Outcomes of Clinical Trial: Tinnitus Masking versus ...€¦ · efficacy of tinnitus masking (TM)...

J Am Acad Audiol 17:104–132 (2006)

104

*VA RR&D National Center for Rehabilitative Auditory Research, Portland VA Medical Center, Portland, Oregon;†Department of Otolaryngology, Oregon Health and Science University, Portland, Oregon; ‡VA Audiology Clinic, Portland VAMedical Center, Portland, Oregon; §Tinnitus and Hyperacusis Center, Department of Otolaryngology, Emory University, Atlanta,Georgia; **School of Nursing, Oregon Health and Science University, Portland, Oregon

James A. Henry, Ph.D., VA Medical Center (NCRAR), P.O. Box 1034, Portland, Oregon 97207; Phone: 503-220-8262, ext 57466; Fax: 503-402-2955; E-mail: [email protected]

Outcomes of Clinical Trial: Tinnitus Maskingversus Tinnitus Retraining Therapy

James A. Henry*†Martin A. Schechter*‡Tara L. Zaugg*Susan Griest*†Pawel J. Jastreboff§Jack A. Vernon†Christine Kaelin*Mary B. Meikle†Karen S. Lyons**Barbara J. Stewart**

Abstract

A controlled clinical study was conducted to evaluate prospectively the clinicalefficacy of tinnitus masking (TM) and tinnitus retraining therapy (TRT) inmilitary veterans having clinically significant tinnitus. Qualifying patients wereplaced into the two groups in an alternating manner (to avoid selection bias),and treatment was administered at 0, 3, 6, 12, and 18 months. Outcomes oftreatment were evaluated using three self-administered tinnitus questionnaires(Tinnitus Handicap Inventory, Tinnitus Handicap Questionnaire, TinnitusSeverity Index) and the verbally administered TRT interview forms. Findingsare presented from the three written questionnaires, and from two of theinterview questions (percentage time aware of, and annoyed by, tinnitus).Outcomes were analyzed on an intent-to-treat basis, using a multilevel modelingapproach. Of the 123 patients enrolled, 118 were included in the analysis. Bothgroups showed significant declines (improvements) on these measures, withthe TRT decline being significantly greater than for TM. The greater declinesin TRT compared to TM occurred most strongly in patients who began treatmentwith a “very big” tinnitus problem. When patients began treatment with a“moderate” tinnitus problem, the benefits of TRT compared to TM were moremodest.

Key Words: Clinical trial, hearing disorders, military veterans, rehabilitation,tinnitus

Abbreviations: GHI = General Hearing Instruments; HLM = HierarchicalLinear Modeling; MLM = multilevel modeling; MML = minimum masking level;PVAMC = Portland Veterans Affairs Medical Center; THI = Tinnitus HandicapInventory; THQ = Tinnitus Handicap Questionnaire; TISI = Tinnitus-ImpactScreening Interview; TM = tinnitus masking; TRT = tinnitus retraining therapy;VA = Veterans Affairs; VHA = Veterans Health Administration

Sumario

Se condujo un estudio clínico controlado para evaluar prospectivamente laeficacia clínica del enmascaramiento del acúfeno (TM) y de la terapia de re-entrenamiento para acúfenos (TRT) en militares veteranos con acúfenos

TRT vs. Masking Clinical Trial/Henry et al

105

Subjective tinnitus is the sensation ofhearing a sound that exists only insidethe head. Persistent sensorineural

tinnitus is experienced by millions of peoplein this country and is symptomatic of auditorysystem pathology. Exposure to loud noise cancause tinnitus and hearing loss (Axelssonand Barrenas, 1992; Penner and Bilger, 1995).Noise exposure is common in the military;thus, tinnitus is a particular concern forveterans. Tinnitus has become increasinglyproblematic both for veterans and for theVeterans Health Administration (VHA).Statistics provided by Veterans Affairs (VA)Office of Policy, Planning and Preparednessreveal that over 289,000 veterans withservice-connected tinnitus received a totalof over $345 million in 2004 for their tinnitusdisability compensation (Henry et al, 2004).These amounts represent a one-year increaseof $60 million for over 46,000 veterans whoreceived a new tinnitus service connection. In2001 tinnitus was the most common new

disability for veterans (Veterans BenefitsAdministration Annual Report, 2001).

Direct effects of problematic tinnitusinclude cognitive, emotional, and sleepdisorders—all of which can impactperformance of everyday activities(Erlandsson et al, 1992;Tyler, 1993;Axelsson,1998).Veterans with tinnitus may report anyor all of these disabling conditions, whichvary considerably in severity. Veterans withmore severe effects require structured,individualized treatment, which is notpresently available at most VA medicalcenters (VAMCs) (Henry et al, 2004). Becauseof the VHA’s emphasis on evidence-basedclinical practice (e.g., VHA Directive 2002-039), any new implementation of treatmentfor tinnitus will require compelling researchevidence. Dobie (1999, 2002, 2004) reviewedthe randomized clinical trials that haveassessed the efficacy of various treatments fortinnitus. He concluded that none of thesestudies showed a reduction or elimination of

clínicamente significativos. Los pacientes fueron colocados en los dos gruposen una forma alternante (para evitar un sesgo de selección) y el tratamientode administró a los 0, 3, 6, 12 y 18 meses. Los resultados del tratamiento seevaluaron utilizando tres cuestionarios auto-administrados sobre el acúfeno(el Inventario de Impedimento con Acúfeno, el Cuestionario de Impedimentocon Acúfeno, y el Indice de Severidad del Acúfeno) y las fórmulas de entrevistade TRT administradas verbalmente. Se presentan los hallazgos de los trescuestionarios escritos y de dos de las preguntas de la entrevista (porcentajede tiempo conciente del acúfeno y porcentaje de tiempo de molestia por elacúfeno). Los resultados fueron analizados de acuerdo al intento de tratar, conun enfoque de modelaje multinivel.De los 123 pacientes involucrados, 118 fueronincluidos en el análisis. Ambos grupos mostraron una disminución significativa(mejorías) en estas mediciones, con una declinación para el TRTsignificativamente mayor que para el TM. La mayor disminución del TRTcomparado con el TM ocurrió con mayor énfasis en aquellos pacientes queiniciaron con un problema de acúfeno “muy grande”. Cuando los pacientesiniciaron tratamiento con un problema moderado, los beneficios comparativosde estas terapias fueron más modestos.

Palabras Clave: Estudio clínico, trastornos auditivos, militares veteranos,rehabilitación, acúfeno

Abreviaturas: GHI = General Hearing Instruments; HLM = modelaje linealjerárquico; MLM = modelaje multinivel; MML = nivel mínimo deenmascaramiento; PVAMC = Centro Médico de Asuntos de Veteranos dePortland; THI = Inventario de Impedimento con Acúfeno; THQ = Cuestionariode Impedimento con Acúfeno; TISI = Entrevista de Tamizaje sobre el Impactodel Acúfeno; TM = enmascaramiento del acúfeno; TRT = terapia de re-entrenamiento para acúfenos;VA = Asuntos de veteranos;VHA = Administraciónde Salud de Veteranos.

tinnitus sensation more than placebo, nordid they demonstrate replicable, long-termbenefit in tinnitus impact compared toplacebo. Surprisingly, there were norandomized clinical trials to report for twowidely used methods of tinnitus treatment:tinnitus masking (TM) and tinnitusretraining therapy (TRT).

TM and TRT have been used for manyyears in tinnitus clinics around the world.These two methods differ significantly withregard to both their rationale for treatmentand their specific treatment protocols.Implementation of treatment with thesetechniques is accomplished most efficientlyby audiologists who have received the propertraining. A number of studies have reportedgood results with these methods (reviewed inHenry et al, 2002b). Most of these studies,however, were retrospective reports of clinicaldata.

VA is committed to implementingnationally developed, evidence-based practiceguidelines to improve health-care outcomesand efficiency in the veteran patientpopulation (Management Decision andResearch Center, Department of VeteransAffairs, 1998; Feussner, 1998). For veteranswith tinnitus, the present need is for evidencedemonstrating the efficacy of implementablemethods of tinnitus treatment. In responseto this need, we have conducted a controlledclinical trial to evaluate the efficacy of TM andTRT in veterans.

METHODS

Design Overview

A two-group by five-time-period repeatedmeasures experimental design was used toevaluate the effectiveness of the twotreatment conditions, TM and TRT, inimproving five patient outcomes reflectingtinnitus difficulty: the Tinnitus HandicapInventory (THI), Tinnitus HandicapQuestionnaire (THQ),Tinnitus Severity Index(TSI), and percentage ratings of awareness oftinnitus (AWARE) and annoyance by tinnitus(ANNOY). A multilevel modeling approachwas used to analyze trajectories of 118patients, of the 123 who were enrolled in thestudy, on each outcome over an 18-monthperiod after study entry (Raudenbush andBryk, 2002). In addition to estimating the

average patient trajectory in terms of baseline(intercept) and rate of change (linear slope)on each outcome, significant variation inpatient intercepts and slopes was predictedby seven predictor variables includingtreatment condition, three baseline patientcharacteristics (hearing loss, duration oftinnitus, extent of tinnitus problem), andinteraction of treatment condition with eachpatient characteristic.

Recruitment of Patients

Recruitment of patients involved a three-stage process to ensure that only veteranswith clinically significant tinnitus (i.e., atinnitus condition warranting 18 months ofindividualized treatment) were enrolled intothe study: (1) telephone screening; (2) baselinequestionnaires and audiologic assessment;and (3) an appointment with a tinnitusspecialist. Most of the veterans who inquiredabout the study responded to anadvertisement placed in the local newspaper.Some veterans responded to a poster placedat the Portland Veterans Affairs MedicalCenter (PVAMC) Audiology Clinic, or werereferred from the clinic. The use of humansubjects for this research purpose wasapproved by the Institutional Review BoardCommittee at the PVAMC. Each participantsigned an informed consent form prior tostudy enrollment.

Interested veterans were screened overthe telephone using the Tinnitus-ImpactScreening Interview (TISI, see Appendix A)that we developed to rapidly assess tinnitusseverity (Henry et al, 2004).The TISI containssix questions that were derived from the TRTinitial interview (Jastreboff and Jastreboff,1999; Henry et al, 2003) and from the TinnitusSeverity Index (Meikle et al, 1995). (Note: thepresent version of the TISI is revised fromwhat is shown in Appendix 1.) Telephonescreening was performed primarily by thestudy audiologist (T. Zaugg), with someassistance by other staff who were trained forthis purpose. If both the interviewer and theveteran felt that the tinnitus was so severeas to warrant long-term treatment, theveteran was scheduled for an evaluationappointment with the study audiologist.

At the evaluation appointment,candidates first signed informed consent.They then underwent audiologic and tinnitustesting, and answered questions from the

106

Journal of the American Academy of Audiology/Volume 17, Number 2, 2006


107

TRT initial interview (all performed by thestudy audiologist). This appointment servedas a second “filter” to determine if theveteran’s condition warranted long-termtreatment. If the study audiologist consideredthe condition to be sufficiently severe, and ifthe veteran was willing to comply with thestudy protocol, then the veteran was assignedto either the TM group or the TRT group.Thefirst qualifying patient was placed into atreatment group by random selection. Eachsubsequent qualifying patient was placed byalternating between groups.This alternatingapproach to group assignment ensured thatthere was no group-selection bias forindividual patients.

Following group assignment, patientswere scheduled for an appointment with theirrespective treatment specialist to review thequestionnaire and testing data and to discussthe specifics of the treatment protocol. Thisappointment served as the third and finalstage of screening. Continuation in the studywas contingent on the patient and thespecialist reaching agreement regarding thetreatment plan and the requirements to fulfillthe study’s objectives.

Over 800 veterans were screened initiallyby telephone, of whom 171 were scheduled foran initial evaluation. During the second andthird stages of screening, 48 of these 171veterans were excluded from the study.Therewere, therefore, 123 veterans enrolled intotreatment, including 59 (53 males and 6females) in the TM group and 64 (all males)in the TRT group. (There were different groupsizes because more veterans were excludedfrom the TM group than from the TRT groupduring the third stage of screening, i.e., afterbeing assigned to a group but prior toinitiating treatment. Also, the alternatinggroup assignment method resulted in all sixof the qualifying females being placed in theTM group.) Mean ages for the patients were61.0 years (sd = 9.6) for the TM group and 58.7years (sd = 10.5) for the TRT group. Anindependent samples t-test to compare meanages between groups found no significantdifference (p = 0.20).

Procedures

All clinical activities for this study wereconducted at the PVAMC Audiology Clinic.The study audiologist performed audiologicand tinnitus testing, collected and checked the

patients’ written questionnaires, andadministered the TRT interviews (to allpatients in both groups). Treatment wasperformed by M. Schechter for TM and by J.Henry for TRT.

In a recent article (Henry et al, 2002b),the rationale and procedures for performingTM and TRT were described in detailconsistent with the protocols of, respectively,J. Vernon and P. Jastreboff—founders of thetwo techniques. These individuals served asconsultants for this project, and theprocedures adhered closely to theirrecommendations. Both methods utilizedspecific variations of both counseling andsound therapy. Importantly, the 18-monthtreatment schedule was based on the schedulenormally utilized with TRT. Thus, TMtreatment, which is most typicallyaccomplished in a single session, wasextended to include return visits on the sameschedule as for TRT. This modification to theTM protocol ensured that all patients weretreated on the same time schedule.

Counseling for TRT

The patients treated with TRT receivedstructured educational counseling (Jastreboff,2000; Jastreboff and Hazell, 2004). The TRTcounseling protocol is modeled at thesemiannual TRT training seminars that areconducted by P. Jastreboff. The TRTcounseling protocol used in this study wasbased on the protocol as presented at theTRT seminars, including the use of patientcounseling materials (diagrams, charts, etc.).The purpose of the counseling is to“demystify” the tinnitus by explainingmechanisms that underlie tinnitus and itsannoyance. The basis of TRT counseling is P.Jastreboff’s “neurophysiological model,” whichdescribes tinnitus from the standpoint of aneural signal and how the signal causesemotional and stress reactions. Patients arealso explained the rationale and importanceof “sound therapy,” that is, maintaining an“enriched” environment of nonannoyingsounds to maximize treatment effects.Successful counseling removes negativeassociations with the tinnitus, which isnecessary to facilitate habituation to theannoyance caused by the tinnitus (Jastreboffand Hazell, 1998). The TRT counselingamounts to a “teaching session.” The initialsession requires about one hour, with

subsequent sessions requiring progressivelyless time to cover essentially the samematerial.

Counseling for TM

The structured and repeated counselingfor TRT contrasts with the informalcounseling approach that is characteristic oftreatment with TM. Because of its informaland unstructured nature, a specific counselingprotocol for TM has not been described inthe literature. The TM specialist for thepresent study, however, has described ingeneral terms the counseling points that areconsistent with a TM approach (Schechterand Henry, 2002). TM patients receivecounseling that focuses mainly on effectiveuse of sound for providing a sense ofimmediate relief from the tinnitus. Dependingon the patient’s particular set of tinnitus-related problems, counseling can also include(1) providing reassurance to allay fearsconcerning the potential health/psychologicalramifications of their tinnitus; (2) basicprinciples for preventing exacerbation of theirtinnitus; (3) hearing loss and its relation totinnitus; and (4) reducing stress in their lives.It needs to be emphasized that counseling forTM is variable, both in terms of content andduration, and is provided to a patient inaccordance with the patient’s particularcomplaints. Thus, although patients in bothof the treatment groups received counseling,the TRT counseling was structured andrepeated during every session while the TMcounseling was informal and variable.

Sound Therapy for TRT

Sound therapy is considered an adjunctto the structured counseling that is usedwith TRT. The sound therapy protocol isintended to modify auditory processing atsubconscious levels so that neural changesfacilitated by sound therapy will promotehabituation to the tinnitus perception. Thecounseling must be successful in removingnegative thoughts about tinnitus in orderfor habituation to take place (Jastreboff andHazell, 1998).Thus, hierarchically, counselingis the most critical aspect of TRT, withoutwhich sound therapy cannot be effective.Theconcepts of sound therapy are incorporatedinto the TRT counseling protocol. Patients areinstructed to “enrich” their sound

environment at all times with comfortable,nonannoying types of background sound.Treatment with TRT thus always includessound therapy, with or without the use ofear-level devices (Jastreboff, 2000; Jastreboffand Jastreboff, 2000). For patients whorequire long-term treatment with TRT,bilateral ear-level devices are normallyrecommended. All patients in the presentstudy received ear-level devices.

Ear-level devices used for TRT arebasically the same as for TM, except that onlycertain models of sound generators andcombination instruments are approved foruse with TRT. “Approved” means that P.Jastreboff has determined that the devicesmeet certain performance criteria, includingstability of the wideband noise, precisionvolume adjustment at low levels, and open-earconfigurations (Jastreboff, 1994; Jastreboff etal, 1996). Models provided by General HearingInstruments (GHI, New Orleans, LA) meetthese criteria, and only GHI sound generatorsand combination instruments were used for theTRT patients in this study. Hearing aids canalso be used for sound therapy with TRT ifamplification needs cannot be met with thecombination instruments.

Patients in the TRT group who usedsound generators or combination instrumentswere instructed to adjust the output of theirdevices to a level below the “mixing” or“blending” point, and always below any levelthat would cause annoyance. The mixingpoint was described to patients as the pointat which the sound and the tinnitus juststart to mix, or blend, together. Below themixing point the tinnitus can be hearddistinctly. It is a premise of TRT that abovethe mixing point, the tinnitus perceptchanges, and this must not occur becausepatients cannot habituate to their “usual”tinnitus if the tinnitus percept is changedduring sound therapy (Jastreboff and Hazell,2004). The patients were asked to “set andforget” their devices (i.e., to not readjust themduring the day) and to wear them at leasteight hours per day. They were told that theyshould not expect improvement within thefirst 3 to 6 months but that gradualimprovement (i.e., habituation) should benoticed beyond this initial period.

Sound Therapy for TM

TM employs wearable ear-level devices,


108


109

which can include any brand/model of (1) tinnitus masker (sound generator),(2) hearing aid, and (3) tinnitusmasker/hearing aid combination instrument.Each of these three types of devices has beentermed a “masker” when used primarily totreat tinnitus with the TM approach (Vernon,1988). The purpose of maskers is to provideimmediate relief from tinnitus, either bycompletely covering the patient’s tinnitus(“complete masking”) or by reducing theperceived loudness of the tinnitus (“partialmasking”) (Vernon et al, 1990). Vernon (pers.comm.) has also pointed out that maskersserve to alter the intrusiveness or disturbingsound quality (pitch) of some tinnitus sounds.As an adjunct to therapy, other means ofattaining relief are recommended to somepatients, using nonwearable devices such asbedside maskers, specially recorded maskingtapes and CDs, and “sound pillows.”

Selection of ear-level devices for the TMgroup was made on the basis of the resultsof trial use of in-clinic stock devices andpatient feedback from trial use. The clinicianworked with the patient to make the selection,but the patient ultimately selected whichdevice he or she felt was most effective andacceptable in providing some level of relief.Amplification was the first option attemptedif the patient was an appropriate hearingaid candidate. After trial use of the hearingaids in a variety of conditions, the patientwould provide feedback as to its effectivenessas a masker.

The next trial would typically includeuse of a combination instrument for assessingthe effectiveness of both amplification andmasking. The hearing aid would be set to acomfortable level, and noise would beintroduced to determine if the combination ofamplification plus noise was a better optionthan amplification alone. (Frequently, lessnoise is needed in combination withamplification.) Many patients who weremarginal hearing-aid candidates (i.e., forwhom benefit from hearing aids wasequivocal) were given the opportunity to trycombination instruments as well.

Patients who had normal hearing levelswere given the opportunity to try a varietyof tinnitus maskers with in-clinic adjustmentof the filter potentiometers to determine themost acceptable type of noise. A sound boothis not representative of typical acousticenvironments and could bias against fitting

of devices that may perform more effectivelyin everyday listening conditions. Therefore,these devices were tested both in a soundbooth and in areas where there was areasonable level of ambient noise.

TM patients were always carefullyinstructed about the trade-off betweencomplete masking and tolerable levels ofnoise, and that they should use no more noisethan necessary. For the TM approach, thelevel of adequate stimulation (from themasker) is essentially chosen by the patient.Complete masking is the ideal choice if thiscan be accomplished with a well-toleratedlevel of noise.

Patients were not required to wear thedevices consistently throughout the day;however, overall sound stimulation and soundenrichment with other devices (radio, CDplayer, tabletop sound generator, etc.) wereencouraged. If patients felt that tinnitus wasnot bothersome on a particular day, theywere free to not wear the maskers. This is adistinct difference from the TRT approachfor which continual use of ear-level devicesis required.

Ear-Level Devices Worn by Patients

Table 1 shows the numbers of thedifferent types of ear-level devices used by allof the 123 patients. For some patients, thedevices were changed during treatment.Table1 thus shows numbers for devices that werefitted initially, and for the final deviceconfigurations.The different ear-level devicesinclude different makes and models of hearingaids, sound generators, and combinationinstruments (as described above). The soundgenerators and combination instrumentspresented various spectra of wideband noisehaving a high-frequency emphasis. Theacoustic spectrum and the acoustic level of thenoise within the external auditory meatus, ofany given listener, was affected by theimpedance of the trapped volume of air, by thesize and the length of vents to theatmosphere, and by masker filter setting onthe devices.

In the TM group, all fittings werebilateral, except for six of the patients whowere fitted unilaterally: two with a singletinnitus masker (sound generator), and fourwith a single combination instrument. Allsix of the TM patients who were fittedunilaterally had bothersome tinnitus in one

ear, while the other ear had no tinnitus ormild occasional tinnitus. The ear withbothersome tinnitus was fitted with a maskeror a combination instrument to provide relieffrom tinnitus and/or to aid withcommunication. These patients had normalor near-normal hearing in the ear with notinnitus (or mild occasional tinnitus), and soa hearing aid would have been inappropriatein that ear. In addition, one patient (the onlyone with an anacusic ear) was fitted with aBICROS hearing aid, which was laterchanged to a CROS aid to reduce the occlusioneffect.

Bilateral fittings for the TM patientsincluded 11 patients fitted with hearing aids,two with tinnitus maskers, and 39 withcombination instruments. All but one set ofthe tinnitus maskers were Starkey Corp.(Eden Prairie, MN) model TM1s. One TMpatient was fitted with a set of tinnitusmaskers (Tranquils) from GHI. Combinationinstruments were either Starkey TMCs orTMLs. Hearing aids were various makes andmodels. Forty-seven of the 59 TM patientswere provided with devices capable ofproducing masking noise.

For the TRT group, 39 of the patientswere fitted initially with sound generators,one with combination instruments, and 22with hearing aids. Two of the TRT patientswere fitted with a sound generator in oneear and a hearing aid in the other ear.Combination instruments that wereeventually fitted to 12 of the TRT patientswere GHI Harmony (custom fit), which werenot available at the start of the study but

which became available toward the end of thepatient recruitment period. One TRT patientwas fitted with a Harmony combinationinstrument initially, and 11 were refittedwhen the devices became available. Of the 11who were refitted with combinationinstruments, eight were fitted initially withsound generators, and three were fittedinitially with hearing aids.

Sound generators used for TRT wereeither GHI Tranquil (custom fit) or SimplyTranquil (noncustom fit) models. The customTranquil sound generators were fittedinitially to 17 of the TRT patients. Of those17, three were switched to combinationinstruments. Noncustom Simply Tranquilswere fitted initially to 22 of the TRT patients.Of those 22, one switched to bilateral hearingaids, nine switched to custom Tranquils, fiveswitched to combination instruments, andthe remaining seven continued to use theSimply Tranquils. It should be noted thatthe noncustom Simply Tranquil soundgenerators have not been formally endorsedfor use with TRT.

Further details concerning the ear-leveldevices used by the study patients are beyondthe scope of this article. A forthcomingpublication will focus on specific differencesin device use between groups, and how thesedifferences might have influenced outcomesof treatment.

Baseline Questionnaires

Prior to attending their firstappointment, patients were mailed a packet


110

Table 1. Numbers of Patients Wearing Different Ear-Level Devices for Each of the Two TreatmentGroups

Initial Fitting Final Configuration

Ear-Level Devices TM TRT TM TRT

Bilateral Hearing Aids 11 22 12 18Bilateral Sound Generators 2 39 2 32Bilateral Combination Instruments 39 1 38 12Unilateral Sound Generator 2 0 2 0Unilateral Combination Instrument 4 0 4 0Sound Generator and Hearing Aid 0 2 0 2BICROS Hearing Aid 1 0 0 0CROS Hearing Aid 0 0 1 0

TOTAL 59 64 59 64

Note: Numbers are shown separately for the initial fittings and for the final configurations of devices (differences resulting from devicechanges during treatment).


111

of self-administered questionnaires. Thequestionnaires included comprehensivemedical, noise, and tinnitus histories thatare similar to those used at the OregonTinnitus Clinic (Johnson, 1998). Since thereis no universally accepted tinnitus outcomesinstrument, three instruments were utilized:Tinnitus Handicap Inventory (THI) (Newmanet al, 1996; Newman et al, 1998), TinnitusHandicap Questionnaire (THQ) (Kuk et al,1990; Tyler, 1993), and Tinnitus SeverityIndex (TSI) (Meikle et al, 1995). Theseparticular instruments were selected becausethey have each been validated for use withtinnitus patients. Also, of the many tinnitusinstruments available, they are the mostoften used and cited. Patients in this studyprovided their completed baselinequestionnaires to the study audiologist atthe initial evaluation appointment.

For each of the three questionnaires,higher scores indicate greater perceivedtinnitus handicap. The THI has 25 items,and response choices are “no” (0 points),“sometimes” (2 points), and “yes” (4 points).The index score can thus range from 0 to100. The THI has been documented forinternal consistency reliability (Cronbach’s α = .93) and test-retest stability (r = .92)(Newman et al, 1998).The THQ has 27 items,each providing response choices between 0(“strongly disagree”) and 100 (“stronglyagree”) for a total possible maximum score of2700. The THQ has been documented forinternal consistency reliability (Cronbach’s α = .95) (Kuk et al, 1990) and test-retestreliability (r = .89) (Newman et al, 1995).The THI and THQ each provide threesubscales, but only the total index scoreswere utilized for this analysis. The TSI is a12-item instrument that provides fiveresponse choices for each question (“never” =0, “rarely” = 1, “sometimes” = 2, “usually” =3, “always” = 4) with a possible index-scorerange of 0 to 48. It has been shown to havea Cronbach’s α of .92 (Meikle et al, 1995) andtest-retest reliability of .88 (Henry et al,2005b).

TRT Initial and Follow-Up Interviews

Administration of the TRT initialinterview is essential for any patient beingassessed for treatment with TRT (Jastreboffand Jastreboff, 1999; Jastreboff and Hazell,2004).The TRT follow-up interview is similar

to the TRT initial interview, but is revised toevaluate specifically outcomes of treatment.For this study it was determined that theTRT interviews would be administered topatients in both the TRT and TM groups fortwo reasons: (1) administering the interviewsrequired significant time, and patient contacttime needed to be equalized between groupsas much as possible; and (2) to enableintergroup comparison of the interviewoutcome data. Each of the TRT interviewforms was modified for use with this projectto ensure consistent administration (Henryet al, 2003). Instructions to patients and theinterview questions were scripted, and aclosed set of response choices was providedfor most questions. The study audiologistadministered all of the interviews for bothtreatment groups. The initial interviewrequired 30–60 minutes to complete, and thefollow-up interview required 20–40 minutes.

Two of the interview questions were usedas outcomes for the data analysis: “Whatpercent of your total awake time, over the lastmonth, have you been aware of your tinnitus?”(AWARE) and “What percent of your totalawake time, over the last month, were youannoyed by your tinnitus?” (ANNOY). Patientresponses for each of these questions wereobtained using a scale of 0% to 100%.

Baseline Audiometric and TinnitusEvaluation

Patients received a complete audiometricand tinnitus evaluation at their initialappointment with the study audiologist. Theaudiologic assessment was performedfollowing completion of the TRT initialinterview, and included otoscopy, pure toneand speech audiometry, immittance measures,and loudness discomfort levels (LDLs) at .5,1, 2, 3, 4, 6, and 8 kHz. Audiometric testingwas carried out with calibrated equipment ina conventional sound-treated booth meetingestablished standards. Mean hearingthresholds for each treatment group areshown in Table 2. The tinnitus evaluationincluded tinnitus loudness and pitchmatching, minimum masking levels, andresidual inhibition testing (Henry, 2004).Equipment for tinnitus testing was custombuilt by the computer engineer (R. Ellingson)who has built different versions of thesesystems for our research. The tinnitusevaluation was consistent with a protocol

described previously (Henry et al, 2002a),except for the addition of residual inhibitiontesting. Space constraints do not allow acomplete description of the audiologic andtinnitus testing, nor of the data resultingfrom those tests. This information will beprovided in a publication that is inpreparation.

Initial Appointment with TinnitusSpecialist

Following the initial evaluation with thestudy audiologist, patients were scheduled foran initial appointment with their respectivetinnitus specialist. The tinnitus specialistreviewed the questionnaire and testing dataand discussed the treatment protocol with thepatient. For patients who entered treatment,ear-level devices were selected and earmoldimpressions were made to order customdevices. Patients were scheduled for anappointment approximately one month laterfor device fitting and initial counseling (i.e.,the first treatment appointment). Patientswere also referred to an otolaryngologist fora medical evaluation and to receive medicalclearance to wear the ear-level devices.

Results of the tinnitus psychoacousticassessment have no bearing on the treatmentprotocol for TRT. These measurements areimportant, however, for selecting ear-leveldevices for treatment with TM. The MML(minimum masking level) is a particularlyimportant measurement for the TM approach,since it can serve as a predictor of successwith application of maskers (Schechter andHenry, 2002). Because of the importance ofthis measurement, the tinnitus specialist forTM made more precise measures in 1 dBsteps (as opposed to the 5 dB steps used by

the study audiologist).The repeated MML testoften resulted in a different measurementvalue; thus, the residual inhibition test, whichwas based on this value, was also repeated.

First Treatment Appointment

At the initial treatment appointment,ear-level devices were fitted, and treatment-specific counseling was administered. Whenamplification was used, real ear testing wasperformed routinely at the time of the fittingto verify target gain settings.

Continuing Treatment Appointments

Continuing treatment appointmentswere scheduled at 3, 6, 12, and 18 monthsafter the first treatment appointment.Approximately two weeks prior to eachtreatment visit, patients were mailed a packetof questionnaires that included anabbreviated version of the medical, noise,and tinnitus histories questionnaire, and theTHI,THQ, and TSI.At each appointment, thestudy audiologist: (1) checked the writtenquestionnaires for completeness;(2) administered the TRT follow-up interview(Henry et al, 2003); and (3) performedaudiologic and tinnitus testing (audiologicand tinnitus testing were not performed at the3-month visit). The treatment specialist thenmet with the patient to check the performanceand proper usage of the ear-level devices,and to administer counseling as describedabove.

Data Analysis

Many studies that focus on change intinnitus outcomes have been limited by cross-


112

Table 2. Mean Audiometric Hearing Thresholds, in dB HL, for the Patients Enrolled in This Study

Tinnitus Masking Tinnitus Retraining Therapy

Left ear Right ear Left ear Right ear

Frequency (Hz) Mean SD Mean SD Mean SD Mean SD

250 19.0 13.3 19.2 15.9 19.2 13.6 17.6 9.6500 19.9 13.6 19.4 18.8 20.2 13.9 18.3 11.7

1,000 21.8 14.3 21.9 17.5 21.6 14.4 20.8 13.72,000 33.1 23.2 32.3 25.6 28.1 22.5 24.9 20.03,000 48.6 23.9 45.8 28.5 45.4 25.2 41.2 22.84,000 58.5 21.4 55.3 25.8 55.0 23.3 51.0 22.16,000 64.3 23.0 62.9 26.7 62.7 22.3 56.4 24.08,000 63.3 20.2 64.8 23.4 62.9 25.6 58.7 24.3


113

sectional and pretest-posttest designs usingmethods such as repeated measures ANOVA.These studies have generally met Guyatt etal’s (1993) minimal criteria for properlyconducting randomized clinical trials (Dobie,2004). There are, however, numerous designissues that have limited the knowledge gainedfrom these studies. In many cases, dropoutshave not been handled on an intent-to-treatbasis, which is now generally agreed to be arequirement of statistical analysis foroutcome studies. Repeated measures ANOVAassumes that all of the patients can besummarized by the same mean profile overtime and any individual variation ischaracterized as error.

Alternatively, innovative approaches suchas multilevel modeling (MLM) capitalize onmultiple times of measurement to explicitlyexamine (1) individual trajectories of tinnitusoutcomes over time (both the averagetrajectory and the individual variation aroundthis average trajectory); (2) reliabilityestimates of change in tinnitus outcomes;(3) the correlation between baseline and rateof change in tinnitus outcomes; and (4) predictors of change in tinnitus outcomes.The individual patient becomes his or herown reference, and thus, rate of change ismore informative and relative.

MLM is considered hierarchical as itinvolves data that vary at two levels: withinindividuals (level 1) and between individuals(level 2). At level 1, each individual’s changeis represented by an individual trajectorycaptured by a set of unique growthparameters: intercept and slope.These latentgrowth parameters are corrected formeasurement error. At level 2, these latentparameters become outcome variables to beexplained by a set of individual-levelvariables.

MLM provides a powerful and flexibleframework for analyzing individual change.The key statistical advantage above othermethods is that it controls for thedependencies among the repeatedmeasurements. The key design advantage isthat it allows for differences in the numberof times of measurement across individualswhen random “missingness” exists. This ispossible as the multiple observations overtime are considered nested within theindividual. For a more detailed discussion ofmultilevel modeling and its applications seeRaudenbush and Bryk (2002).

Outcome Measures

Of the five outcome measures analyzed,three had a potential range of 0 to 100 (THI,AWARE, and ANNOY). For a fourth outcome,the THQ, we used an average score of the 27THQ items rather than the usual total THQscore, in order for the potential range of theTHQ to be from 0 to 100, the same as therange of the THI, AWARE, and ANNOYoutcomes. The fifth outcome, the TSI, had apotential range of 0 to 48.

Predictors

Seven predictors were used. Onepredictor was treatment condition,TM (coded0) and TRT (coded 1).The other six predictorswere three baseline patient characteristicsand the interaction of each characteristicwith treatment condition.

“Hearing loss” was measured using thefirst unrotated factor from a principalcomponents factor analysis of seven right-earand seven left-ear hearing levels (.25, .5, 1,2, 3, 4, and 8 kHz) from audiologic testing.Thefirst unrotated factor, representing extent ofhearing loss, explained 58% of the variancein the principal components analysis and itsz-score form (M = 0.0; sd = 1.0) was used inthe MLM analysis.

“Duration of tinnitus” was the self-reported length of time the patient hadtinnitus and was coded into three categoriesfor analysis: -1 (0–10 years), 0 (11–20 years),or 1 (≥21 years). The percentages of TMpatients in the three respective durationcategories were 32%, 18%, and 51%; thepercentages of TRT patients, respectively,were 28%, 15%, and 57%. The distribution ofduration did not differ by group, χ2 (2, N = 118)= 0.51, p = .78.

“Extent of tinnitus problem” was a self-report item in which patients rated the extentof their tinnitus problem; responses werecoded -1 (moderate problem), 0 (big problem),or 1 (very big problem).The percentages of TMpatients in the three respective problemcategories were 28%, 51%, and 21%; thepercentages of TRT patients, respectively,were 30%, 46%, and 25%. The distribution oftinnitus problem at baseline did not differ bygroup, χ2 (2, N = 118) = 0.33, p = .85.

Three interaction terms were computedfor use in the analysis: treatment X hearingloss, treatment X duration, and treatment X

extent of tinnitus problem. The dichotomousvariable for treatment condition (TM = 0,TRT = 1) was multiplied by each of the threepatient characteristics (z-score for hearingloss; -1, 0, 1 for duration; -1, 0, 1 for extent oftinnitus problem) to obtain the interactionterm predictors.

Overall MLM Analysis

We used a two-level MLM approach todetermine the extent to which variation inpatient trajectories on each of five outcomesover an 18-month period could be explainedby treatment condition and the other sixpredictors. The HLM (Hierarchical LinearModeling) software Version 6.0 was used(Raudenbush et al, 2004). Underlying anMLM approach in analyzing treatmenteffectiveness is the assumption that atreatment condition does not necessarily havea uniform effect on all patients but thattreatment effectiveness can vary acrosspatients. Using HLM we estimated eachpatient’s individual trajectory (baseline andlinear slope) on each of the five outcomes(THI, THQ, TSI, AWARE, ANNOY) over 18months after entering treatment. Then, foreach outcome, we estimated the averagetrajectory across all patients. If there wassignificant variation around the averagebaseline and average slope, we examined theextent to which seven predictors explained

such variation.

Estimating Patient Trajectories at Level 1

At level 1 of HLM, for each patient,ordinary least squares (OLS) estimates ofbaseline and rate of change every six months(linear slope) for each of the five outcomeswere derived where time of assessment (0, 3,6, 12, and 18 months) was centered at zero,and therefore coded as 0, 0.5, 1, 2, and 3,respectively.The questions to be answered atlevel 1 were:

1. What is the average trajectory(baseline and slope) of patients on thetinnitus outcomes over the 18months after entering treatment?The average baseline and slope arereferred to as the level-1 fixed effects.

2. Is there significant variation aroundthe average trajectory, above andbeyond what would be expected bychance? Variation around theaverage intercept and average slopeare referred to as level-1 randomeffects. If there had been nosignificant variation around thegrowth parameters, we would nothave gone on to a level-2 modelwhere we examined predictors ofvariation in individual patienttrajectories.

For illustration purposes, in Figure 1,


114

Figure 1. Average THI trajectory for the overall sample and individual trajectories for six patients. The tra-jectory with a dashed line indicates a patient whose THI scores increased (worsened).


115

we have shown the average THI trajectory,Y = 49.5 + (-6.3 x Time), and individual THItrajectories for six patients. As shown, five ofsix patients declined on the THI, reflectingimprovement, but one patient (with thedashed line) worsened and had increases onthe THI. Of these six patients, their startingplaces (baseline) differed as did their rate ofchange (slope), with some changing rapidlyand others changing more slowly. It is thesevariations in baseline and slope that we wereinterested in predicting.

Predicting Variation in PatientTrajectories at Level 2

At level 2 of HLM, predictors ofsignificant variation in patient trajectories foreach outcome were examined. The questionsof main interest at level 2 were:

1. To what extent does receiving TMversus TRT explain variation inpatient trajectories around theaverage trajectory for each outcome?

2. To what extent do three baselinepatient characteristics—hearingloss, duration of tinnitus, andperceived problem due to tinnitus—and the interaction between thetreatment condition and each ofthese characteristics explainvariation in patient trajectoriesaround the average trajectory foreach outcome?

Rationale for Predictors

Treatment condition, TM versus TRT,was the predictor of main interest. Althoughwe did not hypothesize directional differencesbetween TM and TRT on patient trajectories,comparison of effectiveness of the twotreatment conditions was the main goal of thestudy.

We used hearing loss and duration oftinnitus, including their interactions withtreatment condition, as predictors of patienttrajectories because we wanted to explorethe extent to which they were associatedwith baseline levels of the tinnitus outcomesand to evaluate whether TM or TRT wasdifferentially effective, depending on the levelof a patient’s hearing loss or duration oftinnitus. We selected extent of tinnitusproblem as a predictor of variation in patienttrajectories, because we thought it could serve

as a common baseline measure of tinnitusdifficulty across all the outcomes, rather thanusing each outcome’s baseline score as acovariate. We hypothesized that extent oftinnitus problem would be positivelyassociated with variation in patient baseline.In addition, because patients with moretinnitus difficulty have more room to changeas a result of treatment, we expected a biggerimprovement (larger slope) in patientsstarting out worse (i.e., patients with a “verybig” tinnitus problem) compared to patientswith a “big” or “moderate” tinnitus problem.Finally, although we did not hypothesize aninteraction between treatment condition andextent of tinnitus problem, we were interestedin whether TM or TRT was differentiallyeffective, depending on whether patientsstarted treatment at very big, big, or moderatelevels on the extent of tinnitus problem.

Sample Attrition and Missing Data

Of the 123 patients who entered thestudy, five were missing data on one or moreof the predictor variables and were thereforeremoved from the sample, resulting in a finalsample of 118 patients for the MLM analyses(n = 57 for TM, n = 61 for TRT).When patientsdo not have complete data on predictorvariables, they are dropped from the HLManalysis.

One advantage of MLM is that missingdata over time in the outcome variables (level-1 missingness) can be addressed if the dataare assumed to be missing at random. HLMuses maximum likelihood to estimateparameter values based on all existing dataacross waves of measurement and data atlevel 2. Maximum likelihood is an iterativeprocess that uses all available information inthe data to obtain unbiased parameterestimates that have been adjusted formissingness (Schafer and Graham, 2002). Inthe current study, random missingnessoccurred, in part because the THI and THQwere added as outcome measures after 13 and29 patients, respectively, had begun the study;however, these patients were administeredthe THI and THQ at later data collectionperiods. Further, at the 3-month follow-upperiod, the different outcome measures werenot completed by between 17 and 57 patients.Finally, not all patients participated at eachfollow-up period. See Table 3, which indicatesthe number of patients who have data for each

outcome at each data collection period.Outcome data were reasonably complete.

Of the 118 patients, 46 had complete data onall five outcomes at all five data collectionperiods, resulting in 25 usable scores for eachpatient. Another 62 patients had 15 to 24usable outcome scores of the 25 possible.Only three of the 118 patients had baselinedata only, with one score on each of the fiveoutcomes; the remaining 10 patients had 10to 14 of the possible 25 outcome scores. Datawere most complete for three outcomes—TSI,AWARE, and ANNOY—with 72% to 76%of the patients having outcome scores for allfive periods.

RESULTS

Descriptive statistics for the five outcomesat all five time points are presented in

Table 3. Descriptive statistics for the predictorvariables are presented in Table 4. Theintercorrelations among the treatmentcondition and the three baseline predictors(hearing loss, duration of tinnitus, and extentof tinnitus problem) were low, ranging from -.09 to .24.At baseline, three outcome variables(THI, THQ, and TSI) were highly

intercorrelated (r = .76 to .78), and two otheroutcome variables (AWARE and ANNOY) hada moderately high correlation (r = .60). Atbaseline, the correlations of ANNOY with THI,THQ, and TSI were modest (r = .34 to .36, p <.05), and AWARE was not significantlycorrelated with THI,THQ,or TSI (r = .11 to .20).

Level-1 Models: Average Trajectoryover Time and Variation around theAverage

As shown under the Fixed Effectscolumns in Table 5, both the baseline andlinear slope for each of the outcomes weresignificantly different from zero. Using theTHI as an example, on average, the level oftinnitus handicap for the patients at baselinewas 49.50, and the rate of change was -6.29THI points every six months, or nearly 19 THIpoints decline across the 18 months. Theaverage THI trajectory is shown in Figure 1.

Using the percent Awake Time Aware ofTinnitus (AWARE) as another example, onaverage, the level of AWARE for the patientsat baseline was 71.24, and the rate of changewas -16.05 AWARE points every six months,or more than 48 AWARE points decline across


116

Table 3. Outcome Variables: Means, Standard Deviations, and Ranges

Outcome Variable Baseline 3 Months 6 Months 12 Months 18 Monthsn = 118 n = 101 n = 110 n = 108 n = 107

Tinnitus Handicap Inventory (THI)n 105 61 94 101 106M 52.54 47.27 41.01 35.48 30.97SD 22.41 24.92 25.34 23.47 24.73Range 12.0–98.0 6.0–100.0 6.0–100.0 0.0–98.0 0.0–100.0

Tinnitus Handicap Questionnaire (THQ)n 89 62 95 101 106M 56.22 51.18 45.54 39.95 36.39SD 21.02 22.51 23.35 22.57 22.56Range 16.1–97.0 8.0–93.5 6.3–96.3 3.7–94.4 0.0–92.6

Tinnitus Severity Index (TSI)n 118 97 107 104 107M 28.28 25.16 23.79 20.96 19.21SD 8.72 8.86 9.05 9.92 10.87Range 8.7–48.0 1.3–48.0 5.1–48.0 1.3–48.0 2.0–46.7

Percent Awake Time Aware of Tinnitus (AWARE)n 117 101 110 108 107M 72.51 63.47 52.13 38.37 24.38SD 22.78 28.42 30.13 28.72 22.55Range 6.5–100.0 2.5–100.0 0.0–100.0 0.0–80.0 1.0–100.0

Percent Awake Time Annoyed by Tinnitus (ANNOY)n 116 101 110 108 107M 51.12 35.39 31.31 21.20 14.50SD 28.20 27.06 27.68 24.60 18.53Range 1.0–100.00 0.0–100.0 0.0–100.0 0.0–100.0 0.0–80.0


117

the 18 months. Similar interpretations ofbaseline and 6-month rate of change can bemade to describe the average trajectory forTHQ, TSI, and ANNOY. Over 18 months, theaverage patient declined about 18 THQ pointsfrom a baseline of 54, declined about 8 TSIpoints from a baseline of 27, and declinedabout 34 ANNOY points from a baseline of 46.

One advantage of MLM over traditionalmethods of analyzing change is that the truecorrelation between baseline and rate of

change can be estimated. The correlationsbetween baseline and rate of change intinnitus outcomes were -.36 for THI, -.21 forTHQ, .01 for TSI, -.34 for AWARE, and -.84for ANNOY.The magnitude of the correlationfor ANNOY was large and indicates thatpatients who reported higher annoyance fromtinnitus at baseline were the ones whoshowed the greatest declines in annoyance;change in annoyance can be predicted verywell from initial annoyance status. The

Table 4. Predictor Variables: Means, Standard Deviations, and Ranges (n = 118)

Simple Predictors Interaction Variables: Treatment X Baseline Characteristics

TM (0) v. TRT (1)

M 0.52SD 0.50Range 0–1

Hearing Loss Treatment X Hearing Loss InteractionM 0.01 -0.04SD 1.00 0.69Range -1.81–3.37 -1.68–3.37

Duration of Tinnitus Treatment X Duration of Tinnitus InteractionM 0.25 0.15SD 0.89 0.65Range -1–+1 -1–+1

Extent of Tinnitus Problem Treatment X Extent of Tinnitus Problem InteractionM -0.06 -0.03SD 0.72 0.53Range -1–+1 -1–+1

Table 5. Results of HLM Level-1 Models of Fixed and Random Effects for the Five Outcomes

Fixed Effects Random Effects

Outcome B SE Variance df χ2 Reliability of Growth Unstandardized Parameters

Tinnitus Handicap Inventory (THI) Intercept 49.50** 2.13 459.45 113 952.29** .85 Linear slope -6.29** 0.73 42.60 113 348.20** .64

Tinnitus Handicap Questionnaire (THQ) Intercept 53.87** 1.94 313.92 107 520.86** .72Linear slope -6.01** 0.78 35.05 107 239.51** .50

Tinnitus Severity Index (TSI)Intercept 27.27** 0.73 48.52 114 482.49** .78Linear slope -2.76** 0.31 5.61 114 244.54** .51

Percent Awake Time Aware of Tinnitus (AWARE)Intercept 71.24** 2.17 361.93 114 329.50** .65Linear slope -16.05** 0.91 27.21 114 170.22** .26

Percent Awake Time Annoyed by Tinnitus (ANNOY)Intercept 45.65** 2.29 433.83 114 370.89** .70Linear slope -11.18** 0.80 10.40 114 140.93* .14

*p < .05 **p < .001

correlations between baseline and rate ofchange for THI, THQ, and AWARE weremodest, and there was no relationshipindicated between baseline and rate of changefor TSI.

For each outcome there was significantvariation in the baseline and slope to beexplained in a level-2 model, as shown inTable 5 by the χ2 test results under theRandom Effects columns. Reliabilitycoefficients, which indicate the proportion ofvariance in the OLS estimates of baseline andslope that is true variance, provided furtherevidence that systematic variance existed inthe level-1 baseline and slope coefficients tobe explained at level 2. In examining thereliability of the growth parameters in the lastcolumn of Table 4, we concluded thatreliability of the baseline estimates was veryhigh for all five outcomes. The reliability ofthe slope was very high for the THI,THQ, andTSI but quite low for AWARE (reliability =.26) and below generally acceptable levelsfor ANNOY (reliability = .14).

Due to the low reliability for the ANNOYlinear slope (14% systematic or true variance)and the fact that the individual variationaround the average ANNOY slope was almostnonsignificant (p = .045), we did not proceedwith level-2 analysis of ANNOY. Instead, wefelt a more appropriate analysis of change inANNOY should occur at the group levelthrough repeated measures analysis ofvariance (ANOVA). Thus, at the end of this“Results” section, we report ANOVA resultscomparing TM and TRT across 0, 6, 12, and18 months; the 3-month ANNOY scores werenot included in order to minimize missingdata.

Level-2 Models: Explaining Variationin Patient Trajectories

Overall Level-2 Models

The seven predictors were entered intoa level-2 model to examine their associationwith interindividual differences in thebaseline and linear slope (6-month rate ofchange) in each outcome. As shown in Table6, the seven predictors explained largeamounts of variance (38.9% to 43.4%) in thebaseline for THI, THQ, and TSI, but arelatively small amount of variance (13.9%)in the baseline for AWARE. Very large

amounts of variance (46.0% to 63.3%) in thelinear slopes were explained for THI, THQ,and TSI, with a modest amount of variance(14.3%) in the linear slopes explained forAWARE.

Explaining Variation in THI, THQ,and TSI Trajectories

Examination of the predictors of thebaseline showed that extent of tinnitusproblem was the predictor most stronglyassociated with the baseline, as hypothesized.Controlling for all other predictors, theunstandardized coefficients indicate that foreach one-unit increase in extent of tinnitusproblem from -1 to 0 (moderate problem to bigproblem) and from 0 to +1 (big problem to verybig problem), the THI baseline was 14.53points higher, the THQ baseline was 12.96points higher, and the TSI baseline was 4.57points higher. Duration of tinnitus was alsoa significant predictor of variation in patientbaselines for the THQ and TSI, indicating thatlonger duration of tinnitus was associatedwith more tinnitus difficulty at baseline.Specifically, for each one-unit increase intinnitus duration, from -1 to 0 (0–10 years to11–20 years) and from 0 to +1 (11–20 yearsto 20+ years), the THQ baseline was 5.50points higher and the TSI baseline was 2.17points higher. Other important findings werethat the TM and TRT groups were notsignificantly different at baseline on the THI,THQ, or TSI, suggesting that assignment totreatment conditions resulted in roughlycomparable groups.

There were three significantunstandardized coefficients for rate of 6-month linear change: (1) the intercept forlinear slope (i.e., the value of the slope whenall predictors are 0); (2) TM versus TRT; and(3) treatment X extent of tinnitus probleminteraction. Generally, as shown by theseunstandardized coefficients in Table 6, TMpatients exhibited a significant 6-month rateof decline on each outcome (-3.83 on THI; -2.65on THQ; -1.37 on TSI), but TRT patientsexhibited these declines plus an additionalsignificant 6-month rate of decline (-5.99 onTHI; -7.49 on THQ; -3.03 on TSI), whichtranslates into 6-month declines of -9.82,-10.14, and -4.4 on THI, THQ, and TSI,respectively, for TRT patients.

The nature of the significant treatmentX extent of problem interaction is displayed


118


119

in Figures 2, 3, and 4 for THI, THQ, and TSI,respectively. In each figure, four prototypicaltrajectories of the outcome are plotted. Twotrajectories are for TM patients who begantreatment with either a very big or amoderate tinnitus problem; the other twotrajectories are for TRT patients who begantreatment with either a very big or amoderate tinnitus problem. In the trajectories,hearing loss and duration of tinnitus arecontrolled for, with trajectories computed forpatients with average hearing loss (z = 0)and the median duration of tinnitus (11 to 20years, coded as 0). The most prominentpattern in these figures is that for patientswho began treatment with a very big tinnitus

problem, those receiving TRT exhibited 6-month declines of -15.8,-14.7, and -6.2 on THI, THQ, and TSI,respectively, compared to 6-month declines of-4.5, -2.2, and -1.2 for patients receiving TM.For patients who began treatment with amoderate tinnitus problem, those receivingTRT exhibited declines of -3.8, -5.6, and -2.6on THI,THQ, and TSI, respectively, comparedto 6-month declines of -3.1, -3.1, and -1.5 forpatients receiving TM. The declines on theoutcomes were much larger for TRT comparedto TM for patients who began treatment witha very big tinnitus problem, whereas thebenefits of TRT over TM were more modestfor patients who began treatment with a

Table 6. Parameter Estimates and Associated Standard Errors from the Level-2 Model of the Effects ofthe Predictors on Baseline and Linear Change in Tinnitus Handicap, Severity, and Awareness

Baseline Rate of 6-Month Linear Change (Slope)

Unstandardized Variance Unstandardized VarianceB SE Explained B SE Explained

Tinnitus Handicap Inventory (THI)Intercept 50.29** 2.57 -3.83** 0.96TM vs. TRT 0.78 3.64 -5.99** 1.34Hearing Loss -2.38 2.47 0.47 0.89Duration 4.20 2.90 1.50 1.09Extent of Problem 14.53** 3.58 -0.69 1.29Tx X Hearing Loss 0.50 3.63 0.12 1.31Tx X Duration -5.30 4.10 -0.56 1.51Tx X Extent of Problem 6.39 5.00 38.9% -5.32** 1.83 46.0%

Tinnitus Handicap Questionnaire (THQ)Intercept 52.07** 2.34 -2.65** 0.98TM vs. TRT 5.08 3.36 -7.49** 1.38Hearing Loss -0.89 2.32 0.69 0.94Duration 5.50* 2.70 0.92 1.15Extent of Problem 12.96** 3.21 0.42 1.30Tx X Hearing Loss 0.24 3.36 0.12 1.36Tx X Duration -7.03 3.83 0.56 1.58Tx X Extent of Problem 2.95 4.50 43.4% -4.95** 1.85 56.7%

Tinnitus Severity Index (TSI)Intercept 27.21** 0.89 -1.37** 0.38TM vs. TRT 0.34 1.27 -3.03** 0.53Hearing Loss -1.67 0.86 0.84 0.36Duration 2.17* 1.01 0.43 0.43Extent of Problem 4.57** 1.24 0.13 0.51Tx X Hearing Loss 0.81 1.26 -0.66 0.53Tx X Duration -1.95 1.42 -0.05 0.60Tx X Extent of Problem 2.45 1.72 42.3% -1.98** 0.72 63.3%

Percent Awake Time Aware of Tinnitus (AWARE)Intercept 72.17** 3.07 -14.16** 1.34TM vs. TRT -3.41 4.35 -3.97* 1.88Hearing Loss -3.74 2.99 0.16 1.29Duration 9.00* 3.46 0.80 1.52Extent of Problem 5.66 4.30 -0.74 1.83Tx X Hearing Loss 1.67 4.37 -0.36 1.88Tx X Duration -6.94 4.88 0.78 2.10Tx X Extent of Problem -1.28 5.94 13.9% -0.58 2.54 14.3%

*p < .05 **p < .001


120

Figure 2. Treatment X tinnitus problem interaction for the THI outcome. Shown are four THI trajectories across18 months. Two trajectories are for TM patients who began treatment with either a very big or a moderate tin-nitus problem; the other two trajectories are for TRT patients who began treatment with either a very big or amoderate tinnitus problem. Hearing loss and duration of tinnitus are controlled for, with trajectories computedfor patients with average hearing loss (z = 0) and the median duration of tinnitus (11 to 20 years, coded as 0).

Figure 3. Treatment X tinnitus problem interaction for the THQ outcome. Shown are four THQ trajectoriesacross 18 months. Two trajectories are for TM patients who began treatment with either a very big or a mod-erate tinnitus problem; the other two trajectories are for TRT patients who began treatment with either a verybig or a moderate tinnitus problem. Hearing loss and duration of tinnitus are controlled for, with trajectoriescomputed for patients with average hearing loss (z = 0) and the median duration of tinnitus (11 to 20 years,coded as 0).


121

moderate tinnitus problem.Also of note was the lack of significant

association of duration of tinnitus and hearingloss with rate of change on the THI,THQ, andTSI. Further, as shown by the nonsignificanttreatment X duration and treatment Xhearing loss interaction terms, theeffectiveness of TM and TRT did not differdepending on the patient’s duration oftinnitus or hearing loss.

Explaining Variation in AWARETrajectories

Only duration of tinnitus explainedvariation in the AWARE baseline. Controllingfor other predictors in the model, theunstandardized coefficients show that for eachone-unit increase in duration of tinnitus, theAWARE baseline was 9.0 points higher.Therewere two significant unstandardizedcoefficients for rate of 6-month linear change:(1) intercept for linear slope (i.e., the value ofthe slope when all predictors are 0); and (2) TMvs. TRT. On average, as shown by theunstandardized coefficients in Table 6, TMpatients exhibited a significant 6-month rateof decline on AWARE (-14.16),and TRT patients

exhibited this decline plus an additionalsignificant -3.97 points in 6-month rate ofdecline. Although the treatment X extent ofproblem interaction was not significant forAWARE, for illustration purposes andcomparison with Figures 2, 3, and 4, fourprototypical trajectories of AWARE are plottedin Figure 5. The most noticeable feature inFigure 5 is that the four trajectories havesimilar slopes, with the slope for the TRTpatients with a very big problem (B = -19.5)being 1.3 times larger than the slope for the TMpatients with a very big problem (B = -14.9).

To conclude, the level-2 models with sevenpredictors for THI, THQ, TSI, and AWAREwere a significantly better fit than anunconditional model with no predictors (THI,χ2 = 77.49, df = 14, p < .001; THQ, χ2 = 76.11,df = 14, p < .001; TSI, χ2 = 90.85, df = 14, p <.001; AWARE, χ2 = 29.30, df = 14, p < .01).

Repeated Measures ANOVA forANNOY

A 2 X 4 repeated measures ANOVA forn = 101 patients with complete ANNOY datarevealed significant main effects (p = .001) fortreatment group, F (1, 99) = 11.80 and time,

Figure 4. Treatment X tinnitus problem interaction for the TSI outcome. Shown are four TSI trajectories across18 months. Two trajectories are for TM patients who began treatment with either a very big or a moderate tin-nitus problem; the other two trajectories are for TRT patients who began treatment with either a very big or amoderate tinnitus problem. Hearing loss and duration of tinnitus are controlled for, with trajectories computedfor patients with average hearing loss (z = 0) and the median duration of tinnitus (11 to 20 years, coded as 0).


122

F (3, 297) = 71.57, as well as a significantgroup X time interaction, F (3, 297) = 2.79, p= .041 (see Fig. 6). Generally, we expected tosee declines in ANNOY over time. Therefore,after obtaining the significant F test for theinteraction, we computed a priori comparisonsusing the Bonferroni adjustment for multipletests to see if ANNOY scores were decliningdifferently for each group. Simple effects testscomparing mean changes over time for eachgroup separately showed that within-groupchange on the ANNOY outcome over time wasdifferent for TM compared to TRT. The TMGroup showed a large and significant declineof 19.7 points in ANNOY from Month 0 toMonth 6 (p < .001); however, the 4.3-pointdecline in ANNOY from Month 6 to Month 12and the 4.4-point decline from Month 12 toMonth 18 were not significant (p = .40 and .50,respectively); however, the 8.7-point declinefrom Month 6 to Month 18 was significant (p = .03). Similar to the TM Group, the meanANNOY score for the TRT Group showed alarge and significant decline of 22.4 pointsfrom Month 0 to Month 6 (p < .001). Unlike theTM Group, the TRT group showed continuedsignificant declines from Months 6 to 12 (11.8-point decline, p = .002) and Months 12 to 18

(6.8-point decline, p = .002). Over the entire 18-month period, the annoyance from tinnitusdeclined 28.3 points in TM patients, on average,compared to TRT patients whose declineaveraged 41.0 points. So, the overall picture isthat both the TM and TRT patients showedimprovement over time, but the TRT patientscontinued to show greater improvement afterMonth 6, compared to TM.

DISCUSSION

This report provides an initial analysis ofoutcome data from a controlled clinical

study that was conducted to compare theefficacy of two tinnitus treatment methods inU.S. military veterans.Veterans with clinicallysignificant tinnitus received treatment witheither TRT or TM for a period of 18 months.All patients who were enrolled in the studyunderwent the same three-stage screeningprocess.The screening was designed to ensurethat veterans were enrolled in the study onlyif they had tinnitus of sufficient severity tojustify 18 months of individualized treatment.Patients in both groups were generallymotivated to comply with the study protocol,resulting in a low attrition rate (about 10

Figure 5. Treatment X tinnitus problem interaction for the AWARE outcome. (Note: The treatment X tinnitusproblem interaction was not significant, p > .05.) Shown are four AWARE trajectories across 18 months. Twotrajectories are for TM patients who began treatment with either a very big or a moderate tinnitus problem;the other two trajectories are for TRT patients who began treatment with either a very big or a moderate tin-nitus problem. Hearing loss and duration of tinnitus are controlled for, with trajectories computed for patientswith average hearing loss (z = 0) and the median duration of tinnitus (11 to 20 years, coded as 0).


123

percent—only 12 of the 123 patients enrolledwere lost to attrition).

Summary of Findings—Groups Combined

For both groups combined, the averagepatient entered treatment for tinnitus with anestimated baseline score around 50 on theTHI and THQ and 27 on the TSI (which is justabove midway on the TSI scale). The averagepatient reported awareness of his/her tinnitusabout 71 percent of the time (“waking hours”)and annoyance by it 46 percent of the time.With each additional decade of having tinnitus,patients entered treatment with higher scorelevels on the THQ (5.50 points), TSI (2.17points), and percent of Awake Time Aware ofTinnitus (AWARE) (9.00 percentage points).By the end of treatment, 18 months later, theaverage patient had declined 18 to 19 pointson the THI and THQ and 8 points on the TSI,and was aware of his/her tinnitus 23% of thetime and annoyed by it only 12% of the time.

Summary of Findings—Groups Compared

To compare findings between groups,outcome data were analyzed on an intent-to-treat basis, using multilevel modeling (MLM)

analysis.The analysis revealed that both TMand TRT groups resulted in declines intinnitus handicap and severity (THI, THQ,TSI), but the decline in TRT patients wasconsiderably greater than the decline in TMpatients. However, the greater declines inTRT compared to TM occurred most stronglyin patients who began treatment with a “verybig” tinnitus problem (as measured by asingle baseline item on which patients ratedtheir tinnitus problem as moderate, big, orvery big). For these patients, the rate ofimprovement was considerably faster in TRTcompared to TM. When patients begantreatment with a moderate tinnitus problem,the benefits of TRT compared to TM weremore modest.

On percentage awareness of tinnitus(AWARE), the advantage of TRT over TMwas of a smaller magnitude:TM showed a bigdecline on AWARE (about 14 points everysix months or 42 points over 18 months),with TRT adding an additional four-pointdecline every six months, for a total of 54points decline over 18 months.

Repeated measures ANOVA wasperformed to evaluate outcomes of thepercentage annoyance (ANNOY) of thepatients’ tinnitus. The early improvement(first six months) in TM and TRT patients wasfairly similar. Over the long run (6 to 18

Figure 6. Summary of TM (n = 48) and TRT (n = 53) mean profiles for the repeated measures ANOVA. Stan-dard deviations are in parentheses.

months), however, TRT patients improvedmore than TM, although TM patients stillimproved modestly. Over the 18 months,ANNOY was reduced in TM patients by 26percentage points, but it was reduced in TRTpatients by an additional 14 percentage pointsfor a total of 40 points.

Screening for Study Patients

It has been observed that manyindividuals who experience chronic tinnitusare motivated to participate in any type ofstudy concerning tinnitus. Most estimates oftinnitus prevalence, however, indicate thatonly about 20 percent of individuals whoexperience chronic tinnitus have a clinicallysignificant condition that would warrantclinical intervention. For example, theAmerican Tinnitus Association reports thatapproximately 40–50 million individuals inthe United States experience chronic tinnitus(Davis and Refaie, 2000), but only 10–12million of these seek professional help, and2.5 million of these are debilitated by theirtinnitus. Thus, for the present study, it wascritical to screen for tinnitus severity theover 800 veterans who expressed interest inreceiving treatment for their tinnitus.

Patient screening was a multistageprocess that started with initial contact overthe telephone.The Tinnitus-Impact ScreeningInterview (Henry et al, 2004; and Appendix1) was developed for the study as the firstfilter to ensure that patients had a tinnituscondition that warranted treatment. Forscreening purposes, the questions are shortand concise with an emphasis on issues thatwould reflect the need for treatment. Thestudy audiologist who conducted most of thetelephone screening was trained to properlyinterpret veterans’ responses to the questionsand to make rapid decisions regarding theireligibility for study inclusion.

The screening process worked efficiently,as attested to by the relatively small numberof veterans who did not feel the need topursue treatment following the initialevaluation. Of the original 800 veterans whocalled about treatment, only 123 (15 percent)decided that they required treatment afterscreening.This information will be importantto other VA medical centers as theyimplement clinical management programsfor veterans with tinnitus. There is clearly aprocess that must be defined and documented

to categorize veterans according to their needfor different levels of treatment. Based onour findings, it appears that about 85 percentof veterans with the complaint of tinnitusmay only require some minimal informationto acquire an understanding that theirtinnitus does not pose a significant problemfor them. This information can usually beconveyed over the telephone or during a briefappointment.We intend to follow up on thesefindings to develop a structured screeningtechnique to appropriately respond to allveterans who report difficulties with chronictinnitus.

“Intervention” Provided at the InitialAppointment

The initial evaluation, which wascompleted with 171 veterans, required aboutfour hours to complete all testing and toadminister the TRT initial interview.Although the study audiologist was nottrained in tinnitus treatment techniques, shenonetheless could answer many of theveterans’ questions about their test resultsand about tinnitus in general.The session wasthus interactive, and the information providedto the veterans would equate to “educationalcounseling.” Results of audiologic testingprovided information to explain any hearingdifficulties.The tinnitus testing quantified thetinnitus perception and helped to provideanswers to many questions about tinnitus.Administering the TRT initial interviewinvolved up to an hour of interactive dialogue.

It has been our experience thatuninformed patients are often confused aboutthe differential effects of tinnitus, hearingloss, and sound tolerance. The initialinterview is effective in clarifying these issuesfor patients (Henry et al, 2002a; Henry et al,2003). Of the 171 veterans who completed theevaluation visit, 48 decided that they did notrequire any further treatment visits. Manyof these 48 individuals came to the realizationthat their hearing loss was their primaryproblem and that it was not caused by theirtinnitus. Moreover, their tinnitus was notenough of a problem to warrant any furtheraction. As with the telephone screening, theinitial evaluation thus provided all theintervention these patients needed to besatisfied that further tinnitus interventionwas not necessary.


124


125

Potential Confounding Variables

Although this was a prospective,controlled clinical trial, certain factors couldhave influenced the results.

Service Connection for Tinnitus

Veterans who incur or aggravate adisease or injury during their military servicecan apply for a disability award from VA (seeHenry et al, 2004). If the claim is approved,the veteran becomes “service connected” forthe disabling condition with an assignedrating between 0 and 100 percent. A ratingof at least 10 percent entitles the veteran toVA health-care services as well as monetarycompensation that increases with the percentrating.A 0 percent rating provides health-careservices only. Veterans can claim tinnitus asa service-connected disability, and approvedawards are rated routinely at 10 percent. Inour clinical trial, 38 (18 in TM and 20 inTRT) of the 123 veteran patients were serviceconnected for tinnitus. Although this mightbe considered a potentially confoundingvariable, the disability award is not based onthe severity of the condition but only on thepresence of tinnitus that was incurred oraggravated during military service.Therefore,status of service connection should not haveprovided any incentive for veterans to eitherbenefit or not benefit from treatment. In fact,outcomes between patients who were andwere not service connected for tinnitus did notshow any consistent differences across alloutcome measures.

Types of Ear-Level Devices

Although all patients were fitted withsome type of ear-level devices, the devicesused were different between groups. Withrespect to the sound generators andcombination instruments, only GeneralHearing Instruments (GHI) devices wereused with the TRT patients, and mostlyStarkey devices were used with the TMpatients. The TRT patients encounterednumerous difficulties with their GHI devices(GHI has provided assurance that theseproblems have been remedied), especiallywith the noncustom sound generators thatwere eventually abandoned in favor of customsound generators. The device problems thatwere encountered with the TRT patients

were resolved by the study audiologist.Therewere very few device problems for the TMpatients, and these were handled by the TMspecialist.

Use of Amplification with Ear-LevelDevices

An important distinction between thetwo treatment groups was the disparate useof ear-level amplification (hearing aids orcombination instruments). In the TM group,93 percent of the patients were fitted initiallywith amplification, compared to 36 percent ofthe TRT group (Table 1). For the TRT patients,sound stimulation solely with broadbandnoise was largely the preferred choice (61percent of the TRT patients were fitted withbilateral sound generators—two patientswere fitted with a hearing aid in one ear anda sound generator in the other).Amplificationwas clearly used much more often for theTM patients. One reason for this differenceis that the TM specialist’s clinical experiencehas demonstrated that marginal hearing aidcandidates (e.g., hearing thresholds of 25–35dB HL at 3 kHz, and 40–45 dB HL at 4 kHz)with tinnitus often do well with the additionof high-frequency amplification. The use ofsound generators alone was observed tofrustrate these patients and to exacerbatetheir marginal hearing handicap.

Use of Wideband Noise

Although some patients used hearingaids only, the majority of patients in eachgroup were fitted with ear-level devices thatproduced wideband noise.Table 1 shows that,by the end of the 18-month treatment protocol,46 patients (78%) in TM and 46 patients (72%)in TRT were using devices that producedwideband noise. As explained above (inProcedures), many differences existed betweentreatment groups regarding the prescribeduse of the noise. Although patients receivedspecific instructions as to how to utilize theirdevices, it cannot be known for certain exactlyhow the devices were used on a daily basis foreach patient. The TM patients were allowedto use their devices in any manner thatafforded the greatest tinnitus relief. It seemslikely that some of these patients would haveadjusted the noise output at a level “below themixing point” as for the TRT patients. If thedevices were adjusted this way and worn for

126


at least eight hours per day, then the deviceusage would have been the same as for theTRT patients. This issue points out the needfor the development of technology to monitordevice usage for each patient, with respect toboth hours per day of use and output levels.The availability of such technology wouldenable investigation of these device-usagevariables as factors affecting outcomes oftreatment.

Clinician Variables

There was only one treatment specialistfor each of the two treatment methods.Theseclinicians had in common their clinicalcertification as audiologists and theirspecialized expertise in the treatment oftinnitus. Any differences between clinicians,however, could have influenced outcomes.Unique characteristics of personality, attitude,and professional demeanor can affect apatient’s perception of the quality of carereceived (Isenberg, 1998; Jensen et al, 2005).The TM specialist was a full-time clinicianwho provided clinical audiology services tofive to seven patients per day. The TRTspecialist was a full-time clinical researcherwho only saw patients who were associatedwith his research (4–5 patients per week onaverage during this study). It is possible thatsuch a difference in patient workload couldhave created differences in either the qualityof service provided, the enthusiasm/energylevel of the provider, or the manner in whichthe clinician was perceived by the patient.

Clinical Contact Time

Each patient’s participation in this studyinvolved at least seven clinical appointmentsover an 18-month period. The appointmentschedule was the same for both groups, andall patients had about the same contact timewith the study audiologist. The studyaudiologist reviewed the writtenquestionnaires, administered the TRTinterviews, and performed the audiometricand tinnitus assessments. Other aspects ofthe study, however, resulted in differencesin the amount of patient-clinician contacttime (see Table 7).

The structured TRT counseling wasadministered to patients by the TRTtreatment specialist at every appointment(counseling time was generally shortened atlater appointments as patients became morefamiliar with the concepts). Counseling for theTM patients was not structured, and thesepatients received counseling only as deemednecessary by the TM audiologist (and as timepermitted).TM patients thus had less contacttime with their treatment specialist than didthe TRT patients. Between-groups differencesin counseling time were most pronounced atthe 3-, 6-, and 12-month appointments (Table7). Some of this differential was offset by theadded time required for the TM specialist toperform additional tinnitus tests and to selectand fit devices.

TRT patients had many more problemswith their ear-level devices than did the TMpatients, especially during their first sixmonths of treatment. The TRT patientsrequired an average of 30 minutes to correct

Table 7. Estimated Clinician Contact Time for Patients Who Completed This Study

Average Contact Time (hours)

Appointment TM TRT

Baseline evaluation Assessment and questionnaires (study audiologist) 4.00 4.00Determine treatment plan Device evaluation and earmold impressions (tinnitus specialist) 1.00 0.50Initial treatment Device fitting and counseling (tinnitus specialist) 1.50 1.503 month Questionnaires (study audiologist) 0.75 0.75

Treatment (tinnitus specialist) 0.33 1.006 month Assessment and questionnaires (study audiologist) 1.75 1.75



Treatment (tinnitus specialist) 0.33 0.50Extra for device problems 0.15 0.50

TOTAL 13.97 15.50


127

device problems, while TM patients requiredless than ten minutes on average (Table 7).The device problems experienced by the TRTpatients may have diminished their progressduring the first six months. On the otherhand, the additional contact time required tocorrect devices may have been an overallbenefit to the TRT patients.

Considering all of these differences, theTM patients received an average of almost 14hours of total clinician contact time whilethe TRT patients received about 15.5 hours.Stated another way, the TRT patients receivedapproximately ten percent more “intensity ofclinician interaction” than did the TMpatients.

Potential Sources of Systematic Error

There are three potential sources ofsystematic error that could affect the validityof results: (1) observer bias, (2) subject bias,and/or (3) instrument bias (Hulley andCummings, 1988; Johnson and Danhauer,2002). Observer bias could affect the resultsif there was a consistent distortion in thecollection, scoring, or reporting of the data.The potential for observer bias was minimizedin this study by ensuring that all data werecollected by the study audiologist who was notassociated with either of the treatmentmethods. This individual had no previoustinnitus treatment experience and did notreceive any specific training in TRT or TM forthe duration of the study. The two treatmentspecialists were restricted to providingtreatment only and were not involved in thedata collection process. The TRT specialistwas the study audiologist’s supervisor; thus,unconscious bias on the part of the studyaudiologist toward TRT was a possibility.

Subject bias refers to effects that canoccur if study patients consistently distorttheir responses to outcome measures, eitherconsciously or subconsciously (Johnson andDanhauer, 2002). The classic “Hawthorne”phenomenon (Roethlisberger and Dickson,1939) suggests that patients may performdifferently if they are aware that they areparticipating in an experiment. There is thepotential that patients in the present studywere affected by their knowledge of beinginvolved in an experiment, which could havecaused some of them to alter their responsesto the outcome questions. This effect can beexpanded to apply to any patient who is

receiving any form of treatment fromprofessionals. Regardless of the method used,nonspecific therapeutic factors (e.g., effects ofattention and positive expectancies) mightoccur as a function of the amount of contacttime (“dose”) between the patient and aclinician who is perceived by the patient asbeing an “expert.” Isenberg (1998) has listednonspecific effects that can influence resultsin outcomes studies. Those that are mostgermane to this study are the placebo effect(beneficial effect caused by positiveexpectation) and the “halo” effect (beneficialeffect caused by manner, attention, and caringof clinician). These nonspecific benefits couldhave occurred for both groups in the presentstudy, with potentially greater nonspecificbenefit occurring for the TRT patients whoreceived the greater amount of cliniciancontact time (Table 7).

Instrument bias refers to anymeasurement distortions that could takeplace as a result of faulty functioning ofinstruments used to measure outcome results.An example of instrument bias would be anaudiometer that is out of calibration thatwould systematically alter all measurementsobtained with the audiometer. The tinnitusquestionnaires are subjective measurementinstruments, and they might be subject to thesame types of distortions.The use of multipleoutcome instruments in the present studyserved to confirm the results that wereobtained with any single instrument.Instrument bias was therefore not aconfounding factor in this study.

Modification of Treatment Protocolfor TM

Treatment of tinnitus using the methodof TM often involves only a single treatmentappointment (Henry et al, 2002b). At theappointment, patients are fitted with ear-level devices and are instructed as to theirproper use. Counseling is always provided,although a specific counseling protocol has notbeen described. Vernon (1987) advisedclinicians who dispense maskers to contacttheir patients at six months and one yearpostfitting—to check on the proper use ofmaskers and to obtain information concerningtreatment efficacy. TM patients are alsoencouraged to return to the clinic or totelephone the clinician, but only if necessaryto resolve problems or to answer questions.


128

For the present study, patients in both groupsattended ongoing treatment appointmentson the same schedule: 3, 6, 12, and 18 months.This schedule was determined by theappointment schedule that is typically usedwith TRT. The usual TM protocol of a singletreatment appointment was thereforemodified so that all patients in this studyreceived treatment on the same schedule.

Cost/Benefit Considerations

It is of interest to clinicians to considerthese results from a cost/benefit perspective.Cost in this case refers to hours of contacttime since the monetary cost of ear-leveldevices is about equal for TM and TRT. Theassessment of benefit can be based on any ofthe outcomes instruments. For illustrativepurposes we will use the TSI since more TSIquestionnaires were completed at each of thetime points than for the other measures.

Using the TSI (which provides an indexscore between 0 and 48), the TM patientsimproved by an average of 4.6 points whenthey were first evaluated 3 months afterinitial treatment. Based on the contact timerequired to perform the evaluation and initialtreatment for TM (6.5 hours—see Table 7),each hour of contact time resulted in a .7-pointaverage improvement in the TSI (asmeasured at 3 months). This samecomputation done for the TRT patients showsa .4-point improvement in the TSI per contacthour following initial treatment. Cost/benefitassessed at 3 months thus shows anadditional .3-point improvement per contacthour for the TM patients. When this analysisis done for 6 months, there is a .6-pointaverage improvement, per contact hour, forboth TM and TRT. Thus, cost/benefit is seenas equivalent for TM and TRT patients at 6months. For 12 months, the average TSIimprovement per contact hour for TM is .5,and .9 for TRT. By 18 months, the TSIimprovement per contact hour for TM is .4,and 1.1 for TRT. Thus, there was a steadydecrease in cost/benefit for TM compared toa steady improvement for TRT at the 12-and 18-month time points.

This cursory analysis suggests thepossibility that cost/benefit is better in theshort term for TM compared to TRT. Theremay be no difference in cost/benefit betweengroups after about 3 to 6 months of treatment.

If treatment is continued beyond 6 months,however, cost/benefit may improveincrementally for TRT compared to TM.

Outcome Instruments

Ideally, there should be a standardizedtreatment-outcome instrument that wouldaccurately and reliably reflect the severity ofa patient’s tinnitus condition, and that couldbe used to assess the efficacy of treatmentover time. There are many instruments thathave been developed for this purpose (mostlyself-assessment questionnaires), but no singleinstrument has achieved universalacceptance. We chose to employ threedocumented and commonly used tinnitusoutcome instruments, the Tinnitus HandicapInventory (Newman et al, 1996; Newman etal, 1998), Tinnitus Handicap Questionnaire(Kuk et al, 1990; Tyler, 1993), and TinnitusSeverity Index (Meikle et al, 1995). Thepresent report provides overall results ofeach instrument, revealing that theseinstruments functioned similarly in providingconsistent global measures of tinnitusseverity. In addition, use of the verballyadministered TRT interview (Jastreboff andJastreboff, 1999; Henry et al, 2003) wasrequired for performing treatment with TRT.To enable a direct comparison of treatmentoutcomes using the TRT interview questions,the TRT interview was also used with the TMpatients.

The completion of all four of theseassessment instruments at each visit affordsthe opportunity to compare the instrumentswith regard to content, patient responses,and sensitivity. An important analysis willthus involve the detailed comparison oftreatment outcomes from these differentinstruments that have each been developedindependently. An in-depth analysis of theseinstruments and their relative outcomes fromthis study is presently underway, and a follow-up publication is planned to report results ofthat analysis.

Factors Associated with ProblematicTinnitus

To improve clinical services for tinnitusmanagement, it is important to understandthe factors that dictate why so many peoplewith permanent tinnitus seem to be


129

unaffected by it, while a relatively smallpercentage require clinical intervention.These factors most likely consist of somecombination of personality traits, lifecircumstances, and characteristics of thetinnitus sound (Henry et al, 2005a). In thepresent study, we administered an extensiveset of questionnaires at the initial visit toprofile individuals with regard to personalityand life-experience factors that couldinfluence the impact of tinnitus on their lives.Additionally, each patient’s tinnitusperception was carefully matched usingsophisticated instrumentation andtechniques. A comprehensive analysis willbe conducted to determine how these differentfactors might be associated with differentlevels of treatment efficacy, and results ofthat analysis will be presented in a futurepublication.

Conclusions

This prospective study involved acomparison of both TM and TRT in acontrolled fashion and employing severaloutcome measurements to increase thevalidity of the findings. No prior study of thistype has been conducted.The Veterans HealthAdministration is increasingly turning toevidence-based treatment methods (Feussner,1998), and this study is expected to provideresearch evidence that will support theprovision of effective treatment for tinnitusat VA medical centers.

1. The results of this study suggest thatboth TRT and TM are effectivetherapies for amelioration of tinnitusand may be more effective if patientsare staged into different treatmentgroups. In particular, it appearedthat (a) when patients have serious(high degrees of) difficulty with theirtinnitus, they may benefit most fromTRT, which will help them reach the(relatively) lowest levels of handicapand severity; (b) however, when thetinnitus difficulty being experiencedis only moderate, TM may be a moreefficient treatment because it issimpler and the appointments canbe shorter, but it can still be nearlyas effective as TRT for these patients.

2. Significant improvements for eachof the treatment groups were noted,

but in varying degrees. Additionally,the time course of improvementappeared distinctly different. Ingeneral, for the TM group, most of theimprovement was observed in thefirst 3 to 6 months of treatment andappeared to level off over the courseof 18 months. For the TRT group,however, there appeared to be asteady improvement over time withthe greatest level of improvementnoted in the latter stages oftreatment. This observation is ingood agreement with the Jastreboffmodel, which predicts and assertsthat benefit of TRT treatment shouldcontinue to increase over time. Thepresent study showed that the TRTgroup showed substantial improvementbetween 12 and 18 months.The studyprotocol required that treatment endfor all patients at 18 months; thus, itis unknown if further treatmentmight have produced further benefit.One clinical study has reported thatpatients treated with TRT continuedto improve after the cessation offormal treatment, and that sustainedbenefit was observed after five years(Lux-Wellenhof and Hellweg, 2002).

3. Although longer duration of tinnituswas associated with higher THQ,TSI, and AWARE scores (reflectinggreater tinnitus severity), at baseline,duration was not associated withimprovement on these measures orwith differential TM versus TRTtreatment effectiveness. Likewise,the level of hearing loss was notassociated with improvement ordifferential treatment effectiveness.Thus it appears that TM and TRTcan be recommended, regardless ofthe patient’s duration of tinnitus orhearing loss at the intake assessment.

4. There was good agreement betweenthe written (self-administered)tinnitus instruments in assessingthe initial and final degree of thetinnitus problem for our studypatients. Results of individualquestions from the TRT interview,however, did not provide the samelevel of agreement with the writtenquestionnaires. Because the THI,THQ, and TSI were multiitem


130

measures, they may have been moresensitive than AWARE and ANNOYin detecting differential treatmenteffectiveness depending on the extentof a patient’s tinnitus problem atbaseline. The AWARE and ANNOYitems from the TRT interviews,however, were very sensitive indetecting changes in the group as awhole.

5. This study is one of the first intinnitus and audiology research toapply multilevel modeling to comparetreatment groups and determinepredictors of variation in patienttrajectories. The many advantagesthat MLM affords may be ones thatother researchers will want toconsider.

6. One limitation of this study is thatthere was no control group thatreceived comparable attention andtime to determine if these nonspecificvariables might account for similarimprovements in outcomemeasurements. The study was alsolimited by having only two providers.At the present time there is a studyunderway at four VA medical centersto attempt to address these issues.The study will evaluate results ofseveral providers who areadministering TM, TRT, and“Audiologic Tinnitus Management”(standard audiologic care for tinnitus)(Henry et al, 2005c, 2005d).

Acknowledgments. The authors acknowledge sup-port from the Veterans Health Administration andVeterans Affairs Rehabilitation Research andDevelopment (RR&D) Service (C2887R and C3214R).Special thanks go to Sara Ruth Oliver, Au.D., Chief,Speech Pathology and Audiology Service, PortlandVA Medical Center, who provided the clinical facili-ties to conduct this study. Thanks also to MarciaCollins, Stephen Fausti, Ph.D., and Kimberly Owens,B.S., for their contributions to this study.

REFERENCES

Axelsson A. (1998) How to evaluate treatments fortinnitus. In: Tinnitus Treatment and Relief. VernonJA, ed. Needham Heights, MA: Allyn and Bacon, 1–7.

Axelsson A, Barrenas M-L. (1992) Tinnitus in noise-induced hearing loss. In: Noise-Induced Hearing Loss.Dancer AL, Henderson D, Salvi RJ, Hamernik RP,eds. St. Louis: Mosby-Year Book, 269–276.

Davis A, Refaie AE. (2000) Epidemiology of tinnitus.In: Tinnitus Handbook. Tyler R, ed. San Diego:Singular Publishing Group, 1–23.

Dobie RA. (1999) A review of randomized clinical trialsin tinnitus. Laryngoscope 109:1202–1211.

Dobie RA. (2002) Randomized clinical trials for tin-nitus: not the last word? In: VIIth InternationalTinnitus Seminar Proceedings. Patuzzi R, ed. Perth:Physiology Department, University of WesternAustralia, 3–6.

Dobie RA. (2004) Clinical trials and drug therapy fortinnitus. In:Tinnitus: Theory and Management. SnowJB, ed. Lewiston, NY: BC Decker, 266–277.

Erlandsson SI, Hallberg LRM, Axelsson A. (1992)Psychological and audiological correlates of perceivedtinnitus severity. Audiology 31:168–179.

Feussner J. (1998) Clinical research in the Departmentof Veterans Affairs: using research to improve patientoutcomes. J Invest Med 46:264–267.

Guyatt GG, Sackett DL, Cook DA. (1993) Users’ guidesto the medical literature, II: how to use an articleabout therapy or prevention. Part A: are the resultsof the study valid? Evidence-based medicine work-ing group. J Am Med Assoc 270:2598–2601.

Henry JA. (2004) Audiologic assessment. In: Tinnitus:Theory and Management. Snow JB, ed. Lewiston, NY:BC Decker, 220–236.

Henry JA, Dennis K, Schechter MA. (2005a) Generalreview of tinnitus: prevalence, mechanisms, effects,and management. J Speech Lang Hear Res48:1204–1234.

Henry JA, Jastreboff MM, Jastreboff PJ, SchechterMA, Fausti SA. (2002a) Assessment of patients fortreatment with tinnitus retraining therapy. J AmAcad Audiol 13:523–544.

Henry JA, Jastreboff MM, Jastreboff PJ, SchechterMA, Fausti SA. (2003) Guide to conducting TinnitusRetraining Therapy (TRT) Initial and Follow-upInterviews. J Rehabil Res Dev 40:157–178.

Henry JA, Schechter MA, Loovis C, Zaugg TL, KaelinC, Montero M. (2005b) Clinical management of tin-nitus using a “progressive intervention” approach. JRehabil Res Dev 42(4, Suppl. 2):95–116.

Henry JA, Schechter MA, Nagler SM, Fausti SA.(2002b) Comparison of tinnitus masking and tinni-tus retraining therapy. J Am Acad Audiol 13:559–581.

Henry JA, Schechter MA, Regelein RT, Dennis KC.(2004) Veterans and tinnitus. In: Tinnitus: Theoryand Management. Snow JB, ed. Lewiston, NY: BCDecker, 337–355.

Henry JA, Zaugg TL, Schechter MA. (2005c) Clinicalguide for audiologic tinnitus management I: assess-ment. Am J Audiol 14:21–48.

Henry JA, Zaugg TL, Schechter MA. (2005d) Clinicalguide for audiologic tinnitus management II: treat-ment. Am J Audiol 14:49–70.

Hulley SB, Cummings SR. (1988) Planning the meas-urements: precision and accuracy. In: Designing


131

Clinical Research. Hulley SB, Cummings SR, eds.Baltimore: Williams and Wilkins, 31–52.

Isenberg SF. (1998) Managed Care, Outcomes, andQuality: A Practical Guide. New York: Thieme.

Jastreboff PJ. (1994) Instrumentation and tinnitus:a neurophysiological approach. Hear Instrum 45:7–31.

Jastreboff PJ. (2000) Tinnitus habituation therapy(THT) and tinnitus retraining therapy (TRT). In:Tinnitus Handbook. Tyler RS, ed. San Diego: SingularPublishing Group, 357–376.

Jastreboff PJ, Gray WC, Gold SL. (1996)Neurophysiological approach to tinnitus patients. AmJ Otol 17:236–240.

Jastreboff MM, Jastreboff PJ. (1999) Questionnairesfor assessment of the patients and their outcomes.In: Proceedings of the Sixth International TinnitusSeminar 1999. Hazell JWP, ed. London: Tinnitus andHyperacusis Centre, 487–491.

Jastreboff PJ, Hazell JWP. (1998) Treatment of tin-nitus based on a neurophysiological model. In:Tinnitus Treatment and Relief. Vernon JA, ed.Needham Heights: Allyn and Bacon, 201–217.

Jastreboff PJ, Hazell JWP. (2004) Tinnitus RetrainingTherapy: Implementing the Neurophysiological Model.New York: Cambridge University Press.

Jastreboff PJ, Jastreboff MM. (2000) TinnitusRetraining Therapy (TRT) as a method for treatmentof tinnitus and hyperacusis patients. J Am Acad Audiol11:162–177.

Jensen PS, Weersing R, Hoagwood KE, Goldman E.(2005) What is the evidence for evidence-based treat-ments? A hard look at our soft underbelly. Ment HealthServ Res 7:53–74.

Johnson RM. (1998) The masking of tinnitus. In:Tinnitus Treatment and Relief. Vernon JA, ed.Needham Heights, MA: Allyn and Bacon, 164–186.

Johnson CE, Danhauer JL. (2002) Handbook ofOutcomes Measurement in Audiology. Clifton Park,NY: Delmar Learning.

Kuk FK, Tyler RS, Russell D, Jordan H. (1990) Thepsychometric properties of a tinnitus handicap ques-tionnaire. Ear Hear 11:434–445.

Lux-Wellenhof G, Hellweg FC. (2002) Longterm follow-up study of TRT in Frankfurt. In: Proceedings of theSeventh International Tinnitus Seminar. Patuzzi R,ed. Crawley, Western Australia: PhysiologyDepartment, University of Western Australia,277–279.

Management Decision and Research Center,Department of Veterans Affairs. (1998) ClinicalPractice Guidelines. Boston: Management Decisionand Research Center; Washington, DC: VA HealthServices Research and Development Service in col-laboration with Association for Health ServicesResearch.

Meikle MB, Griest SE, Stewart BJ, Press LS. (1995)Measuring the negative impact of tinnitus: a briefseverity index. In: Midwinter Meeting: Association for

Research in Otolaryngology. Ryan A, ed. Des Moines:Association for Research in Otolaryngology, 167.

Newman CW, Jacobson GP, Spitzer JB. (1996)Development of the Tinnitus Handicap Inventory.Arch Otolaryngol Head Neck Surg 122:143–148.

Newman CW, Sandridge SA, Jacobson GP. (1998)Psychometric adequacy of the Tinnitus HandicapInventory (THI) for evaluating treatment outcome.J Am Acad Audiol 9:153–160.

Newman CW,Wharton JA, Jacobson GP. (1995) Reteststability of the tinnitus handicap questionnaire. AnnOtol Rhinol Laryngol 104:718–723.

Penner MJ, Bilger RC. (1995) Psychophysical obser-vations and the origin of tinnitus. In: Mechanisms ofTinnitus. Vernon JA, Moller AR, eds. NeedhamHeights, MA: Allyn and Bacon, 219–230.

Raudenbush SW, Bryk AS. (2002) Hierarchical LinearModels: Applications and Data Analysis Methods.Thousand Oaks, CA: Sage Publications.

Raudenbush SW, Bryk A, Cheong YF, Congdon R.(2004) HLM 6: Hierarchical Linear and NonlinearModeling. Lincolnwood, IL: SSI Scientific SoftwareInternational.

Roethlisberger F, Dickson W. (1939) Management andthe Worker. Cambridge: Harvard University Press.

Schafer JL, Graham JW. (2002) Missing data: ourview of the state of the art. Psychol Methods 7:147–177.

Schechter MA, Henry JA. (2002) Assessment andtreatment of tinnitus patients using a “maskingapproach.” J Am Acad Audiol 13:545–558.

Tyler RS. (1993) Tinnitus disability and handicapquestionnaires. Semin Hear 14:377–383.

Vernon J. (1987) The common errors of masking forrelief of tinnitus. In: Proceedings III InternationalTinnitus Seminar, Münster. Feldmann H, ed.Karlsruhe: Harsch Verlag, 229–238.

Vernon JA. (1988) Current use of masking for therelief of tinnitus. In: Tinnitus. Pathophysiology andManagement. Kitahara M, ed. Tokyo: Igaku-Shoin,96–106.

Vernon J, Griest S, Press L. (1990) Attributes of tin-nitus and the acceptance of masking. Am JOtolaryngol 11:44–50.

Veterans Benefits Administration, Department ofVeterans Affairs. (2002) Veterans BenefitsAdministration Annual Report, 2001. Washington,DC: Department of Veterans Affairs.


132

APPENDIX 1

TINNITUS-IMPACT SCREENING INTERVIEW

1. Do you have tinnitus that is constant? (i.e., it’s usually or always there, whetheryou are consciously aware of it or not) YES NO

2. How much has tinnitus annoyed you, on average, over the last month? (“0”would be “not annoying at all”; “10” would be “as annoying as you can imagine.”)0 1 2 3 4 5 6 7 8 9 10

3. How much did tinnitus affect or impact your life, on average, over the lastmonth? (“0” would be “not at all”; “10” would be “as much as you can imagine.”)0 1 2 3 4 5 6 7 8 9 10

4. Does tinnitus affect your sleep? Always/Often/Sometimes/Never

5. Does tinnitus affect your concentration? Always/Often/Sometimes/Never

6. If your tinnitus is a problem for you, what is the major reason your tinnitus isa problem?

Note: The TISI has been revised since this version.

Outcomes of Clinical Trial: Tinnitus Masking versus ...€¦ · efficacy of tinnitus masking (TM)...

Documents

Transcript of Outcomes of Clinical Trial: Tinnitus Masking versus ...€¦ · efficacy of tinnitus masking (TM)...