SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ROLE OF THE STUDY SPONSOR
  8. AUTHOR CONTRIBUTIONS
  9. REFERENCES

Objective

To evaluate the reliability and validity of a novel ultrasound (US) imaging method to measure metacarpophalangeal (MCP) and proximal interphalangeal (PIP) finger joint cartilage.

Methods

We examined 48 patients with rheumatoid arthritis (RA), 18 patients with osteoarthritis (OA), 24 patients with unclassified arthritis of the finger joints, and 34 healthy volunteers. The proximal cartilage layer of MCP and PIP joints for fingers 2–5 was bilaterally visualized from a posterior view, with joints in ∼90° flexion. Cartilage thickness was measured with integrated tools on static images. External validity was assessed by measuring radiologic joint space width (JSW) and a numeric joint space narrowing (JSN) score in patients with RA.

Results

Precise measurement was possible for 97.5% of MCP and 94.2% of PIP joints. Intraclass correlation coefficients for bilateral total joint US scores were 0.844 (95% confidence interval [95% CI] 0.648–0.935) for interobserver comparisons and 0.928 (95% CI 0.826–0.971) for intraobserver comparisons (using different US devices). The US score correlated with JSN for both hands (adjusted R2 = 0.513, P < 0.001) and JSW of the same finger joints (adjusted R2 = 0.635, P < 0.001). Reduced cartilage shown by US allowed discrimination of early symptomatic OA versus early RA and healthy joints. In patients with RA, US scores correlated with duration of treatment-resistant, progressive RA.

Conclusion

The US method of direct visualization and quantification of cartilage in MCP and PIP joints is objective, reliable, valid, and can be useful for diagnostic purposes in patients with arthritis.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ROLE OF THE STUDY SPONSOR
  8. AUTHOR CONTRIBUTIONS
  9. REFERENCES

Loss of cartilage reflects irreversible joint destruction and contributes to impaired joint function in osteoarthritis (OA) and rheumatoid arthritis (RA) (1, 2). Loss of cartilage in OA is generally indirectly evaluated by radiologic assessment of joint space distance, together with evaluation of pathologic bone formation (1). A joint space narrowing (JSN) score can be calculated from visual impression of joint space distances across multiple joints. JSN has been calculated in various ways and is a validated estimate of progressive loss of cartilage in RA (3–6). Computerized measurement of actual joint space width (JSW) on radiographs might be a more precise measure than JSN, but faces some unresolved problems (7).

An alternative to radiographic imaging, ultrasound (US) is an established tool for measurement of cartilage in large joints (8). Modern high-frequency transducers with theoretical axial resolutions of ∼0.1 mm could potentially satisfy an urgent need in early arthritis diagnosis (9) by detecting minimal synovitis and erosions in both metacarpophalangeal (MCP) and proximal interphalangeal (PIP) joints, which are important indicator regions of RA. However, US is validated for synovitis assessment in RA only after exclusion of OA (10).

Although the use of US is widely accepted in the medical community, several reports have emphasized the importance of validation studies before broad application of new measures (11, 12). We postulated that loss of cartilage is intimately connected with symptomatic OA, but occurs, if at all, later in the course of RA. We further hypothesized that US would perform as well as radiographic technology in the direct measurement of cartilage. Potential advantages of US over radiography for this purpose would be the simultaneous evaluation of cartilage and synovitis within a single examination, immediate availability at bedside combined with lower investment costs, and ability to do repetitive joint imaging without ionizing radiation.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ROLE OF THE STUDY SPONSOR
  8. AUTHOR CONTRIBUTIONS
  9. REFERENCES

Patients and healthy volunteers.

A total of 124 adult patients and healthy volunteers (Table 1) were included in the study after giving informed consent to participate. The study protocol was approved by the Responsible Ethics Committee of Berne. Fifty patients were consecutively recruited from referrals to the early arthritis unit at the University Hospital Berne (Inselspital, Berne, Switzerland). Patients were admitted by board-certified rheumatologists for US images of finger joint synovitis that were clinically observed for a maximum of 12 months. However, patients could have been experiencing pain in their hands for a significantly longer period of time. Twelve of these 50 patients, as well as 36 additional consecutive patients with persistent RA, were classified according to American College of Rheumatology (ACR; formerly the American Rheumatism Association) diagnostic criteria (13). Twenty-five patients with early RA had a disease duration <24 months; 23 patients with persistent RA for >2 years were refractory to at least 2 conventional disease-modifying antirheumatic drugs, and 9 of them also had failure of therapy with least 1 tumor necrosis factor–blocking agent. According to the ACR criteria (14), 18 patients could be classified as having OA; among them were 14 patients with synovitis duration comparable to that of early RA (median 4 months, range 1–22 months). Also included in the study were 34 healthy volunteers.

Table 1. Demographic and patient characteristics*
CharacteristicDiagnosis groupHealthy volunteers (n = 34)
RA <2 years (n = 25)RA >2 years (n = 23)OA (n = 18)Other diagnosis (n = 24)
  • *

    Values are the median (range) unless otherwise indicated. RA = rheumatoid arthritis; OA = osteoarthritis; RF = rheumatoid factor; NA = not applicable; Anti-CCP = anti–cyclic citrullinated peptide.

  • Including psoriatic arthritis (n = 2) and other spondylarthritis (n = 1), systemic lupus erythematosus and other inflammatory connective tissue disorders (n = 3), calcium pyrophosphate crystal arthritis (n = 4), vasculitis (n = 1), and unclassifiable conditions (n = 13).

Sex, no. women/men19/617/616/215/923/11
RF positive, no.1413NANANA
Anti-CCP positive, no.1214NANANA
Age, years45 (18–81)57 (24–76)59 (39–78)50 (26–77)42 (18–61)
Height, cm167 (152–180)166 (151–184)164 (154–178)169 (155–188)168 (151–190)
Weight, kg66.7 (46–95)66 (44–100)73 (44.5–90)76 (48–104)70 (43–94)
Disease duration, years0.33 (0–1.8)8 (2.5–25)1 (0–11)0 (0–11)NA

US technique.

US scanning was performed by rheumatologists (BM and H-RZ) and an experienced technical US assistant (MR) on the ATL 5000 supply (Advanced Technologies Laboratories, Bothell, WA) with a 10–15 MHz linear array hockey stick transducer. Participants were seated in front of the investigator, with hands positioned on the padded examination desk. Cartilage measurements for the second through fifth fingers of each hand were performed on a static longitudinal scan of MCP and PIP joints on a dorsal view of the proximal cartilage layer, with the joints in the best possible, but maximum, 90° flexed position. This approach, in contrast to that of another recent study (15), allowed direct visualization of the central force–bearing cartilage area (Figure 1). Exact perpendicular incidence of sound waves was evident from precisely sharp reflection of the subchondral bony surface and a less intensive interface reflex artifact at the cartilage surface. The distance between these lines (representing cartilage thickness) was measured by integrated tools. In the absence of both the surface artifact and synovitis (10), an alternative measurement was used: the largest distance of anechoic tissue in a direction perpendicular to the bone surface. If synovitis was present in a joint, but no interface artifact was apparent, that joint was not evaluated. All static US images were measured and saved in maximal applicable magnification by an investigator blinded to the clinical data.

thumbnail image

Figure 1. Ultrasound (US) results from a longitudinal view of flexed joints on the proximal cartilage layer of the metacarpal head and corresponding digital radiograph images in dorsopalmar projection of 2 patients at different stages of rheumatoid arthritis. The upper panels are images of metacarpophalangeal joints with a normal quantity of cartilage shown in the US image (A) and by normal joint space distance on the radiograph image (B). The solid arrow in A highlights the interface artifact at the cartilage surface, and the open arrow highlights reflexion at the base of the cartilage. The lower panels show greatly reduced cartilage on the US image (C) and advanced joint space narrowing on the radiologic image (D). The interface artifact is missing in C. Almost complete absence of anechoic cartilage, unrounding, and a central defect of subchondral bone (open arrow) indicate serious cartilage damage. Hypoechoic synovitis (*) complicates evaluation of this joint.COLOR

Download figure to PowerPoint

To assess interobserver reliability and intraobserver reliability with the additional challenge of using different devices, US measurements were performed by 2 blinded raters in 10 healthy subjects on 2 devices 1 week apart. The US device applied second was an Esaote MyLab 70 XVision (Esaote, Genoa, Italy), with a 4-cm 6–18 MHz linear array transducer (L435, Esaote).

Radiographic imaging.

Separate conventional radiographs of each hand and wrist were taken in the standard posteroanterior projection in every patient without a radiograph taken within the previous 12 months (n = 50 total participants, n = 29 patients with RA). The MCP and PIP joints of the second through fifth fingers were magnified to a level of 1:1 on high-brightness monitors (5k matrix, DIN V 6868-57; Totoku Electric, Tokyo, Japan). A board-certified radiologist (HB) who specialized in musculoskeletal radiology quantified the shortest distance between the subchondral bone plates (JSW) along the force-bearing axis of the joint using the standard measurement tool of the applied picture archiving and communication systems (Easyvision; Philips Medical Systems, Eindhoven, The Netherlands). All picture archiving and communication systems–documented and archival–conventional radiographs of the last 12 months from 36 patients with RA were scored according to established JSN procedures (6).

Statistical analysis.

Descriptive statistics.

Descriptive statistics included the median and range or the mean and SD for Gaussian distributions. All comparisons of medians for 2 independent groups were done by nonparametric Wilcoxon's matched pairs tests. Correlations were described by Spearman's correlation coefficient (rho). For all analyses, an alpha level of 0.05 was used.

Individual and summed US scores.

Because cartilage measurement in a single joint may not be representative of all affected joints in a multijoint disorder, several US scores that were sums of multiple MCP and PIP joints were calculated specifically: bilateral MCP joints (sum of the second through fifth MCP joints of both hands), bilateral PIP joints (sum of all PIP joints of both hands), left hand (all 8 examined joints of the left hand), right hand (all 8 examined joints of the right hand), total joints (sum of all 16 examined joints), and fingers 2 + 5 (sum of the examined joints of the second and fifth fingers of both hands [8 joints]). If data were missing for any joint used to calculate a summary score, the case was excluded from this analysis. An additional set of analyses was performed after extrapolating missing data using the expectation maximization method.

Regression analyses.

Regression analyses were primarily performed with the curve estimation tool. The goodness of fit of linear models was estimated by correlation coefficients (R2), and contribution of multiple independent parameters was expressed by the standardized coefficient beta.

Reliability tests.

Interobserver and intraobserver reliabilities were assessed by calculating intraclass correlation coefficients (ICCs). We also calculated the smallest detectable difference (SDD) and depicted Bland-Altman plots for intraobserver and interobserver comparisons.

Validity tests.

The relationships between the US cartilage scores and both the JSW and JSN scores were evaluated using regression analysis. Discriminative validity of US cartilage scores was estimated by the area under the receiver operating characteristic (ROC) curve. All statistical analyses were performed with Statistical Package for the Social Sciences software, version 15.0 (SPSS, Chicago, IL).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ROLE OF THE STUDY SPONSOR
  8. AUTHOR CONTRIBUTIONS
  9. REFERENCES

Results for individual joints.

Sample US images and corresponding digital radiographic images of 2 patients with RA at different disease stages are shown in Figure 1. Cartilage could not be accurately measured in 2.5% of MCP and 5.8% of PIP joints. Cartilage thickness of the proximal layer in single MCP joints in the entire sample of subjects varied between 0.0 and 0.7 mm. Correlations (Spearman's rho) between contralateral MCP counterpart joints ranged from 0.650 to 0.723 (n = 118–119; P < 0.001 for all correlations). The cartilage layer in PIP joints varied between 0.0 and 0.5 mm. Correlations between contralateral PIP counterpart joints ranged from 0.396 to 0.564 (n = 111–117; P < 0.001 for all correlations). The cartilage thickness in single joints of healthy volunteers was normally distributed, with maximum thickness in the left second MCP joint and minimum thickness in the right fifth PIP joint.

Descriptive results for summary scores.

The unilateral (left or right hand) 8-joint scores ranged from 0.3 to 3.5 mm (median 1.9 mm), and the bilateral total joint scores ranged from 0.9 to 6.9 mm (median 3.8 mm). The fingers 2 + 5 scores ranged from 0.5 to 3.9 mm (median 1.9). By excluding cases with any missing values, the total joint score was calculable in 73 individuals and the fingers 2 + 5 score was calculable in 107 of 124 individuals. These 2 US scores (total joint score and fingers 2 + 5 score), however, were closely correlated among all evaluated individuals (n = 73; ρ = 0.956, P < 0.001) and among healthy volunteers (n = 27; ρ = 0.895, P < 0.001).

RA-related and disease-independent factors.

In healthy volunteers, correlations were calculated for the relationships between height and US summary scores (left hand, right hand, MCP joints, PIP joints, total joints, and fingers 2 + 5); Spearman's rho values ranged from 0.355 to 0.538 (P < 0.001 for all correlations). Age was negatively correlated with US scores, resulting in Spearman's rho values ranging from −0.340 to −0.560 (P < 0.001 for all correlations). Healthy male volunteers had US scores that were on average higher than the scores for healthy female volunteers, with statistically significant differences for the MCP joints (P = 0.032) and fingers 2 + 5 scores (P = 0.048). Weight and body mass index were not correlated with cartilage scores.

In healthy subjects, linear regression modeling was statistically significant for all of the defined US scores and depended predominantly on height and age. Lowest prediction was found for the MCP joint score (R2 = 0.449, P = 0.008) and highest prediction was found for the PIP joint score (R2 = 0.663, P < 0.001).

In patients with RA, after adding disease duration, the linear regression was also predictive; however, in this case, disease duration was the dominant factor (β = −0.870 and −0.479, P = 0.004 and 0.006, respectively) in models of the total joint score (R2 = 0.645, P = 0.018) and the fingers 2 + 5 score (R2 = 0.547, P = 0.001).

Reliability tests for US scores.

ICCs (95% confidence intervals [95% CIs]) for estimating interobserver reliability were as follows: bilateral PIP joints 0.766 (0.499–0.900), bilateral MCP joints 0.890 (0.743–0.955), left hand 0.791 (0.545–0.912), right hand 0.820 (0.600–0.925), total joints 0.844 (0.648–0.935), and fingers 2 + 5 0.880 (0.722–0.951) (n = 20; P < 0.001 for all comparisons). Intraobserver reliability was also high, with ICCs (95% CIs) as follows: bilateral PIP joints 0.807 (0.575–0.919), bilateral MCP joints 0.946 (0.870–0.978), left hand 0.866 (0.693–0.945), right hand 0.931 (0.835–0.972), total joints 0.928 (0.826–0.971), and fingers 2 + 5 0.916 (0.800–0.966) (n = 20; P < 0.001 for all comparisons). The SDD for interobserver comparison was 0.09 mm for the total joint score and 0.04 mm for the fingers 2 + 5 score. The SDD for intraobserver comparison was 0.06 mm for the total joint score and 0.03 mm for the fingers 2 + 5 score. Some systemic differences in scores occurred when different US devices were used. The corresponding results for the fingers 2 + 5 score are depicted in Bland-Altman plots in Figure 2.

thumbnail image

Figure 2. Bland-Altman plots for intra- (top) and interobserver (bottom) retests of ultrasound fingers 2 + 5 scores. The mean of paired results is depicted on the x-axis and the difference between the paired results is shown on the y-axis. The observed variability includes all steps of scanning and measuring and is higher for interobserver than for intraobserver comparison. Intraobserver comparison, however, shows systematic deviation between the 2 devices, indicated by deviation of the mean from 0 on the y-axis (dotted line). Solid lines represent 1.96 SDs from the mean.

Download figure to PowerPoint

Validity tests.

Regression analyses of US total joint scores with the JSN score (combining both hands) (6) showed a linear relationship (n = 20; adjusted R2 = 0.513, P < 0.001). Significant results were also obtained when the radiographic analysis was limited to the same finger joints also examined by US (n = 20; adjusted R2 = 0.624, P < 0.001) (Figure 3A) and when the US fingers 2 + 5 score was used (data not shown). We also compared the US total joint score with JSW and found a close correlation between US and radiographic results in the same joints (n = 27; adjusted R2 = 0.635, P < 0.001) (Figure 3B).

thumbnail image

Figure 3. A, The linear regression modeling of the total ultrasound (US) score and the Sharp joint space narrowing (JSN) score modified by van der Heijde (6) for the same joints. The JSN score depicted here is limited to evaluation of metacarpophalangeal and proximal interphalangeal joints (adjusted R2 = 0.624, F = 31.54; P < 0.001). B, The linear regression modeling of the total US score and the quantitative joint space width measured from digital radiographs (adjusted R2 = 0.635, F = 43.4; P < 0.001).

Download figure to PowerPoint

The US total joint scores and fingers 2 + 5 scores for each patient group are shown in Table 2. US scores were significantly lower in patients with recent-onset synovitis with OA than in those with early RA (P < 0.001 for the total joint score and P = 0.002 for the fingers 2 + 5 score). Patient characteristics did not explain the differences in these cartilage measurement results (patients with recent-onset synovitis OA had a median age of 59 years [range 39–78], median height of 164 cm [range 154–178], and median weight of 75 kg [range 44.5–97]). In contrast, there was no significant median difference between healthy volunteers and patients with early RA. ROC curves for discrimination of patients with OA from a group of subjects with early RA and healthy volunteers gave an area of 0.891 for the total joint score (lower bound of 95% CI 0.771, P = 0.001) and 0.871 for the fingers 2 + 5 score (lower bound of 95% CI 0.719, P = 0.002).

Table 2. Summary of ultrasound total joint scores and fingers 2 + 5 scores for all patients and healthy volunteers
GroupTotal joint score, mmFingers 2 + 5 score, mm
  • *

    Persistent rheumatoid arthritis (RA) is >2 years of disease duration.

  • Early RA is <2 years of disease duration.

Persistent RA*  
 No.1015
 Median (minimum, maximum)2.95 (0.90, 3.90)1.60 (0.50, 2.20)
Early RA  
 No.1321
 Median (minimum, maximum)4.20 (3.40, 5.20)2.30 (1.30, 3.20)
Osteoarthritis  
 No.916
 Median (minimum, maximum)2.50 (1.40, 4.10)1.40 (0.70, 2.30)
Other diagnosis  
 No.1422
 Median (minimum, maximum)3.55 (1.30, 6.40)1.90 (0.70, 3.60)
Healthy volunteers  
 No.2733
 Median (minimum, maximum)4.30 (2.70, 6.90)2.30 (1.30, 3.90)

Extrapolation of missing data.

Data estimation for single missing values in complex scores might be an alternative to the evaluation of fewer joints. Linear regression analyses after inclusion of the estimates into the total joint US score were significant for JSN of finger joints (n = 34; adjusted R2 = 0.419, P < 0.001), as well as for JSN of both hands combined (adjusted R2 = 0.322, P < 0.001). Linear regression modeling on JSW scores was also significant (n = 38; adjusted R2 = 0.688, P < 0.001). Similar to the results for data without imputation, the area under the ROC curves for identifying OA after inclusion of estimates (Figure 4) was significant for all US scores.

thumbnail image

Figure 4. Receiver operating characteristic curves for ultrasound scores. Calculated areas under the receiver operating characteristic curve after inclusion of estimates were between 0.772 (metacarpophalangeal [MCP] joints P = 0.005) and 0.862 (total joint score P < 0.001) for separating patients with synovitis in osteoarthritis from those with early rheumatoid arthritis and healthy subjects. Diagonal segments in the figure are produced by ties. PIP = proximal interphalangeal.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ROLE OF THE STUDY SPONSOR
  8. AUTHOR CONTRIBUTIONS
  9. REFERENCES

This study explored the reliability, validity, and practical usefulness of a novel method for finger joint cartilage measurement using US. Experienced examiners completed the entire process of scanning and documentation of cartilage in 16 finger joints within 5 minutes. Additional time was required for precise measurement, which could be done in the patient's absence. To avoid potential variability in measurement caused by inconsistent scanning positions, a precise, standardized scanning procedure was used, followed by cartilage measurements on static images. Retest reliability achieved by this method was higher than that reported for magnetic resonance imaging studies on finger cartilage (16).

The current standard for estimating cartilage, JSN from radiographic images, varies considerably even among expert readers (17). Reproducibility of JSW in patients with early RA was also recently examined in detail (18), with an excellent ICC of 0.99. However, this study involved correlations between JSW estimates from different raters using the same static images. In contrast, reliability tests of US in the current study with ICCs of 0.84 and 0.93 for the total joint score measured correspondence between raters and US supplies on the entire procedure of positioning, scanning, documentation, and measurement. SDD for JSW with repetitive imaging under ideal conditions (19) is in the same range as reported here for US, but SDD for measuring JSW on radiographs has generally tended to increase considerably when used in clinical settings. Therefore, the introduction into clinical research is still ongoing (7, 20). In this study, variable joint flexion between 60° and 90° had no measurable impact on the US result and allowed examination in a relaxed joint position for almost all patients.

We validated US measurements in this study by comparing them with radiographic measurements of JSN and JSW. Involvement of MCP joints in OA without node formation is less apparent but is evidenced with different methods (21). JSW of MCP and PIP joints is normal in early RA (18), but is reduced in the case of coexisting OA (22). Reduced cartilage is an early and symmetrically distributed sign of magnetic resonance imaging in OA also detectable by US that could be used for diagnostic purposes (23, 24). Close correlations between the US and radiographic methods indicated that the US method has acceptable validity. Because we anticipated that JSW would correspond to the sum of cartilage layers of the proximal and distal surface of diarthrodial joints, it was unexpected to find that the proximal cartilage layer in any joint measured by US was always only approximately one-quarter of the radiologic joint space distances. The radiographic measurements of our study were very close to previously reported results (18), and retests with US excluded major systemic measurement error as an explanation. We believe that the difference between US and radiographic measurements is real, and we hypothesize, based on observations of pathologic cartilage calcification, that the source of the systematic difference between the 2 methods is that US reflects the calcified basal cartilage layer that is still transparent to radiography. Another possible reason for the larger radiographic joint distance would be elevated intracapsular pressure in the presence of effusion or synovial proliferation.

The presence of an interface artifact at the cartilage tissue surface facilitates the US evaluation, but an interface artifact is not always present. Analysis of missing data in the absence of the surface US reflection revealed that this limitation could be handled by mathematic extrapolation, a procedure that is also used for radiographic analyses (25, 26). Alternatively, the fingers 2 + 5 US score could be used; this score was closely correlated with the total joint US score but had fewer missing data points.

Overall, the US method of cartilage measurement was shown to be a promising alternative to radiographic methods for estimating cartilage. In addition to being reliable and valid, this method has several practical advantages (e.g., efficiency and cost) that make it valuable for use in both research and clinical practice.

ROLE OF THE STUDY SPONSOR

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ROLE OF THE STUDY SPONSOR
  8. AUTHOR CONTRIBUTIONS
  9. REFERENCES

Early symptomatic patients in this study were recruited from the Early Arthritis Clinic at Berne University Hospital, which was sponsored by an unrestricted research grant from Abbott Laboratories. Abbott Laboratories had no influence on the topic or study design, or the acquisition, analysis, or interpretation of data.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ROLE OF THE STUDY SPONSOR
  8. AUTHOR CONTRIBUTIONS
  9. REFERENCES

Dr. Möller had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Möller, Bonel, Ziswiler.

Acquisition of data. Möller, Bonel, Rotzetter, Ziswiler.

Analysis and interpretation of data. Möller, Bonel, Villiger, Ziswiler.

Manuscript preparation. Möller, Bonel, Ziswiler, Michael A. Nissen, ELS (nonauthor; Abbott Laboratories, Abbott Park, IL), Ellen R. Stoltzfus, PhD (nonauthor; JK Associates, Inc., Conshohocken, PA).

Statistical analysis. Möller.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. ROLE OF THE STUDY SPONSOR
  8. AUTHOR CONTRIBUTIONS
  9. REFERENCES