In present study, the incidence of cage subsidence was 29.4%, and HUm could be a more effective predictor of postoperative cage subsidence with larger AUC compared with BMD-l/BMD-f.
Lumbar interbody fusion was widely performed in managing lumbar degenerative diseases. Stable fusion and restoration of disc height contributed to satisfactory clinical results. However, cage subsidence may appear following lumbar fusion surgeries and lead to unsatisfactory results because of decrease of disc height and restenosis of the foramen regions [17]. Some previous literatures reported no negative impact of cage subsidence on clinical outcomes [5, 18, 19]. While recent studies pointed out that in patients with cage subsidence, the improvement of ODI index was worse than those without cage subsidence [6]. Overall, the relationship between clinical outcomes and cage subsidence remained controversial.
From the results, we found that improvement of clinical outcomes in non-CS group was better. A possible explanation was that in our study and from our experience, most of the patients were implanted cage less than 12 mm. The patients may be more sensitive to the loss of disc height as the degree of height restoration was relatively lower. We seldom selected cage with height of 14 mm or higher, for cage with large height may result in more preparation of disc space, which increase the risk of endplate injury. Bach et.al summarized disc height in healthy individuals and proposed that a cage greater than 10 mm cage height will result in excessive restoration and potential risks of complications [20]. An obvious characteristic of our patients was that the course of disease was long (mean 4.4 years), which led to severe degenerative conditions of lumbar spine and the disc height decreased obviously. We believed specific conditions of the patients should be considered when selecting appropriate cage height.
The rate of cage subsidence after TLIF ranged around 35% [21]. Our findings went along with previous results at a rate of 29.4%.
A number of factors may contribute to cage subsidence following lumbar spine surgeries. Patient-related parameters consisted of age, gender, BMI, and bone quality [22]. Surgery-related factors including cage height, size and position were also reported. Cage position was considered as an important factor. Several methods have been reported to describe the position. Landham et.al utilized center point ratio (CPR) and posterior gap ratio (PGR) to illustrate cage position and the two parameters were correlated with gain of lumbar lordosis [23]. More anteriorly located cage was recommended to restore lumbar lordosis more effectively and avoid cage subsidence [6, 24]. In our study, in consideration of the design of cage, the position in CS group was determined by modified Taillard index, which was similar with PGR. A different point should be noted that the posterior metallic marker instead of border of cage was selected to draw “c” line, as the cage template we utilized was non-visualized through X-ray [14]. Still, the results confirmed that cage was located more anterior in non-CS group.
The correlation between bone quality and cage subsidence has been assessed in some studies. Choi et.al compared HU value and DXA to assess bone quality in 80 non-lumbar degenerative and 30 lumbar degenerative patients. There was a strong correlation between HU value and T-score. Real bone quality may be overestimated in degenerative patients, for the osteophytes could increase the value of T-score [25]. Ullrich et.al investigated the HU value in 81 patients underwent posterior-anterior stabilization because of thoracolumbar spine fractures and found HU value was strongly correlated with cage subsidence and additional treatment strategies should be considered in patients with a HU value less than 180 [26]. A recent study by Wang et al. analyzed the correlation between cage subsidence in ACDF and HU value in cervical spine. They proposed there was a negative correlation between HU value and segmental height loss at the surgical level [27]. Overall, the value of HU in lumbar spine surgeries with fusion has not been fully investigated. In our study, we measured the HU values from L1 to L4, which were consistent with the reports of BMD. A mean value of the 4 vertebrae was recorded as HUm. The results showed HU value was strongly correlated with T score of BMD of lumbar spine, which was similar with previous reports [28]. Besides, BMD from femoral neck was also taken into account. Spearman analysis confirmed that BMD-f still correlated with HU and the r value (0.661) seemed to be higher than that of BMD-l. Osteophyte formation and bone sclerosis may increase BMD and make bone quality overestimated. We concluded BMD-f provided a more authentic reflection of bone quality compared with BMD-l.
The differences of BMD, BMI, cage position, age and HU values were significant between the two groups in our study. Binary logistic regression analysis showed BMD and HU were risk factors for cage subsidence. Cage position failed to reach a statistical significance. A potential explanation was that the surgeries were performed by the same surgeon, the procedures and intraoperative habits were consistent, which means the variation of cage position was relatively small. Different predicting models were performed. AUC of BMD-f seemed to be larger than BMD-l, while statistical difference was not achieved (P = 0.052). However, the specificity was higher in BMD-f than that of BMD-l. It could be accepted that BMD-f performed better than BMD-l in predicting cage subsidence. Compared with BMD, HUm could be a more effective predicting parameter with larger AUC. HUm was calculated as the mean value of 4 vertebrae, which was time-consuming with 12 times measurements required. To investigate and possibility of simplifying procedures, we separately analyzed the value per vertebra in predicting cage subsidence. The results showed HU4 alone may act as a substitute for HUm with a similar AUC. This could be easily applied in clinical experience as three times of measurements were enough. The results above indicated that HU values could be utilized as an easily obtained and effective value in predicting cage subsidence. CT scans were essential and regularly examined preoperatively at our center. Considering the cost of treatment, HU value could be regarded as an important parameter in surgical plan. Zhao et.al proposed that HU measurement may be used as a predictor of cage subsidence after unilateral fixation [13]. In our study, a larger sample (279 VS 36) was evaluated with bilateral fixation. Although unilateral fixation may achieve satisfactory clinical results in lumbar spinal fusion surgeries in some literatures [29, 30], bilateral fixation was more widely accepted and utilized. Therefore, the results of the study could be a more practical guidance in clinical experience.
Fusion rate was also taken into consideration. Numerous methods have been proposed to assess conditions of fusion. Gruskay et.al analyzed different ways in evaluating fusion and concluded that thin-cut CT and dynamic plain films should be regarded as imaging modalities [31]. Sugiyama et.al proposed that neither plain static nor dynamic radiographs were able to evaluate fusion outcome accurately compared with CT-based assessment [32]. In our center, all patients at one-year follow-up or longer received thin-cut CT scans and dynamic films were obtained to assess fusion condition regularly. Thus, we enrolled the two methods together to assess fusion rate. Fusion rate in CS group was lower but achieved no significance. The possible correlation between cage subsidence and fusion rate required a more detailed exploration.
This study has several limitations. Firstly, the cohort was relatively small. All patients received one level TLIF below L2/3 level and the same type of PEEK cage, which might mean selection bias. A second limitation is that 66 patients (23.7%) received BMD detection of femoral neck preoperatively. The other BMD-f values were obtained at follow-ups. Although no significant difference was found between preoperative values and values at follow-ups in the 66 patients, a more accurate comparison could be implemented using complete preoperative parameters. A third limitation is that mean value of BMI was around 25Kg/m2 and the whole cohort should be regarded as normal population. The potential difference in overweight or obese sample is worth further studying. Lastly, the follow-ups in this study were relatively short. A longer and prospective investigation is needed in the future. Lastly, the retrospective study proposed HU at L4 may act as a convenient predictor of cage subsidence with fewer measurements compared with HUm. A prospective study with larger sample is essential to validate whether this conclusion still hold.