Pedicle screw loosening has been widely reported as one of the concerning complications after spine instrumentation surgery, which may require revision surgery [21]. But the reported data regarding screw loosening were traditionally based on radiological observation, which could be subjective and lead to a considerable variation. Our study reported a 33% loosening rate according to mechanical measurement of extraction torque during instrumentation removal surgery. Meanwhile, we compared our extraction torque data with X-ray and CT scan findings, the result showed that the specificity of both imaging approaches were excellent, and the peri-screw osteolysis present both in X-Ray and CT scan could indicate low extraction torque of screw anchor, but their sensitivities were less than satisfactory (24% and 22% respectively), regarding detection of screw loosening. Furthermore, to our knowledge, for the first time we used extraction torque during screw removal surgery to analyze risk factors for screw loosening, our findings demonstrated that screws in non-fusion spine and fractured vertebrae had significantly lower extraction torque, while BMD and age showed low, but significant linear correlation with extraction torque.
A number of factors have been reported to be related to screw loosening. Excessive strain between the screw and bone interface is considered to be the primary cause for screw loosening [12, 22], which could be deteriorated when fusion is failure or the anterior support is inadequate. Meanwhile, stress shielding can lead to a decrease of stress transferred through the bone tissue, which can reduce bone mineral density and remodel the bone surrounding the screw. The presence of wear debris [23] was reported to induce osteolysis leading to screw loosening. The debris elicited an inflammatory cytokine-mediated particulate-induced response through increased expression of intracellular TNF-alpha, increased osteoclastic activity, and cellular apoptosis. Other factors that could cause bone loss or destruction, such as infection surrounding the implant, bone tumor, metabolic diseases, and microfracture due to excessive loading, are risk factors of pedicle screw loosening. Screw loosening may become a worsening problem due to the aging of population and the increasing number of osteopenic and osteoporotic patients. Wu et al. [17] reported higher occurrence of screw loosening in osteoporotic bone. In our study, we also found a significant linear correlation between BMD and extraction torque, indicating pedicle screws in aged patients or patients with lower BMD might be less stable due to lower extraction torque.
Our findings showed that X-ray had a sensitivity of 24% and a specificity of 98%, while CT scan had a sensitivity of 22% and a specificity of 95% regarding extraction torque as criterion of screw loosening. This result indicated that both radiological examinations were effective to confirm loosening screws; however, the low sensitivity implied a considerable number of loosening screws could be neglected by imaging study. Based on the torque data, the loosening rate was 33%, while both X-ray and CT scan only detected less than 30% of all loosening screws. Sanden [20] reported a 64% sensitivity of X-ray in 79 screws and a 35% rate of screw loosening, but their definition of a loosed screw was an extraction torque of 0.4 Nm or less as there were no screws with and extraction torque between 0.4 Nm and 0.75 Nm. We could not detect such a clear gap in our data. However, by using Twostep Cluster Analysis, we found the distribution of extraction torque data could be clustered into 3 clusters, with the cut-off value of 1.02 Nm and 2.22 Nm respectively. We also tested the average torque of newly implanted screws and found that 1.02 Nm was lower than the low limit of 95% confidence interval of newly implanted screws. Therefore, we set our cut-off torque at 1.02 Ncm for screw loosening. Although the torque values for screw loosening were different, the loosening rate of Sanden’s study based on extraction torque was similar to our findings, around 30–35%.
Ohtori employed both CT scan and X-ray and the results showed that CT scan was more sensitive than X-ray [11]. In our study, there was no significant difference between X-ray and CT scan regarding both sensitivity (p = 0.863) and specificity (p = 1.00). It was observed in one case that CT scan failed to show a clear gap around screw, even though an obvious double halo was found on X-ray and the extraction torque was 0.06 Nm (Fig. 6). This could be resulted from the metal artifact that seriously interfered CT reconstruction and led to incorrect image surrounded metal instrumentation. Therefore, based on our result, CT might not be superior to X-ray in assessment of screw loosening, especially considering its higher cost and radiation exposure.
The risk factors of screw loosening we found based on extraction torque were fixation in fractured vertebrae and non-fusion spine. In the fractured vertebrae, the continuity of cortical and structure of trabecula were damaged, which might affect the stability of screw placed in it. Meanwhile, the strain imposed on pedicle screws might significantly increase when fusion had not been obtained or anterior column support was inadequate. We also found that pedicle screws in the lumbosacral junction (L4, L5, and S1) had significant higher extraction torque than those in other segments. A possible explanation could be that surgeries performed in lumbosacral spine were mainly for degenerative disc diseases, which often required spinal fusion, while in the thoracolumbar spine, where operations were more likely for vertebral fracture, fusion were not always necessary.
As expected, linear regression analysis showed that the stability of pedicle screw correlated positively with BMD and negatively with patients’ age. The relatively low related coefficients might be due to the existence of numerous confounding factors, and these results indicated that the failure risk of instrumentation increased with age and osteopenia/osteoporosis, which was widely credited but rarely proven in vivo with mechanical measurement.
Some reports showed that increasing length and diameter could increase the stability of pedicle screw [24, 25], which had not been observed from our results. The reasons that no significant different extraction torque was found among screws with different length and diameter might be due to the relatively small sample size and narrow range of length (30–55 mm) and diameter (4.0–6.5 mm). Further study with larger simple size is needed to confirm the effect of screw length and diameter on extraction torque in the human spine. Meanwhile, there was no significant difference in extraction torque regarding the screw design (polyaxial vs. monoaxial) and location (placed at the end segment vs. at the middle segment).
There are several limitations in our study. First, although this was a prospective and blinded designed study, the relatively small number of patients’ enrollment and the heterogeneity of screw size and position might render our findings susceptible to confounding factors. Hence, further studies with larger sample size and stratified data according to different factors will be needed to better understand the prevalent of pedicle screw loosening. Second, the lack of pedicle torque during primary surgeries made it unable to observe the longitudinal change of torque and the influence of inserting torque on screw loosening.