Multilevel examination of minor salivary gland biopsy for Sjögren's syndrome significantly improves diagnostic performance of AECG classification criteria

We retrospectively studied a consecutive series of patients thoroughly investigated at our hospital between 1998 and 2002 for suspected primary SS, including a follow-up of at least 1 year after the diagnostic evaluation. Patients with secondary SS or who had been diagnosed by biopsy as having nonspecific inflammation, fibrosis, and atrophy of the gland were excluded [10‐12]. Less-than-optimal tissue area (biopsy section area less than 4 mm²) was not considered a criterion for exclusion, provided that at least one normotrophic glandular lobule had been sampled.

All patients had undergone thorough clinical and instrumental evaluation [3, 4], including MSGB performed as suggested by Daniels [11]. The diagnosis of SS was established for all patients according to the classification criteria proposed by the AECG [5]. MSBG samples were fixed in formalin, processed, and embedded in paraffin according to standardized laboratory methods. Baseline histopathological slides containing 4-μm-thick sections stained with hematoxylin and eosin were reviewed by a pathologist, blinded to clinical and laboratory data, who recorded for each patient the number of glands, the sample surface area, the presence of alterations suggestive of nonspecific sialoadenitis, and the baseline FS [10, 11]. The lymphocytic focus and the focus score were defined according to Greenspan and Daniels [10, 11]. In individual biopsies, lobules with acinar atrophy and diffuse fibrosis were excluded from diagnostic evaluation. The histological parameter was considered as negative in the absence of any inflammatory infiltrate (FS = 0) and in the presence of less than 1 focus per 4 mm² (0 < FS < 1) [5]; the presence of one or more foci per 4 mm² was considered positive when the adjacent glandular parenchyma was histologically normal. We further classified patients with a positive FS into two groups, those with fewer than two foci per 4 mm² (1 ≤ FS < 2) and those with two or more (FS ≥ 2). The area of the biopsy sections was assessed with video-assisted morphometric software capable of measuring the area of delineated surfaces (ImageDB System, Casti Imaging, Cazzago di Pianiga, Italy). The comparison of automated and manual area measurements of a smaller series of MSGB sections did not show a significant difference (data not shown). This prompted us to choose the automated system to simplify the examination of the large number of samples involved in the study.

Sample blocks were recut at two additional levels, about 200 and 400 μm deeper than the original section. Sections 4 μm thick corresponding to these levels were collected on separate slides and stained with hematoxylin and eosin. Considering that an infiltrate of 50 lymphocytes in our section had a mean diameter of 50 μm, we assumed that the interposition of 200 μm between the evaluated sections was enough to ensure that the FS recorded at each level was independent of the other two and that if the same focus was present in two section levels, the focus itself was large enough to justify repeated scoring. The two new sections were blindly examined by the same pathologist, who again recorded the area and the focus score for each level. For each patient, the total number of foci at all three levels and the total surface area measured at all levels were used to calculate a cumulative FS (cFS) for the three sections.

The cFS obtained after re-evaluation was entered in the AECG criteria set [5], to obtain a re-classification of each patient. To compare the diagnostic performance of the original classification and the reclassification, a 'gold standard' was needed independent of the AECG criteria set. We adopted as reference standard the opinion of experienced clinicians, analogously to what had been done by the European Community Study Group on Diagnostic Criteria for Sjögren's Syndrome when SS and control patients were selected to validate the proposed criteria [3‐5]. Briefly, three experienced rheumatologists, blinded to the results of the histopathological re-evaluation, performed a clinical evaluation of each patient and reviewed the patient's charts including the original clinical, laboratory, and instrumental evaluation, and the subsequent documentation covering at least 1 year of follow-up and treatment response. On this basis they were requested to judge whether individual patients had SS.

Quantitative data are shown as means ± standard deviation (SD). Specificity and sensitivity were assessed with their 95% confidence intervals (CI). Differences in frequencies were evaluated by means of chi-square statistics or the Fisher exact test, as appropriate. Given the known limitations of diagnostic accuracy as a parameter for measuring the diagnostic performance of a test, specificity and sensitivity were compared using receiver operating characteristic (ROC) curves [14]. A P value of less than 0.05 was considered to indicate statistical significance. All tests were two-sided. Analyses were performed with Statistica for Windows (StatSoft Inc, 2002, Tulsa, OK, USA) and MedCalc software.

The study series comprised 138 patients, 65 of whom had a baseline FS = 0, 14 with 0 < FS < 1, 18 with 1 ≤ FS < 2, and 41 with FS ≥ 2. Eighteen patients had incomplete clinical data that hampered either the AECG classification or the clinical re-evaluation. These patients (8 with FS = 0, 3 with 0 < FS < 1, 3 with 1 ≤ FS < 2, and 4 with FS ≥ 2) were excluded from further analysis. The final series included 120 patients, for whom demographic, biopsy, and clinical data and the result of the clinical re-evaluation are presented in Table 1.

In 96 (80%) of the 120 biopsies, the FS group did not change after serial sectioning and calculation of the cFS. In 14 of these biopsies, the FS group changed but this did not affect that patient's negative or positive status. In the biopsies for the other 10 patients, 1 (1.7%) of the 57 with a baseline FS = 0 and 1 (9%) of the 11 with a baseline score of 0 < FS < 1 switched to a FS consistent with SS according to AECG criteria (FS ≥ 1). At clinical re-evaluation, these two patients were considered not to have SS. Seven (46%) of the 15 patients with a baseline score of 1 ≤ FS < 2 and one (3%) of 37 with a baseline FS ≥ 2 switched to a grade inconsistent with SS (FS < 1). On clinical re-evaluation, 7 of these 8 patients were assessed as not having SS.

When the cFSs were entered in the AECG criteria set [5], the baseline classifications of the 63 non-SS patients were not changed, while the classifications of 7 of the 57 patients originally classified as having SS were changed to non-SS (Table 2). The classification was changed in 6% of the 120 patients. Six of these seven patients had a baseline score of 1 ≤ FS < 2 and one had a baseline FS ≥ 2. On clinical re-evaluation, all these seven patients were judged not to have SS. The clinical re-evaluation also refuted 7 of the 113 (6.2%) classifications that had not been changed at biopsy revision. Considering the clinical re-evaluation as the reference gold standard, the number of false-negative AECG classifications did not change (3 of 63 AECG non-SS cases), while the number of false positives was reduced from 11 to 4 (63.6% reduction).

In the present series of 120 patients fully evaluated for SS, the sensitivity and specificity of the baseline AECG criteria set were 93.9% and 84.5%, respectively. Reclassification with cFS did not affect sensitivity, whereas specificity changed to 94.4% (P = 0.056), increasing the accuracy from 88.3% (95% CI 81.2–93.5) to 94.2% (95% CI 88.3–97.6). Pairwise comparison of the ROC curves showed a statistically significant difference between patient classification before and after multilevel FS evaluation (difference between areas: 0.049 [SE 0.021]; 95% CI 0.009–0.089; P = 0.016) (Fig. 1). Sensitivity and specificity did not change for biopsies with FS = 0 or FS < 1 (inconsistent with SS), while specificity increased substantially in biopsies consistent with SS (FS ≥ 1) (Table 2). Pairwise comparison of the ROC curves showed a statistically significant difference (P = 0.013) only in biopsies with 1 ≤ FS < 2 (difference between areas: 0.43 [SE: 0.17]; 95% CI 0.09–0.76; P = 0.013; Fig. 1). The diagnostic accuracy of the MSGB histological analysis considered independently of other criteria changed from 85.8% (95% CI 78.3–91.5) to 90.8% (95% CI 84.2–95.3), but the comparison of the ROC curves did not show a statistically significant difference (P = 0.15).