Evaluating for a correlation between osteopathic examination and ultrasonography on thoracic spine asymmetry

Methods

This study was registered with ClinicalTrials.gov and the study identifier is NCT04823637. The study protocol was approved by the Midwestern University (MWU) Institutional Review Board (IRB) at the Glendale, Arizona campus, and the participants provided written informed consent. This study did not require any funding. The study actively recruited subjects between October 10, 2019 and January 30, 2020. Email announcements were sent to students at the Arizona College of Osteopathic Medicine and the Master’s in Biomedical Sciences program at MWU in Glendale, Arizona.

Setup and landmark marking

Subjects were seated upright in stable chairs with substantial lumbar support, with the back of the chairs below the level of T7 SP. Each subject was placed in an individual room next to an ultrasound machine and could not communicate with each other while the study was in session. Each examiner was blinded and could not communicate with each other or see what the other examiner was performing throughout the entirety of the data collection phase. There was a data collector stationed at each subject area to record all data, which were blinded from the examiners.

Marking of landmarks on the skin were performed by an OMM Scholar. Members of the OMM Scholarship program are osteopathic students selected through a competitive evaluation system to become peer instructors within the Department of Osteopathic Family and Community Medicine. The OMM Scholar in this study was a fourth-year medical student who was in her second year of teaching structural palpation, osteopathic diagnosis, and treatment techniques to first- and second-year medical students. The OMM Scholar palpated subjects’ upper backs on the skin and gave verbal instructions such as “look down” and “look up to the ceiling,” to obtain accurate determination of each SP. First, the C7 SP bony protrusion was identified utilizing a nontoxic, non-permanent skin marking pen. Then, the subsequent thoracic SPs of T1-4 were palpated and marked. To mark the TP, the marker is placed at a distance of one-half inch directly lateral to each SP [7]. Then, each TP is marked with the number of the vertebra below (i.e., next to the T2 SP, the number “3” is written to indicate the TP of T3) until the T5 TPs were marked (Figure 1A, B). All vertebral segments indicated in the study, if not otherwise indicated, refer to the TPs of each vertebra (e.g., T3=T3 TP). We have decided to focus on the T3–T5 vertebrae to limit interference from other anatomical structures [27].

Figure 1:

(A) Demonstration picture showing how the subjects are marked. The C7 SP is marked as “C7” in red, the T1 SP is marked as “T1” in black, and the SPs of T2–T4 are marked with black dots. The T3 TP is marked as “3,” the T4 TP is marked as “4,” and the T5 TP is marked as “5.” (B) Demonstration picture showing that the US probe is placed perpendicular to the SP of the vertebra above and the TPs of the vertebra under examination. (C) US imaging with identifications of landmarks. SP, spinous process of the vertebra above; TP, transverse process under examination. The ribs are also identified. (D) US imaging showing how measurements are achieved. The SP is identified, and the distance between the SP of the vertebra above to the TPs is measured. SP, spinous process; TP, transverse process; US, ultrasound.

Results

The overall demographics of participants showed 31 (47.0%) males and 35 (53.0%) females. Race and ethnicity were not collected for the purpose of this study because they were not variables being evaluated or investigated. The subjects’ age range is 22–35 years of age, with the mean age as 26.61 years (SD, 3.1) for males and 25.71 years (SD, 2.7) for females. The BMI range is 15.9–36.9 kg/m², with the mean BMI as 24.13 kg/m² (SD, 4.2) overall, 25.46 kg/m² (SD, 3.5) for males, and 22.95 kg/m² (SD, 4.5) for females.

The average US measurement differences for both US1 and US2 is 0.207 cm or 2.07 mm (see Supplemental section for raw data). The mean US measurement between US1 and US2 across all participants was found to have no significant difference at T3 (p=0.1523), T4 (p=0.5029), and T5 (p=0.4712). Ultrasonography measurements for T3, T4, and T5 were normally distributed around a 95% confidence interval, with a margin of error of 0.062 at T3, 0.066 at T4, and 0.068 at T5 (Figure 2).

Figure 2:

Histogram of the US measurement difference for each vertebral segment, T3, T4, and T5. The blue line underneath represents the 95% confidence interval for each segment. Within the interval was classified as neutral (N), below as left, and above as right. US, ultrasound.

Our results showed that ultrasonography had fair interexaminer agreement between US1 and US2 for the overall “most prominent” segmental rotation of the thoracic spine, with Cohen’s Kappa at 0.27, with range of (0.09, 0.45), and total agreement percentage at a 51.5% (Figure 3A, B). For osteopathic examination (OMM), there is low statistical interexaminer agreement between OMM1 and OMM2 for the overall thoracic spine, with a Cohen’s Kappa at 0.05 (0.0, 0.27) and 31.8%. Both US1 and US2 had slight agreement with one of the osteopathic examiners (OMM2), with Cohen’s Kappa of 0.14 (0.0, 0.33), 31.8% for US1 vs. OMM2 and 0.11 (0.0, 0.30), 30.3% for US2 vs. OMM2 (Table 1A).

Figure 3:

(A) Bar plot of the total agreement comparing the most prominent (significant) findings as indicated by US measurement and OMM palpatory findings. (B) Bar plot of Cohen’s Kappa values for each pairwise comparison for the most significant/overall findings. Cohen’s Kappa is utilized to examine interexaminer agreement. The error bar is representative of the 95% confidence interval for Cohen’s Kappa. We interpret Cohen’s Kappa as 0.10–0.20 is indicative of slight agreement, 0.21–0.40 of fair agreement; 0.41–0.60 of moderate agreement; 0.61–0.80 of substantial agreement; and 0.81–0.99 of near-perfect agreement. (C) Bar plot of the total agreement between OMM and US examiners, for each pairwise comparison. (D) Bar plot of Cohen’s Kappa between OMM and US examiners, for each pairwise comparison at T3, T4, and T5. The error bar is representative of the 95% confidence interval for Cohen’s Kappa. We interpret Cohen’s Kappa as 0.10–0.20 is indicative of slight agreement, 0.21–0.40 of fair agreement; 0.41–0.60 of moderate agreement; 0.61–0.80 of substantial agreement; and 0.81–0.99 of near-perfect agreement. OMM, osteopathic manipulative medicine; US, ultrasound.

However, individual segmental analysis reveals that at T3, US1 and US2 has slight agreement at 0.20 (0.02, 0.37) and 48.5% calculated via Cohen’s Kappa and total agreement percentage, with a correlation coefficient of 0.23 (−0.02, 0.45) calculated by Spearman’s Rank Correlation. OMM1 and OMM2 showed low agreement at 0.01 (0.0, 0.12) and 43.9%, and a correlation of −0.04 (−0.28, 0.20) correlation coefficient (Figure 3C, D; Tables 1B and 1D). There is slight agreement between US1 and OMM2 (0.10 [0.03, 0.23]) and 21.2%. US1 and OMM1, US1 and OMM2, US2 and OMM1, and US2 and OMM2 all showed low levels of agreement, with Cohen’s Kappa and total agreement percentage as the following: 0.05 (0.0, 0.15) and 21.2%, 0.10 (0.03, 0.23) and 21.2%, 0.05 (0.0, 0.15) and 30.3%, and 0.03 (0.0, 0.12) and 28.8%, respectively. US2 and OMM1 had a Spearman correlation of 0.17 (0.08, 0.40), while US1 and OMM1, US1 and OMM2, OMM1 and OMM2, and US2 and OMM2 all had negative correlation values (Table 1D).

At T4, US has slight interexaminer (US1 vs. US2) agreement at 0.2 (0.02, 0.38) and 48.5%, with 0.24 (0.0, 0.46) correlation coefficient, and osteopathic examine showed low interexaminer (OMM1 vs. OMM2) agreement with 0.08 (0.0, 0.24) and 39.4% with 0.17 (−0.08, 0.40) correlation coefficient. All other pairwise comparisons revealed low levels of agreement: US1 vs. OMM1 showed 0.05 (0.0, 0.19) and 33.3%, US1 vs. OMM2 showed 0.06 (0.0, 0.08) and 27.3%, US2 vs. OMM1 showed 0.03 (0.0, 0.15) and 27.3%, and US2 vs. OMM2 showed 0.05 (0.0, 0.21) and 36.4%. Those pairwise comparisons (US1 vs. OMM1, US1 vs. OMM2, US2 vs. OMM1, US2 vs. OMM2) also showed low levels of correlation (Figure 3C, D; Tables 1B and 1D).

At T5, ultrasonography had moderate interexaminer agreement at 0.44 (0.27, 0.60), or 63.6% and moderate correlation of 0.57 (0.36–0.72). OMM had slight interexaminer agreement at 0.12 (0.0, 0.28) and 42.4%, with a correlation coefficient of 0.23 (−0.02, 0.45). All other pairwise comparisons revealed low levels of agreement—US1 vs. OMM1 showed 0.03 (0.0, 0.19) and 33.3%, US1 vs. OMM2 showed 0.02 (0.0, 0.15) and 34.8%, US2 vs. OMM1 showed 0.06 (0.0, 0.12) and 33.3%, and US2 vs. OMM2 showed 0.04 (0.0, 0.11) and 28.8%. Those pairwise comparisons (US1 vs. OMM1, US1 vs. OMM2, US2 vs. OMM1, US2 vs. OMM2) also showed low levels of correlation (Figure 3C, D; Tables 1B and 1D).

Fleiss’ Kappa compares all raters at once: US1, US2, OMM1, and OMM2 (66 subjects and 4 raters). Interpretation is similar to Cohen’s Kappa. Fleiss’ Kappa is −0.0512 at T3, 0.0287 at T4, and 0.0794 at T5, all of which indicate low levels of agreements (Table 1C).

Statistical analysis of ultrasonography and OMM agreement on the SP position revealed low levels of agreement. We found that US1 vs. OMM1 showed a Cohen’s Kappa of 0 (0,0) and total agreement percentage of 21.2% for T2 SP (Figure 4; Table 1B). For T3 SP, we found 0.03 (0, 0.13) and 30.3%, and 0.05 (0, 0.18) and 33.3% for T4 SP. US1 vs. OMM2 showed agreement of 0.03 (0, 0.19) and 37.9% for T2 SP, 0.08 (0, 0.24) and 37.9% for T3 SP, and 0.03 (0, 0.10) and 34.8% for T4 SP. US2 vs. OMM1 showed agreement of 0 (0,0) and 24.2% for T2 SP, 0.009 (0, 0.06) and 18.2% for T3 SP, and 0.06 (0, 0.17) and 28.8% for T4 SP. US2 vs. OMM2 showed 0.02 (0, 0.19) and 27.3% for T2 SP, 0.02 (0, 0.13) and 28.8% for T3 SP, and 0 (0, 0.05) and 22.7% for T4 SP. For comparison of ultrasonography vs. OMM overall, we found Fleiss’ Kappa of −0.0623 for T2 SP, −0.0119 for T3 SP, and 0.0338 for T4 SP, indicating low levels of agreements (Table 1C). Spearman’s rank correlation coefficients were all below 0.19 and consistent with the low levels of agreement calculated via Cohen’s Kappa and Fleiss’ Kappa (Table 1E).

Figure 4:

(A) Bar plot of the total agreement between OMM and US examiners, for each pairwise comparison of T2–T4 SPs. (B) Bar plot of Cohen’s Kappa between OMM and US examiners, for each pairwise comparison at T2 SP, T3 SP, and T4 SP. The error bar is representative of the 95% confidence interval for Cohen’s Kappa. We interpret Cohen’s Kappa as 0.10–0.20 is indicative of slight agreement, 0.21–0.40 of fair agreement; 0.41–0.60 of moderate agreement; 0.61–0.80 of substantial agreement; and 0.81–0.99 of near-perfect agreement. OMM, osteopathic manipulative medicine; SPs, spinous processes; US, ultrasound.

Discussion

The purpose of this study is to evaluate a preliminary assessment of a protocol that examines the relationship between osteopathic palpation and ultrasonographic measurements of the thoracic spine. Our results show, at least for this particular protocol, that in an asymptomatic and generally healthy student population, a low level of interexaminer agreement between these two modalities for thoracic vertebrae levels T3–T5 and an overall lower range of agreement within each modality. Seeing that previous studies comparing OMM and ultrasonography utilized a single examiner within each modality, two examiners are utilized in our study to decrease single examiner bias and in the hopes of enhancing validity [24], [25], [26]. As expected, the osteopathic examination at T3–T5 indicated a low level of interexaminer reliability between OMM1 and OMM2, which is consistent with the literature when few examiners are utilized [12, 27, 35, 36]. In regard to US examination, the fair level of interexaminer agreement at prominent vertebral levels was reduced to a low level of interexaminer agreement when considering all vertebral levels. However, US examination, within the protocol of this study, did demonstrate slightly higher interexaminer agreement than osteopathic examination, as indicated by the Spearman’s Rank correlation. This suggests that with additional adjustments to the protocol and further studies, a reliable modality including US is feasible.

It is worth noting that there was higher correlation within each modality and better agreement between both modalities at the T5 vertebral level. While this study was focused on asymptomatic participants, the higher correlation and agreement at vertebral level T5 is very promising in regard to utilizing a similar study design in the development of future studies. Pagé et al. [37] found that palpation and US of the spinous and transverse process of T5 had a higher validity and accuracy in vitro. This is consistent with the reported higher reliability for both osteopathic examiners and sonographers at T5. Considering the complexity of the thoracic spine, as well as the asymptomatic state of the participants in this study, it may be worth examining the correlation between the two modalities before and after treatment of symptomatic somatic dysfunctions [17, 28, 32]. Further examination of anatomical changes after treatments will provide greater understanding on the pathophysiology of somatic dysfunctions and the mechanisms utilized to relieve them. However, this study was effective in providing groundwork and considerations when improving a study design that matches the modalities and clinical applications. Our results also suggest that application of this study design to a training protocol for osteopathic medical students might yield benefits at least in the lower vertebral segments where perhaps rotations and somatic dysfunctions are more common [24].

Regarding the vertebral levels that were found to have lower levels of agreement and correlation between US and osteopathic palpatory examination, we have identified a few possible explanations and limitations. First, as shown through teaching palpatory diagnosis in osteopathic medical students, one of the difficulties to palpation is that the students typically learn on other, generally healthy, asymptomatic students. This may lead to some of the difficulties seen in reliability of diagnosis and interexaminer reliability, even among advanced practitioners [12, 38], [39], [40]. In our study, for instance, the difference between the right and left TPs was measured to be an average of 2.07 mm in a healthy, asymptomatic, student population. Osteopathic physicians determined a left or right rotation based on their palpatory feedback, which may not always be prominent because the subjects were mostly asymptomatic. Efforts to standardize palpatory feedback were taken into consideration during protocol creation, and we utilized multiple training sessions prior to data collection to diminish the effects of palpatory feedback discrepancy. Given that Mancini et al. [41] showed statistically significant changes after OMT of 14.5 mm with US, one reason for the difference in interexaminer reliability seen in our study and some previous interexaminer studies may be that skilled osteopaths are often seeking to palpate larger differences than the ones measured in our study [10, 35, 41], [42], [43], [44]. As such, the low interexaminer agreement within osteopathic palpatory examination may be best explained by the specifically chosen asymptomatic participants. However, the use of osteopathic palpatory examination in anatomical significant differences (i.e., somatic dysfunctions) is supported by the higher agreement at the T5 level. This possible connection emphasizes further exploration of osteopathic palpatory examination in the clinical setting—where somatic dysfunctions are common and requiring treatment.

Second, it is possible that the sample population and the size utilized in statistical analysis may not be representative of the efficacy of these two modalities and may have resulted in a diminished power that prevented a more accurate comparison of the modalities. However, the improved agreement at T5, even with a small sample size, is encouraging and does provide enough participants to create a study design that allows for a comparison between the two modalities in a scenario as standardized as possible while similarly overcoming single-examiner bias.

Finally, although every effort was made to limit rotation of the test population by asking them to look forward and limiting motion to flexion, it is certainly possible that the pressure of the palpating physicians’ hands, as well as the pressure from the US probes, may have made slight alterations in the location of the TPs between the various examiners, making slight changes seen in our data decreasing interexaminer reliability. Furthermore, the palpation protocol artificially limited structural examination to palpation of the TPs and flexion, limiting the normal diagnostic ability to passively extend, rotate, and sidebend the patient to increase the confidence of the diagnosis [45, 46]. US, on the other hand, is a static exam that requires the subject to remain motionless when diagnosing vertebral rotation [1, 47, 48]. As such, there may be an inherent and incompatible difference between US and osteopathic diagnostic examinations, which is supported by Snider et al. [26], where musculoskeletal changes in the sacral base identified by osteopathic palpatory examination could not be similarly identified by US examination. Considering the potential of this study design in a more clinically applicable scenario (i.e., before and after treatment in asymptomatic and symptomatic patients), the relationships and considerations identified in this study could be instrumental in illuminating important clinical relationships between osteopathic palpatory examination and ultrasonographic examination while further understanding the physiological mechanisms of somatic dysfunction and their appropriate treatments.

Another point to address is how the superior of two vertebra in a single segment may affect the findings of the vertebra below. A rotated vertebra changes the position of the TPs and SP such that the SP would be located more contralaterally to the posteriorly displaced TP and the direction of the vertebral rotation [34]. For instance, if T2 is rotated to the left, then the left T2 TP is more posterior and T2 SP is contralaterally displaced more to the right. This may lead to an inaccurate designation of rotation for the inferior vertebra, as the distance measured between the SP of the superior vertebra to the TP of the inferior vertebra is now altered. To address this phenomenon, we established an alternative analysis by utilizing the SP of each vertebra as our reference point instead of the TP of the inferior vertebra. To do this, we relabeled the ultrasonography measurement data so that it correlates with the SP of the superior vertebra (e.g., T4 TP data=T3 SP measurement). The OMM findings for SPs are also adjusted from the palpation findings of the TPs for statistical calculations (e.g., a palpated posterior T3 TP on the L=T3 rotated to the L=T3 SP displaced to the R). Then, we compared the agreement for both ultrasonography and OMM findings of each SP of T2–T4. This alternative analysis does not have any effects on the interexaminer reliability within each modality (i.e., US1 vs. US2, OMM1 vs. OMM2) and resulted in no changes from our original findings. As such, further studies are needed to establish a true “neutral” on the patient and to compare all US findings to that static neutral point and may further enhance US as a training tool in the medical school setting.

To our knowledge, this is a novel study examining the thoracic spine utilizing US and structural surface palpation as utilized in osteopathic medicine. This study compares some observational techniques that are utilized clinically and utilizes US as a readily available modality to visualize anatomy in an in vivo, noninvasive, and safe manner. Although US demonstrated slightly higher agreement overall, generally, in a fast-paced clinical setting, it may not be realistic for physicians to evaluate the thoracic spine with US because of the extra time required with potentially low clinical yield. In the end, this study was designed to contribute to the growing literature surrounding osteopathic medicine with the hope that the methods and results reported may help contribute to the foundation of future studies examining osteopathic palpatory examination and treatments, as well as the role of ultrasonographic examination in clinical practice.

Conclusions

In conclusion, this preliminary study of an asymptomatic population revealed, as shown via Cohen’s Kappa and Fleiss’ Kappa, that there is a low-to-moderate interexaminer reliability between sonographers, low-to-slight interexaminer reliability between osteopathic physicians, and low interexaminer reliability between OMM palpatory examination and ultrasonographic evaluation of the thoracic spine. There is, however, a slightly higher reliability at T5. The findings align with the main role of osteopathic palpatory examination as a clinical tool: the diagnosis of asymmetrical somatic dysfunction with the intention of treatment. This study was designed to contribute to the growing literature surrounding osteopathic medicine with the hope that the methods and results reported may help contribute to the foundation of future studies examining osteopathic palpatory examination and treatments, as well as the role of ultrasonographic examination in clinical practice and in teaching diagnostic palpation skills to osteopathic medical students. Future studies looking at US and OMM exam correlation on both asymptomatic and a clinically symptomatic population may prove to be an easier way to analyze validity and calibrate the two relationships between ultrasonography and OMM structural examination. Given that previous studies performed on other regions of the spine utilizing US to assess pre- and posttreatment measurements yielded positive and significant findings, further study of interexaminer reliability on the thoracic spine pre- and posttreatments may also yield significant findings that may prove beneficial to the osteopathic medical field.