Anatomic structures involved in early- and late-stage sacroiliitis in spondylarthritis: A detailed analysis by contrast-enhanced magnetic resonance imaging




To localize inflammatory and chronic changes to defined areas in the sacroiliac joints in patients with early-stage compared with late-stage spondylarthritis (SpA), using magnetic resonance imaging (MRI).


Using MRI, 93 patients with SpA and inflammatory back pain who had radiographs of the sacroiliac joints were examined, comprising 31 patients with ankylosing spondylitis (AS) and 62 with other SpA subsets, including 48 with undifferentiated SpA (uSpA). MRI was performed using T1-weighted, T2*-weighted, STIR, and dynamic contrast–enhanced (gadolinium diethylenetriaminepentaacetic acid) sequences. Two readers retrospectively analyzed the images by differentiating 9 areas of the sacroiliac joints: the ventral and caudal joint capsule, cavum, subchondral bone, bone marrow, ligament entheses, and ligaments; the sacral and iliac sides were tabulated separately.


By MRI, sacroiliitis was more often bilateral in AS (84%) than in uSpA (48%) (P = 0.01). Inflammatory changes were found in a mean ± SD 4.7 ± 2.9 regions/joint, with involvement of 4.5 ± 3.2 regions in early disease versus 5.2 ± 2.3 regions in late disease (P not significant [NS]). Involvement of the iliac side of the sacroiliac joints was found to be more frequent than the sacral side in early disease (58% versus 48%; P < 0.01) as compared with that in late disease (58% versus 63%; P NS). The dorsocaudal parts of the synovial joint and the bone marrow were the most frequently inflamed structures in early disease (P < 0.001 for ventral versus dorsal joint capsule). In contrast, involvement of the entheses was more common in advanced disease (early 43% versus late 86%; P < 0.001). Similarly, the ligaments were more frequently involved in the late stages (early 26% versus late 40%; P = 0.06). Both patterns of bone marrow inflammation (focal and diffuse) were observed in equal frequencies in early and late disease (17% and 42% versus 26% and 43%, respectively; P NS). HLA–B27–positive patients (n = 80) had more entheseal involvement than did HLA–B27–negative patients (n = 13) (60% versus 39%; P = 0.05). HLA–B27–negative patients had a shorter disease duration (2.2 years versus 4.4 years; P = 0.05) and were more often female (62%; P = 0.02). When all pathologic changes were assessed, the STIR sequence (performed in 62 patients) was less sensitive than the contrast-enhanced sequences in that it was not able to show all relevant changes in 27% of these patients (n = 17), failing to reveal inflammation of the cavum in 15 patients and of the bone marrow and joint capsule in 1 patient each.


As visualized by MRI, sacroiliitis in SpA is characterized by involvement of different joint structures. Whereas the iliac and the sacral side of the sacroiliac joints are almost equally affected, the dorsocaudal synovial part of the joint is involved significantly more often than the ventral part, especially in early disease. Sacroiliac enthesitis is not a special feature of early sacroiliac inflammation.

The sacroiliac joints are centrally involved in the pathogenesis of the spondylarthritides (1). It has been proposed that enthesitis is the key change that occurs in ankylosing spondylitis (AS) (2) and possibly even in all spondylarthritides (3, 4). With regard to changes in the sacroiliac joints, the situation is not that clear. Recent work by investigators, including one of us (5), showed that synovitis and bone marrow inflammation, rather than enthesitis, explain the changes in the sacroiliac joints of patients with AS. Shichikawa et al in 1985 reported that inflammation in the subchondral bone may be the earliest event (6).

There is evidence provided by ourselves and others that magnetic resonance imaging (MRI) of the sacroiliac joints, especially if dynamic and STIR sequences are used, is a valuable tool for the diagnosis of sacroiliitis in the early and active stages (7–11). The advantage of MRI is its excellent ability to visualize the anatomy and the degree of inflammation without using ionizing radiation. Taking advantage of this property and our experience with the technique, we retrospectively analyzed 186 sacroiliac joints of 93 patients with spondylarthritis (SpA) in different stages of the disease to determine the relative involvement of relevant structures in and around the joint, with special reference to the sacroiliac ligaments and their entheses.

The ability of MRI to visualize inflammation can be obtained by 2 different techniques; one technique uses contrast agents such as gadolinium diethylenetriaminepentaacetic acid (Gd-DTPA), and the other technique uses a T2 approach and fat suppression, which is referred to as STIR. Whether both techniques have the same diagnostic sensitivity and specificity is not clear. To get some more insight into this clinically and also economically important question, we compared the 2 techniques in this study.


The MRI, radiographs, and in some cases, computed tomographies of 186 sacroiliac joints of 93 patients with inflammatory back pain were evaluated retrospectively (56 men and 37 women; mean age 32.2 years, minimum 16 years) (Table 1). Each patient's medical history was recorded by rheumatologists in the outpatient clinic of the Free University Hospital. The time of onset of inflammatory back pain (12) was considered to be the beginning of the disease; the disease duration was counted in years and months. The HLA–B27 status was measured by the standard lymphocytotoxicity test. All patients fulfilled the European Spondylarthropathy Study Group criteria for SpA (13). AS patients had to fulfill the 1984 modified New York criteria for AS (14). The diagnosis of psoriasis (PsA) had to be confirmed by a dermatologist. Inflammatory bowel disease was diagnosed only if overt clinical symptoms were present and if a gut biopsy yielded positive results. Reactive arthritis (ReA) was diagnosed only in cases with a preceding clinically symptomatic infection in the urogenital or intestinal tract (15).

Table 1. Patient characteristics*
 No. of patientsAge, mean ± SD yearsHLA–B27 positive, %Disease duration, mean ± SD yearsMen, %% of SIJ with active inflammation
  • *

    SIJ = sacroiliac joints; uSpA = undifferentiated spondylarthritis; AS = ankylosing spondylitis.

Total9332 ± 10864.1 ± 4.96082
Early disease group6232 ± 12811.3 ± 1.46177
 uSpA4830 ± 12771.5 ± 1.55676
 Non-AS/non-uSpA1439 ± 10930.8 ± 0.77979
Late disease group (AS)3133 ± 8979.6 ± 4.65892
 Very early disease2831 ± 13820.4 ± 0.16877
 Men5630 ± 13913.8 ± 4.610080
 Women3735 ± 11784.6 ± 5.285

Two groups of patients (Table 1) were defined on the basis of radiographic scoring of the disease according to the 1984 New York criteria (bilateral sacroiliac changes grade II or higher). Group 1 consisted of 62 patients in an early stage of SpA. Thus, radiologic changes did not exceed grade II on one side or grade I bilaterally. Of those 62 patients, there were 48 patients who had either possible early AS and a clinical suspicion of sacroiliitis but who did not meet the criteria for AS, ReA, PsA, or arthritis-associated inflammatory bowel disease (AIBD). These patients were classified as having undifferentiated SpA (uSpA). The remainder of this group were SpA patients with PsA (n = 5), ReA (n = 4), and AIBD (n = 5). This subgroup was defined as non-AS/non-uSpA. Group 2, the later-stage group, comprised 31 patients who fulfilled the modified New York diagnostic criteria for AS. All these patients had a disease duration longer than 4 years, inflammatory back pain, and bilateral radiologic changes of at least grade II. Furthermore, a subgroup with very early disease derived from group 1 was distinguished. These 28 patients (56 sacroiliac joints) had a disease duration of one-half year at maximum.

The MRI techniques used have already been described (16). Briefly, MRI was performed using a 1.5-Tesla superconductive unit (Magnetom Vision, Siemens, Germany). Patients were examined while lying in the supine position in the body coil. An initial survey measurement was obtained with a sagittal view. Then, a paraxial view almost parallel to the long axis of the upper sacrum, in accordance with the irregularly curved articular surface of the sacroiliac joints, was selected. The following sequences were used: T1-weighted turbo spin echo, with a repetition time (TR) of 500 msec, echo time (TE) of 10 msec, slice thickness of 4 mm, 2 acquisitions, matrix 360 × 512 pixels, and duration of 3 minutes 40 seconds; T2*-weighted opposed-phase gradient echo, with a TR of 350 msec, TE of 12 msec, flap angle of 30°, slice thickness of 4 mm, 2 acquisitions, matrix 256 × 256 pixels, and duration of 6 minutes; STIR, with a TR of 4,000 msec, inversion time of 150 msec, TE of 60 msec, slice thickness of 4 mm, 1 acquisition, matrix 242 × 256 pixels, and duration of 3 minutes (because of technical requirements, this has been available only since 1996, and was obtained from 62 patients); and contrast-enhanced T1-weighted dynamic gradient echo (before and after administration of 0.1 mmoles/kg body weight Gd-DTPA [Magnevist; Schering, Berlin, Germany]), with a TR of 50 msec, TE of 7 msec, flap angle of 70°, slice thickness of 5 mm, 2 acquisitions, matrix 256 × 256 pixels, and 7 repetitions of 53 seconds each. The MR images were visually judged by 2 of us (MB and BM) on a consensus basis according to recently published criteria (17). It was assumed that liquid deposits (hyperintense signal in STIR sequence) and enhancement after injection of gadolinium were equivalent to evidence of edema and inflammation, respectively.

As shown in Figure 1, 9 different anatomic regions were differentially detected. Furthermore, a differentiation between focal and diffuse bone marrow inflammation and between sacral and iliac involvement was made. The subchondrium was defined as the region of bone directly proximal to the joint space. For practical reasons, the term entheses was restricted to the ligamentous insertions proximal to the dorsal joint capsule. The localization term ligamentous was used for the interosseous sacroiliac ligaments in the retroarticular space. However, there are no sharp anatomic margins between the dorsal joint capsule, entheses, and ligaments. The space between the cartilage–subchondral bone on both sides of the sacroiliac joints was called cavum. In general, a differentiation by MRI between the cartilage and synovial cavity in the sacroiliac joints is not possible.

Figure 1.

Top, Anatomic sections of sacroiliac joints (SIJ) from a 65-year-old man. Bottom, Scheme of the distribution of the anatomic structures in the SIJ, to illustrate iliac versus sacral changes. Left, Cranial section. Right, More caudal section through the joint. 1 = cavum ventral; 2 = cavum dorsal; 3 = joint capsule ventral; 4 = joint capsule dorsal; 5 = subchondral bone ventral; 6 = subchondral bone dorsal; 7 = bone marrow (diffuse/focal changes); 8 = entheses; 9 = ligaments.

In addition, we assessed the significance and the value of the STIR sequences in relation to the other sequences. Pairs of STIR sequences and gadolinium-enhanced T1 images were available from 62 patients, of whom 31 were classified as having uSpA (50%) and 23 as having AS (37%), while 8 were in the non-AS/non-uSpA group (13%). The STIR sequences and the gadolinium-enhanced T1 images were independently analyzed.

The statistical analyses were performed using the chi-square test, the McNemar test with Bonferroni's correction for multiple testing, and the Mann-Whitney U test, using SPSS software (SPSS, Chicago, IL).


Patients in the early and late disease groups were generally similar, but differed in their disease duration, by definition, and in their HLA status, as shown in Table 1 (P < 0.001 and P < 0.01, respectively). Using the radiographs, only 1 sacroiliac joint in the late disease group appeared normal, whereas 42 of the early disease sacroiliac joints (34%) were graded as normal (stage 0) by radiography.

The majority of the SpA patients in this study (73%) had signs of active inflammation in both sacroiliac joints. All together, there were 34 sacroiliac joints (18% of all sacroiliac joints examined) in 29 patients (47%) of the early disease group and in 5 patients (16%) of the late-stage group (P = 0.01) without actual signs of inflammation detected by MRI (graded as “x” according to the method of Bollow et al [17]). Similarly, in the subgroup with very early disease, 46% of the patients had unilateral changes. Among the AS patients, 84% had bilateral changes, while in comparison, 48% of the uSpA patients had bilateral changes (P = 0.01).

Using MRI, of the 9 sacroiliac joint structures screened, a mean of 4.7 (±SD 2.9) regions per joint were involved, with evidence of active inflammation. Enhancement after application of Gd-DTPA was seen in 4.5 (±3.2) regions in early disease as compared with 5.2 (±2.3) structures in late disease (P not significant [NS]). Thus, in the number of structures that were actually inflamed, there was no difference between early and late disease. Inflammation of all 9 differentiated anatomic structures was found in 20 sacroiliac joints (Figure 2, left sacroiliac joint, and Figure 3), while 27 sacroiliac joints had a maximum of 3 involved structures, and only 5 sacroiliac joints had 1 or 2 inflamed structures.

Figure 2.

Magnetic resonance image of the sacroiliac joints (SIJ) of a 22-year-old male patient with undifferentiated spondylarthritis. A, STIR sequence. Diffuse inflammation of all structures is evident in the left SIJ (especially the cavum, subchondral bone, and joint capsule [large thick arrow]), with diffuse bone marrow involvement (asterisk). In the right SIJ, inflammation of the dorsal focal bone marrow (curved arrow) and of the dorsal subchondral bone and joint capsule (small thin arrow) are shown. B, T1-weighted image. C, T1-weighted image with contrast enhancement.

Figure 3.

Magnetic resonance image of the sacroiliac joints (SIJ) of a 49-year-old female patient with reactive arthritis. A–C, T1-weighted sequence of 3 sections, from cranial to caudal. D and E, Dynamic T1-weighted image D, prior to contrast injection, and E, after contrast enhancement. Even in the unenhanced pictures, a thickening and darkening of the enthesial and ligamentous structures can be seen (open arrows). After administration of gadolinium diethylenetriaminepentaacetic acid, there is enhancement of all differentiated structures in both SIJ (for instance, the ventral joint capsule [solid arrows]). High pre-contrast spots in the bone marrow are a result of hemangiomas and are not correlated with the inflammation.

The dorsocaudal parts of the synovial joint and the bone marrow were the most frequently inflamed structures in early disease (Figure 4). The difference in frequency of involvement between the ventral and the dorsal joint capsule was statistically significant (P < 0.001), while there was only a trend toward differences for the cavum and the subchondrium (after Bonferroni's correction). Predominant inflammation in the dorsal part of the synovial joint is shown in Figure 2 (right sacroiliac joint).

Figure 4.

Involvement of different structures of the sacroiliac joints (SIJ) in patients (pats.) with early versus late spondylarthritis. ∗ = P < 0.05 between early and late disease. (+) = P < 0.001 between ventral (ventr.) and dorsal (dors.) joint capsules in early disease. Subchondr. = subchondrium.

Similarly, in the very early disease subgroup (Figure 5), the dorsocaudal structures were the most frequently inflamed regions. This was also the case in the sacroiliac joints with limited inflammation and those with a maximum of 3 involved structures (27 joints among 15 patients with uSpA, 7 with AS, and 5 in the non-AS/non-uSpA group), in which the dorsal cavum and dorsal joint capsule were most frequently inflamed (63% and 59%, respectively), while, in comparison, inflammation of the ventral joint capsule, ventral subchondral bone, ligaments, and bone marrow was seen in <20% of structures (ventral versus dorsal, for cavum P = 0.05, for joint capsule P = 0.02, for subchondral bone P = 0.07 as a trend).

Figure 5.

Involvement of anatomic structures in the subgroup of patients (n = 26, 52 sacroiliac joints) with very early disease (maximum disease duration of 0.5 years). See Figure 4 for definitions.

Involvement of the iliac side (bone marrow and subchondral bone) of the sacroiliac joints was as frequent in the early as in the late disease stages (58% in both), while, in contrast, sacral inflammation was found to be more frequent in the later disease stages (only 48% of the early-stage versus 63% of the late-stage patients; P < 0.01). Similarly, in the very early disease subgroup (Figure 5), there was more iliac than sacral involvement (54% versus 48%; P NS). Taken together, there was significantly more iliac than sacral inflammation in early disease (58% versus 48%; P < 0.01), while this difference had no role in late disease.

Involvement of the entheseal structures, differentiated by MRI, was more common in advanced disease (Figure 6) as compared with the earlier stages of disease (86% versus 43%; P < 0.001). Similarly, involvement of the ligaments was more frequently found in the late disease stages (40% versus 26% of early-stage patients; P = 0.06).

Figure 6.

Magnetic resonance image of the sacroiliac joints (SIJ) of a 37-year-old male patient with advanced ankylosing spondylitis (disease duration 10 years). A, A more cranial T1-weighted image. B, A more caudal T1-weighted image. C,, Caudal STIR sequence. Widely ankylosed SIJ are shown in A (asterisks), but in more caudal sections (B), strong inflammation of the entheses and ligaments is present (arrows). STIR does not reproduce well, but inflammation was equally visualized (C).

Diffuse bone marrow inflammation (42%) (Figure 2, left sacroiliac joint) was more frequently detected than focal spots (20%) in all disease stages. Both patterns of bone marrow inflammation (focal and diffuse) were observed in equal frequencies in early and late disease (17% and 42% versus 26% and 43%, respectively; P NS). Thus, sacroiliac bone marrow inflammation in SpA is often spread out and does not correlate with disease duration.

A decrease in subchondral involvement was observed between early and late disease (40% and 34% ventral, respectively [P NS] versus 57% and 40% dorsal, respectively [P = 0.04]). However, in late disease, replacement of the subchondral bone (according to our definition) by bone marrow was seen in all cases with advanced ankylosis of the sacroiliac joints.

HLA–B27–positive patients (n = 80) tended to have more entheseal involvement than did those negative for HLA–B27 (n = 13; 60% versus 39%; P = 0.05), while, in contrast, the other regions did not show much difference according to HLA status. However, although the mean age of the HLA–B27–positive and HLA–B27– negative SpA patients was similar, the mean disease duration and the sex ratio were significantly different; HLA–B27–negative patients had a shorter disease duration (2.2 years versus 4.4 years; P = 0.05) and the relative percentage of female patients was higher among the HLA–B27–negative patients (62% versus 36%; P = 0.02).

In women, we observed less pathologic changes in the ventrocranial parts of the sacroiliac joints than in men (on average 39% versus 53%). This difference was accentuated in the region of the subchondral bone (P = 0.01) and in the ventral capsule (P = 0.05), while the differences for the ventral cavum were not significant (Figure 7). In contrast, women had more ligamentous inflammation than did men (41% versus 24%; P = 0.02). However, the women were somewhat older than the men at the time of examination (35 years versus 30 years; P = 0.01), but the mean disease duration of female and male SpA patients was similar. As already mentioned, men were more frequently HLA–B27 positive than women (91% versus 78%; P = 0.01). The prevalence of AS among all SpA patients was as high in male (32%) as in female (35%) patients (P NS).

Figure 7.

Involvement of anatomic structures in the patient subgroups men versus women. ∗ = P < 0.05 between men and women (for ventral joint capsule P = 0.005). See Figure 4 for definitions.

STIR sequences were less sensitive than the contrast-enhanced sequences. Compared with the findings on contrast-enhanced imaging, acute inflammatory changes were less clearly depicted in 17 of 62 patients (27%) by the STIR technique. Mainly, inflammation of the joint cavum was not detected by fat suppression (15 of 17 patients). Insufficient visualization of pathologic changes at the joint capsule and the bone marrow was found in only 1 patient each. Moreover, in 1 patient each, a signal of increased water content was seen in the cavum and the bone marrow, which could not be reproduced and confirmed in other sequences; thus, these changes were considered to be false positive. Patients with reduced information in the STIR sequences did not differ from the remaining patients by sex, disease duration, age, HLA status, or AS ratio. The relative differences between the groups remained unchanged when only the inflamed joints were assessed.


The aim of this study was to conduct a detailed analysis of sacroiliac inflammation in SpA patients, because the disease usually starts in this region in the majority of cases. This analysis was made possible by using MRI, a technique that is able to simultaneously provide evidence of both active and chronic sacroiliac inflammation. Nine different regions of the sacroiliac joint were identified in this study, which provides new data on the fine specificity of sacroiliitis in SpA.

In general, a mean of 5 structures was actively inflamed. However, all sacroiliac structures differentiated in this study were involved, and all of these were also affected, although to a different degree, in an early stage. This general lack of a major difference between the early and later disease stages is consistent with the view that AS does not burn out and basically needs to be also treated in the later disease stages by appropriate antiinflammatory agents.

The region most frequently involved at an early stage is the dorsocaudal synovial part of the joint and the bone marrow. This was also true in the subgroup with very early disease and in the patients with more limited involvement of <4 anatomic regions. The early predominance of the dorsocaudal part of the sacroiliac joints observed in this study relates to the involvement of the cavum, the capsule, and the subchondrium, which, rather, represent the synovial part of the joint. This is consistent with data from a systematic histologic study on detailed sacroiliac joint pathology in AS (5). Thus, the data from this study suggest that sacroiliac inflammation starts in the synovial part of the sacroiliac joints.

Whether entheses or synovial structures are predominantly and primarily involved in SpA and sacroiliitis has been a matter of recent debate. This question has been a real challenge to us when preparing the evaluation of the MR images. The problem starts, as recently discussed at the First Workshop on Enthesitis and Ankylosis in AS (18, 19), with the definition of enthesis, then continues to the localization of entheseal and synovial structures in the sacroiliac joints and ends with the ability of the MRI technique to differentiate small sacroiliac structures.

An enthesis is the anatomic structure where ligamentous structures attach to bone. There are 2 types: fibrous and fibrocartilaginous entheses. In the sacroiliac joints, both types are present in the attachment of the joint capsule to the bone and in the attachment of the dorsal ligaments. Furthermore, the attachment of the articular cartilage to the subchondral bone could be regarded as an enthesis. There can be no reasonable doubt that the best anatomic visualization of the sacroiliac joint structures is achieved by MRI. However, the image is 2-dimensional, and entheseal structures can, in general, be rather small and it is difficult to separate the joint capsule and the ligaments themselves from their attachment to bone.

Thus, with this in mind, before starting the evaluation, we had to define which structures were differentiated. As already mentioned, 9 different anatomic regions were detected (Figure 1). Clearly, sacral and iliac involvement had to be separately assessed, as discussed above. The differentiation between focal and diffuse bone marrow inflammation was made because we hypothesized that a higher percentage of focal inflammation would be evidence in favor of the bone marrow as the initial localization. However, this was not the case. It seems possible that, in patients with diffuse bone marrow involvement, the overall degree of inflammation is simply more extensive.

The subchondrium, defined as the region of bone directly proximal to the joint space, was differentiated because, in a histopathologic study on open biopsy samples, Shichikawa et al reported several years ago (6) that this might be the structure initially involved. There was no clear evidence in our study that this might be the case. The space between the subchondral bones was called cavum. It must be stressed that a differentiation of joint cartilage and “real” synovial cavum in the sacroiliac joints is not possible, even by contrast-enhanced or STIR MRI. The directly surrounding soft tissue of the dorsocaudal joint where the interosseous ligaments insert into bone was defined as entheses of the sacroiliac joint. The localization term ligamentous was used for the structures originating from the interosseous ligaments in the retroarticular space, which includes ligamentous insertions in its region. However, in MRI there are no sharp margins between the joint, entheses, and ligaments in this area.

Considering our results and keeping in mind that the ability of MRI to analyze these structures in detail is limited, we suggest that there is more evidence for synovial than for entheseal involvement in an early disease stage in this study. This is despite the relative overrepresentation of ligaments and entheses over the synovium in the sacroiliac joints. However, since one of us has histopathologic evidence (François RJ: unpublished data) that there are entheseal structures not only at the ventral and dorsal capsules, but also all along the sacral and iliac surface in the retroarticular space, we are cautious to state that the question of enthesitis in the sacroiliac joints is now completely resolved. The ligaments in that area are clearly exposed to mechanical stress (20). Whether this contributes to sacroiliac inflammation is unknown.

The blood supply of the sacroiliac joints may have a role in the increased prevalence of sacroiliitis in young patients with septic sacroiliitis and AS (21–23). There are no reports to suggest that there is a different blood supply or a different innervation of the dorsal region of the sacroiliac joints.

Although the majority of the SpA patients in this study (73%) had signs of inflammation in both sacroiliac joints, inflammation in the sacroiliac joint at the onset of disease was more often unilaterally detected (47% in the early disease group versus 16% in the late disease group had joints without signs of inflammation). This is only somewhat different from the historical view implying that unilateral sacroiliitis is typical of ReA and PsA. These data argue more in favor of this being a question of time—the longer and more active the disease runs, the higher the probability of bilateral involvement. Possibly, the 53% of patients with bilaterally involved joints from the early group will proceed to develop AS in later disease stages.

Expectedly and consistent with recent histopathologic findings by François et al (5), the iliac side of the sacroiliac joint was more frequently involved in early disease. However, probably due to the relatively small numbers, this was not statistically significant in the very early subgroup who had a maximum of 0.5 years disease duration. It seems possible that the thin iliac cartilage is responsible for this difference. However, there are more differences between the iliac and the sacral cartilage. In the recent study by McLauchlan and Gardner (24), the sacral articular cartilage was thicker than the iliac (1.8 mm versus 0.8 mm; P < 0.001). However, the iliac cartilage cell density in all zones was higher than the sacral (overall mean 31.2 × 10−3/mm3 versus 23.2 × 10−3/mm3; P < 0.001). Furthermore, iliac subchondral bone end-plates were thicker than the sacral (0.36 mm versus 0.23 mm; P < 0.001), and the iliac subchondral cancellous bone was twice as dense as the sacral (22.1% versus 12.1%; P < 0.001). These differences are likely to influence the onset and course of sacroiliac inflammation. Although the sacral cartilage is a rather strong tissue with low cellularity and limited bone metabolism, the contrary seems the case for the iliac cartilage, which may be more easily affected by inflammation and, possibly, infection.

Although the numbers of HLA–B27–negative patients were rather small, the data suggest that HLA–B27 seems to be associated with entheseal changes. This is an interesting and not really expected observation. Indeed, although it is one of the most characteristic features of spondylarthritides, enthesitis has been directly linked to HLA–B27 in only one recent study (25). In that study, on 28 patients with symptoms in the region of the plantar fascia, bone edema in the adjacent calcaneum was evident in both SpA patients (65%) and patients with mechanically induced disease (45%). HLA–B27 was identified in 53% of the SpA patients as compared with none of the controls, and all 6 SpA patients with extensive bone edema were HLA–B27 positive. However, the observed differences in our study may be due to disease severity and duration, rather than being due to HLA–B27 itself, since, although the mean age of HLA–B27–positive and –negative SpA patients was similar, the mean disease duration and the sex ratio were significantly different; HLA–B27–negative patients had a shorter disease duration (2.2 years versus 4.4 years; P = 0.05) and more of these patients were female (62% versus 36%; P = 0.02). Thus, this interesting question needs further study.

A different prevalence of inflammation at certain sites was observed between the sexes in our study. The anterior parts of the sacroiliac joints were less frequently involved in female patients. It can be hypothesized that the lack of ventral sacroiliac inflammation is a good prognostic sign for the outcome of SpA since, in general, women have less severe spinal involvement, as assessed by radiography (26). Furthermore, female SpA patients had some more ligamentous (P = 0.02) and entheseal (P = 0.07) inflammation. This has not been observed before and it is not easily understood. Whether female SpA patients, or even women in general, have more involvement of the ligaments and entheses has not been systematically studied to date. However, there are clear differences in the susceptibility and recognition of rheumatic diseases according to sex. Rheumatoid arthritis (RA), in contrast to AS, occurs more frequently in women. Female AS and RA patients perceive pain more frequently and more intensively than do men (27, 28), and painful tender points are much more prevalent in female than in male patients with fibromyalgia (29). On the other hand, there may be influences from pregnancies and birth that have effects on the sacroiliac joints, such as elevated mobility. These interesting questions need further study.

The STIR technique was not able to detect all of the pathologic changes visualized by Gd-DTPA–enhanced imaging. In particular, inflammation of the joint cavum was not well detected by STIR. An elevated signal intensity seen in the STIR sequence in the cavum of one patient and bone marrow of another was judged to be false positive because of a complete lack of change after contrast application. However, there is little evidence that this is very critical for the diagnosis of sacroiliitis and SpA. This can be possibly clarified in future examinations. Nevertheless, we still consider MRI sequences after contrast application as the gold standard. It seems probable that in studies performed without contrast material, synovitis will be less well detected. Thus, the well-published observations on the predominance of peripheral enthesitis in SpA (30) may need to be repeated with contrast application.

The use of Gd-DTPA in the MRI assessment of sacroiliitis is associated with higher costs and more time expenditure. On the basis of our experience, we have tended to keep on using both procedures in our investigations in accordance with the practice of other colleagues (31–33), until the equal validity of the STIR technique on the basis of a controlled study using blinded reading procedures (STIR versus contrast enhancement) is proven. However, due to the lack of standardization, we do not see a basis to recommend this for daily routine. At least in the present study, the use of STIR seems to be cost and time saving.

Even if the connection between cartilage and subchondral bone is considered as an enthesis, the data of our study suggest that entheseal and ligamentous structures are more frequently inflamed in the later, but not in the early, disease stages. This is in contrast to the reports by others (34) on peripheral arthritis in which the enthesis is regarded as the starting point of SpA. Consistent with recently published histologic findings (5), we did not observe ligament enthesitis in any one of the sacroiliac joints without inflammation of synovial structures.

Our finding that sacroiliac inflammation starts predominantly in the dorsocaudal part of the sacroiliac joints is also relevant to clinical practice, because this region is not well visualized by conventional radiography. More study is needed to assess the relevance of this finding for the early detection of sacroiliitis by radiography.

Because of the demonstrated inflammation in both the early and late stages of AS and other SpA, we conclude that there is reason to treat patients with local or systemic antiinflammatory agents at all time points in correlation with disease activity (35, 36), which can be documented and proven by MRI (Figure 6). STIR and contrast application techniques are almost, but not entirely, equivalent for the detailed evaluation of sacroiliitis.