Hearing and facial function outcomes for neurofibromatosis 2 clinical trials

Feasibility.

Audiology is widely available and it takes about 20 minutes to perform testing that includes a 50-word recognition list. Standardization of testing is critical for use in clinical trials. For this reason, the group recommends use of a recorded list of standardized monosyllables.

Patient characteristics.

Standard audiometry can be administered reliably to individuals aged 5 years and older.¹⁹ Audiometry is valid for important subgroups such as the elderly and people with intellectual disability.²⁰ Since the results are analyzed as repeated measures, a participant may be evaluated using English materials if the audiologist decides that the results reasonably reflect comparative (test-to-test) performance. A recommended approach may be to require that participants be “testable in English” as opposed to the narrower criterion of English as a first language.⁸ There are also valid tests that can be reported using the same mathematical criteria available in several languages.²¹

Use in published studies.

The 95% critical difference has been used as a primary endpoint or as a secondary endpoint in studies of sudden sensorineural hearing loss¹⁵ and in studies of bevacizumab and lapatinib for NF2-related VS.^22,–,24

Baseline documentation of “target” and “nontarget” lesions.

When 2 functional ears are present at baseline, the target ear and nontarget ear should be identified and recorded at baseline. The goal is to target an ear that lends itself to reproducible repeated measures. Target ears should be selected on the basis of the quality of the hearing and the presence of an ipsilateral VS. Hearing loss in the target ear should be attributed to VS (rather than other causes of hearing loss).

The maximum WRS should be recorded for all ears. To address rollover, an empirical search using at least 2 levels (“high” and “low”) should be used to identify maximum WRS at each evaluation. Speech intelligibility should be evaluated for each ear using standard word recognition of monosyllables.²⁵ All tests should use recorded lists (e.g., CID-W22, Ira Hirsh recording, Q/MASS v 2) from a compact disc²⁶ rather than spoken word lists since this results in less variability of results. The initial level should be set where maximum performance is expected for each ear. The “high” level will be the level calculated for maximum audibility, as in the standard method for level setting in nontumor cases. The “low” level will be the maximum level minus 10–15 dB as chosen by the audiologist to reflect a level with less chance of rollover. The maximum WRS is defined as the greater WRS determined at the “high” and “low” levels.

Response criteria.

Hearing response criteria are defined in reference to the baseline WRS at study initiation (table 2). Hearing response (HR) is defined as an improvement in WRS above the 95% critical difference. Stable hearing (SH) is defined as persistence of WRS within the 95% critical difference. Hearing decline is defined as a decrease in WRS below the 95% critical difference.

Special notes on assessment of hearing.

WRS is not a continuous variable: it is a sum of binary responses (correct/incorrect) that ranges from 0% to 100%. The existence of upper and lower boundaries introduces a “ceiling” and “floor” effect for patients with hearing near the top and bottom of the range, respectively. As table 2 shows, patients with WRS ≤ 6% are not eligible for hearing decline since the 95% critical difference includes 0%; similarly, patients with WRS ≥94% are not eligible for hearing improvement since the 95% critical difference includes 100%. In the NF2 Natural History study, the rate of spontaneous increase in WRS was 16% in patients with baseline WRS ≥ 90%.²⁷ For trials using hearing response as the primary outcome, the functional group recommends excluding patients with WRS ≥ 90% to allow for appreciation of hearing improvement and to ensure that test-to-test variation does not obscure true changes in hearing function.

Frequency of reevaluation.

No evidence-based guidelines are available to help determine the interval between evaluations. The functional group recommends reevaluation of maximum WRS every 3 months during initial treatment and every 3–6 months during prolonged treatment.

Confirmatory measurement/duration of response.

In nonrandomized trials where response is the primary endpoint, confirmation of response is desirable. The time between audiograms should be at least 1 month. Responses that are not confirmed by subsequent audiology should be termed “unconfirmed response.” For randomized trials or for trials where time to hearing decline is the primary endpoint, confirmation of response is not required. To declare stable disease, audiograms must show stable disease at least once after study initiation for a period of time that is defined by the study protocol (usually 6 months).

Duration of hearing response.

The duration of HR is measured from the time when study measurement first shows a HR until the first date that maximum WRS is recorded within the 95% critical difference of the baseline WRS.

Duration of stable hearing.

SH is measured from the start of treatment until the criteria for hearing decline are met, taking as reference the 95% critical difference of the baseline maximum WRS. For protocols designed to study the proportion of patients with SH for a minimum period of time, the protocol should specify the interval between audiograms.

It is important to note that the duration of HR/SH and the determination of time to hearing decline is influenced by the frequency of follow-up after baseline evaluation. These intervals should be defined in the study protocol.

Proportion free from hearing loss.

Historically, most patients with NF2 experience profound bilateral hearing loss during the course of life. For young patients and those with good hearing, a therapy that could prevent hearing loss would be desirable, even if this therapy did not improve hearing. In this scenario, the proportion of patients free from hearing loss at landmark time points might be considered appropriate to provide an initial sign of biologic activity. However, in uncontrolled phase II trials this endpoint is subject to bias from known and unknown factors that influence hearing loss. Ideally, this endpoint is best evaluated in the setting of a randomized trial. In the case of NF2, spontaneous hearing improvement is sufficiently uncommon (about 16% for patients with baseline WRS < 90%) that a nonrandomized trial may be justifiable.²⁷

Reporting best response results.

For phase II trials, all patients in the study must be included in the report of the results, even those who are not evaluable or experience major protocol deviations. Patients should be designated as either HR, SH, hearing decline, or not evaluable (e.g., early death, toxicity, assessment not performed). In general, the denominator for calculation of the response rate should include all eligible patients. Conclusions should not be based on selected “evaluable” patients. For phase III trials, hearing response evaluation may be a primary or secondary endpoint. If hearing response evaluation is a primary endpoint (i.e., the study includes only patients who are capable of hearing improvement), the discussion above should apply. If response evaluation is a secondary endpoint (i.e., the study includes all patients regardless of hearing), the hearing response rate may be reported using an “intention-to-treat” analysis (with all randomized patients in the denominator) or using a predefined subset analysis (including the subset of patients with hearing loss at baseline). The study protocol should specify how response results will be reported, including any planned subset analyses.

Feasibility.

SMILE analysis is performed using frontal photographs of the patient at rest and with a full smile. Photos are imported into the software program (Face-gram), which then calculates an objective analysis of smile symmetry and change in commissure excursion. The Java-based SMILE analysis program is free for use and can be downloaded at www.sircharlesbell.org/facial_nerve_programs.html. Complete facial analysis takes less than 5 minutes and may be stored electronically for future use.

Patient characteristics and validity.

SMILE analysis can be reliably performed in individuals capable of standard facial poses.¹⁷ The lower age limit for reliability of measurement has not been determined. In a previous study of 10 normal individuals (20 hemi-smiles), there was a strong correlation between measured distances and the true excursion (class correlation coefficients R = 0.96–0.99). There was strong correlation between the measurements of excursion distances for multiple raters (intraclass correlation coefficients: R = 0.88–0.98). The intratest reliability after 1 week (for a given rater) was 0.99, indicating a strong correlation between measurements over time.

Use in published studies.

SMILE analysis has been used as an endpoint in studies of facial reanimation due to many causes, including NF2.^31,32 To date, the technique has not been used in multicenter studies.

Baseline documentation of “target” and “nontarget” hemi-smiles.

When unilateral facial weakness is present at baseline, the target hemi-smile should be identified and recorded at baseline. In this instance, the contralateral (nontarget) hemi-smile acts as a normal control for the affected side. A history of previous surgery that might affect facial strength (e.g., resection of VS) should be recorded. When bifacial weakness is present at baseline, a target hemi-smile should be identified and recorded at baseline, based on normative smile parameters.³³

Patients should be photographed using at least 2 facial poses, including “at rest” and with a broad smile. Using the software, the true smile excursion from midline and the smile angle from horizontal midline should be recorded for target and nontarget hemi-smiles (figure). Comparison between smile angles and smile excursion on the normal side and the affected side at baseline provides an objective analysis of smile symmetry. Comparison between smile angles and smile excursion of the target hemi-smile before and after intervention provides an objective analysis of change in commissure excursion with treatment.

Response criteria.

Consensus criteria using SMILE analysis (or other facial rating scales) have not been defined to date. Until more data are acquired, the functional committee proposes to define a response as an increase in true smile excursion of 25% compared with the true smile excursion at baseline. Progression is defined as a decrease in true smile excursion of 25% compared with the baseline true smile excursion. Stable function is defined as all other changes in true smile excursion compared with baseline. These response criteria apply to patients with both unilateral facial weakness and bifacial weakness.

Special notes on assessment of facial function.

Patients with facial weakness also experience paralytic lagophthalmos, which can increase risk of complications such as keratopathy. SMILE analysis of facial photos is capable of measuring corneal exposure before and after intervention. Measuring changes in lagophthalmos should be studied with a goal of defining response criteria in the future. Presently, the group recommends a 2-mm increase in lid position as a functional response and a 2-mm decrease in lid position as functional decline.

Frequency of reevaluation.

No evidence-based guidelines are available to determine an appropriate interval between evaluations. The functional group recommends reevaluation of true smile excursion every 3 months during the first year after surgical intervention and yearly afterwards.

Confirmatory measurement/duration of response.

In nonrandomized trials where response is the primary endpoint, confirmation of response is desirable. The minimum time between facial analyses should be at least 1 month. Responses not confirmed by subsequent facial photos are termed “unconfirmed response.” For randomized trials or for trials where time to facial weakness is the primary endpoint, confirmation of response is not required. To declare stable facial strength, facial photos must show stable strength at least once after intervention at a time that is defined by the study protocol.

Duration of facial response.

The duration of facial response is measured from the time when study measurement first shows a facial response until the first date that facial photos demonstrate progressive facial weakness compared to the best change in true smile excursion.

Duration of stable facial function.

Stable facial function is measured from the start of treatment or intervention until the criteria for progressive facial weakness are met, taking as reference the true smile excursion at baseline. For protocols designed to study the proportion of patients with stable facial function for a minimum period of time, the protocol should specify the interval between facial photos.

The duration of smiling response/stable facial function and the determination of time to progressive facial weakness are influenced by the frequency of follow-up after baseline evaluation. These intervals should be defined in the study protocol.

Proportion free from progressive facial weakness.

Historically, patients with NF2 experience some facial weakness over their lifetime. For patients with normal facial function, a novel procedure or technique (e.g., surgery or radiation) that minimizes, delays, or otherwise prevents facial weakness would be desirable. In this scenario, the proportion of patients free from progressive facial weakness at landmark time points could be an appropriate measure of successful intervention. However, in uncontrolled phase II trials this endpoint is subject to bias from known and unknown factors that influence facial function. Ideally, this endpoint is best evaluated in the setting of a randomized trial.

Reporting best response results.

For phase II trials, all patients in the study must be included in results reporting, even those who are not evaluable or experience major protocol deviations. Patients should be designated as either facial response, stable facial function, progressive facial weakness, or not evaluable (e.g., early death, toxicity, assessment not performed). In general, the denominator for calculation of the response rate should include all eligible patients. Conclusions should not be based on selected “evaluable” patients. For phase III trials, response evaluation may be a primary or secondary endpoint. If response evaluation is a primary endpoint (i.e., the study includes only patients who have baseline facial weakness), the discussion above should apply. If response evaluation is a secondary endpoint (i.e., the study includes all patients regardless of facial function), the response rate may be reported using an “intention-to-treat” analysis (with all randomized patients in the denominator) or using a predefined subset analysis (including the subset of patients with facial weakness at baseline). The study protocol should specify how response results will be reported, including any planned subset analyses.