4.1 Results of search strategy
The PubMed search yielded 52 results as shown in Table 3 (Appendix 3). After selecting for ‘published in the last 5 years’ 17 articles remained (Appendix 4). A selection of these 17 articles was performed on the basis of title and abstract (Appendix 5). Articles in this review are included according inclusion criteria described in the methods. Reasons for not including articles are using a small research population (14), using a modified LDIflare method (14), not understanding Japanese (15), not comparing the LDIflare method with the golden standard (16), using the LDIflare method for a different purpose (17), excluding of neuropathy patients (18, 19), studies performed on rats (20, 21), studies performed in pedriatic populations (22) and studying other unmyelinated C-fibers in regions other than the skin (21) (Appendix 5).
In the end, 7 articles were included in this review to answer the research question.
4.2 Critical assessment of included studies
After selection 7 articles (1, 6, 9, 10, 23-25) remain to write this review.
Critical assessment of these 7 articles was performed with the diagnostic CAT criteria. The Dutch Cochrane diagnostic criteria will complement the diagnostic CAT criteria. Both the diagnostic CAT criteria and the Dutch Cochrane diagnostic criteria are shown in Appendix 6.
4.3 Results of critical review
4.3.1 Comparison with golden standard
In summary, in none of the articles the LDIflare method is compared to the golden standard, skin biopsy with quantification of intraepidermal nerve fiber density and/or quantitative sensory testing in combination with the clinical characteristics of the patient. However, Vas et al. (2013) compared the sensitivities and specificities of the LDIflare method with the golden standard. The characteristics, symptoms and stage of SFN were well described in most articles.
188.8.131.52 Work-up bias
Work-up bias was avoided in the study of Lysy et al. (2014), Gibbons et al. (2013), Namer et al (2013) and Gibbons et al. (2010) because both the LDIflare method and reference tests were performed. In the other studies (1, 9, 23), determining work-up bias was not possible because only LDIflare was applied.
184.108.40.206 Expectation bias
Expectation bias was avoided in the study of Vas et al. (2013), Kokotis et al. (2013), Gibbons et al. (2013) and Gibbons et al. (2010) by blinding trained examiners to other test results, for example test results of nerve conduction studies. In the other articles, expectation bias was not to determine because nothing was mentioned about it in the article (6, 9, 10).
Studies have a cross-sectional manner and patients were only examined once. Therefore, the reproducibility and repeatability of the method is unclear. Although this is unknown, all articles presented a sufficient description that was used or they referred to other articles that used the same method.
In general, no confidence intervals of validity outcomes were represented. Only the 95% confidence intervals of the outcome value LDIflare were represented. Therefore the validity of the LDIflare method is hard to determine. In contrast, Kokotis et al. (2013) described both the 95% confidence intervals of the outcome value LDIflare and the 95% confidence intervals of the sensitivity and specificity. The width of the 95% confidence interval of the sensitivity is approximately 38% while the width of the 95% confidence interval of the specificity is approximately 15%. The 95% confidence interval of the sensitivity is too large, which indicates less accuracy and validity
4.3.5 Sensitivity and specificity
In general, sensitivity and specificity of the LDIflare method are lower than the sensitivity and specificity of the golden standard. In contrast, Vas et al. (2013) found higher sensitivity and specificity of the LDIflare method, namely a sensitivity of 75% and a specificity of 85% by means of ROC analysis. In addition, the LDIflare technique using age-specific cut off values has even higher sensitivity and specificity, 77% and 90% respectively. Kokotis et al. (2013) found a sensitivity and specificity of the flare area measured by LDI of 43.5% and 95.3%, respectively. Nabavi et al. (2012) found an AUC of 0.72 when employing the England criteria (9). The optimal threshold value of 1.90 cm2 has limited sensitivity and specificity of 70% and 66%, respectively. When employing the sural nerve criteria for subclinical diabetic sensorimotor neuropathy, the AUC was 0.75 and the sensitivity was 79% for LDIflare. Gibbons et al. (2010) found a best overall sensitivity of 71% and specificity of 43% for axon reflex % change measured by Laser Doppler flowcytometry.
4.3.6 Diagnostic process
In all studies, there is no appropriate cut-off point used because these cut-off points were not decisive for diagnosing SFN. However, in the study of Vas et al. (2013), Namer et al. (2013) and Nabavi et al. (2012) cut-off points were decisive whether the outcome LDIflare was abnormal. In all studies, there was no question of a diagnostic process and therefore the moment of performing the LDIflare method cannot be compared with the moment in practice, when patients experience sensory and autonomic symptoms.
In the overview below results of the critical appraisal of a topic are represented (Table 4). There are points awarded to the plus and minus signs in which ++ = 2 points, + = 1 point, +/- = 0 points, – = -1 point and — = -2 points. The sign X is awarded if the criterion could not be determined. This should be taken into account in the assessment of the article. From the results it appears that few articles are of sufficient quality to answer the research question. The article of Vas et al. (2013) is the most useful to write this review. Additional, the article of Kokotis et al. (2013) and Gibbons et al. (2013) also will be useful to answer the research question. The remaining articles will still be included due to the fact that these articles meet certain criteria (6, 9, 10, 25).
Table 4. Overview of the critical assessment of articles according the CAT criteria