DIFFERENCES IN SUBCHONDRAL BONE SIZE AFTER ONE YEAR IN OSTEOARTHRITIC AND HEALTHY KNEES

Abstract

OBJECTIVE

Increase of subchondral bone area (tAB) in OA has been reported, but it remains unclear if this is specific to OA. We investigated differences in knee tAB after one year in healthy subjects and in those with radiographic OA (rOA).

METHOD

MR images of 899 right knees from the OA Initiative were acquired at baseline and one year follow-up (year-1). Medial and lateral tibial cartilage (MT and LT) and weight-bearing femoral cartilage (cMF and cLF) were segmented and tAB computed. Subjects were stratified into: healthy controls, pre-rOA (K&L grades 0 and 1, with OA risk factors), established rOA (K&L grades 2–4), and independently with regards to joint space narrowing (without, with medial, lateral and bilateral JSN). Primary analysis tested if tAB was different between baseline and year-1 in rOA. Exploratory analyses investigated whether: 1) tAB changes differed between healthy controls and those with rOA; 2) tAB differences were greater in higher K&L grades; and 3) tAB was different between baseline and year-1 in JSN. Significance was set at p<0.0125.

RESULTS

Differences in tAB were found in rOA in MT, cMF and cLF (ranging from +0.2% to +0.4%; p<0.001), but not in healthy controls or pre-rOA. Rates of change did not differ between groups. Within the JSN groups differences of 0.2% to 0.4% were found in the femur (p<0.05).

CONCLUSION

We find that knee tABs differ in rOA between baseline and year-1, but the change was not greater than in healthy knees, and is restricted to the femur in JSN.

INTRODUCTION

Bone remodeling is an ongoing process in which bone adapts to its mechanical environment through addition, removal and rearrangement of bone mass. Importantly, bone remodeling is thought to play a role in disease progression of osteoarthritis (OA)^1–3. An increase of subchondral bone area (tAB) in knee OA has been suggested^4–15, but it remains unclear if this is specific to OA, or occurs independently of OA, possibly due to physiological aging¹⁶ or an altered biomechanical environment^17–19.

Several studies support the hypothesis that the increase in tAB in the course of OA is primarily associated with the principal pathological processes of OA such as osteophytosis, cartilage defects, and general cartilage thinning, represented by joint space narrowing (JSN). Cross-sectionally, an association between larger tibial tAB and osteophytosis of the medial tibia (MT) was reported in early OA⁴, and it was found that Kellgren and Lawrence (K&L) grade explains approximately 20% of medial tAB variance²⁰. A longitudinal study found an incremental increase in tAB for each grade of osteophytosis and JSN⁸. Other prospective evaluations have observed an association between both male sex and higher baseline medial JSN grade with an increase in medial tibial tAB in knee OA⁷. It has been suggested that the increase of tibial tAB may precede cartilage defects⁵ and cartilage swelling¹¹, and may be predictive for the incidence of OA^9,15.

Other studies have found an association between tAB increase in OA and mechanical factors. For example, Howell et al observed that in OA subjects with varus malalignment, the radius of the lateral femoral condyle (a surrogate measure of bone size) is higher than that of the medial condyle, whereas in valgus knees, the relation is reversed¹³. In similar subjects, the annual tAB increase has been reported to range from 0.1% to 0.3%, depending on the compartment with regards to alignment¹². Also, the size of medial tibial plateau bone area has been shown to positively correlate with the knee adduction moment in medial femorotibial OA¹⁴. Similarly, in 52 young subjects an increase in the medial and a decrease in the lateral tibial bone area were found seven years after anterior cruciate ligament reconstruction¹⁹.

Evidence exists that remodeling of tAB also occurs in subjects without OA, potentially as a result of physiological aging, or an adaptation to the mechanical environment. In 50–76 year old healthy females, an increase of approximately 1% per annum in tibial tAB has been reported, with a negative association between baseline tAB and age¹⁶. In comparison, a positive association between age and medial/lateral tibial tAB and patellar volume has been found in a cross-sectional convenience sample of 372 subjects²¹. In postmenopausal women, an increase of tAB over two years in the dominant knee has been shown to positively correlate with the prevalence of meniscal tears¹⁸. In elderly healthy women, the size of the medial, but not of the lateral tibial subchondral bone area has been shown to positively correlate with the knee adduction moment¹⁷.

In summary, it remains unclear whether increasing subchondral bone area is specific to OA, or occurs in association with biomechanical factors or age, independently of OA. The purpose of the current study was to examine differences in tAB between baseline and year-1 in healthy knees and in those with different grades of rOA, and to explore the correlation with JSN as a surrogate marker for malalignment and therefore localization of joint loads. Specifically, the primary hypothesis was:

• Knees with definite rOA show significant larger tABs over one year

Further exploratory analyses tested whether:

• The rate of change in tAB over one year is greater in knees with definite rOA than in healthy reference knees

• Differences in tAB over one year are greater in knees with higher grades of rOA

• Differences in tAB over one year are greater in compartments with JSN than in those without.

METHOD

OAI cohort and MRI sequences

Participants in the OAI (http://oai.epi-ucsf.org/datarelease/) were 45–79 years old (2804 females, 1992 males). General exclusion criteria were rheumatoid or inflammatory arthritis, bilateral end-stage knee OA, inability to walk without aids, and MRI contraindications. The healthy, “non-exposed” control subcohort (n=122) met the following inclusion criteria: 1) no pain, aching or stiffness in either knee in the past year; 2) no femorotibial rOA in either knee using the clinical site readings of the baseline fixed flexion radiographs²²; 3) no risk factors for OA, including obesity, knee injury, knee surgery, a family history of knee replacement in a biological parent or sibling, Heberden’s nodes, or repetitive knee bending. The pre-rOA and rOA cohorts were sampled from the combined “incidence” and “progression” subcohorts of the OAI, following self-defined inclusion criteria. Both groups could have risk factors like frequent knee symptoms, overweight (BMI <= 25), history of knee injury or surgery, family history of knee replacement, and hand OA. The pre-rOA subjects had no symptoms or definite signs of rOA in the target (right), but possibly in the contralateral knee, represented by K&L grades 0 and 1, while the rOA subjects had definite OA (K&L grades 2 to 4) in the target knee in addition to being symptomatic (pain, aching or stiffness on most days of a month in past year). Please refer to the OAI webpage for more detailed information: http://oai.epi-ucsf.org/datarelease/docs/StudyDesignProtocol.pdf.

The datasets for the current longitudinal analysis were drawn from a subsample of 1003 right knees, which were analyzed by a consortium of industry partners (see acknowledgments), the OAI Coordinating Center at University of California San Francisco (UCSF), and an image analysis company (Chondrometrics GmbH) as described previously²³. From this group, 96 datasets were excluded, as they were acquired with a double-echo steady-state MRI sequence, and not with a double oblique coronal FLASH water excitation sequence at 3Tesla according to a standardized protocol^24,25. One dataset was excluded because central rOA grading was missing, and 47 datasets were excluded because they were left knees (some overlap with the 96 from above), leading to a final sub sample of 899 right knees.

Radiographic grading and sample selection

The rOA grading was derived from the central X-ray readings provided by the OAI, which were performed at Boston University by two expert readers, who independently assessed each image, blinded to each other’s reading and to clinical data. Discrepancies between readers were adjudicated in a consensus session with blinded images viewed simultaneously and a third reader participating. The readers assessed each knee for K&L grades (Figure 1)²⁶, and other individual radiographic features^27,28, including medial and lateral JSN grades (Figure 2). Based on these readings, the analyzed sample was stratified into the following cohorts, based on OA status:

Figure 1.

Coronal spoiled gradient echo (FLASH) images of the right knee of four subjects from cohorts with different Kellgren and Lawrence (K&L) grades of radiographic OA: a) K&L grade 0, b) K&L grade 2, c) K&L grade 3, and d) K&L grade 4.

Figure 2.

Coronal spoiled gradient echo (FLASH) images of the right knee of four subjects from different cohorts with joint space narrowing (JSN): a) without JSN, b) with bilateral JSN grade 2, c) with medial unilateral JSN grade 3, and d) with lateral unilateral JSN grade 2.

• Healthy = non-exposed control cohort without rOA in either knee and without risk factors

• Pre-rOA = cohort without signs of rOA in the target knee (K&L grades 0 and 1), but with risk factors and/or rOA in the contralateral knee

• rOA = cohort with established rOA (K&L grades 2 to 4) in the investigated knee

  • KLG 2 = cohort with K&L grade 2 (the KLG cohorts are sub-cohorts of the rOA cohort)

  • KLG 3 = cohort with K&L grade 3

  • KLG 4 = cohort with K&L grade 4

For the exploratory analyses, stratification was also performed based on JSN rather than OA status:

• M0 / L0 = medial and lateral JSN grade 0 (including healthy, pre rOA and rOA cases)

• M0 / L1–3 = medial JSN grade 0 and lateral JSN grades 1 to 3

• M1–3 / L0 = medial JSN grades 1 to 3 and lateral JSN grade 0

• M1–3 / L1–3 = bicompartimental JSN, with medial and lateral JSN grades 1 to 3

Table 1 gives a detailed breakdown of the number of subjects in each of the strata, and the basic demographic variables.

TABLE 1.

Demographic variables for the study sample, the healthy group without risk factors (Healthy Control), the group with risk factors for knee osteoarthritis (pre-rOA), groups with different grades of radiographic osteoarthritis (rOA: KLG2 through KLG4), and independently groups with different configurations of joint space narrowing (without, medial, lateral, and bilateral).

	Females	Males	Age	Height[cm]	Weight[kg]	BMI
Total	539	360	61.6±9.5	167.7±9.0	81.5±16.1	28.9±4.8
Healthy Control	61	40	55.0±7.6	167.8±8.7	69.0±11.9	24.4±3.0
Pre-rOA	137	117	60.7±9.5	168.0±9.3	80.7±15.7	28.4±4.3
rOA	341	203	63.2±9.2	167.5±8.9	84.3±15.9	30.0±4.8
KLG2	220	96	62.4±9.1	166.7±8.5	83.5±16.0	30.1±4.8
KLG3	94	73	64.9±9.1	168.0±9.3	85.3±15.8	30.1±4.7
KLG4	27	34	62.7±9.5	171.0±9.0	85.4±16.1	29.0±4.6
M0 / L0	198	158	59.1±9.2	167.3±8.9	77.5±15.5	27.6±4.6
M0 / L1–3	68	36	63.8±9.5	168.2±9.8	82.6±16.9	29.1±4.8
M1–3 / L0	225	156	63.5±9.2	168.0±8.9	85.2±15.5	30.2±4.8
M1–3 / L1–3	3	10	62.4±10.8	169.6±9.8	88.9±18.8	30.0±3.9

Pre-rOA = Kellgren & Lawrence grades 1 and 2, rOA = Kellgren & Lawrence Grades 2 to 4, M0 = no medial joint space narrowing, M1–3 = medial joint space narrowing grade 1–3, L0 = no lateral joint space narrowing, L1–3 = lateral joint space narrowing grade 1–3

MRI analysis

Acquisition of MRI datasets, transfer logistics and segmentation were performed as described previously^29,30. The total area of subchondral bone (tAB), represented by the tidemark between calcified and non-calcified cartilage, was segmented manually in the medial (MT) and lateral tibiae (LT), and in the weight-bearing (central) part of the medial (cMF) and lateral femoral condyles (cLF). Osteophytes were excluded from tAB by taking into account the borders of the tibial and femoral condyles. The posterior aspects of the femoral condyles were not included, as segmentation in these areas is not supported by the coronal imaging protocol, and because the sensitivity to change in these regions has been shown to be unfavourable³¹. To minimize segmentation errors and deviations between readers, all segmentations were quality controlled by one expert (S.M.), and corrected if necessary. The segmentations were used to generate virtual 3D-reconstructions, from which the size of the tAB was computed by triangulation³². The test-retest precision of the method has been reported as good to excellent (precision error ranging from 0.7–1.7%) previously³³.

Statistical analysis

The primary analysis focused on differences in tAB size in the four femorotibial cartilage plates after one year in the cohort with rOA. The exploratory analyses were to test whether the changes in tAB size over one year were different between rOA and healthy knees, and to compare tAB differences between groups with pre-rOA, different K&L and JSN grades.

The mean change (MC) and standard deviation of change in tAB [mm²] between baseline and year-1 follow-up were determined for all cartilage plates of the femorotibial joint (MT, cMF, LT, cLF). Percent changes were derived by dividing the MC in a group by the mean tAB at baseline for the same cohort. The standardized response mean (SRM defined as mean change divided by the standard deviation of change) was used as a measure of sensitivity to change. Double-sided paired student’s t-test was employed to detect significant differences in tAB between baseline and year-1 follow-up and one-sided non-paired t-tests were used to check whether baseline values or rates of change differed between groups. Double-sided paired t-tests were used to test whether tAB differences were greater in compartments affected by JSN rather than in the non-JSN compartment. The required p-value was set to ≤0.0125 to correct for multiple testing of the four cartilage plates in each analysis at a global error level of 5%. Changes where single comparisons did not reach the global significance level but were p ≤0.05 were deemed trends.

RESULTS

Figure 3 gives an overview of absolute changes in tAB size for each analyzed cartilage plate in the various cohorts. Baseline values of tAB differed significantly between healthy controls, pre-rOA and rOA for all cartilage plates (unpaired t-test). Only the cMF tAB differed between pre-rOA and rOA.

Figure 3.

Graph showing the size of subchondral bone area (tAB) in the medial tibia (MT), lateral tibia (LT), central medial femur (cMF), and central lateral femur (cLF) at baseline (left side of each pair) and at one-year follow up (right side of each pair), for healthy controls, pre-rOA (Kellgren and Lawrence grades 0 and 1) and rOA (Kellgren and Lawrence grades 2 to 4) cohorts. Error bars indicate 95% confidence intervals, *** p<0.001, * p<0.05, please note that only p<0.0125 is considered significant for the primary analysis (Bonferroni correction).

tAB differences and changes in rOA

Between baseline and year-1, the entire rOA cohort displayed significant (p<0.001) differences in tAB of 0.2% in MT, 0.4% in cMF, and 0.2% in cLF. tAB of the LT did not significantly change (p=0.437) (Table 2). The rates of change were not different from those found in the healthy reference cohort.

TABLE 2.

Absolute size and change over one year of the bone-cartilage-interface area (tAB) for the entire study sample, a healthy group without risk factors (Healthy Control), a group the group with risk factors for knee osteoarthritis (pre-rOA), and groups with different grades of radiographic osteoarthritis (rOA: KLG2 through KLG4). p is given for a paired t-test between baseline and 1 year follow-up.

	Area[mm2]	Change[mm2]	Change[%]	95%CI[mm2]	SRM	p
Medial Tibia (MT)
Total	1204.6	0.8	0.1	0.85	0.06	0.062
Healthy Control	1114.3	−1.2	−0.1	2.38	−0.10	0.309
Pre-rOA	1212.8	−0.5	0.0	1.54	−0.04	0.490
rOA	1217.6	1.8	0.2	1.12	0.14	0.002
KLG2	1195.0	1.0	0.1	1.37	0.09	0.133
KLG3	1231.6	2.4	0.2	2.07	0.18	0.023
KLG4	1296.6	4.2	0.3	4.25	0.25	0.055
Central Medial Femur (cMF)
Total	585.3	1.7	0.3	0.65	0.17	<0.001
Healthy Control	532.0	1.2	0.2	2.11	0.11	0.279
Pre-rOA	576.4	0.9	0.2	1.15	0.10	0.099
rOA	599.3	2.1	0.4	0.85	0.21	<0.001
KLG2	589.3	1.7	0.3	1.02	0.18	0.001
KLG3	608.6	3.0	0.5	1.67	0.27	0.001
KLG4	626.2	1.9	0.3	3.03	0.16	0.223
Lateral Tibia (LT)
Total	1024.9	0.4	0.1	0.81	0.03	0.382
Healthy Control	964.0	−0.7	−0.1	2.36	−0.05	0.577
Pre-rOA	1036.6	0.6	0.1	1.44	0.05	0.391
rOA	1030.7	0.4	0.0	1.08	0.03	0.437
KLG2	1017.1	0.6	0.1	1.38	0.05	0.374
KLG3	1038.5	0.5	0.1	1.95	0.04	0.583
KLG4	1079.5	−0.9	−0.1	3.84	−0.06	0.638
Central Lateral Femur (cLF)
Total	633.7	0.8	0.1	0.59	0.09	0.006
Healthy Control	597.4	0.2	0.0	1.72	0.02	0.828
Pre-rOA	637.3	−0.1	0.0	1.07	−0.02	0.816
rOA	638.8	1.4	0.2	0.79	0.15	<0.001
KLG2	630.1	1.4	0.2	1.03	0.15	0.010
KLG3	645.9	1.3	0.2	1.53	0.13	0.103
KLG4	664.7	2.2	0.3	2.13	0.27	0.042

SEM = standard error of the mean, SRM = standardized response mean, Pre-ROA = Kellgren & Lawrence Grades 1 and 2, rOA = Kellgren & Lawrence Grades 2 to 4, KLG = Kellgren & Lawrence Grade

tAB differences and changes in KLG

The healthy reference cohort and the pre-rOA cohort (K&L grades 0 and 1) showed no significant differences in tAB in any of the cartilage plates between baseline and year-1. In the KLG2 sub-cohort differences were seen in cMF (0.3%, p=0.001) and a trend in cLF (0.2%, p=0.010), but not in the tibia. The KLG3 cohort showed tAB differences in cMF (0.5%, p=0.001), and a trend in MT (0.2%, p=0.023), but not in the lateral compartment (Table 2). In the KLG4 cohort trends were found in cMF (0.3%, p = 0.027) and cLF (0.3%, p=0.042). The SRMs for significant tAB changes in these cohorts ranged from 0.14 to 0.27 (Table 2). The rate of change was significantly different in MT between the pre-rOA and the rOA cohort (p=0.009), and a trend existed for MT between the healthy and the rOA cohort (p=0.016).

tAB differences and changes in JSN

The cohort without JSN (M0 / L0), including healthy, pre-rOA and rOA knees, showed significant (p=0.003) tAB differences of 0.3% in cMF between baseline and year-1, and a trend in LT (0.1%, p=0.047). In the cohort with lateral JSN (M0 / L1–3) a tAB difference of 0.4% was found in cLF (p=0.003), and a trend in cMF (0.3%, p=0.043). The cohort with medial JSN (M1–3 / L0) displayed larger femoral cartilage plates at year-1 (cMF: 0.3%, p=0.002 and cLF: 0.2%, p=0.011). In the cohort with bilateral JSN (M1–3 / L1–3) a tAB difference was found in cMF (1.0%, p=0.018), but not in the other cartilage plates. The SRMs for significant tAB changes between baseline and year-1 in unicompartimental JSN ranged from 0.13 to 0.30, being in the same range as for the different OA specific strata (Table 3). The rates of change in tAB in unicompartimental JSN were not significantly different between compartments with and without JSN.

TABLE 3.

Absolute size and rate of change over one year of the bone-cartilage-interface area (tAB) in a group without joint space narrowing (JSN, M0 / L0), groups with unilateral JSN (M1–3 / L0 and L1–3 / M0) and a group with bilateral JSN (M1–3 / L1–3).

	Area[mm2]	Change[mm2]	Change[%]	95%CI[mm2]	SRM	p
Medial Tibia (MT)
M0 / L0	1169.4	0.2	0.0	1.19	0.02	0.727
M0 / L1–3	1221.3	1.7	0.1	2.63	0.13	0.193
M1–3 / L0	1235.1	1.2	0.1	1.38	0.09	0.095
M1–3 / L1–3	1266.3	0.9	0.1	6.82	0.08	0.786
Central Medial Femur (cMF)
M0 / L0	559.7	1.4	0.3	0.92	0.15	0.003
M0 / L1–3	593.3	2.0	0.3	1.98	0.20	0.043
M1–3 / L0	608.6	1.7	0.3	1.05	0.16	0.002
M1–3 / L1–3	626.9	6.2	1.0	4.94	0.76	0.018
Lateral Tibia (LT)
M0 / L0	1003.4	1.2	0.1	1.14	0.01	0.047
M0 / L1–3	1061.1	1.9	0.2	2.63	0.14	0.162
M1–3 / L0	1034.0	−0.8	−0.1	1.29	−0.07	0.206
M1–3 / L1–3	1130.4	−1.3	−0.1	5.64	−0.14	0.633
Central Lateral Femur (cLF)
M0 / L0	617.9	0.1	0.0	0.89	0.01	0.873
M0 / L1–3	661.2	2.6	0.4	1.68	0.30	0.003
M1–3 / L0	641.4	1.2	0.2	0.92	0.13	0.011
M1–3 / L1–3	675.9	1.0	0.1	6.82	0.09	0.764

SEM = standard error of the mean, SRM = standardized response mean, M0 = no medial joint space narrowing, M1–3 = medial joint space narrowing grade 1 to 3, L0 = no lateral joint space narrowing, L1–3 = lateral joint space narrowing grade 1–3

DISCUSSION

The purpose of the current study was to test the primary hypothesis that knees with definite radiographic osteoarthritis (rOA) show significant differences in subchondral bone size (tAB) over one year. Exploratory analyses investigated whether the rate of change in tAB was greater in knees with definite rOA than in healthy reference knees, and compared differences in tAB after one year in various groups with different K&L grades and configurations of joint space narrowing (JSN). The knees with rOA showed significant tAB differences in the medial tibia (MT) and in the central femur (cMF and cLF). The tAB increase in rOA was not significantly different (greater) than in healthy reference knees. Differences in tAB over one year tended to be greater in knees with higher grades of rOA, but differences were not generally greater in compartments with JSN rather than those without.

A limitation of the study is that the differences/changes observed here are smaller, albeit only slightly smaller, than the precision errors that have been reported for MRI-based measures of tAB with this technique³³. However, the analysis was performed in a large cohort, and results were generally consistent between different rOA grades. Future studies with observation periods of more than one year may observe differences/changes that are greater than the reported precision errors. It is also important to be aware of the sample size differences between groups (Table 1); the p-values should not be compared between groups, as these become intrinsically smaller (i.e. reach higher levels of significance) with larger group sizes. The large differences in group sizes are due to the numbers enrolled in each group of the OAI (i.e. the healthy control group was deliberately chosen to be roughly 100 subjects), and factors intrinsic to OA (i.e. medial JSN is more common than lateral³⁴). However, to maximize study power, all eligible participants for each group were included. It should also be noted that the groups differ in age and BMI; the healthy control group is younger and has a lower BMI than other groups. We have shown previously that knee tAB does not differ between younger and older healthy subjects³⁵, which is supported by the lack of difference between baseline and year-1 in the healthy control group. It is possible that increased load due to high BMI is a driving factor behind changes in tAB size in knee OA (see also below), and could possibly also have been used as a surrogate marker for mechanical load, as JSN was in this study. However, both BMI and tAB²⁰ are strongly associated with body height, whereas JSN is probably not. Another limitation is that the association with standard biomechanical factors, such as knee alignment or quadriceps strength was not investigated, because these have not been reported for all of the investigated subjects in the OAI database. However, unicompartimental JSN may be an indicator of the side with greater mechanical stress, as varus and valgus alignments are known to increase load³⁶, cartilage loss³⁷, JSN and osteophyte growth³⁸ in the affected compartment. It is important to note that bias from osteophytes has been minimized in the current study, as the segmentation process explicitly excluded osteophytes from the tAB measurements. Although the change of tAB was not significantly different between the various K&L grades, the amounts and SRMs seem to generally increase with K&L grade in MT and cLF, and stay even in LT and cMF. These findings imply that the tAB differences are slightly more pronounced in MT and in higher grades of rOA, which has also been suggested by others^8,12,14.

A strength of this study is the inclusion of knees from the non-exposed healthy reference cohort free of symptoms femorotibial rOA, and without risk factors for knee OA. The healthy controls and the pre-rOA cohort (K&L grades 0 and 1) displayed no significant differences in tAB over one year, providing evidence that the differences measured in the other groups are associated with rOA, and not scanner drift or other systematic bias. The size of the tAB after one year was lower, albeit non-significant, in the healthy reference cohort in all cartilage plates, except for the medial femur, where a slight but non-significant higher tAB (0.2%) was found. This seems to contradict the reported association between age and bone-size in a mixed sample of people with a family history of knee replacement and a random population based selection²¹. The association remained significant, but decreased after adjusting for rOA status, supporting the hypothesis that a tAB increase might not be linked to rOA directly, but to accompanying or preluding factors like changes in biomechanics or age.

Similar to a previous cross-sectional study²³ the tAB size at baseline in the current study was significantly smaller (7–12%) in the healthy reference cohort than the pre-rOA or rOA cohorts. When considering that the yearly amount of increase of tAB ranges between 0.1% and 0.5%, it is possible that these groups experienced several years of small yearly increases in tAB, while shifting from healthy to pre-rOA or rOA. This would lead to the assumption that the tAB increases due to factors that do not immediately lead to rOA, but may facilitate the pathogenic process, which is supported by several studies suggesting that an increase in tAB is mainly associated with early rOA and may precede more severe pathologies^4,5,7. It is reasonable to think that mechanical loading is at least one of the factors influencing tAB, as some studies imply that mechanical loading correlates with tAB¹² or bone size in general¹³. The differences or changes in tAB or bone size in these studies were also very small but significant, which may be the pattern after which subchondral bone adapts (to the mechanical environment) in rOA, which makes it difficult to detect.

In the current study, the increase of tAB in rOA was predominantly found in the medial tibia and femur, which might represent the high prevalence of varus malalignment in rOA³⁹. However, the tAB increase was not confined to medial unicompartimental JSN (surrogate marker for varus malalignment), and was also found in cohorts with bicompartimental JSN or those without JSN. Only unicompartimental lateral JSN did not account for medial tAB increases, but instead accounted for an increase in the cLF. There is an intriguing possible explanation of the medial dominance of tAB increases in rOA: although it is reasonable that in varus malalignment loads are higher medially than laterally⁴⁰, it has been shown that medial loading is also higher in healthy controls⁴⁰, and even in valgus malalignment⁴¹. A possible connection between tAB increase (and osteophyte growth) on the one hand, and mechanical loading on the other hand, may be a mechanism by which the subchondral bone copes with increased mechanical loads⁴², probably through an increase of the load transferring area.

In conclusion, we find larger subchondral bone areas after one year in the medial tibia and the central femur of subjects with radiographic OA, but not in healthy controls or subjects with pre-radiographic OA. The differences in subchondral bone size tend to increase in the femur with higher grades of radiographic OA, but were not consistently associated with joint space narrowing. These results suggest that an increasing subchondral bone area is associated with radiographic OA in general, but not so much with OA grading and even less with joint space narrowing. These findings may prove useful for understanding the pathogenesis of osteoarthritis.