Visual Pathway Outcomes Based on Sample Size in Clinical Neuroprotection Research

Abstract

This cohort study calculates clinical trial sample sizes powered by visual pathway outcomes of acute optic neuritis in neuroprotection research.

Acute optic neuritis (ON) is a model for neuroprotection in demyelinating disease. While most clinical trials have been powered for structural changes,^1,2,3 regulatory approval of new agents depends on functional preservation. Sample size calculations based on subjective and electrophysiologic outcomes are needed.

Methods

This cohort study used data from the Treatment of Optic Neuritis With Erythropoietin (TONE) randomized, placebo-controlled trial (NCT01962571). The ethics commission of the University of Freiburg approved this study. The study followed the STROBE reporting guideline.

We calculated sample sizes for double-blind, placebo-controlled, parallel-arm trials based on a 2-sided t test (α = .05; 80% or 90% power) with effect estimates ranging from 30% to 60% of the mean disease-induced change using R, version 4.3.2 (R Foundation for Statistical Computing) (eMethods in Supplement 1). For the TONE trial, 108 patients with a first episode of unilateral ON and baseline high-contrast visual acuity (HCVA) below 0.5 decimal (0.3 logMAR) were recruited from 12 German academic research centers between November 2014 and October 2017. As the trial medication (erythropoietin) had no effect on the visual system,⁴ we used treatment-agnostic data (eTable in Supplement 1). Data analysis was performed between April 1 and September 9, 2022.

Results

A total of 103 TONE participants (median [IQR] age, 30 [25-36] years; 71 women [69%]; 32 men [31%]) were included in this analysis. A clinical trial with 80% power, 2-sided α = .05, and treatment effect of 50% of disease-induced change would detect a significant difference in the full-field, pattern-reversal, visual-evoked potential (VEP) P100 peak time by including 113 participants per group. Subjective benefits could be demonstrated via the 2.5% low-contrast letter acuity (LCLA) (168 participants per group) or its intereye delta (IED) (109 participants per group). Detecting a similar effect for high-contrast vision would require 470 participants per group for HCVA or 314 per group for its IED. Structural changes would be detectable by optical coherence tomography (OCT) measuring the mean thickness of the macular ganglion cell and inner plexiform layer (35 participants per group) or peripapillary retinal nerve fiber layer (64 participants per group) (Table).

Table. Sample Size Calculations.

Outcome at 6 mo	Δmax, Mean (SD)	Sample size estimation per treatment group by estimated % of Δmax
		30%	40%	50%	60%
LCLA, 2.5% Sloan chart score
80% power	13.3 (21.6)	463	261	168	117
90% power		620	349	224	156
LCLA-IED, 2.5% Sloan chart score
80% power	16.4 (21.4)	299	169	109	76
90% power		400	225	145	101
HCVA, ETDRS chart score
80% power	5.2 (14.2)	1303	734	470	327
90% power		1744	982	629	437
HCVA-IED, ETDRS chart score
80% power	6.0 (13.5)	870	490	314	219
90% power		1165	656	420	292
VEP P100 peak time, ms
80% power	15.2 (20.3)	310	175	113	79
90% power		415	234	150	105
pRNFL thickness change, μm a
80% power	15.3 (15.2)	175	99	64	45
90% power		233	132	85	59
mGCIPL thickness change, μm b
80% power	8.8 (6.4)	93	53	35	25
90% power		125	71	46	32

Abbreviations: Δmax, disease-induced change; IED, intereye delta; LCLA, low-contrast letter acuity; mGCIPL, macular ganglion cell and inner plexiform layer; pRNFL, peripapillary retinal nerve fiber layer; VEP, visual-evoked potentials.

^a Compared with the unaffected eye at baseline.

^b Compared with the affected eye at baseline.

Discussion

Detecting functional neuroprotection in ON will be essential to achieve regulatory approval for new treatments. We found that VEP P100 peak times and LCLA-IED would be the most size-efficient outcomes, requiring a similar number of participants. Still, enrollment must be 3 times as high as trials measuring macular ganglion cell and inner plexiform layer thickness, the most sample size–efficient structural outcome by OCT.²

The LCLA-IED has been proposed as an alternative to single-eye measurements in visual pathway neuroprotection.^2,5 To date, IEDs have not been used as a primary outcome in ON clinical trials, and we are not aware of previously published sample size estimates. An important finding is that using IEDs diminished size requirements by one-third vs single-eye LCLA and HCVA measurements. The IEDs cancel out learning effects and day-to-day fluctuations but cannot capture drug effects that are conferred equally to both eyes. Additionally, recruitment may be hampered when trials must impose inclusion criteria on the unaffected eye.

Several trials have used VEP conduction latency as their primary outcome, typically enrolling fewer participants than our calculations would suggest.³ This discrepancy could be due to our assumption that the unaffected eye’s value would be a lower bound for drug efficacy, ie, trial agents would mitigate acute demyelination but not affect previously incurred damage. Drugs that also target preexisting demyelination may reduce conduction delay further and therefore require fewer trial participants.⁶

A limitation of our work is that a small number of patients with anti–myelin oligodendrocyte glycoprotein (MOG) positivity might have been included,⁴ as MOG testing was not performed for all patients. Anti-MOG positivity could cause overestimation of sample sizes due to better recovery. Specific estimates are based on our own test-retest reliability. Modestly different sample sizes might be achieved with different test-retest reliability or other baselining modifications (eg, OCT or VEP accuracy, patient factors in subjective visual tests).

Supplement 1.

eMethods

eTable. Pooled Patient and Baseline Characteristics of the TONE Study Population

Supplement 2.

Data Sharing Statement