The central aim of this study was to establish if CRP values measured from DBS samples using an ELISA agree with those from a conventional plasma sample assayed by a hospital reference immunoturbidometric method. This is the first study exploring this approach in a rheumatoid arthritis population with active disease.
Method comparison
The Bland-Altman and Deming regression plots in this study show that there is reasonable agreement between the DBS and reference methodologies, especially in the low to mid-range CRP values. The calibration of the standard curve in ELISA does not cope as well with very high levels of CRP as the immunoturbidometric method does [24]. The majority of DBS ELISA measurements (91% in the baseline samples and 85% in the 6 week) lie between the limits of agreement lines (Figure 2), indicating reasonable agreement between the two collection and assay methods. The weighted Deming regression coefficients for the values within the limits of agreement, which correct for systematic measurement error, indicate good agreement in particular between the DBS ELISA and reference methods.
The imperfect spearman correlation coefficients indicate a proportional error that might have originated from incomplete recovery of CRP in the ELISA methods, particularly with frozen plasma. The non-zero intercepts observed likely reflect a systematic error such as insufficient volume correction in DBS.
The normal probability plots (Supplementary Figure 1) indicate how closely the two data sets agree, when the two cumulative distributions are plotted against each other, whilst also highlighting any skewness of the distribution. If the two measurements showed perfect agreement with each other, then the plotted data would fall on an ideal-diagonal line. The largest and most consistent skew was observed from the 75th percentile and above, therefore indicating that the higher CRP values measured by ELISA did not show good agreement with the immunoturbidimetry data.
Flare detection
Our data does not suggest that rises in CRP alone can predict an impending flare in RA patients, but that longitudinal CRP measures can help to confirm biochemical evidence of a flare which can be used by clinicians alongside clinical measures such as tender and swollen joint counts. Continuous weekly CRP data, collected while study participants were at home, was able to identify flare with greater specificity, than minimal data taken at baseline and follow up hospital appointments. The occurrence of flares often prompts an escalation in therapy [25]. With further sample and analytical refinement, a DBS approach could offer new opportunities to optimally suppress chronic inflammatory episodes, and reduce long term morbidity and mortality risk to people with RA.
Clinical implications
The decision to treat inflammation in RA can be initiated when mean changes in CRP of 3 mg/L or more distinguish active flare [26]. Thus, changes in higher plasma CRP concentrations beyond the capability of the ELISA method should not adversely influence the course of treatment.
If a clinician were to attempt to identify active flares using a DBS CRP assay, the 95% limits of agreement (using 6 week Bland-Altman parameters) between the two methods at the active disease cut-off of 10 mg/L would be 8.45–12.02 mg/L [27, 28]. This could be tolerable from a clinical perspective as the variance observed for reference readings below 20-25mg/L is not appreciable and the positive bias shown with DBS sample readings may compensate to some degree for under measurement. Though the imprecision may mean a small number of patients with borderline flare may be overlooked, it would still allow patients with active disease clearly above a 10 mg/ml CRP threshold to be accurately identified.
Limitations
The agreement between DBS and reference methods was acceptable with CRP concentrations in the low to mid-range, but DBS CRP readings for samples with reference values above 25mg/L were imprecise, with increased variance beyond the 95% confidence limits. It is possible that the use of ELISA assays may have caused a systematic bias, as both ELISA assays showed imprecision with higher levels of CRP. Another potential source of error with DBS samples is changes in haematocrit due to inaccurate volume correction estimates, though dried plasma solutions may mitigate against this [29-31]. It is also likely that using the patient’s subjective ‘self-report’ of flare will not always distinguish an inflammatory flare from an increase of pain due to other causes. However, a recent study comparing patient and clinician reported flares indicates close agreement across 79-93% of joints, particularly on swollen joints [32].