From 1 January 1990 to 31 June 2005 serum samples were collected from consecutive patients, when they first presented to our unit, who fulfilled the inclusion criteria of: a definitive diagnosis of SSc according to the American College of Rheumatology 16; a disease duration less than 4 years from the onset of skin involvement; and no previous treatment with any of the following drugs: azathioprine, chlorambucil, colchicine, cyclophosphamide, cyclosporine, D-penicillamine, methotrexate or mycophenolate mofetil.

Sixty-three patients (51 women, 12 men) with a median age of 54 years (range 23 to 78 years) and a median disease duration of 18 months (range 3 to 42 months) met these criteria. Forty-five patients (38 women, seven men) had limited cutaneous SSc, and 18 patients (13 women, five men) had proximal skin involvement and fulfilled the criteria for diffuse cutaneous SSc 17. Fourteen patients were on treatment with a median (25th and 75th percentiles) of 10 mg (5 to 20 mg) prednisolone daily. Medications that may have influenced the systolic blood pressure at the time of evaluation consisted of calcium channel blockers (n = 26), angiotensin-converting enzyme inhibitors (n = 2), diuretics (n = 3), β-blockers (n = 2) or a combination thereof (n = 4). Smoking summarized all current smokers (n = 14) or previous smokers (n = 6), whereas nonsmokers included the remaining patients that had never smoked (n = 43). Thirty-three age-matched healthy control individuals were also enrolled. Informed consent was obtained from all participants.

All clinical and laboratory data reported in this study were obtained within 1 week of blood sampling. The modified Rodnan skin score was evaluated by standardized palpation of the skin at 17 locations on the body and grading the skin thickness on a scale from 0 to 3, resulting in a maximum skin score of 51 points 18. Nailfold capillaries were analysed quantitatively by direct microscopy counts of capillaries per millimetre on at least two fingers 19. A capillary density below the second standard deviation of a normal population of 80 healthy individuals (<5.8 loops/mm) was regarded as pathologically reduced (A. Scheja, personal communication).

Oesophageal involvement was defined as distal hypomotility on cine radiography. Radiological lung involvement was defined as basal interstitial fibrosis on a plain chest X-ray scan or on high-resolution computer tomography. Lung function tests included assessment of the vital capacity (VC) by dry spirometry and of the diffusing capacity for carbon monoxide (DLCO) by the single-breath test. All values are depicted as the population percentage (p%). DLCO values less than 75 p% were regarded as reduced.

Cardiac involvement was assessed by radiological examination of the chest, by 12-lead electrocardiography and by Doppler cardiography, and was defined as pericarditis, an abnormal electrocardiography or cardiomegaly. Systolic pulmonary artery pressures of 40 mmHg and above determined by Doppler cardiography were regarded as pathologically increased.

Muscle involvement was defined as proximal muscle weakness or serum creatinine kinase levels elevated three times or more above the upper normal limit (3.3 μkat/l). Joint involvement was defined as palpable synovitis.

Renal involvement was defined as a decreased glomerular filtration rate <70 p%, either assessed by ⁵¹Cr-ethylenediamine tetraacetic acid and iohexol clearance 20 or calculated from serum cystatin C levels 21.

Antinuclear antibodies were analysed by an indirect immunofluorescence technique using the human Hep-2 cell line as the substrate 22. Inflammatory activity was defined as either an increased erythrocyte sedimentation rate (≥15 and ≥22 mm/hour for men below and above age 50 years, respectively; ≥24 and ≥32 mm/hour for women below and above age 50 years, respectively), an increased C-reactive protein level (≥3 mg/l), or an increased orosomucoid level (>1.17 g/l).

IL-15 was assayed with a commercial human IL-15 ELISA (R&D Systems, Minneapolis, MN, USA), following the manufacturer's instructions. Each serum sample was tested in duplicate. A linear detection range of the ELISA above 0.1 pg/ml has previously been reported 23. The detection threshold was set above the blank value.

Data were analysed with STATISTICA version 6 (StatSoft, Tulsa, OK, USA) and are depicted as the median (25th and 75th percentiles). The Kruskal–Wallis test was used for multiple group comparison, before the Mann–Whitney test was used for comparison between two groups. The chi-square test for 2 × 3 tables was used when applicable. Frequencies between groups were calculated by Fisher's exact test. All variables and groups as well as residuals in multiple linear regression analyses were analysed for a normal distribution and were tested by the Shapiro–Wilk test. W values >0.93 were accepted as normally distributed. Spearman's (r_s) and partial correlations as well as multiple linear regression analyses were used to estimate correlations.

Data are expressed as the partial correlation coefficient (r), the regression coefficient (β_est), the 95% confidence interval and as the coefficient of determination. For continuous variables, the standardized β_est value and the 95% confidence interval is depicted. Results of multivariable logistic regression are depicted as the odds ratio and the 95% confidence interval. Probability P values (two-sided) were considered significant when <0.05.

Increased serum IL-15 levels were observed in SSc patients (0.63 (0.47 to 0.88) pg/ml) compared with the healthy individuals (0 (0 to 0.46) pg/ml) (Figure 1). This applied to IL-15 levels in patients with limited cutaneous SSc (0.63 (0.45 to 0.83) pg/ml) and diffuse cutaneous SSc (0.72 (0.47 to 1.03) pg/ml) compared with control individuals, but the IL-15 levels did not differ between limited cutaneous SSc and diffuse cutaneous SSc.

Three outliers were identified in the SSc patient group. Two of these outliers were excluded from further analysis since both were derived from patients in renal crisis, one of which was already undergoing haemodialysis, and the reduced renal function might impede IL-15 elimination. The third outlier, however, had no identifiable confounding factors and was therefore included in the analysis.

Serum IL-15 levels of the 61 studied patients fulfilled the criteria for a normal distribution. Fourteen out of 33 control individuals had detectable levels of IL-15. These individuals did not take any medications; neither did they have any accompanying disease or did they differ in age from the other control individuals.

Low, intermediate and high serum IL-15 levels depict three subsets of SSc patients regarding the skin score, the VC, the DLCO and the systolic pulmonary arterial pressure. Table 1 presents the findings of patients with low, intermediate and high levels of IL-15.

The 75th percentile of all SSc patients (0.88 pg/ml) was chosen as the cut-off value for high serum levels of IL-15 because serum IL-15 levels in the control individuals were not normally distributed in accordance with other studies 24 (Figure 1). In the group with high serum IL-15 levels, diffuse cutaneous SSc was twice as prevalent as limited cutaneous SSc. Among patients with low or intermediate serum IL-15 levels, limited cutaneous SSc predominated. The VC was significantly lower among patients with high IL-15 levels (P < 0.05).

We tested the possibility that a lower dividing point between low IL-15 levels and high IL-15 levels could change the characteristics of our patient groups. The median value of the SSc patients (0.63 pg/ml) was chosen as alternative cut-off level. Twenty-eight patients with serum IL-15 levels ≥0.64 pg/ml were compared with 33 patients with serum IL-15 levels <0.64 pg/ml. Lower DLCO levels (P < 0.01) and a higher prevalence of reduced finger capillary density (P < 0.05) were found in patients with serum IL-15 levels ≥0.64 pg/ml. The skin score, the SSc type or the VC showed no difference. Grouping the patients into those with low (<0.64 pg/ml), intermediate (0.64 to 0.88 pg/ml) and high (>0.88 mg/ml) serum IL-15 levels, however, showed that the group with high serum IL-15 levels had both low VC and low DLCO (Table 1). The intermediate group had a tendency to lower DLCO compared with the group with low serum IL-15 levels (Figure 2), although the two groups had similar skin scores or VC levels. The intermediated group also had a higher prevalence of patients with reduced DLCO (<75 p%) compared with the group with low serum IL-15 levels (53% versus 21%, P < 0.05). In addition, the VC/DLCO ratio tended to be higher in the intermediate serum level group, indicating predominating pulmonary vascular disease. Pulmonary hypertension (defined as systolic pulmonary pressure ≥40 mmHg) was found in four out of eleven patients (36%) in the intermediate IL-15 group, compared with one out of 31 patients (3%; P < 0.05) in the low IL-15 level group and zero out of 13 patients (0%; P < 0.05) in the high IL-15 group.

IL-15 correlated negatively with the VC (r_s = -0.31, P < 0.05). This correlation became even more significant after statistical consideration of the presence of cortisone treatment by partial correlation (r = -0.42, P < 0.001). Multiple linear regression analysis of laboratory variables that may influence the VC showed that only IL-15 correlated inversely with the VC, whereas the presence of anticentromer antibodies was associated with high VC (Table 2). All associations remained significant after adjustment for multiple correlations and for smoking.

IL-15 also correlated inversely to the DLCO (r_s = -0.33, P < 0.05), and the correlation remained after adjustment for cortisone (r = -0.32, P < 0.05). Applying multiple linear regression analysis of IL-15 and the DLCO including autoantibodies and inflammatory parameters, the inverse relation did not remain significant after adjustment for multiple correlations.

When the two patients with renal crisis were included in the analyses, however, the multiple linear regression analyses of IL-15 against both the VC and the DLCO became strong (r = -0.54, P < 0.001 for IL-15 versus VC; r = -0.46, P < 0.001 for IL-15 versus DLCO).

Forty-two out of 54 examined patients showed a reduced nailfold capillary density and had significantly increased serum IL-15 levels (0.72 (0.49 to 0.90) pg/ml) compared with the remaining 12 patients (0.47 (0.27 to 0.61) pg/ml) (P = 0.010). Serum IL-15 levels were identified as an important variable for the occurrence of reduced nailfold capillary density using multivariable logistic regression analysis (Figure 3). This correlation remained significant after adjustment for multiple correlations.

IL-15 correlated inversely with the systolic blood pressure by multiple linear regression analysis including variables that may influence the systolic blood pressure (Table 3). Age and the body mass index showed the expected positive correlation with systolic blood pressure. These data remained significant after adjustment for multiple correlations.