Introduction
Nailfold capillary microscopy (NCM) is a cornerstone in the diagnosis of systemic sclerosis (SSc) in adulthood [
1]. Although Raynaud’s phenomenon (RP) is common in childhood with a prevalence of 14.9% in children aged 12–15 years [
2‐
4], studies on diagnostic methods to differentiate between primary RP (PRP) and secondary RP (SRP) at a young age are scarce. In PRP no cause can be identified, whereas SRP occurs as part of an underlying disease, most commonly a connective tissue disease (CTD) [
5]. The disease course of CTDs varies, and is often associated with severe morbidity and sometimes mortality. In children and adolescents, early diagnosing and optimal treatment at an early stage of the disease can improve prognosis significantly [
6]. Therefore, it is important to distinguish between PRP and SRP early.
NCM is a non-invasive method for visualizing the microcirculation of the nailbed. In adults, NCM has been established as an effective method for predicting the development of SSc in RP patients, especially when combined with serology [
1]. Capillaroscopic findings that should alert the physician to the possibility of undetected SSc are hemorrhages, loss of capillaries, widened capillaries and giant capillaries [
7,
8]. These capillaroscopic abnormalities are mainly seen in SSc, but have been reported in other CTDs as well [
9,
10]. For rheumatoid arthritis (RA), it was reported that 20.9% of the patients showed capillaroscopic abnormalities [
11]. There are few studies on capillaroscopic abnormalities in Sjögren’s syndrome (SS) and systemic lupus erythematosus (SLE). However, the studies that have been performed found significantly more capillaroscopic abnormalities in SS and SLE patients compared to the healthy population [
12,
13].
For children with RP, however, guidelines for initial investigation and subsequent follow-up are based on limited evidence [
14,
15]. Only one pediatric prospective study has been conducted previously, and found results similar to studies in adults with RP. In this study, SSc-related capillary changes significantly correlated with development of SSc-spectrum disorders in the near future. Future development of other CTDs could not be predicted by specific capillary changes [
15,
16].
A screening method for CTDs that is often performed in children is serological testing for ANAs. A high ANA titer is suggestive of a CTD. Antibodies against one or more extractable nuclear antigens (ENAs) suggest a specific CTD such as SSc, SS or SLE [
17]. ANA and anti-ENA antibodies are important diagnostic markers of systemic rheumatic diseases, especially because for many diseases they are part of the diagnostic criteria.
The general aim of this study was to determine the value of NCM in differentiating between PRP and SRP in children and adolescents with RP.
Discussion
The aim of this study was to determine the value of NCM in differentiating between PRP and SRP in children and adolescents with RP. Capillary loss shown on NCM was associated with SRP. However, it did not add to the predictive value of ANAs. Other NCM characteristics were not associated with the presence of SRP.
This study showed that whereas SRP in adults is mostly associated with SSc, children and adolescents with SRP seem to mostly develop other CTDs. This is supported by the fact that only 1 of the 18 SRP patients developed SSc. However, 6 (33.3%) of the SRP had an SSc pattern with NCM, all of which showed active changes. Therefore, a study with more children should be performed to confirm whether children develop SSc less frequently than adults.
In this study, 22% of the patients were classified as SRP and 78% as PRP. At the time of referral, none of the patients were diagnosed with SRP yet. Some patients showed other symptoms at presentation, mostly joint complaints. For some patients, a CTD was suspected, but ANA and NCM were requested to further aid in diagnostics. In a few cases, patients showed ANA positivity as well as NCM abnormalities. These patients were still classified as PRP because they did not have symptoms suggestive of a CTD for longer than three years, which was one of the criteria for UCTD [
24]. The mean time that passed between the start of RP and a definite diagnosis of a CTD was 2.1 years. These findings are comparable to those reported in the only other longitudinal study on childhood RP by Pavlov et al., which found SRP in 24% of the patients and PRP in 76% of the patients. The time between onset of RP and the diagnosis of a CTD was 2.4 years [
15]. Both studies indicate that childhood RP is primary in most cases. However, in comparison to adult RP, childhood RP is more often secondary [
29]. Other studies reported even higher proportions of SRP in children, ranging from 31 to 52% [
15,
16,
30]. Pavlov et al. proposed as possible explanation the fact that the aforementioned studies only included children younger than 18 years, whereas their own study investigated children and adolescents younger than 20 years. However, our study only included patients below the age of 18 and also found a lower frequency of SRP. Therefore the cut-off point for age does not seem to be the cause of this difference. It is more likely that the difference in frequencies is partly attributable to differences in criteria chosen to establish RP and to classify CTDs.
The majority of children with SRP were diagnosed with MCTD, UCTD and SLE. SSc, DM and SS were seen less frequently. A possible explanation for this finding is that children with DM and SS are not always referred. Pavlov et al. reported a similar distribution of CTD diagnoses. Noteworthy is that Pavlov et al. found RA or juvenile idiopathic arthritis (JIA) in 4% of the patients and Nigrovic et al. found arthritis in 29% of the patients [
15,
16]. Although the present study did find joint complaints to be a common early presenting symptom of juvenile CTDs, none of the children were diagnosed with RA or JIA. This difference is probably caused by a selection bias, as only patients who had underwent NCM were selected for this study. RA and JIA rarely present with RP as its first symptom. For this reason, NCM is usually not performed at first clinical evaluation [
31]. When RP develops later during the disease process, NCM is not required to differentiate between PRP and SRP, as it is has already been established that a disease is present. Therefore RA and JIA patients may not have been included in this study.
The present study showed that ANA positivity was highly predictive of SRP. Anti-ENA antibodies were also strongly associated with SRP. These associations underline the importance of serology in children with RP. However, for most CTDs ANA or anti-ENA positivity is part of the diagnostic criteria [
32]. Therefore, the involvement of ANA positivity in the diagnostics of CTDs has inevitably influenced the predictive value of ANA positivity for SRP. For this reason, this result should be interpreted with extreme caution.
Capillary loss on NCM was associated with SRP. However, capillary loss did not add to the predictive value of ANAs as shown by the multivariate analysis including ANA positivity. Other NCM parameters were not associated with SRP. It is plausible that the predictive value of ANA positivity was so strong, as it occurs so frequently in childhood rheumatic diseases, that it prevented other variables to be selected as independent predictors in this multivariate model. The multivariate analysis not including ANA positivity shows that this is likely true. In the multivariate analysis without ANA positivity, capillary loss was shown to be an independent predictor of SRP. Noteworthy for the models with and without ANA positivity is that they had high percentages of correctly predicted PRP (92% and 98% respectively), but quite low percentages of correctly predicted SRP (38% and 28% respectively). Because the multivariate model was likely influenced by the strong predictive value of ANA, an ROC analysis was also performed. An ROC analysis cannot have been influenced by ANA positivity. ROC and AUC analysis showed that capillary loss was an insufficient predictor of SRP.
The NCM finding capillary loss as a predictor for SRP had a high specificity and negative predictive value (NPV). However, the sensitivity and positive predictive value (PPV) were low. ANA screening performed better as a predictor for SRP with a higher sensitivity, specificity PPV and NPV. ANA screening and capillary loss combined as a predictor of SRP performed equally to ANA screening only as a predictor. The variations of NCM findings at different ages should be considered when interpreting this result [
33,
34]. Specifically, Piotto et al. showed that age is positively correlated with the amount of capillaries in healthy children [
33]. We were unable to analyze subgroups due to our small study population. In our study, the mean age of the SRP patients was lower than that of the PRP patients. This may have influenced our results. Furthermore, although the SRP patients had significantly fewer capillaries, they had a mean number of capillaries of 6.4 per mm. This is above the definition of capillary loss of < 6 per mm. It should also be noted that the ROC curve for the number of capillaries had an AUC of 0.65 that was not significant (
p = 0.06). This shows that capillary loss has no value in distinguishing between PRP and SRP and this should be kept in mind when interpreting the sensitivity and specificity. Our study did show a trend towards a high specificity and NPV, which suggests NCM shows potential as a first screening method for SRP. However, because of the aforementioned points, definite conclusions about this cannot be made. A larger prospective study is needed to analyze whether NCM is a predictor of SRP.
Whereas studies in adults clearly show a relationship between giant capillaries on NCM and SRP [
27], the present study found no evidence of this relationship in children. A possible explanation for this observation is that giant capillaries are not an early feature of the CTDs that children with RP are mostly at risk for. They are, however, an early feature of SSc, the most common CTD in adults with SRP [
29]. Giant capillaries might still be helpful in defining a small subgroup of children with RP at risk of developing SSc. In our study, the one SSc patient did have giant capillaries, as well as three out of four MCTD patients and one of the two DM patients. This is in line with previous studies, which showed that a SSc pattern is common in MCTD and DM [
27,
35‐
37]. It is also possible that although giant capillaries do not differentiate between PRP and SRP, they could still be predictive of SRP or a specific CTD. This can be seen in a recent study by Schonenberg-Meinema et al., which investigated whether NCM findings are abnormal in childhood-onset SLE and found that children with SLE had significantly more giant capillaries, abnormal capillaries and hemorrhages in comparison to healthy controls [
38]. Although children with SLE had significantly more giant capillaries than healthy controls, the study did not find a significant correlation of RP and giant capillaries in the SLE patients [
38]. The patient numbers of this study are too small to test for a relationship with individual CTDs, but studies with more participants and larger diagnostic subgroups might be capable of detecting such a relationship. More research is needed to determine whether giant capillaries are predictive of specific CTDs such as SSc.
It was established that NCM findings are poor predictors of SRP in children. However, SRP can be due to a large variety of CTDs. Juvenile SSc is rare, with an incidence of 0.27 per million per year [
39]. In comparison, the incidence of childhood SLE is 0.3–0.9 per 100.000 children per year [
40], the incidence of childhood MCTD is 0.1 per million children per year [
41] and the incidence of childhood DM is 2–4 per million children per year [
42]. It is possible that NCM is not as good a predictor in children as it is in adults, because children mostly develop CTDs that are less strongly associated with capillaroscopic abnormalities. Furthermore, our study only looked at SRP, meaning it analyzed the predictive value for all CTDs together. NCM could still be predictive of some individual CTDs. The current study was unable to analyze this hypothesis, because of the small patient group with SRP at follow up (
n = 18). It might be of interest for future research to analyze the predictive value of NCM for each CTD individually when a larger study population of children with RP is available. The aforementioned study by Schonenberg-Meinema et al. investigated NCM findings in childhood-onset SLE in comparison to healthy controls [
38]. Our study adds to theirs, because we compared to PRP patients, whereas their study compared to healthy controls. The study by Schonenberg-Meinema et al. emphasizes the importance of looking at specific NCM abnormalities. Because of the findings of this study, we believe NCM might be predictive of SLE in children with RP. The results of this study emphasize that it is worthwhile to analyze the predictive value of NCM for each CTD individually.
A limitation of this study is that the study population might be skewed towards SRP. All patients were referred to a tertiary center because they were suspected of having SRP. This may have led to overrepresentation of SRP patients, limiting the external validity for first line RP patients. Our study was also limited due to the retrospective design. Patient characteristics such as smoker status and family history of RP were not documented in the electronic patient files at the time of NCM and therefore unavailable to us. Another limitation of this study was that not all NCM were assessed using the same method. For the NCM performed until August 2013, dilated and giant capillaries were judged visually, whereas for the NCM performed after August 2013 they were measured apically. We expect that there are little differences between the visual judgement and measurements of giant capillaries. For dilated capillaries, however, there could be differences as these are visually less prominent. We found no significant association between SRP and dilated and giant capillaries. Therefore this limitation is not relevant for the outcome of our study. The number of capillaries was assessed over a 3 mm grid in both methods. Therefore, the use of different methods is highly unlikely to have affected the significant association of capillary loss with SRP. Lastly, the amount of parameters assessed with NCM was limited. Not all parameters defined by Piotto et al. were assessed during NCM [
33]. More research should be performed to determine the value of NCM as a screening method when using the parameters defined by Piotto et al. [
33].
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.