Radiographic assessment.

Standard radiographs of both wrists in posteroanterior view were made at the baseline, 1-year, and final visits using 3M XDA film and a 3M Trimax T16 cassette (3M Company, St. Paul, MN), with settings of 50 mA, 0.03 seconds, and 43–47 kV. Radiographic damage was scored according to the Poznanski method (9) by 2 independent investigators (FR and FT) who were unaware of the clinical data. This method is based on measurements of the radiometacarpal (RM) width, which is the distance from the base of the third metacarpal bone to the midpoint of the distal growth plate of the radius, and the maximum length of the second metacarpal bone (M2).

Prior to the measurements, the 2 observers were trained in the technique of RM and M2 determination by an experienced radiologist (GB), who supervised all radiographic assessments. In cases in which advanced carpometacarpal erosions made it difficult to define the bone ends, the last film that allowed a reliable assessment was used as the final radiograph. If advanced destruction occurred within the first 2 years of followup, then the patient was excluded from the analysis. Because the Poznanski method cannot be used when there is radiographic closure of the growth plates of the second metacarpus, patients who had apparent radiographic closure of the second metacarpal growth plate within the first 2 years of followup were excluded. In patients who had closure of the growth plate after the second year of followup, the wrist radiograph performed immediately before demonstration of the closure was used as the final radiograph. The RM and M2 measurements were made to the nearest 0.1 mm by using a precision gauge (DialMax; Swiss Precision, Solothurn, Switzerland). The scores assigned by the first investigator (FR) were used for the analyses; the scores assigned by the second investigator (FT) were used to validate the scores of the first investigator.

The measures obtained for RM and M2 were plotted against each other using the normative charts of Poznanski et al (9). For each wrist, the number of standard deviations between the expected and the observed RM width for the measured M2 length was calculated according to the formulas reported by Poznanski et al (9).

The RM/M2 score, which represents the carpal length and will be referred to as the Poznanski score, reflects the amount of radiographic damage in the wrist. The greater the negative value of the Poznanski score, the more severe the radiographic damage. In each pair of wrists, the score for the more damaged side was used in the analyses. We made this choice because we believed that separate assessments of the right and left wrists would be less meaningful clinically and that the use of the arithmetic sum or mean of the 2 wrists could affect the reliability of the estimation of the true amount of damage in patients with a marked difference in the severity of radiographic changes between the 2 wrists (i.e., with positive versus negative Poznanski scores).

Radiographic progression during the first year and radiographic progression during the entire study period were then determined by calculating the change in the Poznanski score between the radiographs obtained at baseline and those obtained at 1 year as well as those obtained at the final visit, respectively. Because the duration of followup varied among the study patients, the radiographic change between the baseline and the final radiographs was divided by the number of years of followup for each patient to yield the yearly radiographic progression. A positive value for radiographic progression indicates improvement, whereas a

negative value indicates worsening.

To investigate the reproducibility of the radiographic scoring method we used, we assessed the intraclass interreader correlation coefficients by comparing the values obtained for the RM width and the M2 length by the 2 observers for all radiographs. We assessed the intraclass intrareader correlation coefficients by asking 1 of the 2 examiners (FR) to interpret all radiographs for a second time in a blinded manner, 3 months after the previous review. The intraclass interreader and intrareader correlation coefficients for the RM and M2 measurements were very high, ranging from 0.97 to 0.99. The bias estimate and the 95% confidence interval (95% CI) between the 2 observers, calculated according to the method of Bland and Altman (17), for the Poznanski score were − 0.041 and − 1.72, 1.72, respectively, in the less damaged wrist and − 0.071 and − 3.72, 3.57, respectively, in the more damaged wrist. The bias estimate and 95% CI between the 2 observers for the progression scores from the first to the last radiograph, calculated according to the same method, were 0.009 and − 0.32, 0.34, respectively (Figure 1).

Figure 1. Progression scores plotted according to the method of Bland and Altman. Broken lines show 95% limits of agreement.

To verify whether our patients developed an abnormal shortening of M2 in comparison with the impairment of the patient's height, which often occurs in JIA, we compared the M2 length in the worse wrist with the reported norms and evaluated the relationship between the shortening of the M2 length and the reduction in height (as compared with the norms) at both the baseline assessment and the final assessment. Furthermore, we evaluated whether a shortening of the M2 length was related to age at disease onset, sex, disease duration, amount of radiographic damage, and rate of radiographic progression. For these purposes, Z scores (units of standard deviation above or below the normal mean for each measurement) were derived for the M2 length (18) and the patient's height (19) by interpolation from published norms.