The association between CRP and SAA was high in our population (rho = 0.755, p < 0.001) (Table 2), in accordance with other studies done in similar specific populations: patients with FMF [24], patients with hidradenitis suppurativa [25] and patients with infectious diseases [14]. The association between CRP and SAA was significant across the ethnicities in this study, but seems to be a little higher in the Caucasian population compared to the Armenian and Arabic population. In most studies we know of in this field, a minority of patients were Caucasian, since most studies were done in FMF patients and most of the FMF patients are not Caucasian. The difference in association across ethnicities might be explained by more or less prevalent diseases within ethnic groups. Other markers with a significant positive correlation with SAA were ESR, ferritin and number of granulocytes. This may be explained by the fact that these parameters increase during activation of the innate immune system, similar to SAA. The association between ESR and SAA that we found (rho = 0.550, p < 0.001) is in agreement with previous research [24]. The number of lymphocytes was weakly inversely associated with SAA (rho = − 0.173, p = 0.003). These finding may be explained by the fact that in autoinflammatory diseases the innate immune system is more active than the adaptive immune system. Although significant, the association between age and BMI with SAA in this study was very weak (rho = 0.221, p < 0.001; rho = 0.255, p = 0.017). This positive association between age and BMI with acute phase reactants is also observed in other populations [7]. With advancing age chronic antigen stimulation and accumulation of cellular residues affects both the innate and the adaptive part of the immune system. For the innate immune system the consequence is inflammaging (low grade clinically undetectable inflammation with increased production of pro-inflammatory cytokines) and for the adaptive immune system it is immunosenescence (immune dysfunction that progresses with age). Inflammaging is considered as a basis of most age-related diseases and might explain the weak correlation between age and SAA levels [7, 26]. The association between BMI scores and SAA levels can be explained by the widely accepted relation between adiposity and inflammation, through IL-6 secretion by adipose tissue [27].

Based on our study, if a 100% sensitivity of SAA levels below 4 mg/L is desired, a CRP cut-off value of 0.45 mg/L should be used (Fig. 2). The best CRP cut-off value to detect low SAA levels is still a discussion: chosen values range from 0.5–10 mg/L, largely depending on the desired sensitivity and specificity [4, 7, 13, 14]. In our opinion further decrementation of CRP levels below the value 5 mg/L is not clinically relevant as CRP levels under 5 mg/L are used as normal based on the supposed absence of an acute phase reaction below this level [18]. When using the predetermined cut-off value of 5 mg/L for CRP, 85.4% sensitivity and 83.6% specificity were found to predict a SAA level < 4 mg/L, which in our opinion are acceptable values for use day to day. The sensitivity and specificity of a cut-off value for CRP changes when different levels of SAA are pursued. In relevant literature different cut-off values for SAA are used. Choices are based on possible damage to organs and/or increased risk of AA amyloidosis and vary between 4 and 10 mg/L [4, 7, 13, 14]. We have chosen a cut-off value of SAA below 4 mg/L, because of known favourable outcomes in terms of mortality and renal disease progression in patients with already existing AA amyloidosis [4]. However, the best cut-off value is still debatable.

When comparing the false positives to the true positives and the false negatives to the true negatives, the number of patients in remission was higher in the false positive group compared to the true positives (77% versus 59.2%) and all patients were in remission during the visits in the false negative group compared to 92% of the true negatives. The correlation between CRP and SAA was lower in the group in remission compared to the group with a flare (rho = 0.690 versus rho = 0.911). These results are in line with other research in FMF patients where CRP seems to be a less suitable marker to predict elevated SAA levels in FMF patients in an attack-free period. The frequency of colchicine use was lower in the false negative group compared to true negatives (30.8% versus 47.1%) and slightly higher in the false positive group compared to the true positive group (41.7% versus 35.5%). Colchicine decreases or even inhibits the production of SAA, as reported in mice studies [13, 28]. The higher frequency of colchicine use in the false positive group and lower frequency of colchicine use in the false negative group might be explained by the assumption that SAA levels are more influenced by colchicine than the CRP levels.

The frequency of prednisone use was higher in the false negative group compared to the true negatives (23.1% versus 5.3%)Age, BMI, use of medication and disease activity state may vary over time in the included patients. Therefore, we decided to include all available combined SAA and CRP measurements for all patients in the analysis shown in Table 4. But some patients have more frequent visits than other and as a consequence, their results may have weighed more than the results of other patients, thus resulting in a bias. For this reason, we also looked at the correlation between SAA and CRP and the SAA/CRP ratios, when only looking at the first visits of patients with and without prednisone. The correlation between CRP and SAA seems to be lower and the SAA/CRP ratio seems to be higher in the group with prednisone compared to the group without prednisone. Apart from that, the multivariate linear regression analysis revealed CRP, prednisone usage and ESR as independent predictors of log10SAA (Table 5). The function of prednisone, and other corticoids, in relation to inflammation is still not fully understood. Glucocorticoids seem to have both suppressing and stimulatory effects on the immune system, depending on the concentration of glucocorticoids in the body [29]. A study in healthy canines describes non-significant changes in CRP and SAA levels after prednisone administration [30]. No other research has looked into the effect of prednisone on acute phase reactants in humans to the best of our knowledge. Our results suggest that clinicians should be careful with interpretation of CRP levels to predict SAA levels in patients using prednisone.

The association of increased ESR levels with SAA might be the result of the presence of more proteins in the blood increasing the erythrocyte sedimentation rate. All of this should be taken into account by physicians interpreting CRP values to predict SAA levels. If a patient shows elevated ESR levels and/or uses prednisone it is likely that increased SAA levels are present, even with a low CRP.

A limitation of this retrospective study design is that, besides CRP and SAA, various other markers for inflammation (such as interleukin-18, interleukin-6, ferritin, and body temperature) were not investigated. Another limitation of this study is that we didn’t gather information about the exact time point of the blood test in time relation to the disease episode. Previous research showed that, although CRP and SAA are both sensitive biomarkers for inflammation in the early phase of an inflammatory disease, the kinetics of CRP and SAA levels herein are not completely similar. The half-life times of SAA levels are significantly shorter than that of CRP [31] and this will influence the correlation between SAA and CRP levels at different time points throughout the disease period.

Another limitation of this study was that we did not include genetic test results in the analysis of the relation between CRP and SAA since this might also play a role [7]. Furthermore, we included a relative small population (n = 99), with a larger variety of autoinflammatory diseases, making it a very heterogenous group. These limitations should be taken into account when interpretating the results of this study.