Determination of the smallest detectable change (SDC) and the minimal important difference (MID) for the Small Fiber Neuropathy Screening List (SFNSL) in sarcoidosis.

Small fiber neuropathy (SFN) is a disabling generalized sensory nerve disorder with structural and functional abnormalities of small fibers. It is characterized histopathologically by degeneration of small fiber nerve endings, selectively involving thin myelinated A∂- and unmyelinated C-fibers ( 1 ). Small fibers are associated with thermal and noci-ceptive sensations (somatic) and autonomic function, resulting in symptoms of neuropathic pain and autonomic dysfunction ( 2 , 3 ). The widespread spectrum of symptoms is associated with substantial negative effect on patients’ quality of life (QoL) ( 4 , 5 ). To date, SFN has been difficult to diagnose, as a gold standard is still lacking. The diagnosis is based on the presence of characteristic clinical features in combination with abnormalities on neurophysiological tests and/or reduced numbers of small fibers in skin biopsy ( 3 , 6 - 8 ). SFN can be an epiphenomenon in many diseases ( 9 ), including sarcoidosis. Sarcoidosis is a multisystem disorder that represents a burden on patients’ lives. The clinical manifestation, natural history, and prognosis of sarcoidosis are highly variable, and its course is often unpredictable ( 10 ). In addition to specific organ-related symptoms with functional impairments, sarcoidosis patients have disabling non-specific symptoms, including fatigue, physical impairments, everyday cognitive failure, and pain ( 5 , 11 ). Previous studies found that 40-60% of the sarcoidosis patients suffer from SFN ( 12 - 15 ). However, there is as yet insufficient awareness of this among clinicians responsible for the follow-up of sarcoidosis patients. There is a clear need for an easily administered small fiber neuropathy (SFN) screening instrument to identify sarcoidosis-associated SFN in general clinical practice and for research purposes. To meet this need, we have previously developed the Small Fiber Neuropathy Screening List (SFNSL) ( 16 ). The quality of a (screening) questionnaire is determined by its validity, reliability and responsiveness ( 17 , 18 ) as well as interpretability ( 19 ). Interpretability is an important property of questionnaires that are regularly used in daily practice; it refers to what the SFNSL score means for the patient ( 19 ). Previously, Sun et al. showed the validity of the SFNSL in a different cohort, reporting a high correlation between intraepidermal nerve fiber density and the SFNSL, with a sensitivity of 94.1% and a specificity of 90.9% ( 20 ). For evaluative purposes, interpretability is assessed by change in scores: it is important to know when it can be said that a patient has improved or when their situation has worsened. Interpreting change in score requires two benchmarks: the measurement error, expressed as the smallest detectable change (SDC), and the minimal important difference (MID). The SDC represents the minimal change that a patient must show on the scale to ensure that the observed change is real and not just measurement error ( 19 ). The MID is the smallest change score that a patient perceives to be important ( 21 ). Establishing the SDC and MID will help clinicians interpret the clinical meaning of changes in SFNSL scores over time at individual level. At group level, both measures can be used to perform methodologically sound research and for the clinical interpretation of research findings. Several previous studies have used the SFNSL scores as an outcome measure to determine therapeutic response, and have found significant differences between placebo versus study drugs in patients suffering from SFN ( 22 - 24 ). Although these differences were statistically significant, they might not be relevant to a patient, since no MID and SDC were known for the SFNSL at the time. Till now, no SDC or MID has been established for this questionnaire in patients with sarcoidosis. It is important to determine changes in the SFN-related complaints, and knowledge is therefore needed regarding the measurement properties of the SFNSL. The present study aimed to determine the measurement error (SDC) and interpretability (MID) of the Small Fiber Neuropathy Screening List SFNSL in patients with sarcoidosis.

Curavista (a certified eHealth platform) provided the online platform for this study and maintained the website. The data were collected online by Curavista and were exported to a database. Invitation to complete demographic questions and questionnaires were sent by email via the online Dutch Neurosarcoidosis Registry to all patients every 6 months. The following demographic and medical characteristics of patients were collected (self-reported): sex, age, time since diagnosis of sarcoidosis, medications use, use of prednisone, use of azathioprine (Imuran), use of methotrexate, use of infliximab (Remicade), use of adalimumab (Humira), use of cyclophosphamide and use of other medication. Patients received usual care and medication for their sarcoidosis treatment. SFN-associated symptoms are likely to remain stable or progress in time, since no curative treatment is as yet available. At baseline, the whole cohort was asked to complete the Fatigue Assessment Scale (FAS) ( 30 , 31 ), King’s sarcoidosis questionnaire (KSQ) ( 32 , 33 ), World Health Organization Quality of Life (WHOQOL) Bref ( 34 ) and a visual analogue scale (VAS) for pain ( 35 ). Furthermore, they were asked to complete the SFNSL at baseline (t0) and at 6-month follow-up (t1). An anchor question was added at t1.

The anchor question asked at t1 was: Compared to six months ago, do you think that – regarding the questions about SFN complaints (SFNSL) – you are doing worse, the same or better?

Data from t0 were used to determine the measurement error. The change score distribution was tested for normality (skewness and excess kurtosis), as it was important that the change scores should be normally distributed and close to zero ( 43 ). Measurement error can be expressed as the standard error of measurement (SEM) or as SDC ( 19 ). The SEM represents the standard deviation of repeated measures in one patient, and was calculated using the intraclass correlation coefficient (ICC): SEM = SD * √(1-ICC agreement) ( 21 ). The ICC was calculated using a two-way mixed effects model for absolute agreement. The SDC, the smallest change in a score that that you can detect with the SFNSL above measurement error ( 19 ), is also known as the minimal detectable change when using its 95% confidence interval (MDC 95 ), and was calculated using the SEM: 1.96 × √2 × SEM ( 44 ). These values were expressed in the unit of measurement of the SFNSL scale.

The flowchart of the study is displayed in Figure 1 . A total of 150 patients subscribed to the online Dutch Neurosarcoidosis Registry study. Twelve of them never completed the SFNSL and were therefore excluded from the analysis. The response rate at t0 (baseline), with 138 respondents, was 92%. Of these 138 patients, 89 (64%) completed the SFNSL and the anchor question at t1, and could thus be analyzed. Fig. 1. Flowchart of study patients Demographical and medical characteristics of the 89 patients are presented in Table 1 . Neurosarcoidosis was established in 44 patients (49.4%) and SFN-associated complaints in 81 patients (91%). The organ involvement presented in table 1 was self-reported by completing the KSQ. Table 1. Demographic and medical characteristics of neurosarcoidosis patients and/or patients suffering from SFN symptoms (n=89) Characteristics Number (%) or median (25 th -75 th percentile) or mean±SD Demographics Sex: male 45 (51) Age (yr) 53.2±10.5 Medical Variables Time since diagnosis (yr ) 6(2-12) Age at diagnosis neurosarcoidosis (yr ) 44 (49.4) SFN-related complaints (SFNSL >11) 81 (91) Medication use 62 (70) • prednisone 32 (36) • azathioprine (Imuran) 3 (3) • methotrexate 30 (34) • infliximab (Remicade) 12 (14) • adalimumab (Humira) 4 (5) • cyclophosphamide 0 (0) • other medication 14 (16) Psychological variables FAS score 32.1±8.3 WHOQOL-BREF Physical Health 10.5±2.7 Psychological Health 13.2± 2.7 Social Relationships 13.5±2.9 Environment 15.2±2.2 KSQ General Health Status 55±19.2 Lung (N=60, %) 60.9±23.8 Skin (N=20) 66.7±22.5 Eyes (N=41) 54.3±22.6 Medication (N=62) 65.7±26.7 SFN=small fiber neuropathy, SFNSL=small fiber neuropathy screening list, FAS = fatigue assessment scale, WHOQOL-BREF=World Health Organization Quality of Life BREF, KSQ=King’s sarcoidosis questionnaire The mean scores on the FAS, SFNSL and VAS pain were 32.1, 34.9 and 4.4, respectively (see Tables 1 and 2 ). The VAS correlates well with the SFNSL score (R=0.542; p-value=<0.0001). Table 2. Small Fiber Neuropathy Screening List (SFNSL) and VAS pain scores (N=89) Mean±SD or number (%) SFNSL t0 score (0-84) 34.9±18.7 SFNSL t0, definitely no SFN (<11) 8 (9) SFNSL t0, probable SFN (11-48) 56 (63) SFNSL t0, definitely SFN (>48) 25 (28) VAS pain score 4.4±2.8 VAS=visual analogue scale A statistically significantly higher age was found in the follow-up group compared to the age in the drop-out group (t(136)=2.75, p <0.01). No other differences were found in the variables mentioned in table 1 nor the SFNSL at t0 between the patients who dropped-out of the study before t1 (i.e. the 6-month follow-up) and the patients remaining in the study. Smallest detectable change (measurement error) The t1 questionnaires were completed within an average of 5.9 months (SD=0.57) after t0, without skewness but with a leptokurtic distribution (z-score>3.29), implying a higher peak and thus more clustered data around the mean. The change scores of the SFNSL were normally distributed (no statistically significant skewness or excess kurtosis) and close to zero, with an average of -2.44 (SD=8.12). The intraclass correlation coefficient (ICC agreement) was 0.95 for the SFNSL, with a SEM of 4.26 as calculated by the abovementioned equation. The calculated value of SDC or MDC 95 was 11.8.

The SFNSL is a validated questionnaire to assess the presence of SFN-related symptoms in sarcoidosis. The present study established the SDC and MID for the SFNSL in a group of sarcoidosis patients. A MID of 3.5 was found to distinguish between change of complaints and no change of complaints. To avoid any misconception, the SFNSL was developed to assess symptoms which might be related to SFN. The results might offer a clinician a first suspicion of SFN and stimulate him or her to refer this patient to a neurologist for further evaluation of the problem related to or probably associated with the symptoms. Even when the symptoms are not related to SFN they cause a substantial burden for the patients and have impact on their QOL. It is always hard to measure and follow-up non organ related symptoms. Moreover, so far no gold standard for diagnosing SFN exists. Therefore, the MID was calculated just to evaluate whether from the patient’s perspective the symptoms changed or not. Hence, it is not meant to assess improvement of the underlying cause. The MID is the smallest change score that patients perceive to be important ( 21 ). As such, it is the smallest change score that one wants to detect with the SFNSL. The SDC is the smallest change score that one can detect with the SFNSL in individual patients. The SDC we found was larger than the MID for detecting change of complaints and slightly larger than the MID for detecting improvement. This means that it is more difficult to distinguish a clinically important change from the measurement error with a large degree of certainty in individual cases, compared to patient-related outcomes (PROMs), where the SDC is smaller than the MID ( 19 , 45 ). In this case: if the individual change in score is 3.5 points or higher, there may be a clinical relevant change on the SFNSL. However, it may also be due to a measurement error on the scale. A change of 11.8 points or higher gives a high degree of certainty that it indicates a clinically relevant change in SFN related symptoms. Nevertheless, the similar results for both methods (MID for improvement 8.5 vs 8.0) might suggest that our estimation for MID is robust. SFN is a major concern in sarcoidosis, leading to decreased QoL ( 4 ). SFN complaints are difficult to manage, since no curative treatment currently exists ( 3 ).The treatment of SFN is at present merely symptomatic (neuropathic pain medications) with poor to moderate responses. The validated SFNSL is a practical clinical screening tool that can guide clinicians in assessing whether patients are suffering from SFN-associated symptoms, but the interpretability of change scores (MID) over time was unclear. The MID is of great clinical importance, since it will help to manage these patients better. In clinical practice, a positive change in symptoms scores is not to be expected for many patients. This is why we decided to determine the MID, in order to enable doctors in clinical practice to distinguish changes in complaints. For research purposes, we also determined the MID for improvement in SFNSL scores. The higher SDC is less relevant for use in clinical trials at group level than it is for use in individual patients. In clinical research, it is important to look at clinically relevant improvements of SFN complaints and not merely at a statistically significant change. The MID could therefore be used as a useful outcome measure in clinical trials. TNF-alpha inhibition and immunoglobulins have been reported to be effective treatment options for sarcoidosis-associated SFN ( 46 - 49 ). Whether these expensive treatments should be initiated for SFN is unclear. These treatment modalities, as well as new promising drugs such as Cibinetide® (ARA290) ( 24 , 50 ) also need to be evaluated using the SFNSL and MID, to assess whether there is a clinically relevant effect, rather than a mere statistically significant increase in the number of small fibers. Our group has previously determined the MID for another validated questionnaire in sarcoidosis, the fatigue assessment scale (FAS) ( 31 ). Fatigue is also a major concern in sarcoidosis, and determining the MID for the FAS has made a significant contribution to clinical practice in terms of improving the clinical interpretation of fatigue scores. The present study used the same methods (anchor-based and distribution-based methods). The area under the ROC curve (AUC) for both analyses (improvement and change in complaints on the SFNSL) was 0.6, which is poor. The low AUC can be explained, amongst other things, by the heterogeneity of sarcoidosis and the wide organ- as well as non-organ related symptoms. For instance, the MCID of the FAS has an AUC of 0.6 and the MCID of the VAS pain was 0.7 ( 31 , 51 ). Therefore, an AUC of 0.6 cannot be considered a poor performance and may be used in clinical practice, especially since no other estimate is available’. This study had some limitations. One limitation may be that the patients we