	-	-	A	693	(63.2)	404	(36.8)	1.00	(ref.)
				P	20	(41.7)	28	(58.3)	1.97	(1.08–3.60)
+	-	+	-	A	628	(65.6)	329	(34.4)	1.00	(ref.)
				P	85	(45.2)	103	(54.8)	1.95	(1.41–2.71)
+	+	+	-	A	687	(63.2)	400	(36.8)	1.00	(ref.)
				P	26	(44.8)	32	(55.2)	1.73	(0.99–2.99)
+	-	-	+	A	705	(62.9)	416	(37.1)	1.00	(ref.)
				P	8	(33.3)	16	(66.7)	2.16	(0.89–5.23)
+	+	-	+	A	708	(62.8)	419	(37.2)	1.00	(ref.)
				P	5	(27.8)	13	(72.2)	3.41	(1.17–9.92)
+	-	+	+	A	704	(63.1)	412	(36.9)	1.00	(ref.)
				P	9	(31.0)	20	(69.0)	2.87	(1.25–6.58)
+	+	+	+	A	708	(63.1)	414	(36.9)	1.00	(ref.)
				P	5	(21.7)	18	(78.3)	4.08	(1.47–11.3)

In a logistic regression analysis, the dependent variable is pain [without pain: VAS = 0, with: VAS ≥ 1], and the independent variable is each lesion [without subchondral bone lesions: score 0, with: score 1–3; without cartilage lesions: score ≤ 1, with: score ≥ 2; without synovitis: score 0, with: score 1–3] adjusted for age, gender, and BMI. OP, osteophyte; BML, bone marrow lesion; SBA, subchondral bone attrition; SBC, subchondral bone cyst; CL, cartilage lesion; A, absent; P, present; ref., reference

Association between knee pain and MRI-detected OA structural changes in the weight bearing region of the knee joint

As knee pain in knee OA is closely associated with accelerated concentration of mechanical load to a local area [13], we hypothesized that the association between knee pain and OA-structural changes in the loading region of the knee joint (medial femorotibial joint) may be stronger than those in the whole knee joint. When limited to 5 regions of the loading site, knee pain was associated with BML (OR: 2.14 [95% CI: 1.52-3.00]), and SBC (OR: 1.69 [95% CI: 1.29–2.21]) as well as cartilage lesion (OR: 1.36 [95% CI: 1.04–1.77]) and meniscus lesion (OR: 1.35 [95% CI: 1.04–1.75]) (Table 4). On the other hand, knee pain was not associated with osteophyte (OR: 1.44 [95% CI: 0.85–2.43]) among the structural changes of knee OA detected on MRI (Table 4).

Table 4.

Association between knee pain and structural changes of knee joint in the loading site (medial femorotibial joint) of the subjects

		Prevalence, n (%)				Adjusted OR (95%CI)
		Pain: − n = 713		Pain: + n = 432
Cartilage	A	522	(66.0)	269	(34.0)	1.00	(ref.)
	P	191	(54.0)	163	(46.0)	1.36	(1.04–1.77)
OP	A	54	(71.1)	22	(28.9)	1.00	(ref.)
	P	659	(61.6)	410	(38.4)	1.44	(0.85–2.43)
Meniscus	A	299	(68.3)	139	(31.7)	1.00	(ref.)
	P	414	(58.6)	293	(41.4)	1.35	(1.04–1.75)
BML	A	639	(65.7)	333	(34.3)	1.00	(ref.)
	P	74	(42.8)	99	(57.2)	2.14	(1.52–3.00)
SBA	A	52	(46.4)	60	(53.6)	1.00	(ref.)
	P	661	(64.0)	372	(36.0)	0.54	(0.36–0.81)
SBC	A	538	(67.8)	256	(32.2)	1.00	(ref.)
	P	175	(49.9)	176	(50.1)	1.69	(1.29–2.21)

In a logistic regression analysis, the dependent variable is pain [without pain: VAS = 0, with: VAS ≥ 1], and the independent variable is each lesion [without subchondral bone lesions: score 0, with: score 1–3; without cartilage lesions: score ≤ 1, with: score ≥ 2; without synovitis: score 0, with: score 1–3] adjusted for age, gender, and BMI. OP, osteophyte; BML, bone marrow lesion; SBA, subchondral bone attrition; SBC, subchondral bone cyst; A, absent; P, present; ref., reference

When the associations of coexisting lesions in the weight-bearing region were further analyzed, knee pain was associated with neither cartilage lesion without coexistent with subchondral bone changes (OR: 0.96 [95% CI: 0.70–1.32]) nor BML alone regardless of cartilage lesion (OR: 1.53 [95% CI: 0.95–2.48]) (Table 5). On the other hand, knee pain was associated with BML coexisting with SBA (BML + SBA; OR: 3.87 [95% CI: 1.76–8.51]), SBC (BML + SBC; OR: 3.01 [95% CI: 1.92–4.74]), or both SBA and SBC (BML + SBA + SBC; OR: 3.23 [95% CI: 1.59–6.53]) regardless of cartilage lesion.

Table 5.

Association between knee pain and BML with or without coexistent subchondral bone changes in the presence or absence of cartilage lesion above the same BML region in the loading site (medial femorotibial joint) of the subjects

Subchondral bone changes			CL		Prevalence, n (%)				Adjusted OR (95%CI)
BML	SBA	SBC			Pain: − n = 713		Pain: + n = 432
-	-	-	+	A	582	(62.6)	347	(37.4)	1.00	(ref.)
				P	131	(60.6)	85	(39.4)	0.96	(0.70–1.32)
+	-	-	+/-	A	676	(63.3)	392	(36.7)	1.00	(ref.)
				P	37	(48.1)	40	(51.9)	1.53	(0.95–2.48)
+	+	-	+/-	A	704	(63.4)	406	(36.6)	1.00	(ref.)
				P	9	(25.7)	26	(74.3)	3.87	(1.76–8.51)
+	-	+	+/-	A	680	(64.9)	368	(35.1)	1.00	(ref.)
				P	33	(34.0)	64	(66.0)	3.01	(1.92–4.74)
+	+	+	+/-	A	701	(63.5)	403	(36.5)	1.00	(ref.)
				P	12	(29.3)	29	(70.7)	3.23	(1.59–6.53)
+	-	-	-	A	680	(63.0)	399	(37.0)	1.00	(ref.)
				P	33	(50.0)	33	(50.0)	1.42	(0.85–2.39)
+	+	-	-	A	709	(63.0)	416	(37.0)	1.00	(ref.)
				P	4	(20.0)	16	(80.0)	5.37	(1.75–16.5)
+	-	+	-	A	684	(64.2)	382	(35.8)	1.00	(ref.)
				P	29	(36.7)	50	(63.3)	2.64	(1.62–4.31)
+	+	+	-	A	705	(62.8)	418	(37.2)	1.00	(ref.)
				P	8	(36.4)	14	(63.6)	2.71	(1.10–6.71)
+	-	-	+	A	706	(62.9)	417	(37.1)	1.00	(ref.)
				P	7	(31.8)	15	(68.2)	2.35	(0.92–5.99)
+	+	-	+	A	708	(62.8)	420	(37.2)	1.00	(ref.)
				P	5	(29.4)	12	(70.6)	3.19	(1.08–9.42)
+	-	+	+	A	707	(63.1)	414	(36.9)	1.00	(ref.)
				P	6	(25.0)	18	(75.0)	3.88	(1.48–10.2)
+	+	+	+	A	709	(63.0)	417	(37.0)	1.00	(ref.)
				P	4	(21.1)	15	(78.9)	3.91	(1.26–12.2)

In a logistic regression analysis, the dependent variable is pain [without pain: VAS = 0, with: VAS ≥ 1], and the independent variable is each lesion [without subchondral bone lesions: score 0, with: score 1–3; without cartilage lesions: score ≤ 1, with: score ≥ 2; without synovitis: score 0, with: score 1–3] adjusted for age, gender, and BMI. OP, osteophyte; BML, bone marrow lesion; SBA, subchondral bone attrition; SBC, subchondral bone cyst; A, absent; P, present; ref., reference

Discussion

In the present study, we have examined relationships between knee pain and combined subchondral bone changes in elderly populations from the large-scale cohort, and demonstrated to the best of our knowledge for the first time that knee pain is significantly associated with BML coexistent with SBA and/or SBC, but not BML without the coexistence. Our study has also indicated that this association is not affected by the cartilage lesion present above the BML region, and that the association becomes stronger in the loading site of the knee joint.

BML has been shown to be associated with knee pain and progression of knee OA [23, 24, 25]. In addition, greater size of BML is known to be associated with greater pain and progression [4, 24, 38]. As shown in the previous [10, 21, 22, 25] and present studies, BML is commonly accompanied by SBA and/or SBC in knee OA joints. However, associations between knee pain and combined subchondral bone changes remained to be investigated. In the present study, when the association was independently assessed without consideration of the combined lesions, BML was associated with knee pain, supporting the previous studies [10, 15, 18, 19, 20, 23, 24, 25]. Importantly, however, our study has demonstrated that BML coexisted with SBA and/or SBC shows significant association with knee pain, but BML alone has no such association. Previous longitudinal studies revealed that BML disappears without any treatments such as joint unloading or reduction of joint loading due to alignment changes of lower limb by osteotomy around the knee [39, 40] and suggested the possibility that SBA and SBC may be generated by interactions with BML based on the findings that SBA and SBC develop in the areas close to BML [13, 21, 22, 25]. These data indicate reversibility and variability of BML, and thus it is possible to speculate that knee pain may emerge when SBA and/or SBC develop around the areas of BML or when SBA and/or SBC induce reversible BML around the lesions.

SBA is well known to be associated with knee pain in knee OA patients [10, 19, 22, 38]. SBA appears to be a pathological condition of subchondral bone, which suffers from porosity by increased bone resorption pits, reduction of bone strength, and microfracture [41]. Bone resorption pit is characterized by vascular invasion of subchondral bone plate and articular cartilage through the tidemark [42]. In osteochondral channels of the bone resorption pits, osteoclast density and activation and nerve growth factor expression are frequently increased in symptomatic knee OA [43, 44]. Thus, these cellular and molecular reactions in the subchondral bone region may be involved in acceleration of inflammation and induction of pain, and suggest one of the mechanisms for the result in the current study that BML coexisted with SBA is associated with knee pain.

SBC is commonly identified in the load-bearing regions of OA knee joints, and the formation of SBC is explained by two prevalent theories [45]. In the synovial breach theory, SBC is speculated to be created by the penetration of synovial fluid and/or tissue into the subchondral bone marrow due to microfractures of the subchondral bone plate in the articular surface, leading to the induction of inflammation and the development of myxomatous tissue in the subchondral bone marrow. In the bone contusion theory, SBC is thought to be formed by microfracture of subchondral bone and bone marrow edema without penetration of synovial fluid, resulting in the activation of osteoclasts and osteoblasts in response to microfracture. Although pathological data on these theories are still limited, microfracture of subchondral bone and subsequent tissue reactions involve osteoclastogenesis and nerve distribution, both of which are reportedly increased in the joint tissues obtained from the OA patients showing BML with SBC as compared to those having BML without SBC [46]. In knee OA, osteoclasts in the subchondral bone region are known to secrete netrin-1 and induce sensory nerve axonal growth in subchondral bone [47]. Therefore, it seems likely that these tissue and cellular reactions might be implicated for the significant association between knee pain and BML coexisted with SBC in the present study.

In early-stage knee OA, cartilage lesion is a risk factor for subsequent cartilage loss [17]. However, loss of cartilage thickness is only weakly associated with worsening of knee pain [4, 48]. Since articular cartilage is aneural and avascular, it is generally accepted that articular cartilage is not the tissue, which directly generates pain. Our findings showing that significant association between knee pain and BML coexisted with SBA and/or SBC is not affected by the presence or absence of cartilage lesion strongly suggest that the knee pain is depending mainly on the subchondral bone changes, but not cartilage lesion.

In the present study, osteophytes were not associated with knee pain. The finding appears to contradict the data of previous cohort studies [49, 50, 51], which showed significant associations between them. However, relationship between knee pain and osteophyte is controversial. One study reported that MRI-detected osteophytes are not associated with knee pain [52] and a systematic review concluded that there was limited level of evidence for association between MRI-detected osteophytes and knee pain [18]. The discrepant results between the previous and present studies may be ascribed to the different evaluation method for osteophyte. In the present study, we used WORMS for osteophyte evaluation, in which osteophyte was semi-quantitively evaluated and defined to be present when osteophyte was the WORMS grade ≥ 1. On the other hand, the previous studies evaluated osteophyte quantitively and defined to be positive when they were ≥ 3 mm or ≥ 5 mm in width [49, 50, 51]. In addition, the associations between osteophytes and knee pain in a cohort study on older adults were reported to become non-significant after adjustment for BML [51]. Therefore, it is possible to speculate that osteophytes may cause OA symptoms such as knee pain, if any, through other structural changes [51]. In this context, we have provided the data suggesting that medial and anterior tibial osteophytes contribute to development of medial and anterior meniscus extrusion, respectively [31, 32, 53, 54]. However, the involvement and role of osteophytes in subchondral bone changes warrant further study.

BML is one of the treatment targets for knee OA. However, a clinical trial targeted to BML resulted in failure [55]. This may be, at least in part, due to the insufficient understanding of the pathophysiology of BML in knee OA without attention to BML combined with SBA and/or SBC, both of which are the irreversible subchondral bone changes [56]. Thus, therapeutic strategies to treat and/or reduce knee pain in knee OA patients may include development and application of new remedies targeting SBA and SBC, in addition to BML. However, deliberate and detailed analyses on the characteristics and developmental processes of the subchondral bone changes in knee OA are necessary.

Although individuals of our BHS cohort were not the subjects with higher risk of knee OA, the prevalence of knee OA was greater than that of previous large cohort studies from Japan [57] or USA [58]. The reason is not clear at present, but this may be related to the different methods for evaluation of knee OA by K/L grade. In the present and previous studies, the radiographic knee OA was evaluated by bilateral standing extended antero-posterior view. However, since evaluation of joint space narrowing is occasionally difficult due to unclear joint line by extended antero-posterior view, we estimated joint space narrowing by using 45° of flexion view, i.e. Rosenberg view, which is known to have high sensitivity of both radiographic change and clinical symptoms of knee OA compared to an antero-posterior view [59, 60, 61]. Thus, it is plausible to speculate that the relatively high prevalence on knee OA in the present study may be due to the evaluation method of the radiographic severity of knee OA, although further study is needed to confirm our speculation.

Limitations

The present study has some potential limitations. First, our study focused only on the relationships between knee pain and structural changes of knee OA, such as cartilage lesion, subchondral bone changes and synovitis. However, the determinants of pain are associated with multiple interactive pathways, such as psychological and social factors [62]. In addition, not only nociception but also central sensitization are suggested to affect worsening of chronic pain in OA patients [63, 64]. Therefore, these factors needed to be assessed by further study. Second, since this was a cross-sectional study using baseline data from the BHS, the association between a risk of OA progression and OA-related structural changes was unable to be assessed. Third, although larger BML is reportedly associated with greater knee pain [44], presence or absence of BML was only evaluated and the size of BML was not assessed in the present study. Fourth, the use of a 0.3-Tesla MRI scanner in the present study may affect image quality due to its inherent limitations in spatial resolution, contrast resolution, and signal-to-noise ratio (SNR) compared to higher field strength systems, such as 1.5T or 3T. Previous studies have shown that 0.3T MRI can reliably detect hand OA lesions, particularly BML, synovitis, and subchondral bone changes [65]. On the other hand, lower spatial resolution may limit the detection of low-grade cartilage defects [66]. To mitigate these limitations, we employed several image optimization techniques. The field of view was adjusted from the usual 160 mm to 200 mm to enhance SNR, although this inevitably resulted in a slight reduction in spatial resolution. Zero fill interpolation was applied by reconstructing pixel size from 0.98 mm to 0.39 mm, which helps smooth the images but may slightly reduce sharpness. Despite these adjustments, low-field MRI has advantages, such as reduced susceptibility artifacts and geometric distortion, making it particularly useful in assessing subchondral bone structures.

Conclusions

In conclusion, we have demonstrated that in elderlies from the large-scale cohort, knee pain is associated with BML which coexists with SBA and/or SBC, but not BML without the coexistence. Our study has also shown that the association is present regardless of the presence or absence of cartilage lesion above the BML region and is stronger in the loading site of the knee joint. These results suggest that knee pain is closely related to concentration of mechanical load to the subchondral bone during joint moving and is associated with subchondral bone changes rather than cartilage changes.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

BHS

Bunkyo Health Study

BMI

Body mass index

BML

Bone marrow lesion

CI

Confidence interval

K/L

Kellgren-Lawrence

MRI

Magnetic resonance imaging

OA

Osteoarthritis

OR

Odd ratio

PDFS

Proton density-weighted fat-suppressed

SBA

Subchondral bone attrition

SBC

Subchondral bone cyst

SNR

Signal-to-noise ratio

TFA

Tibiofemoral angle

VAS

Visual Analogue Scale

WORMS

Whole Organ Magnetic Resonance Imaging Score

Author contributions

JT, HK, AA, YN, YS, YT, and MI conceived and designed the study. JT, AA, TA, LL, YN, HA, SW, JS, YM, TY, TS, and KS collected and registered data of the subjects. HK, KY, and KK evaluated knee and full-limb radiographs. JT, AA, YN, HK, TA, LL, SH, MK, SN, TNK, YO, and MI had the major role in analysis and interpretation of the data, and contributed to drafting the report. SN supervised the statistical analysis. All authors have read and approved the final manuscript.

Funding

This study was supported in part by Grant-in-Aid for Scientific Research from the Japanese Society for the Promotion of Science (JSPS) to MI (15K10494, 18K09082, and 24K02572), HK (15K20019, 18K09083, and 21K09259), SH (16K20069 and 23K08705), TA (19K17209), JS (20K18075 and 23K15750), and YO (16H05454, 19H03788, and 22K07029). This study was also funded in part by a High Technology Research Center Grant and the Program for the Strategic Research Foundation at Private Universities (2014–2019) from the Ministry of Education, Culture, Sports, Science and Technology of Japan (MEXT) and by a Center of Innovation (COI) program, one of the main funding programs under the Center of Innovation Science and Technology based Radical Innovation and Entrepreneurship Program (COI STREAM), which was launched in 2013 by the MEXT and operated by Japan Science and Technology Agency (JST), to Juntendo University. This study was also supported in part by the Strategic International Collaborative Research Program (SICORP) from the Japan Agency for Medical Research and Development (AMED) and the NordForsk (the Academy of Finland, the Research Council of Norway, and the Swedish Research Council) and by a JOA (Japanese Orthopaedic Association)-Subsidized Science Project Research 2022–2024 to MI.

Data availability

Datasets and images used and/or analysed during the current study are available from the corresponding authors on reasonable request.

Declarations

Ethics approval and consent to participate

The study protocol of BHS has been approved by the ethics committee of Juntendo University from November in 2015 (No. 2015078, 2016138, 2016131, and 2017121). The study is conducted in accordance with the principles outlined in the Declaration of Helsinki. At the orientation meeting, participants were told that they have the right to withdraw from the trial at any time and, then, written informed consent was obtained from all the participants.

Consent for publication

This manuscript has not been published and is not under consideration for publication elsewhere.

Competing interests

The authors declare no competing interests.