Abstract
Objective
In patients undergoing bariatric surgery or medical management for obesity, we assessed whether those experiencing substantial weight loss had changes in innervated knee structures or in cartilage.
Methods
Severely obese patients (BMI ≥35) with knee pain on most days were seen before bariatric surgery or medical weight management and at 1-year follow-up. Examinations included 3T MRI acquired at both time points for semi-quantitative scoring of bone marrow lesions (BML), synovitis, cartilage damage, and for quantitative measurement of cartilage thickness. Association of ≥20% vs. <20% weight loss with change in semi-quantitative scores was evaluated using linear mixed-effects models, and that with cartilage thickness change used non-parametric and parametric methods. Sensitivity analyses tested different thresholds for weight loss, weight loss as a continuous measure, examined those with and without bariatric surgery, and with worse osteoarthritis (OA).
Results
75 subjects (median age 49 years, 92% women) were included. At baseline, 61 subjects (81%) had Kellgren and Lawrence (KL) grade >0, and 16 (21%) had KL grade ≥3; 69 (92%) had cartilage damage. For BML, synovitis, and cartilage damage, the majority of knees had change in semi-quantitative scores of 0, and there was no difference between those with and without ≥20% weight loss. Similarly, in terms of cartilage thickness loss, in 14 of 16 sub-regions thickness loss was not associated with weight loss. Sensitivity analyses showed similar findings.
Conclusion
In middle-aged persons with mostly mild radiographic OA, structural features changed little over a year and weight loss was not associated with effects on structural changes.
Keywords: osteoarthritis, weight loss, obesity, cartilage, magnetic resonance imaging, knee joint
Introduction
Osteoarthritis (OA) is a leading cause of disability and a major public health problem in the United States. Nonetheless, no definitive intervention exists to diminish the development or progression of OA. Bariatric surgery in patients with severe obesity has been shown to provide substantial improvement in OA symptoms and also has been reported to reduce cartilage loss and have beneficial effects on cartilage composition (i.e. the dGEMRIC index)1. Several studies have shown that many persons with knee pain undergoing bariatric surgery experience a dramatic reduction in pain coincident with the marked and rapid weight loss2,3, a finding we recently confirmed4. The mean durable reduction in knee pain severity scores for all such persons has been approximately 50%, with some persons experiencing a remission of their pain and others not noting much change1,3. Bariatric surgery may constitute a model of an effective treatment for chronic knee pain and understanding the structural mechanisms underlying the marked and rapid reduction in knee pain might provide guidance as to how knee pain might be reduced by other treatments.
MRI-based studies suggested that two intra-articular structural features are associated with knee pain and its severity, synovitis and bone marrow lesions (BML)5. Further, studies have suggested that a reduction in synovial thickening or the volume of BML is associated with a reduction in knee pain severity6,7. While hyaline cartilage does not contain nociceptive fibers, its loss may be indirectly associated with pain by cartilage degradation products irritating the synovium, or by exposure of pain-sensitive subchondral bone in areas denuded of cartilage8.
To our knowledge, there have been no studies of persons with knee pain undergoing bariatric surgery to investigate if pain improvement was accompanied by a decrease in synovial thickening or BML size. Only one study has examined the association of weight loss from bariatric surgery with change in cartilage thickness1.
We assessed whether those experiencing substantial weight loss had changes in innervated knee structures or in cartilage. To assess whether those experiencing substantial weight loss had changes in structures linked with knee pain and in cartilage, we carried out a longitudinal study of persons with obesity and knee pain in a weight loss program, some of whom underwent bariatric surgery.
Methods
Subjects were recruited sequentially and by convenience from the Nutrition and Weight Management Center at Boston Medical Center in Boston, Massachusetts. Both surgery and medical management groups had to provide informed consent and meet the body mass index (BMI) eligibility criteria for bariatric surgery of ≥ 35 kg/m2 with a serious comorbidity including type 2 diabetes, sleep apnea, or hypertension at baseline to be included in our study. Subjects also had to have knee pain on most days of the past month and be 25-60 years old. Patients with prior knee surgery or inflammatory arthritis were excluded. Subjects also had to be eligible to obtain an MRI.
Subjects undergoing surgery had either laparoscopic roux-en-y-gastric bypass or laparoscopic sleeve gastrectomy. Persons in the medical management group were given dietary and exercise prescriptions with or without a combination of medications including phentermine, lorcaserin, phentermine/topiramate, bupropion/naltrexone, or liraglutide. The dietary prescription consisted of a high-protein, low-fat diet of 1200-1500 kilocalories/day for women and 1500-1800 kilocalories/day for men with or without the use of meal replacements used to substitute one or two meals per day. Both groups were advised to walk at least 30 minutes/day and perform resistance exercise at least twice a week. Assessments occurred at baseline which, for those undergoing surgery, was up to two weeks prior to their surgery, and approximately 12 months after the baseline examination for all subjects.
Measurements
Knee pain was assessed using the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC)9. Based on Angst et al10, a reduction of ≥18% was considered the minimal clinically important difference (MCID) for the WOMAC pain subscale. All subjects obtained a PA view of knees using a Synaflexer frame at baseline, and these were read for the purpose of Kellgren and Lawrence grading by an experienced rheumatologist reader (DTF), a Framingham Study and Osteoarthritis Initiative (OAI) Study radiograph reader.
All subjects obtained an MRI of their more painful knee at baseline and follow-up visits. If both knees were equally painful, we selected the knee with more advanced radiographic findings. MRIs were acquired using a Philips Achieva 3.0T X Series MRI (Philips Medical Systems, Best, The Netherlands). The coil used was an 8-channel PMS SENSE Hi Res Knee MRI Coil – Philips Achieva 3.0T (or Cardiac/torso coil for knees that did not fit into the knee coil), with a protocol involving a sagittal water selective cartilage scan (WATSc) (TR = 20.0 ms, TE = 5.15 ms, Flip angle = 15 degrees, FOV= 160 mm, slice thickness = 1.5 mm, resolution= 512×512 pixels). The WATSc sequence is a high-resolution sequence used for cartilage morphometric and thickness assessment. We also obtained spectral attenuated inversion recovery (SPAIR) (3D Fat saturated sequences) (TR = 1800 ms, TE = 30 ms, Flip angle = 90 degrees, FOV = 160 mm, slice thickness = 1.0 mm, resolution = 384×384 pixels). The SPAIR sequence was used for MRI semi-quantitative morphological grading of BML, synovitis, and cartilage.
MRIs for baseline and follow-up images were read paired using Osteoarthritis Knee Scores (MOAKS) for BML, synovitis, and cartilage morphology by AG, an experienced musculoskeletal radiologist who was a primary knee MRI reader for the Framingham OA Study and for the Multicenter Osteoarthritis (MOST), and OAI studies. Detailed descriptions of the MOAKS protocol including scoring BML, synovitis and cartilage morphology have been published11. Cartilage scoring consists of two scores, one for the percentage of the subregion area eroded and a second for the maximal depth of the lesion (e.g. 1.1 would signify a lesion encompassing 1-10% area of subchondral bone with 1-10% full thickness loss). To calculate the worst (i.e. maximum) and sum scores, the scoring system for cartilage morphology was re-coded from 0 to 9. Because the depth of cartilage erosions to bone has been associated with knee pain irrespective of the area of these lesions12,13 and because we wanted to use an ordinal scale, we re-coded the MOAKS score, prioritizing depth over area as follow: normal with score = “0.0” to “0”, 1-10% area of subchondral bone with no full thickness loss with score = “1.0” to “1”, 1-10% area of subchondral bone with 1-10% full thickness loss with score = “1.1” to “2”, 10-75% area of subchondral bone with no full thickness loss with score = “2.0” to “3”, and so forth. The >75% total area of subchondral bone with >75% full thickness loss with score = “3.3” corresponded to “9”.
We also assessed cartilage thickness using Chondrometrics Works software with the medial tibial, medial weight-bearing femoral, lateral tibial, and lateral weight-bearing femoral cartilage being segmented manually14. Baseline and follow-up images were read as pairs by seven readers with about a decade or more years of experience in cartilage segmentation, with blinding the reader to the order of acquisition. All segmentations were quality controlled by one expert reader (FE). Cartilage thickness was then determined by the software in sixteen sub-regions in medial and lateral femorotibial compartment that included the central (ccMF), external (ecMF), and internal (icMF) sub-regions of medial femoral cartilage plate, and the central (cMT), external (eMT), internal (iMT), anterior (aMT), and posterior (pMT) sub-regions of medial tibial cartilage plate, and similarly was read in the comparable plates in the lateral compartment that included ccLF, ecLF, icLF, cLT, eLT, iLT, aLT, and pLT for lateral femoral and tibial cartilage plates. Cartilage loss was quantified by assessing mean cartilage thickness over a total area of subchondral bone, and was computed as difference in thickness for compartments, plates, and sub-regions. To evaluate change in cartilage thickness, we used both region-based and also an extended ordered values (eOV) approach, which is location-independent15,16. This assesses the ordinal (i.e. rank-based) correlation of cartilage thickness loss at a particular sub-region (for region-based) or a particular rank (for ordered values).
The change in cartilage thickness was calculated relative to baseline as [(follow-up -baseline) / baseline]. Weight change across visits was calculated similarly. Thus, a positive value indicated a gain in cartilage thickness or weight, whereas a negative value indicated cartilage thickness loss or weight loss. All readers (DTF, AG, FE) were blinded to treatment received and weight loss experienced.
Statistical analysis
Participant characteristics at baseline were stratified by the percentage of weight loss at 1-year follow-up visit. We selected a threshold of ≥20% to represent substantial weight loss as there has been no widely-used threshold to compare against, and because this captured roughly half of our subjects, and we felt that it was substantial enough to reveal effects of weight loss on structural findings. Further, it is a threshold of weight loss that is achieved by most persons undergoing bariatric surgery and is rare in those on medical management. Thus, the study population consisted of participants who lost ≥20% bodyweight by the 1-year follow-up and the comparator cohort of those who lost <20% (or gained) bodyweight by the follow-up visit.
We also carried out sensitivity analyses as follows: testing ≥10% weight loss; testing absolute kilograms of weight loss; examining those with and without bariatric surgery; examining percent weight loss as a continuous variable; and lastly limiting analysis to knees with radiographic OA defined as knees with Kellgren and Lawrence grade ≥2.
To assess the effect of weight loss on BML, synovitis, and cartilage damage, we developed linear mixed-effects models17. Mean of the worst (i.e. maximum) MOAKS scores as well as the scores' sum across all sub-regions for BML, synovitis, and cartilage morphology were compared between the group of patients who lost ≥20% body weight by 1-year follow-up and those who did not. The models included random-effects for patients to appropriately adjust for the correlation between observations at the baseline and follow-up visits on the same individuals.
To assess the effect of weight loss on cartilage thickness, we carried out both a non-parametric rank-based correlation test and also parametric regression modeling. We assessed the ordinal correlation of cartilage thickness loss at a given sub-region (for region-based) or a given rank (for ordered values) with the percentage of weight change across baseline and follow-up visits by the non-parametric Kendall's tau statistic. Extended ordered values were created based on the sixteen femorotibial sub-regions, where the highest rank (i.e. eOV 1) consisted of the sub-region with the largest thickness loss (i.e. most negative value) or the smallest gain (i.e. least positive value). The lowest ordered value (i.e. eOV 16) was the sub-region with the smallest thickness loss (i.e. least negative value) or the largest thickness gain (i.e. most positive value). A high value for Kendall's tau coefficient suggests high similarity in the rank of two sets of ordered observations, i.e. change in cartilage thickness and weight change percentage.
Additionally, we separately calculated the sum of “only negative” and of “only positive” extended ordered values to define thinning and thickening sum scores, respectively16. Finally, the absolute values of thinning and thickening scores were summed to calculate the total change score as described in Eckstein et al16. In addition to assessing Kendall's tau non-parametric correlation between cartilage thickness change and percentage of weight change, we fitted linear regression models to these sum scores to assess their association with weight change percentage.
Analyses of cartilage thickness to this point were sub-region specific and therefore, each of them used limited data. To incorporate data on all sub-regions and better evaluate the relation of weight change percentage with cartilage thickness change, we carried out an analysis we shall refer to as global. The global model consisted of all sub-regions in a single linear mixed-effects model that included nested and crossed random-effects to account for spatial and repeated measurements correlations17. We specified nested random intercepts for sub-regions within a knee and a crossed random intercept for visits to quantify the association of cartilage thickness across all sixteen medial and lateral sub-regions and across both visits with weight loss of ≥20%. In addition to comparing the group with ≥20% weight loss vs. the group without it, we carried out sensitivity analyses listed above.
Finally, to assess whether the pain reduction experienced was correlated with change in any of the structural parameters, we calculated a Kendall's tau statistic between changes in worst score for BML, synovitis, cartilage damage, and also the cartilage thickness loss total change score with change in WOMAC pain score across visits.
Results
Baseline data
Characteristics of the study participants are presented in Table 1. Of 87 original participants at the baseline visit, 75 obtained follow-up MRI and completed the study. Twelve participants who did not complete the study had a median age of 44 years, 91.7% (all except 1) were women, 50.0% were white, and all were from the medically-managed weight loss group. The reasons for failure to participate in the follow-up included 1 with total knee replacement in the index knee, 7 who were lost to contact, and 4 who refused to participate in the follow-up visit. Participants' median age at baseline was 49 years, 92.0% (69/75) were women, 52.0% (39/75) had Kellgren and Lawrence grade ≥2 in their most painful knee, but only 22.7% (17/75) had Kellgren and Lawrence grades ≥3. The median weight loss in the group with ≥20% weight loss was -34.9 kg (min loss = -22.2 kg, max loss = -56.5 kg), whereas for the group with <20% weight change, median change was -6.2 kg (max loss = -25.5 kg, max gain = 22.2 kg; 5 persons in the medical management group gained weight). Of those who lost ≥20% body weight by follow-up, 97.5% (39/40) underwent bariatric surgery vs. 22.9% (8/35) who lost <20% (Table 1). An MCID reduction in WOMAC pain subscale occurred in 75.0% (30/40) of patients who lost ≥20% weight by follow-up, compared to 34.3% (12/35) with < 20% weight loss by follow-up. Baseline BML, synovitis, and cartilage damage were present in 64.0% (48/75), 56.0% (42/75), and 92.0% (69/75) of patients, respectively (Table 1). In those who eventually lost ≥20% body weight by 1-year follow-up, 60.0% (24/40), 60.0% (24/40), and 85.0% (34/40) had BML, synovitis, and cartilage damage at baseline, respectively. In the group of patients who had <20% weight loss by follow-up, 68.6% (24/35), 51.4% (18/35), and 100.0% (35/35) had BML, synovitis, and cartilage damage at baseline. Participants who eventually lost ≥20% weight by the follow-up, compared to those who did not, had a slightly thicker cartilage in most tibiofemoral compartment sub-regions at baseline (Table 1).
Table 1.
Participant characteristics at baseline visit grouped by percent weight loss at 1-year follow-up visit.
| Characteristic at Baseline | <20% Weight Loss at 1-Year N = 35 | ≥20% Weight Loss at 1-Year N = 40 |
|---|---|---|
| Age [mean (SD), median] | 47.3 (8.3), 49.0 | 42.5 (9.6), 49.0 |
| Female sex | 30 (85.7) | 39 (97.5) |
| African-American | 27 (77.1) | 19 (47.5) |
| Bariatric surgery | 8 (22.9) | 39 (97.5) |
| BMI [mean (SD), median] | 40.9 (4.5), 40.0 | 42.3 (4.5), 41.6 |
| College/graduate education | 15 (42.9) | 14 (35.0) |
| Employed | 9 (25.7) | 9 (22.5) |
| WOMAC pain (0-24 scale) [mean (SD), median] | 12.5 (4.9), 13.0 | 11.8 (4.2), 12.0 |
| Kellgren-Lawrence grade (%) | ||
| 0 | 6 (17.1) | 8 (20.0) |
| 1 | 11 (31.4) | 11 (27.5) |
| 2 | 8 (22.9) | 14 (35.0) |
| 3 | 9 (25.7) | 7 (17.5) |
| 4 | 1 (0.0) | 0 (0.0) |
| BML | 24 (68.6) | 24 (60.0) |
| Synovitis | 18 (51.4) | 24 (60.0) |
| Semi-quantitatively measured baseline cartilage damage (%) | ||
| No partial or full thickness loss | 0 (0.0) | 6 (15.0) |
| Any partial thickness loss | 32 (91.4) | 32 (80.0) |
| Any full thickness loss | 24 (68.6) | 23 (57.5) |
| Quantitatively measured baseline cartilage thickness at sub-regions[mean (SD), median] | ||
| ccMF | 1.971 (0.548), 1.968 | 2.033 (0.465), 2.000 |
| ecMF | 1.433 (0.374), 1.375 | 1.418 (0.318), 1.403 |
| icMF | 1.840 (0.402), 1.844 | 1.937 (0.350), 1.875 |
| cMT | 1.930 (0.489), 1.943 | 2.100 (0.363), 2.048 |
| eMT | 1.375 (0.351), 1.341 | 1.442 (0.290), 1.447 |
| iMT | 1.940 (0.535), 1.812 | 2.019 (0.450), 1.893 |
| aMT | 1.480 (0.325), 1.509 | 1.522 (0.231), 1.524 |
| pMT | 1.355 (0.232), 1.353 | 1.475 (0.232), 1.438 |
| ccLF | 2.154 (0.581), 2.145 | 2.222 (0.525), 2.112 |
| ecLF | 1.699 (0.438), 1.662 | 1.704 (0.377), 1.631 |
| icLF | 1.705 (0.479), 1.643 | 1.799 (0.459), 1.670 |
| cLT | 2.857 (0.960), 2.933 | 2.815 (0.651), 2.929 |
| eLT | 1.588 (0.437), 1.577 | 1.613 (0.396), 1.621 |
| iLT | 1.929 (0.484), 1.982 | 1.928 (0.395), 1.981 |
| aLT | 1.618 (0.413), 1.607 | 1.591 (0.284), 1.626 |
| pLT | 1.716 (0.491), 1.745 | 1.710 (0.303), 1.700 |
BMI - body mass index, BML - bone marrow lesion, SD - standard deviation, xcMF - central, external, or internal sub-regions of medial femoral cartilage plate, xMT - central, external, internal, anterior, or posterior sub-regions of medial tibial cartilage plate, xcLF - central, external, or internal sub-regions of lateral femoral cartilage plate, xLT - central, external, internal, anterior, or posterior sub-regions of lateral tibial cartilage plate, WOMAC - Western Ontario and McMaster Universities Osteoarthritis Index.
Longitudinal change in BML, synovitis, and cartilage damage
There were no significant changes in study knees from baseline to follow-up in BML score (difference in score for largest BML = 0.04; 95% confidence interval [CI]: -0.12, 0.20), synovitis (difference = -0.03, 95% CI: -0.15, 0.09), or cartilage damage (difference = 0.04, 95% CI: -0.07, 0.15) (Table 2). Further, change was not different between those who lost <20% vs. those who lost ≥20% of their weight (Table 2). Similar null results were seen when we examined the sum scores for BML, synovitis and cartilage morphology across all knee sub-regions (Table 2) and when we limited analyses to knees with OA.
Table 2.
Longitudinal changes in structural features of knees from baseline to follow-up measured by semi-quantitative magnetic resonance imaging Osteoarthritis Knee Score (MOAKS) grouped by weight loss at 1-year follow-up visit.
| Structural Feature | <20% Weight Loss at 1-Year1 [Mean (SD), Median] | ≥20% Weight Loss at 1-Year [Mean (SD), Median] | Coefficient for Mean Difference in Scores (95%CI) | |||
|---|---|---|---|---|---|---|
| Baseline | Follow-up | Baseline | Follow-up | Across Visits | Between Weight Loss Groups (<20% and ≥20%) | |
| Bone marrow lesion | ||||||
| Worst2 score | 1.14 (0.94), 1.00 | 1.26 (1.09), 2.00 | 1.10 (1.02), 1.00 | 1.15 (1.10), 1.00 | 0.04 (-0.12, 0.20) | 0.04 (-0.41, 0.50) |
| Sum3 of scores | 2.63 (2.84), 2.00 | 2.64 (2.88), 2.00 | 2.28 (2.47), 1.50 | 2.06 (2.30), 2.00 | -0.19 (-0.64, 0.25) | -0.16 (-1.27, 0.95) |
| Synovitis | ||||||
| Worst score | 0.89 (0.99), 1.00 | 0.85 (1.05), 0.50 | 0.79 (0.73), 1.00 | 0.80 (0.69), 1.00 | -0.03 (-0.15, 0.09) | -0.01 (-0.39, 0.38) |
| Sum of scores | 1.23 (1.44), 1.00 | 1.09 (1.38), 0.00 | 1.05 (1.22), 1.00 | 1.05 (0.99), 1.00 | -0.09 (-0.26, 0.08) | -0.02 (-0.58, 0.54) |
| Cartilage morphology | ||||||
| Worst score | 4.74 (1.93), 4.00 | 4.79 (1.82), 5.00 | 4.00 (2.28), 4.00 | 4.20 (2.31), 4.50 | 0.04 (-0.07, 0.15) | -0.37 (-1.33, 0.59) |
| Sum of scores | 16.83 (15.20), 12.00 | 16.11 (14.37), 13.00 | 16.13 (13.25), 13.50 | 16.88 (13.74), 14.00 | -0.84 (-2.51, 0.82) | 3.19 (-2.60, 8.97) |
Worst score is the maximum score across all sub-regions.
The scores sum is calculated using scores across all sub-regions.
CI - confidence interval, SD - standard deviation. The confidence intervals that included the null value of zero suggested lack of statistical significance.
When percentage weight change was treated as a continuous measure, its effects on the change in worst score across visits for BML (coefficient for a unit difference [coef] = 0.01, 95% CI: -0.01, 0.02), synovitis (coef = 0.01, 95% CI: 0.00, 0.02), and cartilage damage (coef = 0.01, 95% CI: 0.00, 0.01) were not statistically significant (Table 2). Similarly, no statistically significant differences were observed in the effect of weight change percentage on the change in sum of the scores across visits for BML, synovitis, and cartilage morphology (Table 2).
Longitudinal change in cartilage thickness
Of sub-regions in the knee, there were no significant changes in thickness noted. Further, we found little if any correlation between cartilage thickness change and weight change percentage, (Table 3), for both region-based and location-independent approaches. The only predefined sub-regions that showed a statistically significant positive correlation with weight change percentage were the anterior medial and lateral tibial cartilage sub-regions (aMT and aLT). As shown in Fig. 1, more weight loss was correlated with a relative thinning at anterior medial tibial (aMT) and anterior lateral tibial (aLT) sub-regions. The rank-based correlation for the thickening score also suggested a weak positive correlation, in which less weight loss correlated with more thickening) (Table 3). Linear regression models of these ordered values-based sum scores also suggested no significant association between the percentage of weight change and the thinning (coeff = 0.01, 95% CI: -0.01, 0.04), thickening (coeff = 0.02, 95% CI: 0.00, 0.04), or total change sum scores (coeff = 0.00, 95% CI: -0.03, 0.02).
Table 3.
Non-parametric rank-based association between the region-based and location-independent of the quantitatively-measured cartilage thickness change and the percentage of weight change from baseline to follow-up visits.
| Approach | Sub-Region/Extended Ordered Value | Kendall's Tau Coefficient | 95% Bootstrap CI |
|---|---|---|---|
| Region-Based | |||
| Δ ccMF | -0.041 | (-0.214, 0.136) | |
| Δ ecMF | -0.088 | (-0.255, 0.085) | |
| Δ icMF | -0.070 | (-0.228, 0.087) | |
| Δ cMT | 0.147 | (-0.009, 0.293) | |
| Δ eMT | 0.009 | (-0.180, 0.191) | |
| Δ iMT | 0.062 | (-0.090, 0.206) | |
| Δ aMT | 0.204 | (0.057, 0.354) * | |
| Δ pMT | 0.011 | (-0.158, 0.165) | |
| Δ ccLF | 0.134 | (-0.062, 0.317) | |
| Δ ecLF | 0.142 | (-0.031, 0.314) | |
| Δ icLF | 0.049 | (-0.140, 0.239) | |
| Δ cLT | 0.140 | (-0.066, 0.324) | |
| Δ eLT | 0.165 | (-0.009, 0.337) | |
| Δ iLT | 0.083 | (-0.091, 0.250) | |
| Δ aLT | 0.200 | (0.035, 0.362) * | |
| Δ pLT | -0.017 | (-0.173, 0.163) | |
| Location-Independent | |||
| eOV 1 | -0.022 | (-0.206, 0.150) | |
| eOV 2 | 0.043 | (-0.130, 0.229) | |
| eOV 3 | 0.082 | (-0.098, 0.260) | |
| eOV 4 | 0.129 | (-0.048, 0.298) | |
| eOV 5 | 0.125 | (-0.066, 0.286) | |
| eOV 6 | 0.155 | (-0.012, 0.338) | |
| eOV 7 | 0.138 | (-0.029, 0.301) | |
| eOV 8 | 0.134 | (-0.030, 0.289) | |
| eOV 9 | 0.120 | (-0.039, 0.271) | |
| eOV 10 | 0.124 | (-0.022, 0.269) | |
| eOV 11 | 0.108 | (-0.060, 0.246) | |
| eOV 12 | 0.127 | (-0.047, 0.296) | |
| eOV 13 | 0.136 | (-0.029, 0.308) | |
| eOV 14 | 0.132 | (-0.046, 0.318) | |
| eOV 15 | 0.120 | (-0.062, 0.296) | |
| eOV 16 | 0.066 | (-0.135, 0.244) | |
| Summary Scores | |||
| Thinning Score | 0.090 | (-0.075, 0.262) | |
| Thickening Score | 0.165 | (0.008, 0.340) * | |
| Total Change Score | -0.047 | (-0.211, 0.121) |
Confidence intervals that excluded the null value of 0 suggested statistical significance.
CI – confidence interval, eOV - extended (i.e. medial and lateral) ordered value, Δ xcMF -change in quantitatively-measured cartilage thickness in central, external, or internal sub-regions of medial femoral cartilage plate, Δ xMT - change in quantitatively-measured cartilage thickness in central, external, internal, anterior, or posterior sub-regions of medial tibial cartilage plate, Δ xcLF - change in quantitatively-measured cartilage thickness in central, external, or internal sub-regions of lateral femoral cartilage plate, Δ xLT - change in quantitatively-measured cartilage thickness in central, external, internal, anterior, or posterior sub-regions of lateral tibial cartilage plate.
Fig. 1.
Scatterplot of the anterior medial tibial (aMT), anterior lateral tibial (aLT), and central lateral tibial (cLT) cartilage thickness change from baseline to follow-up against percentage of weight change.
Dots above the horizontal line on the vertical axis at 0 represents cartilage thickness gain, whereas those below 0 represents cartilage thickness loss across visits. The figure suggests that bigger weight loss results in an increased (i.e. more negative) cartilage loss in those who experienced thickness loss or a reduced (i.e. less positive) cartilage gain in those who experienced cartilage gain.
We examined cartilage thickness change in each sub-region and its relation to weight loss (Table 4) on both a continuous (i.e. per 1% change in weight) and categorical (i.e. ≥ 20% weight loss at 1-year vs < 20%) scales. The sub-regions that showed statistically significant positive associations with weight change on a continuous scale included the anterior medial tibial cartilage (aMT) and central lateral tibial cartilage (cLT) (Table 4, Fig. 1) such that greater weight loss resulted in an increased cartilage loss in those who experienced thickness loss or a reduced cartilage gain in those who experienced cartilage gain.
Table 4.
Association of weight change on a continuous and categorical (≥20% vs <20%) scales with quantitatively-measured cartilage thickness change across baseline and 1-year follow-up visits.
| Thickness Change Sub-Region | Coefficient (95% CI) | |
|---|---|---|
| A Unit Change (1%) in Weight | ≥20% Weight Loss at 1-Year (Reference Level: <20%) | |
| Δ ccMF | -0.0014 (-0.0044, 0.0016) | 0.0277 (-0.0610, 0.1164) |
| Δ ecMF | -0.0016 (-0.0038, 0.0006) | 0.0293 (-0.0371, 0.0957) |
| Δ icMF | -0.0008 (-0.0029, 0.0013) | 0.0011 (-0.0612, 0.0635) |
| Δ cMT | 0.0029 (-0.0001, 0.0059) | -0.0332 (-0.1236, 0.0572) |
| Δ eMT | 0.0012 (-0.0015, 0.0038) | -0.0287 (-0.1080, 0.0506) |
| Δ iMT | 0.0004 (-0.0025, 0.0033) | 0.0324 (-0.0538, 0.1185) |
| Δ aMT | 0.0031 (0.0010, 0.0053) * | -0.0789 (-0.1434, -0.0144) * |
| Δ pMT | -0.0004 (-0.0022, 0.0013) | 0.0054 (-0.0464, 0.0572) |
| Δ ccLF | 0.0013 (-0.0009, 0.0036) | -0.0098 (-0.0784, 0.0588) |
| Δ ecLF | 0.0011 (-0.0015, 0.0038) | -0.0121 (-0.0916, 0.0674) |
| Δ icLF | 0.0003 (-0.0016, 0.0023) | 0.0335 (-0.0255, 0.0924) |
| Δ cLT | 0.0052 (0.0017, 0.0088) * | -0.0808 (-0.1943, 0.0328) |
| Δ eLT | 0.0023 (0.0002, 0.0045) | -0.0467 (-0.1142, 0.0209) |
| Δ iLT | 0.0008 (-0.0014, 0.0031) | 0.0120 (-0.0561, 0.0800) |
| Δ aLT | 0.0019 (0.0000, 0.0037) | -0.0459 (-0.1028, 0.0110) |
| Δ pLT | -0.0005 (-0.0026, 0.0017) | 0.0455 (-0.0189, 0.1100) |
Confidence interval excluded the null value, i.e. zero, indicating statistical significance at p<.0.05.
CI - confidence interval, Δ xcMF - change in quantitatively-measured cartilage thickness in central, external, or internal sub-regions of medial femoral cartilage plate, Δ xMT - change in quantitatively-measured cartilage thickness in central, external, internal, anterior, or posterior sub-regions of medial tibial cartilage plate, Δ xcLF - change in quantitatively-measured cartilage thickness in central, external, or internal sub-regions of lateral femoral cartilage plate, Δ xLT - change in quantitatively-measured cartilage thickness in central, external, internal, anterior, or posterior sub-regions of lateral tibial cartilage plate.
Finally, in the global model including all sub-regions in one analysis, we found no significant associations between cartilage thickness change in patients who lost ≥20% bodyweight at 1-year follow-up, compared to those who did not (difference = 0.02, 95% CI: -0.14, 0.16).
In sensitivity analyses examining ≥10% weight loss (rather than ≥20%) and where we tested absolute weight loss (rather than percentage weight loss) on a continuous scale, results were the same as those presented (not shown). Further sensitivity analyses comparing surgically-vs. medically-treated patients and analyses where the study sample was confined to those with radiographic OA at baseline, similarly showed null findings.
While the majority of those with weight loss experienced a marked reduction in knee pain, testing suggested no significant correlation between change in WOMAC pain and change in worst score for BML (Kendall's tau = -0.06, p = 0.50), synovitis (Kendall's tau = 0.09, p = 0.35), cartilage damage (Kendall's tau = 0.10, p = 0.33), or total change score in cartilage thickness loss (Kendall's tau = -0.12, p = 0.15).
Discussion
Our study showed that neither the anatomic structures in the knee that are innervated and possible sources of pain in OA (i.e. BML and synovitis) nor cartilage changed significantly at a time when subjects were experiencing both weight loss and knee pain reduction. Moreover, except for the anterior medial tibial (aMT), anterior lateral tibial (aLT), and central lateral tibial (cLT) cartilage sub-regions, thickness loss or gain measured quantitatively were not associated with weight loss of ≥20% or with percent weight change when measured on a continuous scale.
In general, we found little effect of weight loss on structural features of OA. One explanation could be that our sample size was small and that our failure to find an effect of weight loss on BML, synovitis, cartilage damage, and cartilage thickness loss represents a type 2 error. We believe this is not the case as the majority of knees in both weight loss groups (i.e. ≥20% vs <20%) showed no change at all in structural findings. Further, the confidence bounds around our estimates were narrow, so that the mean change in BML could not have exceeded the upper 95% confidence bound of 0.2 (for a 0-3 score); for synovitis, it could not have exceeded 0.09 (0-3 score), and for cartilage damage, it could not have exceeded 0.15 (a 0-9 score). For cartilage thickness measures, 95% confidence bounds were even narrower given the nearly universal absence of change over a 1-year period. Further, we found the usual marked reduction in knee pain after massive weight loss. Even when we limited analyses to knees with radiographic OA, the amount of change seen was minimal. We had too few subjects with grade 3 disease to examine this subset.
While we report that weight loss led to a slight thinning of cartilage in two sub-regions, the aMT and aLT, these are not major sub-regions affected by disease. We are uncertain of the conclusiveness of these findings. Some may suggest that these represent positive findings, but we doubt it. Many sub-regions were examined and many analyses performed, making it possible, as expected, given multiple testing to find positive results by chance. We note that in early OA, cartilage swelling and thickening may be the first pathologic changes of disease18,19 and it is possible that these sub-regions were swollen or edematous before the weight loss occurred and that the thinning represents a return to a nonedematous state.
Our finding on the relationship between weight loss and BML is consistent with a previous study by Gudbergsen et al20 that reported that changes in BML score during a 16-week weight loss program were not affected by changes in weight.
Our findings that weight loss was associated with an increased cartilage thickness loss or a reduced thickness gain within a 1-year period at a few sub-regions (Table 4, Fig. 1) was not fully consistent with those suggested by Anandacoomarasamy et al1, who carried out a similarly designed and sized study in Australia examining percent weight loss and cartilage thickness change (but not BML or synovitis) which, even more than our study, focused on knees without OA. In that study, in an analysis nearly identical to ours, the only significant region they found affected by weight loss was the cMF sub-region, where there was 0.007mm (p=0.009) less cartilage loss with each percent weight loss. In the ccMF (the central sub-region within cMF), we found a result in the opposite direction, and at a substantially diminished slope: a 0.0014mm increase in cartilage loss with each percent weight loss. The Australian study did not examine aMT or aLT sub-regions but interestingly, reported that weight loss led to a non-significant increase in thickness in lateral compartment cartilage.
Our study focused on those undergoing bariatric surgery who lost weight rapidly, most of whom experienced a marked reduction in knee pain, and we focused here on whether these persons had preservation of joint structure. Gersing et al21 asked a different question, that is, whether consistent weight loss over 4 years among older persons with OA was associated with cartilage preservation. Using data from the OAI, they reported that those who lost at least 10% body weight had a reduction in cartilage damage compared with those with stable weights. The Gersing et al study focused on OAI participants who were, on average, 63 years old at baseline (vs. 49 years in our study) with 50% having radiographic disease of Kellgren-Lawrence grade 3 at baseline and all had WORMS cartilage lesions. It is unclear whether pain improved in these persons. The difference in findings between our study and that of Gersing et al suggests that older persons with radiographic OA may experience structural benefits in their knees from a slow but consistent weight loss over a 4-year period.
While we found little effect of this weight loss on knee structure at 1-year, caveats include that we might have found more changes with longer follow-up or in those with more severe radiographic disease. There may not be a sufficiently high rate of cartilage loss in middle aged persons with mild OA to detect any effects of weight loss at 1-year. These are potential limitations, although we note that the aforementioned Australian study reported differences in cartilage loss within a year. While cartilage loss may not be easily detectable in this time frame, both synovitis and BML change have been shown to occur in 6 weeks or less7,22,23. Our study did not have a large number of subjects, although our estimates of effect were precise, suggesting that we did not miss an effect. Our study sample was similar in size to one that has reported some positive findings. The results of our examination of cartilage thickness remained consistent when analyzed using different methodology, namely parametric regression modeling and non-parametric Kendall's tau correlation for region-based and ordered values approaches that were described in Buck et al15 and Eckstein et al16.
In conclusion, rapid reduction in pain with marked weight loss was seen but was not accompanied by structural change in cartilage or in the structures that have nociceptive innervation.
Acknowledgments
Role of the funding source: This study was supported by the National Institutes of Health (grant NIH P60-AR-47785)
Footnotes
Contributions: Conception and design: Felson.
Acquisition of data: Felson, Guermazi, Apovian, Clancy, Li, Eckstein.
Analysis and interpretation of data: Jafarzadeh, Felson.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.Anandacoomarasamy A, Leibman S, Smith G, et al. Weight loss in obese people has structure-modifying effects on medial but not on lateral knee articular cartilage. Ann Rheum Dis. 2012;71(1):26–32. doi: 10.1136/ard.2010.144725. [DOI] [PubMed] [Google Scholar]
- 2.Peltonen M, Lindroos AK, Torgerson JS. Musculoskeletal pain in the obese: a comparison with a general population and long-term changes after conventional and surgical obesity treatment. Pain. 2003;104(3):549–557. doi: 10.1016/S0304-3959(03)00091-5. [DOI] [PubMed] [Google Scholar]
- 3.Hooper MM, Stellato TA, Hallowell PT, Seitz BA, Moskowitz RW. Musculoskeletal findings in obese subjects before and after weight loss following bariatric surgery. Int J Obes 2005. 2007;31(1):114–120. doi: 10.1038/sj.ijo.0803349. [DOI] [PubMed] [Google Scholar]
- 4.Stefanik JJ, Felson DT, Apovian CM, et al. Changes in Pain Sensitization after Bariatric Surgery. Arthritis Care Res. 2018 Jan; doi: 10.1002/acr.23513. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Felson DT. Imaging abnormalities that correlate with joint pain. Br J Sports Med. 2011;45(4):289–291. doi: 10.1136/bjsm.2010.081398. [DOI] [PubMed] [Google Scholar]
- 6.Talk: Developmental Trajectories of WOMAC Knee Pain and Their Predictors Over 3 Years Among Subjects of Osteoarthritis Initiative (2011 ACR/ARHP Annual Scientific Meeting) [Accessed May 16, 2017]; https://acr.confex.com/acr/2011/webprogram/Paper22608.html.
- 7.Callaghan MJ, Parkes MJ, Hutchinson CE, et al. A randomised trial of a brace for patellofemoral osteoarthritis targeting knee pain and bone marrow lesions. Ann Rheum Dis. 2015;74(6):1164–1170. doi: 10.1136/annrheumdis-2014-206376. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Cotofana S, Wyman BT, Benichou O, et al. Relationship between knee pain and the presence, location, size and phenotype of femorotibial denuded areas of subchondral bone as visualized by MRI. Osteoarthritis Cartilage. 2013;21(9):1214–1222. doi: 10.1016/j.joca.2013.04.001. [DOI] [PubMed] [Google Scholar]
- 9.Bellamy N, Buchanan WW, Goldsmith CH, Campbell J, Stitt LW. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol. 1988;15(12):1833–1840. [PubMed] [Google Scholar]
- 10.Angst F, Aeschlimann A, Michel BA, Stucki G. Minimal clinically important rehabilitation effects in patients with osteoarthritis of the lower extremities. J Rheumatol. 2002;29(1):131–138. [PubMed] [Google Scholar]
- 11.Hunter DJ, Guermazi A, Lo GH, et al. Evolution of semi-quantitative whole joint assessment of knee OA: MOAKS (MRI Osteoarthritis Knee Score) Osteoarthritis Cartilage. 2011;19(8):990–1002. doi: 10.1016/j.joca.2011.05.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Torres L, Dunlop DD, Peterfy C, et al. The relationship between specific tissue lesions and pain severity in persons with knee osteoarthritis. Osteoarthritis Cartilage. 2006;14(10):1033–1040. doi: 10.1016/j.joca.2006.03.015. [DOI] [PubMed] [Google Scholar]
- 13.Moisio K, Eckstein F, Chmiel JS, et al. Denuded subchondral bone and knee pain in persons with knee osteoarthritis. Arthritis Rheum. 2009;60(12):3703–3710. doi: 10.1002/art.25014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Eckstein F, Yang M, Guermazi A, et al. Reference values and Z-scores for subregional femorotibial cartilage thickness--results from a large population-based sample (Framingham) and comparison with the non-exposed Osteoarthritis Initiative reference cohort. Osteoarthritis Cartilage. 2010;18(10):1275–1283. doi: 10.1016/j.joca.2010.07.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Buck RJ, Wyman BT, Hellio Le Graverand MP, et al. Using ordered values of subregional cartilage thickness change increases sensitivity in detecting risk factors for osteoarthritis progression. Osteoarthritis Cartilage. 2011;19(3):302–308. doi: 10.1016/j.joca.2010.12.006. [DOI] [PubMed] [Google Scholar]
- 16.Eckstein F, Buck R, Wirth W. Location-independent analysis of structural progression of osteoarthritis-Taking it all apart, and putting the puzzle back together makes the difference. Semin Arthritis Rheum. 2017;46(4):404–410. doi: 10.1016/j.semarthrit.2016.08.016. [DOI] [PubMed] [Google Scholar]
- 17.Bates D, Mächler M, Bolker B, Walker S. Fitting linear mixed-effects models using lme4. J Stat Softw. 2015;67(1):1–48. doi: 10.18637/jss.v067.i01. [DOI] [Google Scholar]
- 18.Buck RJ, Wyman BT, Le Graverand MPH, et al. Osteoarthritis may not be a one-way-road of cartilage loss--comparison of spatial patterns of cartilage change between osteoarthritic and healthy knees. Osteoarthritis Cartilage. 2010;18(3):329–335. doi: 10.1016/j.joca.2009.11.009. [DOI] [PubMed] [Google Scholar]
- 19.Eckstein F, Boudreau R, Wang Z, et al. Comparison of radiographic joint space width and magnetic resonance imaging for prediction of knee replacement: A longitudinal case-control study from the Osteoarthritis Initiative. Eur Radiol. 2016;26(6):1942–1951. doi: 10.1007/s00330-015-3977-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Gudbergsen H, Boesen M, Christensen R, et al. Changes in bone marrow lesions in response to weight-loss in obese knee osteoarthritis patients: a prospective cohort study. BMC Musculoskelet Disord. 2013;14:106. doi: 10.1186/1471-2474-14-106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Gersing AS, Schwaiger BJ, Nevitt MC, et al. Is Weight Loss Associated with Less Progression of Changes in Knee Articular Cartilage among Obese and Overweight Patients as Assessed with MR Imaging over 48 Months? Data from the Osteoarthritis Initiative. Radiology. 2017 May;:161005. doi: 10.1148/radiol.2017161005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Felson DT, Niu J, Neogi T, et al. Synovitis and the risk of knee osteoarthritis: the MOST Study. Osteoarthritis Cartilage. 2016;24(3):458–464. doi: 10.1016/j.joca.2015.09.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.O'Neill TW, Parkes MJ, Maricar N, et al. Synovial tissue volume: a treatment target in knee osteoarthritis (OA) Ann Rheum Dis. 2016;75(1):84–90. doi: 10.1136/annrheumdis-2014-206927. [DOI] [PMC free article] [PubMed] [Google Scholar]