Introduction

Temporomandibular joint (TMJ) osteoarthritis (OA) is a degenerative disease that is accompanied by articular tissue deterioration, underlying subchondral bone remodeling, immobility, and pain1,2. According to previous studies, the overall prevalence of TMJ OA is reportedly 18.0–84.7% among all patients with temporomandibular disorder (TMD)3. Computed tomography (CT) or cone beam CT has been widely used in the diagnosis of OA to observe bone changes4,5.

Recently, with the development of various imaging techniques, several methods have been used to diagnose TMJ OA, including panoramic view, ultrasonography, magnetic resonance imaging (MRI) and nuclear imaging6,7,8,9. Among these, nuclear imaging, such as bone scintigraphy, provides relatively early detection compared with other radiographical modalities that show changes only after decalcification has progressed to some extent10.

Single-photon emission computed tomography (SPECT)–CT, another type of nuclear imaging, can be useful for simultaneously diagnosing and predicting the prognosis of TMD by using SPECT and CT images simultaneously11,12. In particular, technetium-99 m methylene diphosphonate (99mTc-MDP) reflects the rate of local osteometabolic activity and provides more localized information on bone metabolism related to bone changes10.

The standardized uptake value (SUV) is used to quantify the degree of uptake on the SPECT–CT images. However, in clinical practice, there are some discrepancies between the clinical diagnosis and SUV. In other words, some patients are diagnosed with TMJ OA but do not have a high SUV, whereas others are not thought to have TMJ OA but have a high SUV. Moreover, exact diagnostic criteria for TMJ OA using SPECT–CT have not yet been established.

This study aimed to investigate the cutoff value and accuracy of using SPECT–CT for TMJ OA diagnosis based on MRI and clinical examination.

Materials and methods

Participants

This retrospective study was approved by the hospital’s institutional review board (approval No. #3-2023-0320), and a waiver of written informed consent was granted. This study was also complied with the tenets of the Declaration of Helsinki. All patients visited the Department of Oral and Maxillofacial Surgery, Gangnam Severance Hospital, Yonsei University (Seoul, Korea) with complaints of discomfort and pain in the TMJ between January 2018 and April 2022. After clinical examination, the patients underwent MRI and SPECT–CT at the Department of Radiology and Nuclear Medicine of the same hospital. Patients were excluded if they met the following criteria: (1) scan interval of > 6 months between MRI and SPECT–CT; (2) history of radiation therapy in the head and neck area; (3) a diagnosis of rheumatoid arthritis; or (4) history of trauma or surgery on the maxillofacial bone, including the TMJ. Finally, 56 patients (112 joints) were included in this study.

MRI

MRI was performed using a 3.0-T Magnetom scanner (Achieva; Philips Medical Systems, Best, The Netherlands) with 3-inch surface coils for the TMJs. For T1-weighted imaging, the following parameters were used: repetition time, 450 ms; echo time, 20 ms; slice thickness, 3 mm; field of view, 120 mm; and acquisition matrix size, 240 × 240. The parameters for T2-weighted imaging were as follows: repetition time, 2900 ms; and echo time, 90 ms13.

Clinical diagnosis of TMJ OA

Clinically, TMJ OA was diagnosed based on previous studies14,15. Patients who fall into stage IV (intermediate/late) or stage V (late) of the Wilkes classification are diagnosed with OA16,17. In brief, TMJ OA was diagnosed when structural changes in the TMJ condyle were observed on MRI, and the patients experienced pain. Patients with structural changes without pain or structural changes were diagnosed as non-OA.

SPECT–CT

SPECT–CT examinations were performed using a Symbia Intevo16 (Siemens Healthineers, Erlangen, Germany). SPECT was performed 3 h after intravenous administration of 740 MBq of 99mTc-HDP from the top of the head to the lung apex with the following parameters: 3° rotation per step, 20 s acquisition time per projection, and 512 × 512 matrix size. Low-dose CT images were acquired at 110 kV and 120 reference mA, using adaptive dose modulation. The images were reconstructed using xSPECT Quant and xSPECT Bone (Siemens Healthineers, Erlangen, Germany) for high-resolution analysis and quantification.

Image analysis was performed by two nuclear medicine physicians (J.H. Lee and Y.H. Ryu) using an open-source LIFEx software version 7.4.618. The SUV was calculated as follows: SUV = (decay-corrected activity [kBq] per mL of tissue volume)/(injected activity [kBq] per gram of body mass). A spherical region of interest (ROI) was manually placed centered on the mandibular condyle and adjusted to encompass all relevant TMJ structures based on axial and coronal CT images. SUVmax was defined as the maximum voxel value within the ROI. SUVpeak was defined as the average SUV within a 1 cm3 sphere centered on the maximum value voxel of the ROI. To account for variations between patients, SUVmax and SUVpeak were normalized by the average SUV of the clivus (SUVref), producing SUVmax/SUVref and SUVpeak/SUVref, respectively.

Statistical analysis

McNemar’s test was used to determine whether there was a difference in the rates of diagnosis using SPECT–CT and clinical diagnoses. Cohen’s kappa was calculated to confirm the overall agreement between SPECT–CT and clinical diagnosis. Using clinical diagnosis as the outcome variable, each cutoff value was calculated using the Youden index, and diagnostic performance was evaluated. The area under the ROC curve, sensitivity, specificity, accuracy, PPV, and NPV was calculated. A Generalized estimating equation (GEE) was used to adjust the estimate for correlated observations and to determine the differences in SPECT–CT variables between patients with and without OA based on clinical diagnosis. Finally, a GEE for all variables was used to determine how SPECT–CT values affected the diagnosis. All statistical analyses were performed by statistical experts using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results

In total, 53 patients (106 joints) were included in this study. Of these, five were male and 48 were female, and the average age was 45.47 ± 16.30 years. Among the 106 joints, 42 (39.62%) were diagnosed as non-OA and 64 (60.38%) were diagnosed as OA based on clinical examination and MRI. Among 42 joints diagnosed as non-OA, 28 joints showed neither pain nor bone change, 12 joints showed bone change without pain, and two joint showed pain without bone change. Forty joints (37.74%) were diagnosed with non-OA and 66 joints (62.26%) were diagnosed as OA based on SPECT–CT examination. There was no significant difference between the clinical and SPECT–CT diagnoses (p = 0.6374). The results are summarized in Table 1. Cohen’s kappa value between the two diagnoses was 0.642 (95% confidence interval [CI] 0.492–0.792).

Table 1 Comparison between clinical and SPECT–CT diagnosis (McNemar’s test)—Number of joints (%).
Full size table

The SUVmax cutoff value was 5.15. The area under the curve (AUC) was 0.797 (95% CI 0.736–0.858) with sensitivity, 59.375 (95% CI 47.342–71.408); specificity, 100.000 (95% CI 100.000–100.000); accuracy, 75.472 (95% CI 67.281–83.663); positive predictive value (PPV), 100.000 (100.000–100.000); and negative predictive value (NPV), 61.765 (95% CI 50.214–73.315). The cutoff value for SUVpeak was 3.635. AUC was 0.781 (95% CI 0.720–0.842) with sensitivity, 56.250 (95% CI 44.096–68.404); specificity, 100.000 (95% CI 100.000–100.000); accuracy, 73.585 (95% CI 65.192–81.978); PPV, 100.000 (100.000–100.000); and NPV, 60.000 (95% CI 48.523–71.477) (Table 2).

Table 2 Cutoff value, AUC, sensitivity, specificity, accuracy, PPV, and NPV for various SPECT–CT variables.
Full size table

There were significant differences between the non-OA and OA groups in all investigated variables (Table 3). SUVmax was 3.286 ± 0.780 and 6.623 ± 3.442 in the non-OA and OA groups, respectively (p < 0.0001). SUVpeak was 2.324 ± 0.688 and 4.913 ± 2.749 in the non-OA and OA groups, respectively (p < 0.0001). Thirty-two of the 42 (76.19%) joints were diagnosed as non-OA both on SPECT–CT and clinically. Similarly, 56 of 64 (87.50%) joints were diagnosed as OA both on SPECT–CT and clinically. Among the patients clinically diagnosed as non-OA, none showed SUVmax and SUVpeak greater than the cutoff values of 5.15 and 3.635, respectively.

Table 3 Comparison of variables between non-OA and OA (independent t-test).
Full size table

In Generalized estimating equation (GEE), the odds ratio (OR) was 3.184 (95% CI 2.040–4.969) for SUVmax and 3.679 (95% CI 2.260–5.990) for SUVpeak (p < 0.0001). When SUVmax was greater than the cutoff value (5.15), the OR was 66.307 (95% CI 7.880–557.920), which was statistically significant (p = 0.001). Similarly, the OR was 46.884 (95% CI 6.909–318.180) when the SUVpeak was greater than the cutoff value (3.635), which was significant (p < 0.0001). The results are summarized in Table 4.

Table 4 Generalized estimating equation for all variables.
Full size table

The receiver operating characteristic (ROC) curves for all variables are shown in Fig. 1. There was no significant difference between SUVmax and SUVpeak (p = 0.3684), and SUVmax/SUVref and SUVpeak/SUVref (p = 0.6096), respectively. Other variables such as SUVmax vs SUVmax/SUVref (p = 0.0005), and SUVpeak vs SUVpeak/SUVref (p = 0.0007) showed significant difference.

Fig. 1
figure 1

Receiver operating characteristic (ROC) curve in all variables.

Full size image

Discussion

SPECT–CT has been widely used to diagnose TMJ OA, and many studies have been conducted9,19. Recently, Jeon et al. reported significantly higher value of 99Tc-MDP uptake ratio in OA compared with a normal group20. Although changes in the mandibular condyle can also be determined with CT, SPECT can reflect functional aspects21. Therefore, SPECT–CT has the advantage of being able to identify functional information and specify areas of disease. In addition, we believe it is helpful for determining whether to undergo surgery and identify the surgical site more precisely because SPECT–CT can reflect bone inflammation.

In the present study, the average value of SUVmax was also significantly higher (6.623 ± 3.442) in patients with OA compared with non-OA individuals (3.286 ± 0.780). The SUVmax value in the non-OA group was similar with that observed by Ogura et al22. (3.57 ± 0.46 in normal mandibles). Although there is a significant difference in SUVmax values in cases of OA compared with non-OA, the range is wide, making diagnosis occasionally difficult. Accordingly, this study was conducted to determine the cutoff value of SPECT–CT and to analyze its accuracy in diagnosing TMJ OA.

In this study, OA was diagnosed when the morphology of the condyle was changed with joint pain on palpation, based on the diagnostic criteria for temporomandibular disorders reported in 2014 14. According to the result of this study, the cutoff value of SUVmax for diagnosis of TMJ OA was 5.15, with 59.375 (95% CI 47.342–71.408) sensitivity, 100.000 (95% CI 100.000–100.000) specificity, 100.000 (95% CI 100.000–100.000) PPV, and 61.765 (95% CI 50.214–73.315) NPV. This means that even if the patient has joint pain and bony changes on MRI, the SUVmax on SPECT–CT may show relatively diverse values. However, if the patient's SUVmax exceeds 5.15, it can be interpreted that the patient is likely to have OA. In addition, as shown by the 100% specificity, no patients had SUVmax value exceeding 5.15 on SPECT–CT when clinically diagnosed as non-OA (Table 3).

A Generalized estimating equation was performed to determine how SPECT–CT values affected the diagnosis (Table 4). In this study, the OR of SUVmax was 3.184, with statistical significance (p < 0.0001), indicating that SUVmax affects clinical diagnosis. When the SUVmax increased by 1, the possibility of clinical OA increased by 3.184 times. Likewise, if SUVmax and SUVpeak are greater than the cutoff value (5.15 and 3.635, respectively), we can interpret that the possibility of clinical OA increases by 66.307 and 46.884 times, respectively. Arvola et al. reported a strong correlation between SUVmax, SUVpeak and SUVmean23. Thus, we believe that SUVmax and SUVpeak showed similar results.

In this study, the sensitivity and NPV were relatively low, whereas the specificity and PPV were high for all variables, which may be related to the location of the pain. Even if there is pain in the joint, this pain may originate from surrounding tissues such as in the cases of capsulitis or retrodiscitis, or may be referred pain originating from muscles such as the masseter or temporalis24. Obvious masticatory muscle pain can be distinguished from joint pain, but the distinction may not be easy when the pain becomes chronic. According to Palconet et al.25, the degree of bone change in patients with TMJ OA has a poor correlation with pain. In addition, radiographic imaging of TMD does not necessarily reflect aggressive disease activity26. Therefore, joint pain itself may not necessarily be related to condyle problems.

OA is known to be a self-limiting disease; however, in some cases, patients suffer from pain and may also develop skeletal deformities, including malocclusion, anterior open bite, and facial asymmetry26,27,28. We believe that early detection and treatment can reduce secondary problems that may occur. Using bone scintigraphy, it is possible to detect bone changes earlier than with conventional radiography, because the former demonstrates bone metabolism29,30. In addition, the fusion image of SPECT–CT can increase both sensitivity and specificity compared with SPECT alone or planar scintigraphy31.

This study had some limitations. First, sex and age were not considered in this analysis. It has been reported that the value may vary due to differences in bone metabolic rate depending on age and sex32. In this study, most patients were women, so we believe that the influence of sex is relatively small. However, further research considering age is needed in a larger number of patients. Second, the composition of the non-OA group was another limitation. In this study, we categorized contralateral TMJ to the non-OA group. However, the authors also recognize that such cases cannot be true negative controls.

Despite these limitations, this study attempted to determine a cutoff SUV for the diagnosis of TMJ OA, and the cut-off SUV showed high sensitivity and PPV. Conducting a prospective study with more patients and considering onset of pain in the future will help in a more accurate diagnosis of TMJ-OA.

In conclusion, SPECT–CT can be helpful for the diagnosis of patients clinically suspected of having OA. Clinicians should also keep in mind that patients with SUVmax values higher than the cutoff value should be managed with a higher possibility of OA.