Inter-observer agreement was “substantial” for ASAS score (κ = 0.63); “good” for sclerosis score (ICC = 0.82) and sclerosis depth (ICC = 0.83); and “excellent” for global assessment score (ICC = 0.93) and total BME score (ICC = 0.95).
Based on the recruitment criteria (Fig. 1), a total of 349 female patients with a mean age of 47 ± 14 years (range 18-87 years) were included. Two hundred thirty of 349 (66%) subjects have had at least one childbirth in their history as opposed to 119 of 349 (34%) without childbirth. Fifty-three of 349 (15%) included patients have SpA (SpA+), 296 of 349 (85%) do not have SpA (SpA−), 207 of 3494 (59%) patients had a 1.5 T MRI, and 142 of 349 (41%) had a 3 T MRI of the sacroiliac joints. Four subgroups were formed based on the criteria SpA and childbirth: SpA−CB− (n = 103), SpA−CB+ (n = 193), SpA+CB− (n = 16), and SpA+CB+ (n = 37), respectively. The magnetic field strength ratio (1.5 T/3 T MRI) was evenly spread in all subgroups: SpA−CB− 59%/41%, SpA−CB+ 59%/41%, SpA+CB− 62%/38%, and SpA+CB+ 59%/41%, p = 0.99.
The portion of patients with a BME score of ≥ 1 in the entire cohort was 51.3% (179/349), and the percentage with a sclerosis score of ≥ 1 was 59.9% (209/349). In the four subgroups, a BME score of ≥ 1 was seen in 42.7% for group SpA−CB−, in 51.8% for SpA−CB+, in 68.7% for SpA+CB−, and in 64.9% for SpA+CB+. Accordingly, a sclerosis score of ≥ 1 was given in 39.8% (group SpA−CB−), 64.2% (SpA−CB+), 75.0% (SpA+CB−), and 86.5% (SpA+CB+), respectively.
Patients in subgroup SpA−CB+ were 51 ± 12 years old and in regard to childbirth, 69.4% of patients (134/193) had exclusively vaginal deliveries, 18.7% (36/193) only C-sections, and 11.9% (23/193) at least one of each. The mean number of children in subgroup SpA−CB+ was 2.1 ± 1.0 (range 1-7), and the average time between MRI and last childbirth was 254.7 ± 155.4 months (range 27-753 months). In subgroup SpA+CB+, the mean age was 48 ± 12 years, and 59.5% of patients (22/37) had only vaginal deliveries, 24.3% (9/37) only C-sections, and 16.2% of patients (6/37) at least one of each. The number of children in subgroup SpA+CB+ was 2.4 ± 1.4 (range 1-7) and the average time between MRI and last childbirth was 220.6 ± 143.0 months (range 25-549 months).
Correlation analysis in the entire cohort (n = 349)
Table 2 shows the influence of age, CB, and SpA diagnosis on BME, sclerosis, overall assessment score, and ASAS score.
Age and CB did not correlate with BME score (p = 0.33; p = 0.38, respectively), whereas SpA diagnosis was positively associated with BME score (r = 0.31, p < 0.001). Age showed no correlation with sclerosis score (p = 0.24) or sclerosis depth (p = 0.21); however, both CB and SpA diagnosis were positively associated with sclerosis score (r = 0.21; r = 0.33; p < 0.001, respectively) and sclerosis depth (r = 0.20; r = 0.30; p < 0.001, respectively) (Fig. 3). Age (r = 0.05; p = 0.31) and CB (r = 0.1; p = 0.07) did not correlate with global assessment score, as opposed to a significant positive correlation between SpA diagnosis and global assessment score (r = 0.61; p < 0.001). Age showed no association with ASAS score (r = 0.1; p = 0.06), whereas both CB (phi = 0.13; p = 0.02) and SpA diagnosis (phi = 0.23; p < 0.001) were associated with a positive ASAS score.
Detailed numerical data for all 6 subgroup comparisons are shown in Table 3. BME score, sclerosis score, sclerosis depth, global assessment score, and ASAS score in each subgroup are illustrated in Figs. 4 and 5.
Age was significantly lower in group SpA−CB− compared to both SpA−CB+ (p < 0.001) and SpA+CB+ (p = 0.004). Additionally, patients in group SpA−CB+ were significantly older compared to group SpA+CB− (p = 0.02). The other subgroup analysis yielded no significant differences.
The subgroup SpA−CB− showed a significantly lower BME score compared to SpA+CB− (p = 0.01) and SpA+CB+ (p = 0.007). The remaining subgroup comparisons showed no significant differences (p ≥ 0.16).
Both sclerosis score and sclerosis depth were significantly lower in SpA−CB− compared to the other three groups (p ≤ 0.008). The sclerosis score and depth were significantly lower in SpA−CB+ compared to SpA+CB+ (p = 0.001). The remaining group pairs showed no significant differences (p ≥ 0.89).
Global assessment score
Both SpA+ groups showed significantly higher scores compared to both SpA− groups (p < 0.001). No significant difference was seen between SpA+CB− and SpA+CB+ (p = 1.0) as well as between SpA−CB− and SpA−CB+ (p = 0.42).
Both SpA+ groups showed significantly more ASAS positive SIJ-MRI compared to both SpA− groups (p ≤ 0.03). Additionally, SpA−CB+ had significantly more ASAS positive SIJ-MRI compared to SpA−CB− (p = 0.003).
Correlation analysis in subgroup SpA−CB+ (n = 193)
Table 4 illustrates the influence of age and childbirth-related factors on BME, sclerosis, overall assessment score, and ASAS score.
No association was found between age and BME score (r = 0.028, p = 0.70), time since last childbirth and BME score (r = 0.063; p = 0.38), birth method and BME score (eta = 0.11; p = 0.69), and number of children and BME score (r = 0.10, p = 0.17).
A weak correlation was observed between age and total sclerosis score (beta = − 0.23, p = 0.001; regression coefficient = − 0.038; R-squared 0.054) as well as age and total sclerosis depth (beta = − 0.25, p < 0.001; regression coefficient = − 0.088; R-squared 0.062). In addition, there was a weak correlation between time since last childbirth and total sclerosis score (beta = − 0.18, p = 0.01; regression coefficient = − 0.002; R-squared 0.033) as well as between time since last childbirth and total sclerosis depth (beta = − 0.19, p = 0.01; regression coefficient = − 0.005; R-squared 0.035).
There was no significant association between birth method and total sclerosis score (eta = 0.055; p = 0.75) or total sclerosis depth (eta = 0.026; p = 0.74). However, the number of children showed a significant correlation with total sclerosis score (beta = 0.22, p = 0.003; regression coefficient = 0.45; R-squared 0.05) and with total sclerosis depth (beta = 0.16, p = 0.02; regression coefficient = 0.73; R-squared 0.027).
Tables 5 and 6 show results of multivariable regression analysis: Higher age was an independent predictor of both a lower sclerosis score (beta = − 0.43; p = 0.005) and a lower total sclerosis depth (beta = − 0.47; p = 0.002). Additionally, number of children was an independent predictor of a higher sclerosis score (beta = 0.25; p = 0.001) and a higher total sclerosis depth (beta = 0.20; p = 0.005). However, both models poorly predicted the effect on total sclerosis score with R-squared = 0.12 and total sclerosis depth with R-squared = 0.11 (independent variables: age, time since latest childbirth, number of children).
Global assessment score
No association was detected between age (r = 0.002, p = 0.98), time since last childbirth (r = 0.02; p = 0.76), birth method (eta = 0.08; p = 0.74), number of children (r = 0.008, p = 0.92), and global assessment score.
No correlation was found between age and ASAS score (r = 0.05, p = 0.45), time since last childbirth and ASAS score (r = 0.09, p = 0.20), and birth method and ASAS score (phi = 0.09, p = 0.42) or between number of children and ASAS score (r = 0.01, p = 0.88).
Distribution of BME and sclerosis along the SIJ
BME score was equally distributed in the upper and lower half along the cartilaginous portion of the SIJ for all four subgroups (p > 0.22), whereas subchondral sclerosis revealed a marked predominance in the upper half of the SIJ in all four subgroups (p < 0.006) (Table 7 and Fig. 6).