Baseline new bone formation does not predict bone loss in ankylosing spondylitis as assessed by quantitative computed tomography (QCT) - 10-year follow-up

In 1999 thirty seven males fulfilling New York modified AS criteria, with no concomitant diseases and drugs affecting bone metabolism, underwent X-ray (spine), DXA (spine and hip) and QCT (spine) measurements [7]. In 2009 fifteen patients underwent X-ray, DXA and QCT. Of the 22 excluded patients, 14 were treated with bisphosphonates and/or corticosteroids, 3 refused to participate, 1 lost contact. Three patients died. In one patient QCT was not feasible due to unsatisfied positioning.

We obtained standard radiographs of the lumbar and thoracic spine at baseline and follow-up.

DXA was performed at L2-L4 and left hip using Lunar DPX-IQ (1999) and Lunar Prodigy (2009, Lunar, USA). Right hip was scanned if the left was difficult to position or hip replacement had been performed. In one patient with new L2 and L4 fractures L3 was analyzed. The coefficient of variation (CV) for repeated measurements in vivo and least significant change at 95% confidence level (spine) were 2,2% and 0,062 g/cm² for DPX-IQ and 1,6% and 0,043 g/cm² for Prodigy, respectively. T-scores and g/cm² were based on manufacturer-supplied reference values.

QCT (L1-L5) was determined at baseline and follow-up with single-energy QCT (Twin Flash, Marconi, USA). Standard calibration phantom (Picker) recommended by Marconi was used for each scan. First, from the lateral lumbar spine scannogram the midportions of L1-L5 vertebral bodies parallel to the endplates were localized manually. Then, scans of 10 mm thickness were carried out in each vertebrae and other scanning parameters were set according to scanner manufacturer. Data from scanning elliptical region of interest (ROI, Figure 1B) including only trabecular bone, was analyzed using software CirsCT Bone Densitometry of CIRS company supplied by scanner manufacturer. T-scores were calculated using manufacturer's reference data. The fractured vertebrae were excluded from analyses.

We compared means using t-test and proportions using chi-square test, where applicable. To check the prognostic value of studied variables on follow-up BMC, we fitted multiple regression models, both without and after adjustment for baseline values of age, body mass index (BMI), erythrocyte sedimentation rate (ESR), smoking, QCT derived BMC stratum, radiologic score and duration of disease prior to baseline visit. To enable visual inspection of individual patient data, we plotted the baseline and follow-up values of measures of bone-density in strata of baseline BMC based on QCT. To formally test the between-subject (BMC baseline strata) differences in studied variables, we fitted the two-way ANOVA models. Finally, using linear regression approach, we assessed interaction between BMC and radiographic scoring at baseline, in their influence on measures of bone density at follow-up.

Mean ± SD baseline age of 15 included patients was 45.6 (± 7.3) years and disease duration averaged 16.5 (± 8.6) years. ESR averaged 41.8 mm/hr (± 27.5), BMI 27.4 (± 4.2), 33,3% patients were smoking. Three patients experienced four new vertebral fractures during follow-up (Th7, Th12, L2, L4). Baseline and follow-up densitometry results are shown in table 1.

We used stratification of patients based on baseline QCT according to Kalender et al. [8](normal >120; osteopenia 80--120; osteoporosis < 80; mg/cm³) to follow changes in QCT and DXA. In QCT at baseline there were 4 patients in normal stratum, 5 had osteopenia and 6 osteoporosis. To follow changing bone density data were plotted for individual patients (Figure 2). In spine QCT a statistically significant (p = 0,001) decrease of bone density (change ± SD: 18.0 ± 7.3) was observed, and it was universal across all strata (Figure 2). There were no clear-cut trends across strata in DXA that would resemble those noted in spine QCT. In spine DXA, a significant BMD change was noted (p = 0,0009) with trend towards increased density (change ± SD: -0.15 ± 0.14). Neither neck nor Ward DXA changed from baseline to follow-up. There was no correlation between QCT and DXA, both in density units and T- scores - at baseline and follow-up. Additionally, to test the between-subject (BMC baseline strata) differences in studied variables, we fitted the two-way ANOVA models. We found that the baseline stratification by BMC significantly influenced the BMC values both at baseline (p < 0.0001) and at follow-up (p = 0.0009), but had no impact on the indices derived from DXA (all p > 0.07). We found no interaction between baseline BMC stratum and time-related change in all studied variables (all p > 0.31).

To determine the effect of radiological severity on BMD/BMC change X-rays were assessed and patients were split into two groups based on baseline syndesmophyte scores according to Devogelaer [9] (table embedded in Figure 3); early group (grade 0-I, ie. no definite syndesmophytes) - 7 patients; and advanced group (grade II-IV, ie. bridging syndesmophytes) - 8 patients (Figure 3). At follow-up 6 patients progressed into advanced group.

The mean BMC, both at baseline and follow-up was significantly lower (p = 0.02 and p = 0.005, respectively) in advanced group (77.2 ± 21.5 and 55.4 ± 18.9 mg/cm³, baseline and follow-up, respectively) as compared to early group (113.6 ± 31.7 and 99.8 ± 31.1 mg/cm³, baseline and follow-up, respectively).

However, in multiple regression model after adjustment for baseline BMC, the baseline radiological scoring did not influence the progression of bone loss as assessed with QCT (p = 0.22, p for BMC*X-ray syndesmophyte scoring interaction = 0.65, p for ANOVA-based X-ray syndesmophyte scoring*time interaction = 0.39).

To discover the predictive value of other variables possibly influencing 10-year changes in BMC similar multiple regression model was used. Only baseline QCT values (mg/cm³) predicted 10-year decrease in QCT; with no influence of baseline DXA values, disease duration, age, ESR, smoking and BMI.