This retrospective single-center study was carried out at the radiology department of Balgrist University Hospital in Zurich, and approved by the institutional review board. The study was conducted according to the principles of the Declaration of Helsinki and national ethical standards. All patients included in the study have given written informed consent that allows their health-related data to be used for research purposes.

The institutional picture archiving and communication system (PACS) and clinical information system (CIS) were reviewed for uni-segmental, bilateral spondylolysis patients ≥ 18 years of age, with chronic lower back pain (CLBP, ≥ 3 months) refractory to physiotherapy alone, who had received CT-guided lumbar pars infiltrations between July 2015 and July 2024.

Exclusion criteria included (a) lack of pain assessment, (b) missing pre-infiltration MRI or MRI > 6 months prior to pars injection, uni-lateral pars injection and/or concomitant contralateral facet joint infiltration or nerve root block, (d) incomplete MRI protocols or insufficient MR imaging quality, and (e) lumbar spine surgery before pars injection, and pars injection without iodine-based contrast agent due to possible prior allergic reaction.

Lumbar spine MRI was conducted on either a 1.5-T unit (Magnetom Avanto Fit and Magnetom Sola, Siemens Healthineers) or a 3-T unit (Magnetom Vida and Magnetom Skyra Fit, Siemens Healthineers) using a 32-channel phased-array spine coil. All patients underwent routine lumbar spine imaging protocols in supine and feet-first positions. All examinations were performed as non-contrast MRI. The sequence parameters varied depending on the scanner (1.5 T vs 3 T) (Supplementary Table 1).

All lumbar pars injections were performed by board-certified, fellowship-trained musculoskeletal radiologists with 5–19 years of experience. The pars injection protocol was standardized to ensure consistency of the intervention. The injections were guided by a 64- or 128-detector row CT (Philips Brilliance, Philips Medical Systems; SOMATOM Definition AS, SOMATOM Definition AS+, or SOMATOM Edge Plus, Siemens Healthineers). Patients had to lie still in a prone position during the procedure. CT imaging was acquired over 1–2 lumbar spine segments according to the previously obtained lateral scout view and confirmed the presence of a bilateral, uni-segmental spondylolysis in all patients. The best access approach for the needle (Sterican, 20 G 7 cm or 21 G, 12 cm, BRAUN) to the pars defect was chosen by the radiologist. Following skin disinfection and subcutaneous application of a local anesthetic, the needle was introduced under CT guidance from posterior until its tip touched the bone at the posterior aspect of the pars defect (Supplementary Fig. 1). For each pars defect, an injection of 0.5 mL iopamidol (Iopamiro 300, 300 mg of iodine per milliliter; Bracco) was performed to verify correct needle tip positioning. Following contrast agent injection, 2 mg (0.5 mL) of the crystalline corticosteroid triamcinolone (Triamcort Depot; Zentiva) was slowly injected on each side. Then 1 mL of 0.2% ropivacaine (Naropin; Astra-Zeneca) was slowly injected into each pars defect.

Pain levels were assessed right before the pars injection (baseline), 15 min after, and at follow-up 1 month (4–5 weeks) post-procedure (either in clinical consultation or by phone consultation). For assessment, the numerical rating scale (NRS) was used, where 0 = no pain, and 10 = worst pain imaginable. Percent change in NRS was calculated from pre-procedure to 15 min and to the first time follow-up post-procedure. The percentage of pain reduction was used to evaluate the primary categorical outcome of successful pain relief for both time points. Based on previous literature, the threshold for success was set at ≥ 50% pain reduction prior to data analysis [11, 17, 18]. All other responses were considered as not improved for our analysis.

Image analysis was performed using a commercially available PACS workstation (Merlin, Phoenix-PACS). All CT and MRI studies were anonymized and independently reviewed by two board-certified and musculoskeletal fellowship-trained radiologists from the same institution (G.W.K. and S.S.G., with 7 years and 6 years of experience, respectively). No further study-specific pre-readout training was undertaken by any of the readers. Examinations were analyzed in random order, whereby both readers were blinded to clinical data.

For each CT scan, readers had to indicate the pattern of contrast distribution during pars injection as either immediately around the pars defect (peri-defect), within the defect (intra-defect) or with contrast distribution through the defect reaching the epidural space (trans-defect) (Fig. 1). As the pars defect injection was bilateral, it was always the deeper reaching contrast distribution pattern that was chosen for classification.

Three different contrast distribution patterns during CT-guided lumbar pars injections in three different patients. Contrast distribution within the soft tissues immediately surrounding the pars defect, coined peri-defect (yellow arrowhead), contrast distribution (A). Contrast distribution within the pars defect, coined intra-defect (yellow arrowhead), contrast distribution (B). If the contrast agent reached through the pars defect into the epidural space (yellow arrowhead), the pattern was coined trans-defect contrast distribution (C). In case different patterns were observed on each side, the deeper contrast distribution pattern was chosen for case classification. Thus, image (C) represents a case of trans-defect contrast distribution

On MRI, readers assessed concomitant signs of degeneration of the lumbar spine, potentially related to lower back pain [19]: The presence of Modic Typ I changes [20] (yes or no), isthmic bone marrow edema (BME) (Supplementary Fig. 2) around the pars defect (yes or no) [21,22,23] and soft tissue edema [24] and/or synovial cysts [25, 26] in the surrounding soft tissue (yes or no) around the pars defect (Fig. 2 and Supplementary Fig. 3) were included in our analysis. Disc degeneration was graded according to Pfirrmann et al [27], facet joint degeneration according to Weishaupt et al [28], lateral recess stenosis according to Bartynski et al [29], and lumbar foraminal stenosis according to Lee et al [30]. Spondylolisthesis was measured in millimeters, as well as according to Meyerding [31]. Readers evaluated whether degeneration was limited to the segment of the pars defect or affected adjacent lumbar segments (predominantly uni-segmental vs multi-segmental). To qualify for classification as predominantly uni-segmental lumbar degeneration, bone edema, disc degeneration > Pfirrmann Grade 3, and facet joint degeneration > Weishaupt Grade 1 in more than one lumbar segment had to be excluded.

Forty-one-year-old male spondylolysis patient with CLBP, with 33% pain reduction at 15 min, and 100% pain reduction at 1-month follow-up after CT-guided bilateral pars injection. The sagittal (A) and axial (B) CT images show chronic bilateral spondylolysis at the L3 level with typical sclerotic margins of the lysis borders, which are also identified on the T2w TSE and T1w TSE images as hypointense bands along the pars defect (white arrowheads, A–E). Sagittal STIR images (E) reveal minimal isthmic BME (pink arrowhead) and adjacent soft tissue edema around the spondylolysis (yellow arrow). CLBP, chronic lower back pain; TSE, turbo spin echo

All statistical analyses were performed in SPSS Statistics (v. 29, IBM Corporation).

Normal and non-normal distribution of data was assessed graphically and analytically using Quantile–Quantile plots and the Shapiro–Wilk test. In addition to descriptive statistics, the Wilcoxon signed-rank test was used to assess significant differences in NRS at the different follow-up time points. To ensure consistency and increase methodological transparency, univariate and binary logistic regression analysis of the assessed imaging findings was performed for each reader individually, aligning with established practices [32], as follows: The Mann–Whitney U-test and the Chi-square-test were performed to assess significant imaging differences individually between successful pain relief patients and non-successful pain relief patients at the immediate and at the 1-month follow-up. For a more comprehensive model, binary logistic regression analysis was executed to evaluate significant imaging predictors for successful pain relief at both the 15 min and the 1-month follow-up. Considering their non-normal distribution, Spearman’s correlation coefficient (rs) was used to assess the correlation between NRS pain scores at the 15 min and 1-month follow-up. Inter-reader agreement (IRA) was analyzed by kappa statistics (Cohen’s κ) for all categorical variables. Intraclass correlation coefficient (ICC) was used for continuous variables. For the ICC, a two-way mixed-effects model with absolute agreement definition was applied. Single measures were used, as the raters‘ scores were not averaged in subsequent analysis. The level of agreement for Cohen’s κ was categorized as follows [33]: 0.0 = poor, 0.01–0.20 slight, 0.21–0.40 = fair, 0.41–0.60 = moderate, 0.61–0.80 = substantial, 0.81–1.00 = almost perfect agreement. The level of agreement for the ICC was categorized as follows [34]: 0–0.49 = poor, 0.5–0.74 = moderate, 0.75–0.89 = good, 0.9–1.0 = excellent agreement. All statistical tests were performed two-sided, and a level of significance (α) of 0.05 was used.