ADVERTISEMENT

Upcoming Website Maintenance Notification

On Sunday October 20th, 2019 from 5:00 am cst to 8:00 am cst there will be approximately 3 hours of scheduled downtime while we make upgrades to our servers. During this time-window the JCSM website will be unavailable. Thank you for your patience.

gears

Issue Navigator

Volume 13 No. 03
Earn CME
Accepted Papers





Scientific Investigations

Intravenous Immunoglobulin Therapy in Pediatric Narcolepsy: A Nonrandomized, Open-Label, Controlled, Longitudinal Observational Study

Michel Lecendreux, MD1,2; Johanna Berthier, Medical Student3; Jennifer Corny, PharmD4; Olivier Bourdon, MD, PhD4,5; Claire Dossier, MD6; Christophe Delclaux, MD, PhD1,7
1AP-HP, Pediatric Sleep Center, Hospital Robert-Debré, Paris, France; 2National Reference Centre for Orphan Diseases, Narcolepsy, Idiopathic Hypersomnia and Kleine-Levin Syndrome (CNR Narcolepsie-Hypersomnie), Paris, France; 3Carol Davila University of Medicine and Pharmacy, Bucharest, Romania; 4Pharmacy Department, Hospital Robert-Debré, Paris, France; 5Pharmacy Faculty, Paris Descartes, Paris, France; 6Pediatric Nephrology Department, Hospital Robert-Debré, Paris, France; 7Paris Diderot University, Sorbonne Paris Cité, Paris, France

ABSTRACT

Study Objectives:

Previous case reports of intravenous immunoglobulins (IVIg) in pediatric narcolepsy have shown contradictory results.

Methods:

This was a nonrandomized, open-label, controlled, longitudinal observational study of IVIg use in pediatric narcolepsy with retrospective data collection from medical files obtained from a single pediatric national reference center for the treatment of narcolepsy in France. Of 56 consecutively referred patients with narcolepsy, 24 received IVIg (3 infusions administered at 1-mo intervals) in addition to standard care (psychostimulants and/or anticataplectic agents), and 32 continued on standard care alone (controls).

Results:

For two patients in each group, medical files were unavailable. Of the 22 IVIg patients, all had cerebrospinal fluid (CSF) hypocretin ≤ 110 pg/mL and were HLA-DQB1*06:02 positive. Of the 30 control patients, 29 were HLA-DQB1*06:02 positive and of those with available CSF measurements, all 12 had hypocretin ≤ 110 pg/mL. Compared with control patients, IVIg patients had shorter disease duration, shorter latency to sleep onset, and more had received H1N1 vaccination. Mean (standard deviation) follow-up length was 2.4 (1.1) y in the IVIg group and 3.9 (1.7) y in controls. In multivariate-adjusted linear mixed-effects analyses of change from baseline in Ullanlinna Narcolepsy Scale (UNS) scores, high baseline UNS, but not IVIg treatment, was associated with a reduction in narcolepsy symptoms. On time-to-event analysis, among patients with high baseline UNS scores, control patients achieved a UNS score < 14 (indicating remission) less rapidly than IVIg patients (adjusted hazard ratio 0.18; 95% confidence interval: 95% confidence interval: 0.03, 0.95; p = 0.043). Shorter or longer disease duration did not influence treatment response in any analysis.

Conclusions:

Overall, narcolepsy symptoms were not significantly reduced by IVIg. However, in patients with high baseline symptoms, a subset of IVIg-treated patients achieved remission more rapidly than control patients.

Commentary:

A commentary on this article appears in this issue on page 363.

Citation:

Lecendreux M, Berthier J, Corny J, Bourdon O, Dossier C, Delclaux C. Intravenous immunoglobulin therapy in pediatric narcolepsy: a nonrandomized, open-label, controlled, longitudinal observational study. J Clin Sleep Med. 2017;13(3):441–453.


INTRODUCTION

Narcolepsy is a debilitating and currently an incurable neurological condition with a characteristic constellation of symptoms, including excessive daytime sleepiness (EDS), cataplexy, sleep paralysis, hypnagogic hallucinations, and disturbed nocturnal sleep.

Until recently, and at the time patients in this study were treated, the disorder was categorized into narcolepsy with cataplexy and narcolepsy without cataplexy according to the revised second edition of the International Classification of Sleep Disorders (ICSD).1 ICSD-2-revised has now been replaced by ICSD-3, which classifies narcolepsy into type 1 and type 2 narcolepsy.2 The majority of patients with type 1 narcolepsy have hypocretin deficiency (cerebrospinal fluid [CSF] hypocretin-1 level ≤ 110 pg/ mL), which is considered to result from an autoimmune-mediated loss of hypocretin-secreting neurons in those who are genetically predisposed (most commonly those with human leucocyte antigen [HLA] DQB1*06:02 haplotype). The pathophysiology of type 2 narcolepsy, however, is less well delineated.

BRIEF SUMMARY

Current Knowledge/Study Rationale: Pediatric narcolepsy is a chronic, debilitating condition with no available disease-modifying therapies. Because the disease is thought to result from autoimmune destruction of hypocretin neurons, intravenous immunoglobulin therapy may offer potential benefit, especially in patients with early-onset disease.

Study Impact: This report represents the most extensive evaluation to date of the treatment of narcolepsy in children <18 y of age with IVIg (1 g/kg in 3 monthly infusions plus standard care, n=22) compared with standard care alone (n=30), both in terms of numbers of patients evaluated (n=52) and the length of follow-up (3.9 y). Although, symptoms of narcolepsy were not significantly reduced by intravenous immunoglobulin compared with standard care alone, improvement was more rapid with intravenous immunoglobulin in a subset of patients with high baseline symptoms.

Pediatric narcolepsy usually manifests with higher levels of excessive daytime sleepiness, more frequent spontaneous cataplexy than emotion-triggered cataplexy, and more frequent secondary forms of the disease. In addition to the core symptoms of narcolepsy, affected children have higher rates of obesity,3 depressive symptoms,4 and attention deficit/hyperactivity disorder symptoms,5 which in turn results in substantially reduced quality of life6 and academic performance for these children.7,8

Current treatment options in narcolepsy remain limited to symptomatic treatment with psychostimulants and anticataplectic agents.9 An unmet need exists, therefore, to identify therapies for narcolepsy that target and prevent the assumed underlying autoimmune destruction of hypocretin neurons. In support of the autoimmune hypothesis underlying narcolepsy, autoantibodies against Tribbles homolog 2 (TRIB2), a protein synthesized in hypocretin neurons, are evident in a proportion of patients with narcolepsy, a finding that has been replicated in three separate studies.1012 In addition, the surge of cases of narcolepsy following pandemic H1N1 infection and mass vaccination with Pandemrix (GlaxoSmithKline, Brentford, Middlesex, UK) lends further support to an autoimmune etiology,1316 likely due to serum antibodies to H1N1 nucleoprotein in patients with vaccine-associated narcolepsy cross-reacting with the hypocretin receptor 2 (an effect observed following Pandremix, but not Focetria [Novartis, Novartis International AG, Basel, Switzerland] vaccination).17 However, in the absence of formal identification and detailed understanding of the mechanisms linking autoimmunity with the destruction of hypocretin neurons, no disease-modifying therapies for narcolepsy have yet been developed.

A number of immunomodulatory therapies, including corticosteroids, intravenous immunoglobulin (IVIg) and plasmapheresis, have been evaluated in narcolepsy.9,1825 A single case report of complete remission of cataplexy during treatment with alemtuzumab, a humanized monoclonal antibody targeting CD52 and approved for use in chronic lymphocytic leukemia, has been reported.26 Alemtuzumab has also been recently approved for use in multiple sclerosis, another auto-immune neurologic disorder, but is associated with serious infusion-related side effects and can itself precipitate new autoimmune disorders.27

In pediatric narcolepsy, patients often present close to the onset of the disease, which affords the opportunity to intervene rapidly with immunomodulatory therapies. Among the potential immunomodulatory therapies available, and despite its lack of specificity, IVIg is an attractive option for children with narcolepsy because of established efficacy and good tolerability in a number of other pediatric autoimmune diseases, including Guillain-Barré syndrome,28 Kawasaki disease,29 and Stevens-Johnson syndrome.30

The first published reports on the use of IVIg in patients with narcolepsy have been uncontrolled case studies involving very small numbers of patients. The results of these studies have been contradictory, although with limited evidence that early intervention shortly after disease onset may be more effective than later intervention.1823,25 Here, we present findings in patients who were consecutively referred to a French Pediatric National Reference Center for the treatment of narcolepsy, some of whom were treated with IVIg, and who were followed up over a period of years.

METHODS

Participants

All children (younger than 12 y) and adolescents (12 to 18 y) with a diagnosis of type 1 narcolepsy who were seen consecutively in a pediatric national reference center for narcolepsy (Hospital Robert-Debré, Paris) between March 2007 and November 2012 were included in this longitudinal, nonrandomized, retrospective medical chart review of patients receiving either standard care for narcolepsy (control group) or standard care plus open-label intravenous immunoglobulin therapy (IVIg group). Both parents and children signed a written consent form. The study was approved by the local ethics committee (Comité de Protection des Personnes - Ile de France 06).

According to the diagnostic criteria available at the time of clinical referral, patients were classified as having narcolepsy with cataplexy according to the criteria of the ICSD-2-revised,1 including: (1) complaints of EDS for at least 3 mo; (2) symptoms not better explained by other medical or psychiatric disorders; (3) the absence of secondary narcolepsy; (4) the presence of clear-cut cataplexy and/or (5) mean sleep latency lower than 8 min calculated from Multiple Sleep Latency Tests (MSLT) conducted throughout the day with two or more sleep onset rapid eye movement periods. Patients underwent HLA genotyping and a lumbar puncture for CSF hypocretin measurement.

Clinical Examination

A thorough physical examination was conducted to rule out any possible concomitant disorder, which could have indicated a diagnosis of secondary narcolepsy. Height and weight measurements were conducted at time of referral to the service and at follow-up visits. Body mass index (BMI) was calculated (weight/height2) and a z-score computed representing a standardized measure of weight adjusted for height, sex, and age relative to a smoothed reference distribution.31 Using standardized growth curves, obesity was defined as being on or above the centile trajectory corresponding to a BMI of 30 kg/m2 at the age of 18 y (International Obesity Task Force criteria).32

Interventions

All patients were offered standard treatments following the diagnosis of narcolepsy, including psychostimulants alone or with anticataplectic agents. In addition, a subgroup of referred children was offered IVIg at the discretion of the treating physician (no prespecified inclusion/exclusion criteria were applied). IVIg was administered at a dose of 1 g/kg as three separate infusions at monthly intervals, according to the protocol used in previous investigations.1921

Outcome Measures

Parent and/or adolescent self-report questionnaire measurements of narcolepsy and associated symptoms were conducted on all subjects at the time of referral to the service (baseline) and at follow-up visits to the clinic (up to eight follow-up visits were available). For the IVIg group, up to three measurements were available before IVIg was administered and up to six measurements after the intervention. The date of IVIg administration was set as time zero, with time (in months) calculated for each measurement from this reference point. In order to compare trajectories of symptom change between the IVIg group and the control group, the midpoint between the third and the fourth measurements was taken as time zero for the control group.

The following questionnaire measures were collected. The Ullanlinna Narcolepsy Scale (UNS) is an 11-item scale (range 0–44), which assesses both excessive daytime sleepiness and cataplexy. A cutoff score of 14 reliably distinguishes patients with narcolepsy from patients with other sleep disorders with 100% sensitivity and 98.8% specificity.33 Daytime sleepiness was also evaluated with the Pediatric Daytime Sleepiness Scale (PDSS),34 which has a score range of 0 to 44, and the Child and Adolescent Sleepiness Scale (CASS), which is a 10-item questionnaire that rates the degree of sleepiness while performing various activities throughout the day (the maximum score is 30, with a cutoff score higher than 15 indicating abnormality). The Clinical Global Impression (CGI) scale35 was rated by clinicians and used to capture the severity of daytime sleepiness (CGI-S) and cataplexy (CGI-C).

Statistical Analysis

Descriptive statistics for normally distributed continuous variables were expressed as means and standard deviations or otherwise as medians (min, max). Descriptive statistics for categorical variables were expressed as number (percent). In order to assess the influence of selection bias in this nonrandomized, open-label, observational study, differences in these variables at the referral visit (baseline) between the IVIg and control groups were evaluated using univariate generalized linear models (GLM function from the R statistical software platform version 3.1.2; http://www.R-project.org) for normally distributed continuous variables, using the npar.t.test function from the nparcomp package in R (http://cran.r-project.org/web/packages/nparcomp/index.html) for continuous variables that were not normally distributed, and using Fisher exact tests for categorical variables.

Comparisons of changes in outcomes of interest over time between the IVIg and control groups were analyzed using linear mixed effects (LME) models (nlme package, version 3.1–122) with fixed-effects terms for time, time2, treatment, baseline value, treatment-by-time interaction, and baseline-by-time interaction; and random-effects terms for intercept and slope. Correlations between repeated measurements within each subject were modelled using an autocorrelation structure. Model building used maximum likelihoods estimation with comparison of models using change in log likelihood. Definitive models used restricted maximum likelihoods estimation.

Two forms of adjustment were undertaken to assess the influence of differences in baseline characteristics between the IVIg and control groups on changes in outcomes with IVIg. First, multivariate adjustment with additional fixed-effects terms added to the LME models, including age at referral, sex, and BMI z-score, and other baseline variables identified from the univariate analyses of baseline variables as significantly different (p < 0.05) between the IVIg and control groups. Second, propensity score weighted LME models were conducted to account for selection assignment differences between the IVIg and control groups according to the methods outlined by Olmos and Govindasamy.36 For the generation of propensity scores using logistic regression, age at referral, sex, and BMI z-score were obligatorily included, and other variables were selected if they demonstrated an imbalance between treatment groups (p < 0.2 threshold). Propensity scores were then transformed into weights for use in subsequent analyses. For variables previously identified as imbalanced, weighted regression using the propensity score-derived weights was used to confirm that imbalance had been corrected.

Additional analyses included Cox proportional hazards models for achieving a ≥ 20% reduction in UNS scores or achieving a UNS score < 14 over time assessed using repeated time-to-event data with discontinuous risk intervals according to the methods of Guo et al.,37 a methodology that better reflects fluctuating symptom levels during the course of chronic illness than time-to-first-event analysis.

Statistical significance was determined at a level of alpha lower than 0.05, unadjusted for multiple comparisons in this exploratory, observational study.

RESULTS

Patients

During the course of the study period, 24 patients were recorded in their medical files to have been dispensed IVIg therapy by the hospital pharmacy department (IVIg group) and 32 patients had not received IVIg (control group). For two patients in the IVIg group and two patients in the control group, the medical record was unavailable at the time of the retrospective data collection. Thus, a total of 22 patients in the IVIg group and 30 patients in the control group were available for inclusion in this analysis.

With the exception of one control patient, all were HLADQB1*06:02 positive. All 22 patients treated with IVIg underwent a lumbar puncture and all exhibited a low hypocretin level (≤ 110 pg/mL). Of the 30 control patients, 12 underwent a lumbar puncture, and all of these exhibited a hypocretin level ≤ 110 pg/mL. Therefore, for the purposes of the current analysis, all patients were retrospectively reclassified as having ICSD-3 narcolepsy type 1 (hypocretin deficiency syndrome).2

Age at disease onset, age at referral, sex, BMI z-score, and proportions with obesity were balanced across the IVIg and control groups. However, patients were significantly more likely to have been assigned IVIg therapy if they had a shorter time interval between disease onset and referral, had received H1N1 vaccination, or had a shorter latency to sleep onset on MSLT (Table 1).

Demographics, physical characteristics, and clinical features at referral.

jcsm.13.3.441.t01.jpg

table icon
Table 1

Demographics, physical characteristics, and clinical features at referral.

(more ...)

Mean (standard deviation) length of follow-up was 2.4 (1.1) y for the IVIg group and 3.9 (1.7) y for the control group. Of the 22 IVIg-treated patients, 17 continued psychostimulants or anticataplectic agents from baseline throughout the follow-up period, 4 initiated psychostimulants or anticataplectic agents during the course of the follow-up period, and for 1 patient data on drug status was unknown. Corresponding values for the 30 control patients receiving standard care were 20, 3, and 7, respectively.

Descriptive Summaries of Symptom Scores by Study Visits

For all outcomes assessed, baseline mean scores tended to be higher in the IVIg group, mean scores at each time interval tended to be higher in the IVIg group, and mean changes from baseline scores showed no meaningful differences between the control and IVIg groups (Figure 1 and Figure 2).

Descriptive scores and associated 95% confidence intervals over time periods (in months).

(A) UNS mean score. (B) UNS mean change from baseline score. (C) PDSS mean score. (D) PDSS mean change from baseline score. (E) CASS mean score. (F) CASS mean change from baseline score. BL = baseline, CASS = Child and Adolescent Sleepiness Scale, IVIg = intravenous immunoglobulin, PDSS = Pediatric Daytime Sleepiness Scale, UNS = Ullanlinna Narcolepsy Scale.

jcsm.13.3.441a.jpg

jcsm.13.3.441a.jpg
Figure 1

Descriptive scores and associated 95% confidence intervals over time periods (in months).

(more ...)

Descriptive scores and associated 95% confidence intervals over time periods (in months).

(A) CGI-S mean score. (B) CGI-S mean change from baseline score. (C) CGI-C mean score. (D) CGI-C mean change from baseline score. BL = baseline, CGI-C = Clinical Global Impression – Cataplexy, CGI-S = Clinical Global Impression – Sleepiness, IVIg = intravenous immunoglobulin.

jcsm.13.3.441b.jpg

jcsm.13.3.441b.jpg
Figure 2

Descriptive scores and associated 95% confidence intervals over time periods (in months).

(more ...)

Linear Mixed-Effects Models

Model building

Before estimating treatment effects or covariate adjustment of treatment effects, the best-fitting model was identified that most accurately represented the observed data structure. The process of model building is described in Table 2. The best-fitting model (Model 5) allowed for exponential decay in UNS symptoms, between-subject variation in the initial UNS value (random intercepts) and subsequent change in UNS scores over time (random slopes), and within-subject correlation for repeated measurements of UNS scores (estimated at 0.16).

Model building for linear mixed effects analysis of change from baseline in Ullanlinna Narcolepsy Scale scores.

jcsm.13.3.441.t02.jpg

table icon
Table 2

Model building for linear mixed effects analysis of change from baseline in Ullanlinna Narcolepsy Scale scores.

(more ...)

Effect of treatment over time

Having established the best-fitting data structure model, the effects of treatment with IVIg versus standard care over time were assessed alone and with adjustment for baseline symptom scores and associated interactions (Table 3). The effect of IVIg versus standard care over time was not significant in either analysis. However, higher baseline UNS scores were strongly associated with both greater initial change in UNS scores and in change in UNS scores over time (both p < 0.001). Baseline and baseline-by-time interaction terms were therefore included in subsequent covariate adjusted models.

Linear mixed-effects analyses of change from baseline in Ullanlinna Narcolepsy Scale scores examining the effect of treatment and interactions with baseline symptom scores.

jcsm.13.3.441.t03.jpg

table icon
Table 3

Linear mixed-effects analyses of change from baseline in Ullanlinna Narcolepsy Scale scores examining the effect of treatment and interactions with baseline symptom scores.

(more ...)

Covariate adjusted and propensity score weighted analyses

The covariate adjusted analysis demonstrated similar findings to the analysis adjusted solely for baseline symptoms, notably only baseline UNS scores were associated with initial change and change over time in UNS scores (Table 4). Latency to sleep onset on MSLT could not be added to the covariate adjusted model because of the extent of missing data for this variable.

Covariate adjusted and propensity score weighted linear mixed effects analyses of change from baseline in Ullanlinna Narcolepsy Scale scores.

jcsm.13.3.441.t04.jpg

table icon
Table 4

Covariate adjusted and propensity score weighted linear mixed effects analyses of change from baseline in Ullanlinna Narcolepsy Scale scores.

(more ...)

In addition to age at referral, sex, and BMI z-score that were obligatorily entered into propensity score generation, variables identified as imbalanced were UNS baseline score, H1N1 status, interval between disease onset and referral, and latency to sleep onset on MSLT. Latency to sleep onset on MSLT could not be added to the logistic regression model used to generate the propensity scores owing to the extent of missing data. However, subsequent weighted regression using propensity score-derived weights confirmed that imbalance had been corrected for both the included variables (age at referral, sex, BMI z-score, UNS baseline score, H1N1 status, interval between disease onset and referral) and for latency to sleep onset on MSLT even though this latter variable had not been used to generate the propensity scores. Results of the propensity score weighted linear mixed-effects analysis confirmed the effect of IVIg on change in UNS scores over time was not statistically significant (Table 4).

In order to further explore the relationships between treatment group and baseline UNS, UNS adjusted mean changes from baseline by treatment group and by UNS baseline category (higher versus lower) were explored in an additional linear mixed-effects analysis. The median UNS score for the IVIg group of 25 was chosen as the cutoff point for both the IVIg and control groups to compare equivalent levels of baseline UNS scores across both treatment groups. Four treatment-by-baseline UNS category groups were created for this analysis: Control–Baseline UNS ≥ 25; Control–Baseline UNS < 25; IVIg–Baseline UNS ≥ 25; and IVIg–Baseline UNS < 25. Figure 3A shows the adjusted mean changes over time for the four groups. Estimated mean differences at 36 w were significant for higher versus lower baseline comparisons, regardless of treatment. Thus, adjusted mean differences at 36 w were: −10.3 (95% confidence interval [CI]: −17.6, −3.0; p = 0.003) for Control–Baseline UNS ≥ 25 vs the Control–Baseline UNS < 25; −9.7 (95% CI: −19.0, −0.5; p = 0.036) for Control–Baseline UNS ≥ 25 versus IVIg–Baseline UNS < 25; −13.5 (95% CI: −21.6, −5.3; p < 0.001) for IVIg–Baseline UNS ≥ 25 versus Control–Baseline UNS < 25; and −12.9 (95% CI: −21.9, −3.9; p = 0.002) for IVIg–Baseline UNS ≥ 25 versus IVIg– Baseline UNS < 25. In contrast, estimated mean differences at 36 w were not significant for treatment comparisons within the same level of baseline UNS. Thus, adjusted mean differences at 36 w were: −0.6 (95% CI: −8.5, 7.4; p = 0.998) for IVIg–Baseline UNS < 25 versus Control–Baseline UNS < 25; and −3.2 (95% CI: −12.6, 6.2; p = 0.802) for IVIg–Baseline UNS ≥ 25 vs Control–Baseline UNS ≥ 25.

Ullanlinna Narcolepsy Scale (UNS) adjusted mean changes from baseline and 95% confidence intervals.

(A) Categories of treatment by baseline UNS; and (B) categories of treatment by interval between disease onset and referral. The linear mixed-effects model in (A) was adjusted for age at referral, sex, body mass index z-score, interval between disease onset and referral, H1N1 vaccination, and standard drug therapy initiation. The model in (B) excluded the covariate interval between disease onset and referral. BL UNS = baseline Ullanlinna Narcolepsy Scale score, CI = confidence interval, IVIg = intravenous immunoglobulin.

jcsm.13.3.441c.jpg

jcsm.13.3.441c.jpg
Figure 3

Ullanlinna Narcolepsy Scale (UNS) adjusted mean changes from baseline and 95% confidence intervals.

(more ...)

Given that previous literature has suggested that patients with shorter durations of disease respond more favorably to IVIg,9 UNS adjusted mean changes from baseline by treatment group and by interval between disease onset and referral (longer versus shorter) were explored in an additional linear mixed-effects analysis. The median interval between disease onset and referral for the IVIg group of 3 w was chosen as the cutoff point for both the IVIg and control groups to compare equivalent durations of disease across both treatment groups. Four treatment-by-disease duration category groups were created for this analysis: Control–Interval ≥ 3 w; Control–Interval < 3 w; IVIg–Interval ≥ 3 w; and IVIg–Interval < 3 w. Figure 3B shows the adjusted mean changes over time for the four groups. Estimated mean differences at 36 w were not significant for higher versus lower disease duration comparisons within each treatment group or for control versus IVIg in patients with short disease duration.

Other outcomes

Similar findings in linear mixed-effects analyses, notably a significant effect of baseline value on initial change and change over time but without a significant effect of IVIg treatment over time, were observed for changes in PDSS, CASS, CGI-S, and CGI-C (data not shown).

Time-to-Event Analysis

A time to repeated events analysis by the four treatment-by-baseline UNS categories using Cox proportional hazards models was conducted as an additional analytic approach to explore potential treatment by baseline UNS score interactions and treatment by disease duration interactions. Events were classified in one of two ways: (1) achieving a UNS score < 14 (the cutoff score for significant pathology) to denote symptomatic recovery; or (2) achieving a ≥ 20% reduction in UNS scores from baseline to denote clinically meaningful improvement, but not necessarily to levels associated with symptomatic recovery. The cumulative hazard was plotted for each of four groups and adjusted hazard ratios (HRs) were computed from the Cox model to ascertain whether the cumulative hazard of achieving the outcome was significantly different between the four treatment-by-baseline UNS category groups or treatment-by-disease duration category groups (Table 5).

Multivariate Cox proportional hazard models assessing adjusted hazard ratios.

jcsm.13.3.441.t05.jpg

table icon
Table 5

Multivariate Cox proportional hazard models assessing adjusted hazard ratios.

(more ...)

Figure 4A shows the cumulative hazard of achieving a UNS score < 14 by the four treatment-by-baseline UNS category groups. Among control patients, those with high baseline UNS scores achieved a UNS score < 14 less rapidly than those with low baseline UNS scores (adjusted HR 0.39; 95% CI: 0.22, 0.69; p = 0.001). Among patients with high baseline UNS scores, control patients achieved a UNS score < 14 less rapidly than IVIg patients (adjusted HR 0.18; 95% CI: 95% CI: 0.03, 0.95; p = 0.043).

Cumulative hazard of achieving: (A) A Ullanlinna Narcolepsy Scale (UNS) score < 14 by categories of treatment by baseline UNS; (B) a ≥ 20% reduction from baseline in UNS score by categories of treatment by baseline UNS.

Hazard ratios (HR) were adjusted for age at referral, interval between disease onset and referral, sex, body mass index z-score, H1N1 status, and standard drug therapy initiation for all models. CI = confidence interval.

jcsm.13.3.441d.jpg

jcsm.13.3.441d.jpg
Figure 4

Cumulative hazard of achieving: (A) A Ullanlinna Narcolepsy Scale (UNS) score...

(more ...)

Figure 4B shows the cumulative hazard of achieving a ≥ 20% reduction from baseline in UNS score by the four treatment-by-baseline UNS category groups. Among control patients, those with low baseline UNS scores achieved a ≥ 20% reduction in UNS score less rapidly than those with high baseline scores (adjusted HR 0.53; 95% CI: 0.29, 0.96; p = 0.037). Similarly, among IVIg patients, those with low baseline UNS scores achieved a ≥ 20% reduction in UNS score less rapidly than those with high baseline scores (adjusted HR 0.39; 95% CI: 0.22, 0.69; p = 0.001). Among patients with high baseline UNS scores, although a visual trend was evident, IVIg patients did not achieve a ≥ 20% reduction in UNS score more rapidly than control patients.

Regarding time-to-event analysis by categories of treatment-by-disease duration, no differences between shorter and longer durations within treatment groups were observed (Figure 5A and 5B). In addition, in patients with a short interval between disease onset and referral, no differences between treatment groups were observed for either patients achieving a UNS score < 14 (adjusted HR 0.46; 95% CI: 0.20, 1.03; p = 0.059) or patients achieving a ≥ 20% reduction in UNS score (adjusted HR 0.53; 95% CI: 0.25, 1.12; p = 0.100).

Cumulative hazard of achieving: (A) a Ullanlinna Narcolepsy Scale (UNS) score < 14 by categories of treatment by interval between disease onset and referral; and (B) a ≥ 20% reduction from baseline in UNS score by categories of treatment by interval between disease onset and referral.

Hazard ratios (HR) were adjusted for age at referral, UNS baseline score, sex, body mass index z-score, H1N1 status, and standard drug therapy initiation for all models. CI = confidence interval.

jcsm.13.3.441e.jpg

jcsm.13.3.441e.jpg
Figure 5

Cumulative hazard of achieving: (A) a Ullanlinna Narcolepsy Scale (UNS) score...

(more ...)

Patients needing to initiate standard drug therapy during the follow-up period achieved a UNS score < 14 less rapidly than those continuing standard drug therapy (Tables 5A and 5C).

DISCUSSION

This report represents the most extensive evaluation to date of the treatment of narcolepsy in children with IVIg, both in terms of numbers of patients evaluated and the length of follow-up. Descriptive analyses and linear mixed-effects analyses of changes from baseline in symptom scores over time did not support a treatment effect with IVIg, rather changes in symptoms appeared to be driven by higher baseline symptoms scores within the IVIg group. These higher baseline scores were, in turn, likely driven by investigator choice to offer IVIg treatment to patients who were more severely affected (shorter mean latency to sleep onset on MLST) and who the investigator perceived might have had a greater likelihood of response (shorter interval between disease onset and referral, and disease onset after H1N1 vaccination). In further support of a lack of any specific treatment effect was the finding that mean symptom scores had already begun to fall prior to the administration of IVIg (Figures 1, 2, and 3), which may simply reflect the natural course of untreated narcolepsy over time (sudden onset of symptoms with partial improvement over time).38

However, these data should be interpreted with caution. First, the propensity score adjusted linear mixed-effects analysis demonstrated a treatment estimate for change from baseline in UNS scores that fell short of statistical significance by a small margin (Table 4). Second, the model including all baseline-by-treatment-by-time interactions (i.e. two- and three-way interactions) demonstrated an estimate for the interaction term IVIg-by-baseline UNS-by-time that fell short of statistical significance by a small margin (Table 3), which might suggest a possible signal that treatment with IVIg was only influential for patients with higher baseline UNS scores (this latter hypothesis, however, was not supported by the linear mixed-effects analysis of treatment by baseline UNS categories; see Figure 3A). Third, in contrast to the linear mixed-effects analyses, the time-to-event analysis for patients achieving symptomatic recovery (UNS score < 14) did suggest an interaction between high baseline UNS scores and treatment as demonstrated by the significant IVIg–Baseline UNS ≥ 25 versus Control–Baseline UNS ≥ 25 adjusted HR (Figure 4B). In addition, symptomatic recovery did not begin to occur in the IVIg group until the precise point that IVIg was administered (Figure 4A). Given that selection bias is likely to have led to the investigator allocating more severely affected children for inclusion in the IVIg group, it is tempting to speculate that these children might have fared worse had they not received IVIg. Of note, those treated with IVIg had a much higher exposure to H1N1 vaccination (15 of 22) and a potentially different pathogenesis and disease course39 compared with the control group (2 of 30), which largely comprised patients with sporadic onset. Differing vaccine coadjuvants may modulate immunological response, as illustrated by cross-reactivity between H1N1 nucleoprotein serum antibodies and the hypocretin receptor 2 following Pandremix, but not Focetria vaccination. Thus, high exposure to H1N1 vaccination may have explained earlier symptomatic recovery in some patients receiving IVIg. However, a single case report of early IVIg administration in narcolepsy post H1N1 vaccination did not show benefit.25 Moreover, for the outcome of a ≥ 20% reduction in UNS scores, this appeared to be driven solely by differences in baseline UNS scores rather than by treatment, as evidenced by the non-significant IVIg–Baseline UNS ≥ 25 versus Control–Baseline UNS ≥ 25 adjusted HR, and the fact that patients in the IVIg group had begun to achieve a ≥ 20% reduction in UNS scores prior to administration of IVIg (Figure 4B).

In contrast to previous reports suggesting a greater response to IVIg in patients with a shorter duration of disease, our results did not support this conclusion (Figure 3B, Tables 5C and 5D, and Figures 4C and 4D). It is possible that in the initial stages of disease presentation, symptom burden is highest (high baseline values), which may better explain response than a shorter disease duration per se. The observation that patients needing to initiate standard drug therapy during the follow-up period achieved a UNS score < 14 less rapidly than those continuing standard drug therapy from the outset may reflect the fact that these patients were deteriorating over time having not required drug therapy at the time of referral.

This analysis has major limitations. Inherent within observational studies that are not subject to randomization and blinded allocation to investigational therapies, severe selection bias was evident, making an unbiased assessment of IVIg effect very difficult. Although we attempted to adjust for bias using modern statistical methodologies, including propensity score weighted analysis, it was not possible to fully establish whether an independent effect of treatment existed or whether the observed changes were driven by greater disease severity at baseline or simply reflected the natural course of narcolepsy.38 Although this is the largest reported series to date, the sample size was still very small for adequate evaluation of an investigational product. The extent of missing data was too great to enable inclusion of latency to sleep onset on MSLT into the covariate adjusted or propensity score weighted analyses. However, propensity scores generated using the other variables identified as imbalanced appeared to correct the imbalance observed with latency to sleep onset on MSLT. It is unlikely, therefore, that if data had been more extensive for this variable, the conclusions of our analysis would have changed. Finally, adverse event data were not systematically collected in this investigation, so an assessment of possible harms of IVIg treatment could not be undertaken.

In conclusion, narcolepsy symptoms were not significantly reduced in children with type 1 narcolepsy treated with IVIg plus standard care compared with control children receiving standard care alone. However, in patients with high baseline symptoms, a subset of IVIg-treated patients achieved remission more rapidly than controls. Shorter versus longer intervals between disease onset and referral did not influence IVIg treatment response in any analysis. A substantive randomized, placebo controlled trial is required to establish the efficacy, safety, and tolerability of IVIg in pediatric type 1 narcolepsy that has sufficient power to distinguish treatment responses between H1N1 vaccine-associated narcolepsy versus sporadic narcolepsy and to define the characteristics of patients most likely to benefit from IVIg treatment.

DISCLOSURE STATEMENT

This was not an industry supported study. M Lecendreux has served as a consultant for Bioprojet, Shire, UCB, and Jazz Pharma and has received research support from Shire and UCB and honoraria from UCB and Shire. Other authors have indicated no financial conflicts of interest. Institutions where work performed: Pediatric Sleep Center, Hospital Robert-Debré, APHP, Paris, France. Investigational use: This observational study in a clinical setting describes off-label use of intravenous immunoglobulin in pediatric narcolepsy.

ABBREVIATIONS

BL

baseline

BMI

body mass index

CASS

Child and Adolescent Sleepiness Scale

CGI

Clinical Global Impression

CGI-C

Clinical Global Impression – Cataplexy

CGI-S

Clinical Global Impression – Sleepiness

CI

confidence interval

CSF

cerebrospinal fluid

EDS

excessive daytime sleepiness

GLM

generalized linear models

HLA

human leucocyte antigen

HR

hazard ratios

ICSD

International Classification of Sleep Disorders

IVIg

intravenous immunoglobulins

LLR

log-likelihood ratio

LME

linear mixed effects

MSLT

Multiple Sleep Latency Tests

PDSS

Pediatric Daytime Sleepiness Scale

SD

standard deviation

TRIB2

Tribbles homolog 2

UNS

Ullanlinna Narcolepsy Scale

REFERENCES

1 

American Academy of Sleep Medicine. International Classification of Sleep Disorders. 2nd ed. Westchester, IL: American Academy of Sleep Medicine; 2005.

2 

American Academy of Sleep Medicine. International Classification of Sleep Disorders. 3rd ed. Darien, IL: American Academy of Sleep Medicine; 2014.

3 

Inocente CO, Lavault S, Lecendreux M, et al. Impact of obesity in children with narcolepsy. CNS Neurosci Ther. 2013;19(7):521–528. [PubMed]

4 

Inocente CO, Gustin M-P, Lavault S, et al. Depressive feelings in children with narcolepsy. Sleep Med. 2014;15(3):309–314. [PubMed]

5 

Lecendreux M, Lavault S, Lopez R, et al. Attention-deficit/hyperactivity disorder (ADHD) symptoms in pediatric narcolepsy: a cross-sectional study. Sleep. 2015;38(8):1285–1295. [PubMed Central][PubMed]

6 

Inocente CO, Gustin MP, Lavault S, et al. Quality of life in children with narcolepsy. CNS Neurosci Ther. 2014;20(8):763–771. [PubMed]

7 

Hayes D Jr. Narcolepsy with cataplexy in early childhood. Clin Pediatr (Phila). 2006;45(4):361–363.

8 

Stores G, Montgomery P, Wiggs L. The psychosocial problems of children with narcolepsy and those with excessive daytime sleepiness of uncertain origin. Pediatrics. 2006;118(4):e1116–e1123. [PubMed]

9 

Wozniak DR, Quinnell TG. Unmet needs of patients with narcolepsy: perspectives on emerging treatment options. Nat Sci Sleep. 2015;7:51–61. [PubMed Central][PubMed]

10 

Cvetkovic-Lopes V, Bayer L, Dorsaz S, et al. Elevated Tribbles homolog 2-specific antibody levels in narcolepsy patients. J Clin Invest. 2010;120(3):713–719. [PubMed Central][PubMed]

11 

Kawashima M, Lin L, Tanaka S, et al. Anti-Tribbles homolog 2 (TRIB2) autoantibodies in narcolepsy are associated with recent onset of cataplexy. Sleep. 2010;33(7):869–874. [PubMed Central][PubMed]

12 

Toyoda H, Tanaka S, Miyagawa T, Honda Y, Tokunaga K, Honda M. Anti-Tribbles homolog 2 autoantibodies in Japanese patients with narcolepsy. Sleep. 2010;33(7):875–878. [PubMed Central][PubMed]

13 

Dauvilliers Y, Arnulf I, Lecendreux M, et al. Increased risk of narcolepsy in children and adults after pandemic H1N1 vaccination in France. Brain. 2013;136(Pt 8):2486–2496. [PubMed]

14 

Partinen M, Saarenpaa-Heikkila O, Ilveskoski I, et al. Increased incidence and clinical picture of childhood narcolepsy following the 2009 H1N1 pandemic vaccination campaign in Finland. PLoS One. 2012;7(3):e33723. [PubMed Central][PubMed]

15 

Han F, Lin L, Warby SC, et al. Narcolepsy onset is seasonal and increased following the 2009 H1N1 pandemic in China. Ann Neurol. 2011;70(3):410–417. [PubMed]

16 

Heier MS, Gautvik KM, Wannag E, et al. Incidence of narcolepsy in Norwegian children and adolescents after vaccination against H1N1 influenza A. Sleep Med. 2013;14(9):867–871. [PubMed]

17 

Ahmed SS, Volkmuth W, Duca J, et al. Antibodies to influenza nucleoprotein cross-react with human hypocretin receptor 2. Sci Transl Med. 2015;7(294):294ra105. [PubMed]

18 

Dauvilliers Y. Follow-up of four narcolepsy patients treated with intravenous immunoglobulins. Ann Neurol. 2006;60(1):153. [PubMed]

19 

Dauvilliers Y, Carlander B, Rivier F, Touchon J, Tafti M. Successful management of cataplexy with intravenous immunoglobulins at narcolepsy onset. Ann Neurol. 2004;56(6):905–908. [PubMed]

20 

Lecendreux M, Maret S, Bassetti C, Mouren MC, Tafti M. Clinical efficacy of high-dose intravenous immunoglobulins near the onset of narcolepsy in a 10-year-old boy. J Sleep Res. 2003;12(4):347–348. [PubMed]

21 

Plazzi G, Poli F, Franceschini C, et al. Intravenous high-dose immunoglobulin treatment in recent onset childhood narcolepsy with cataplexy. J Neurol. 2008;255(10):1549–1554. [PubMed]

22 

Valko PO, Khatami R, Baumann CR, Bassetti CL. No persistent effect of intravenous immunoglobulins in patients with narcolepsy with cataplexy. J Neurol. 2008;255(12):1900–1903. [PubMed]

23 

Zuberi SM, Mignot E, Ling L, McArthur I. Variable response to intravenous immunoglobulin therapy in childhood narcolepsy [Abstracts of the ESRS, Prague 2004]. J Sleep Res. 2014;13 Suppl 1:828.

24 

Chen W, Black J, Call P, Mignot E. Late-onset narcolepsy presenting as rapidly progressing muscle weakness: response to plasmapheresis. Ann Neurol. 2005;58(3):489–490. [PubMed]

25 

Knudsen S, Biering-Sorensen B, Kornum BR, et al. Early IVIg treatment has no effect on post-H1N1 narcolepsy phenotype or hypocretin deficiency. Neurology. 2012;79(1):102–103. [PubMed]

26 

Donjacour CE, Lammers GJ. A remarkable effect of alemtuzumab in a patient suffering from narcolepsy with cataplexy. J Sleep Res. 2011;21(4):479–480. [PubMed]

27 

Hersh CM, Cohen JA. Alemtuzumab for the treatment of relapsing-remitting multiple sclerosis. Immunotherapy. 2014;6(3):249–259. [PubMed]

28 

Hughes RA, Swan AV, van Doorn PA. Intravenous immunoglobulin for Guillain-Barré syndrome. Cochrane Database Syst Rev. 2014;9:CD002063.

29 

Oates-Whitehead RM, Baumer JH, Haines L, et al. Intravenous immunoglobulin for the treatment of Kawasaki disease in children. Cochrane Database Syst Rev. 2003;(4):CD004000. [PubMed]

30 

Barron SJ, Del Vecchio MT, Aronoff SC. Intravenous immunoglobulin in the treatment of Stevens-Johnson syndrome and toxic epidermal necrolysis: a meta-analysis with meta-regression of observational studies. Int J Dermatol. 2014;54(1):108–115. [PubMed]

31 

Cole TJ, Bellizzi MC, Flegal KM, Dietz WH. Establishing a standard definition for child overweight and obesity worldwide: international survey. BMJ. 2000;320(7244):1240–1243. [PubMed Central][PubMed]

32 

Rolland-Cachera MF. Childhood obesity: current definitions and recommendations for their use. Int J Pediatr Obes. 2011;6(5-6):325–331. [PubMed]

33 

Hublin C, Kaprio J, Partinen M, Koskenvuo M, Heikkila K. The Ullanlinna narcolepsy scale: validation of a measure of symptoms in the narcoleptic syndrome. J Sleep Res. 1994;3(1):52–59. [PubMed]

34 

Drake C, Nickel C, Burduvali E, Roth T, Jefferson C, Pietro B. The pediatric daytime sleepiness scale (PDSS): sleep habits and school outcomes in middle-school children. Sleep. 2003;26(4):455–458. [PubMed]

35 

Busner J, Targum SD. The clinical global impressions scale: applying a research tool in clinical practice. Psychiatry (Edgmont). 2007;4(7):28–37.

36 

Olmos A, Govindasamy P. A practical guide for using propensity score weighting in R. Practical Assessment, Research & Evaluation. 2015;20(13):1–8.

37 

Guo Z, Gill TM, Allore HG. Modeling repeated time-to-event health conditions with discontinuous risk intervals. An example of a longitudinal study of functional disability among older persons. Methods Inf Med. 2008;47(2):107–116. [PubMed Central][PubMed]

38 

Pizza F, Franceschini C, Peltola H, et al. Clinical and polysomnographic course of childhood narcolepsy with cataplexy. Brain. 2013;136(Pt 12):3787–3795. [PubMed]

39 

Dauvilliers Y, Montplaisir J, Cochen V, et al. Post-H1N1 narcolepsy-cataplexy. Sleep. 2010;33(11):1428–1430. [PubMed Central][PubMed]