Effects and mechanotransduction pathways of therapeutic ultrasound on healthy and osteoarthritic chondrocytes: a systematic review of in vitro studies

The databases PubMed, EMBASE and Web of Science were searched from database inception up to 19^th September 2021 to identify in vitro studies that assessed the effects of therapeutic ultrasound on chondrocytes for OA treatment. The search strategy is provided in Appendix 1. The reference list of the most significant studies was also screened to further identify potential other eligible studies.

All records were exported to an Excel file (Microsoft® Office) and duplicates were removed by the software filter and then manually verified. Two authors (S.O. and R.A.) independently screened the titles and abstracts of all identified studies and assessed them for eligibility criteria of the most relevant studies full texts. Three other reviewers (B.B.H., A.L. and O.C.) were consulted in case of disagreement. The inclusion criteria were: 1) in vitro studies; 2) studies using chondrocytes or articular cartilage explants; 3) studies focusing on the effects of therapeutic ultrasound on chondrocytes or explants. The exclusion criteria were: 1) reviews or meta-analysis, conference proceedings or case reports; 2) non-English language; 3) in vivo studies; or 4) clinical studies. In vitro studies investigating the effect of ultrasound on the chondrogenic differentiation of stem cells were not included since our goal was to understand the ultrasound effects in restoring the normal function of chondrocytes. The Cohen's kappa coefficient was calculated using the IBM® SPSS® statistics software (version 26).

Data from the included studies were extracted by one author (S.O.) and four other authors (R.A., B.B.H., A.L. and O.C.) independently reviewed the extracted data, each revised 25% of the studies. Study characteristics, measured outcomes, mechanotransduction pathways and ultrasound parameters were collected as described in the Supplementary Table S1. Quantitative data on outcome measures were extracted from the studies' figures using the software WebPlotDigitizer. Median, 25% and 75% percentiles, minimum and maximum values were calculated for each ultrasound parameter. We recorded all relevant information that each study reported to allow independent replication (Supplementary Text).

Two authors (S.O. and R.A.) assessed the risk of bias using the Risk of Bias Assessment tool for Non-randomized Studies (RoBANS). It contains six domains comprising cell selection, confounding variables, exposure measurement, blinding of outcome assessment, incomplete outcome data and selective outcome reporting. We adapted these domains criteria to assess the risk of bias to in vitro studies, and added the judgment of bias arising from the experimental conditions, funding and other biases. Supplementary Table S2 reports the criteria to classify the risk of bias domains as low, high or unclear risk of bias. Any disagreements were resolved by consensus.

We quantitatively synthesized the results with standardized mean differences (SMDs) with 95% confidence intervals (CI). Due to the heterogeneity across studies regarding cell culture conditions, comparison conditions and outcome measures, a meta-analysis was not feasible; thus, the SMDs of each study were presented and the pooled SMD was not provided to avoid misleading and non-valid conclusions. The SMDs are provided as illustrative measures to enable comparisons between studies and are based on post-treatment values only. Data synthesis was performed separately for healthy and OA chondrocytes and grouped based on the ultrasound intensity. All main analyses were performed using the post-treatment values between control (cells without stimulation) and intervention (cells with stimulation).

The searches yielded a cumulative total of 3,827 records, from which 431 duplicates were removed. From 3,396 titles and abstracts, 103 full texts were analyzed, from which 31 studies met the eligibility criteria and were included in this systematic review. The reasons for exclusion are reported in the PRISMA flowchart (Fig. 1) and Supplementary Table S3. The Cohen's kappa coefficient was 0.876 (95% CI, 0.77 to 0.98) with an agreement of 94.9%.

The judgment of risk of bias of each study is provided in Fig. 2. Only one study (3%, k = 31) was judged to have a high risk of bias in the “Cell selection” domain due to the selection of chondrocytes from different anatomical sites. Almost half of the studies (48%, k = 31) were judged to have unclear risk of bias due to “Confounding” since they did not report the animal characteristics from which chondrocytes were isolated from or the cell density used, precluding us to assess the potential risk of differences between groups. The “Experimental conditions” domain was judged to have a high risk of bias in 84% of the studies (k = 31) due to the absence of temperature control and/or previous calibration of the setup. Nearly 65% of the studies (k = 31) were judged to have a high risk of performance and detection bias in the “Exposure measurement” and “Blinding outcome assessment” domains respectively; because semi-quantitative analyzes were not conducted by two independent observers or blindly, respectively. Almost one third of studies (32%, k = 31) were judged to have an unclear risk of bias since they did not clearly provide the number of experiments or replicas used, preventing us to ascertain if there was incomplete outcome data reporting. Only two studies (6%, k = 31) were judged to have high risk of reporting bias due to not reporting the results of all measured outcomes. The “Funding bias” domain was judged to have unclear risk of bias in 74% of studies (k = 31) owing to the absence of conflict of interest statements, while high risk of bias was judged in one study that reported important conflict of interests. Roughly 55% of studies (k = 31) were judged to have high risk of bias for pseudoreplication, arising from not clearly reporting if the experiments or replicas were independent. Only one study (3%, k = 31) has performed an a priori sample size calculation. Five studies (16%, k = 31) provided all relevant information to allow independent replication (Supplementary Table S4).

In 61% of studies (k = 31), chondrocytes were isolated from animal sources. Primary chondrocytes were mainly cultured in a monolayer (65%, k = 31). Healthy chondrocytes were more commonly used (79%, k = 28), two studies utilized OA chondrocytes (7%, k = 28) and five studies included both (18%, k = 28). Chondrocyte cell lines were explored in four studies (14%, k = 28). Six studies (19%, k = 31) utilized healthy (66.6%, k = 6), OA (16.7%, k = 6), or both (16.7%, k = 6) explants cultures. In vitro models of OA were established in six studies (19%, k = 31) by administrating interleukin (IL)-1β (66.6%, k = 6) or hydrogen peroxide (16.7%, k = 6) to induce inflammation, or lipopolysaccharide (16.7%, k = 6) to induce oxidative stress. Study characteristics are summarized on the Supplementary Table S5 and described in Table I.

The range of ultrasound parameters is provided in the Supplementary Tables S6 and S7. Ultrasound was applied with a median center frequency of 1.5 MHz (range, 1 to 8.5) at a median power density of 40 mW/cm² (range, 0.14 to 700) for a median stimulation time of 20 min (range, 0.5 to 60) and a median number of sessions of 6 (range, 1 to 80) once daily. In pulsed mode, ultrasound waves were repeated at a median frequency of 1 kHz (range, 0.1 to 1) and median duty cycle of 20% (range, 20 to 40) with a pulse duration of 200 μs. The ultrasound equipment commonly used was the SAFHS® (23%, k = 31).

Three different setups were employed to apply ultrasound to chondrocytes (Fig. 3). More than half of studies applied the ultrasound in direct contact with the culture plate (55%, k = 31) using acoustic gel [Fig. 3(A)]. Almost 20% of studies (k = 31) immersed the ultrasonic transducer directly into the culture media [Fig. 3(B)], while others immersed the transducer in water bath [Fig. 3(C)] and placed the culture plates on the water surface (16%, k = 31). Only 39% of the studies (k = 31) reported the distance between the ultrasound transducer and the cells (Supplementary Tables S6 and S7).

Chondrocyte viability, proliferation or apoptosis (Supplementary Table S5) were assessed in 48% of the studies (k = 31). Viability was either statistically significantly increased^{,  ,}  or unaffected in healthy chondrocytes after ultrasound application^{,  ,}  , but statistically reduced with increasing center frequency. Apoptosis was statistically significantly decreased in both healthy and OA cells. Proliferation of healthy chondrocytes statistically enhanced^,,, following intervention, while seven studies^{,,,  ,  ,}  did not observe statistically significant effects in both healthy and OA chondrocytes (Table I). The effect sizes for each outcome and study are reported in Table II and the forest plots are provided on Supplementary Fig. S1.

About 84% of the studies (k = 31) measured the effect of ultrasound on extracellular matrix (ECM) synthesis (Supplementary Table S5). The expression of ECM proteins including collagen (COL) type II^{,,  ,  ,,,,  ,  ,,,,  ,  ,  ,  ,  ,}  , aggrecan (ACAN)^{,,,,  ,  ,,,,,  ,  ,}  and proteoglycans (PGs)^,,,,,,,,, was statistically significantly increased in nearly all studies after ultrasound in both healthy and OA chondrocytes, even in IL-1β-treated cultures (Table I)^,. Although most studies reported an increase in ECM production, statistically significant decrease^, or no effect^,,,, in the expression of these proteins were also reported. Similarly, the hypertrophic marker COL X statistically significantly increased^, or decreased^,, whereas the chondrocyte fibrocartilage marker, COL I, remained unaltered after ultrasound application^,. The effect sizes for each outcome and study are reported in Table II and the forest plots are provided on Supplementary Fig. S1.

ECM degradation markers were assessed in 13 studies (42%, k = 31) (Supplementary Table S5). Most of these studies exhibited statistically significant reduction in the metalloproteinases (MMPs) expression that degrade ECM, namely, MMP-1^,,, MMP-3, MMP-13^{,,,,,,,  ,  ,}  and the aggrecanase ADAMTS after ultrasound application, while the tissue inhibitors, TIMPs, were statistically significantly increased^,, (Table I). Both statistically significant increased^, and unaltered^,, expression were also noticed for these degradation markers. For vascular endothelial growth factor A, VEGFA, and inflammation, IL-1β and IL-6^,, ultrasound exposure statistically significantly decreased their expression, whereas the oxidation markers have both statistically significantly increased or decreased. The effect sizes for each outcome and study are reported in Table II and the forest plots are provided on Supplementary Fig. S1.

Eleven studies^,,,,,,,,,, investigated the signaling pathways involved in transducing the ultrasound effects (Table III). The mitogen-activated protein kinases (MAPK) subfamilies were the most commonly explored (Supplementary Table S5), including extracellular signal-related kinases (ERK1/2; 27%, k = 11), c-Jun amino-terminal kinases (JNK; 27%, k = 11) and p-38 (45%, k = 11), along with integrin-mediated pathways (45%, k = 11).