Socio-Demographic, Health, and Tinnitus Related Variables Affecting Tinnitus Severity Carlijn E. L. Hoekstra,1,2 Francina M. Wesdorp,1 and Gijsbert A. van Zanten1,2 Objectives: Tinnitus is a highly prevalent symptom with potential severe morbidity. Fortunately, only a small proportion of the population experience problems due to their tinnitus in such a degree that it adversely affects their quality of life (clinically significant tinnitus). It is not known why these individuals develop more burden from tinnitus. It seems likely that the severity of tinnitus can be influenced by different factors, such as socio-demographic or tinnitus characteristics or additional health complaints. It remains unclear from the current literature as to what are the main independent variables that have a bearing on tinnitus severity. This study addresses this problem by investigating variables previously described in the literature as well as additional variables. The aim of this study is to identify sociodemographic, health, and tinnitus variables that independently relate to tinnitus severity the most.
INTRODUCTION Tinnitus is a phantom auditory perception of meaningless sound in the absence of an external or internal acoustic stimulus. It is a common problem that affects 7 to 19% of the adult population (Chung et al. 1984; Coles 1984; Axelsson & Ringdahl 1989; Davis 1989; Nondahl et al. 2002). The majority of these individuals do not experience the tinnitus as problematic, but in up to 5% of the adult population it does lead to annoyance or even interferes with the ability to lead a normal life (Coles 1984; Davis & Rafaie 2000; Nondahl et al. 2002). Many recommendations have been made to recognize these distressed patients or to assess which patients are in need of further counseling (Tyler & Erlandsson 2003; Henry et al. 2005; Cima et al. 2012). It is difficult though, to identify beforehand individuals who may develop problems due to their tinnitus to such a degree that it may negatively affect their quality of life (clinically significant tinnitus). Knowledge about factors that are associated with tinnitus severity could be useful to identify these individuals who may develop problems. If these factors are present they could support an indication for follow-up of these patients to monitor them for development of potential distress. In addition, the knowledge about factors associated with tinnitus severity might provide a useful tool in counseling or psychological therapies and, therefore, in the effectiveness of the care provided. As a next step, the question would arise whether such factors associated with tinnitus severity would also be related to treatment outcomes. It seems likely that the severity of tinnitus can be influenced by different factors, such as socio-demographic, personality, psychological or tinnitus characteristics, or additional health complaints. Many studies have reported variables related to tinnitus severity (Table 1). Among these studies many relationships have only been examined once, contradicting results are reported across studies, and reliable multivariate analyses are scarce. Combined, the 18 studies shown in Table 1 identified 28 variables significantly related to tinnitus severity and 35 variables without such a relationship. Of these 28 variables, only three variables (continuous tinnitus, sleep problems, and vertigo) have been shown to significantly relate to severity in more than one study. Thirteen variables are described to both having and not having a significant relationship to severity (age, anxiety, depression, etiology, sex, hearing loss, location, loudness, loudness variability, maskability, pitch, [psycho]somatic complaints, and tinnitus type). The remaining 12 variables were reportedly significantly related to tinnitus in a single study only, namely, age at onset, avoidance, awareness, change since onset, chronic pain, controllability, hyperacusis, influence of hearing aid, internal locus of control, personality, tolerance, and various questions on quality of life. An important drawback of the studies performed so far is that most do not include a multivariate analysis. Multivariate
Design: This is a retrospective cohort study performed at the Tinnitus Care Group of the University Medical Center, Utrecht, in 309 consecutively seen chronic tinnitus patients. At this care group, patients are examined according to a structured diagnostic protocol, including history-taking by an otorhinolaryngologist and audiologist, physical examination, and audiometry. Based on results from previous research and theoretical considerations, a subset of data acquired through this diagnostic protocol were selected and used in this study. Univariate and multivariate correlations with tinnitus severity were investigated for 28 socio-demographic, health, and tinnitus variables. Tinnitus severity was measured with the Tinnitus Questionnaire (TQ) and the Tinnitus Handicap Inventory (THI). Results: Eighteen variables related univariately with the TQ and 16 variables related univariately with the THI. Among these, 14 variables related univariately with both the TQ and the THI. Multivariate analyses showed three variables with an independent significant effect on both the TQ and the THI: percentage of tinnitus awareness during the day, self-reported depression and/or anxiety, and loudness on a Visual Analogue Scale. Three additional variables contributed independently significantly to the explained variance in either the TQ or the THI: level of education, somatic complaints, and tinnitus variability in loudness and/or pitch on a Visual Analogue Scale. These variables were among the first not to reach significance on the other questionnaire. Conclusions: Tinnitus severity is shown to be strongly related to percentage of tinnitus awareness during the day, self-reported depression and/or anxiety, subjectively experienced loudness, level of education, existence of additional somatic complaints, and subjectively experienced variability in loudness and/or pitch. Further research is needed to investigate the causal relationship between these variables and tinnitus severity. This knowledge may contribute to a better recognition, follow-up, and/or counseling of more vulnerable patients. Key words: Chronic tinnitus, Severity. (Ear & Hearing 2014;35;544–554)
Department of Otorhinolaryngology and Head & Neck Surgery, University Medical Center Utrecht, Utrecht, The Netherlands; 2Rudolf Magnus Institute of Neuroscience, University Medical Center Utrecht, Utrecht, The Netherlands; and 3Faculty of Medicine, University Medical Center Utrecht, Utrecht, The Netherlands. 1
0196/0202/2014/355-0544/0 • Ear & Hearing • Copyright © 2014 by Lippincott Williams & Wilkins • Printed in the U.S.A. 544
5 5
11
112
72
166
6
71
136 68 755
Ooms et al. 2011 Pinto et al. 2010 Schlee et al. 2011
3 3 1
3 5 3
531 281 100
Mazurek et al. 2010 Meric et al.1998 Mondelli & da Rocha 2011 Ooms et al. 2012
13*
127
10
4
2378
4995
3
19
146
39
Holgers et al. 2005
Hiller & Goebel 2006
Andersson et al. 2001 Axelsson & Ringdahl 1989 Halford & Anderson 1991 Henry & Wilson 1995 Hiller & Goebel 1999
Andersson et al. 1999
No. Patients Included
No. Variables Analyzed
THI THI Mini-TQ
THI
TQ TRQ, THQ STSS THI
Own criteria for “tinnitus sufferer”
Mini-TQ
STI global score
TRQ
STSS
Likert scale
TRQ
K&L grading system
Outcome Parameter for Tinnitus Severity
Pitch (A), loudness (A), BDI-II Sex, age, hearing loss
Pitch (A), loudness (A)
Sex, age, duration, hearing loss, BMI, marital status, educational level, income level, professional status, alcohol, tobacco Sex Sex, age, duration Sex, age, hearing loss
Duration, familial status
Pitch (A), minimal masking level, ATQ Location, tinnitus type, pitch (S), pulsatile character, onset, nine types of etiology
Sex, age, variability
Sex
Hearing loss, loudness (A), duration, continuous loudness, influence of background sound, somatic problems, influence of substance or treatment, possibility to do something to ease, situations when less problematic, irritation, psychological factors, anxiety, depression, concentration difficulties Hearing loss, loudness (A)
Variables Not Related to Severity
STAI-state, STAI-trait, somatic anxiety — — Age at onset
Hearing loss, location Hearing loss, MMPI —
Hearing loss, continuous tinnitus, maskability, progressive loudness, SCL90-R scales, two types of etiology (sudden hearing loss, craniomandibular dysfunction) Sex, age, hearing loss, progressive onset, location, continuous tinnitus, vertigo, hyperacusis Various questions on NHP, mental disorders (mostly anxiety)
Loudness (A), BDI
Advancing age (females), hearing loss, sleep problems STAI-trait, DTQ
Maskability
Pitch (A), minimal masking level, noise sensitivity, change since onset, avoidance, tolerance
Variables Related to Severity
Table 1. Characteristics of studies assessing variables related to tinnitus severity (excluding various tinnitus questionnaires as variables)
Not performed Not performed Not performed
Not performed
Not performed Not performed Not performed
(Continued )
Three questions on NHP (find it hard to reach for things, finding live not worth living, bad sleep)
Not performed
Not performed
BDI
Not performed
Not performed
Not performed
Audiometric pitch, minimal masking level, avoidance, tolerance
Variables Related to Severity in Multiple Regression Analyses
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554 545
546
* Thirteen variables were analyzed, of which of four variables the individual components (different questions within a questionnaire or interview or different average hearing levels) were also analyzed separately. (A), audiometric; ATQ, Automatic Thoughts Questionnaire; BDI, Beck Depression Index; BMI, body mass index; DTQ, Depressive Tendency Questionnaire; K&L grading system, Klockhoff and Lindblom grading system; MMPI, Minnesota Multiphasic Personality Inventory; NHP, Nottingham Health Profile; (S), subjective; SCL-90-R, Symptom Checklist-90-Revised; STAI, State-Trait Anxiety Inventory; STI, Structured Tinnitus Interview; STSS, Subjective Tinnitus Severity Scale; THI, Tinnitus Handicap Inventory; THQ, Tinnitus Handicap Questionnaire; TQ, Tinnitus Questionnaire; TRQ, Tinnitus Reaction Questionnaire.
Binaural/central tinnitus, influence of hearing aid, sleep problems, chronic pain, depression, anxiety, somatic symptom severity 4705
15
Mini-TQ
Pitch (S), duration, perception as illness, comorbidity Sex, age, hearing loss, duration, onset, somatic comorbidity Unterrainer et al. 2003 WallhäusserFranke et al. 2012
149
6
THI
Depression, loudness (S), internal locus of control Awareness, location, influence of hearing aid, vertigo, chronic pain, sleep problems, depression, anxiety, somatic symptom severity
Tinnitus type, four psychosomatic complaints (depression, lack of concentration, insomnia, balance disturbance), maskability, presence, controllability, loudness variation Not performed Tinnitus type, nine psychosomatic complaints, maskability, presence, controllability, loudness variation Duration, onset, location, hearing loss Likert scale 18 3372 Scott et al. 1990
Table 1. Continued.
No. Patients Included
No. Variables Analyzed
Outcome Parameter for Tinnitus Severity
Variables Not Related to Severity
Variables Related to Severity
Variables Related to Severity in Multiple Regression Analyses
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
analyses are essential in finding uniquely related variables because these take into account the interdependencies between different variables. Variables that have been found in studies only performing univariate analyses could be based on unknown confounding effects. Eleven of these studies evaluated only a small number of potential variables and were thus not able to perform a contributory multivariate analysis (Axelsson & Ringdahl 1989; Halford & Anderson 1991; Henry & Wilson 1995; Meric et al. 1998; Andersson et al. 2001; Mazurek et al. 2010; Pinto et al. 2010; Mondelli & da Rocha 2011; Schlee et al. 2011; Ooms et al. 2011; Ooms et al. 2012). Three studies that could have performed a multivariate analysis did not include this analysis (Hiller & Goebel 1999; Unterrainer et al. 2003; Hiller & Goebel 2006). Another drawback of previous studies is that the included number of participants is not always in relation to the examined number of variables. As a rule of thumb at least 10 participants are needed per variable in a model for the analysis to be reliable (Peduzzi et al. 1996). Unfortunately, of the five studies including a multivariate analysis, two studies did not reach the 1:10 ratio (Andersson et al. 1999; Holgers et al. 2005). The three studies including a reliable multivariate analysis show 13 variables related to tinnitus severity, of which only depression, (psycho) somatic complaints, and sleep problems were shown in two studies to relate to severity independently (Scott et al. 1990; Henry & Wilson 1995; Wallhäusser-Franke et al. 2012). In sum, although much has been published about variables related to severity, it remains unclear from the current literature as to what are the most important and independent variables related to tinnitus severity. Our study addresses this problem by investigating variables previously described in the literature as well as additional variables, through univariate and multivariate analyses in a large enough sample size for the results to be considered reliable. The aim of this study is to find socio-demographic, health, and tinnitus variables that independently affect tinnitus severity. This study focuses on these three groups of variables because these variables are commonly evaluated in a standard outpatient visit. The results of this study may give clinicians a convenient number of risk variables for (chance for) more severe tinnitus that can easily and quickly be evaluated in the medical office or through audiometry. In addition to knowing risk variables for more severe tinnitus, clinicians will also have knowledge of which variables that are of less interest.
MATERIALS AND METHODS Study Design and Participants This study was performed at the Tinnitus Care Group of the Otorhinolaryngology department of the University Medical Center, Utrecht, as a retrospective cohort study. Patients with chronic tinnitus, defined as more than 2 months in duration, were examined according to a structured diagnostic protocol. Through this diagnostic protocol patients were evaluated by an otorhinolaryngologist, an audiologist, and a psychologist. Audiometry was performed and patients filled out various tinnitus questionnaires, Visual Analogue Scales (VASs) and psychological questionnaires (including the Symptom Checklist-90-Revised [SCL-90-R]). All patient data collected through this protocol were anonymized and entered in an Access database. For this retrospective study, a selection of these routinely collected data (see Independent Variables) from all consecutively
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
547
seen chronic tinnitus patients between June 2007 and November 2012 was included. The study was performed in accordance with the Declaration of Helsinki. Because this is a retrospective study with anonymized data, exemption for a full review from the Local Research Ethics Committee was obtained (12–611/C).
Sample Size
Dependent Variable (Outcome Measure)
Data Analysis
As part of their evaluation by the Tinnitus Care Group patients completed two questionnaires to measure tinnitus severity: the Tinnitus Questionnaire (TQ) and the Tinnitus Handicap Inventory (THI). The TQ is a 52 item self-response questionnaire with three answer possibilities (not true, partly true, true) and a total score ranging from 0 to 84 (Hallam et al. 1988). The THI is a 25-item self-response questionnaire with three answer possibilities (no, sometimes, yes) and a score range of 0 to 100 (Newman et al. 1996). For both the TQ and the THI higher scores account for more severe tinnitus. Since there is no standard definition or measurement for tinnitus severity, both questionnaires are included in this study. This allows to find potential similarities or differences in the results that might arise from using different ways to measure severity.
Statistical analyses were conducted with SPSS 20.0 (IBM, Armonk, NY). Descriptive analyses were calculated for the socio-demographic, health, and tinnitus variables. Univariate effects on tinnitus severity (TQ and THI) were explored through independent-sample t tests for dichotomous variables and through analysis of variance for ordinal variables. Effect sizes were computed using Cohen’s d to examine clinical significance of differences found between groups. Effect sizes ≥0.80 reflected a large clinically relevant difference, those between 0.50 and 0.80 were considered moderately large, and those between 0.20 and 0.49 were considered small (Cohen 1988). Bonferroni post hoc tests were executed as part of the analysis of variance to assess which groups differed significantly. For continuous variables, bivariate correlations were calculated (Pearson correlation for normally distributed data and Spearman correlations for abnormally distributed data). Correlation coefficients 0.50 strong (Cohen et al. 2003). Socio-demographic, health, and tinnitus variables showing a univariate significant relationship with tinnitus severity were entered into two separate stepwise multiple regression analyses to examine unique predictive effects. The first analysis used the TQ as dependent variable, the second analysis used the THI as dependent variable, and p values ≤0.05 were defined as statistically significant.
Independent Variables This study had the aim to investigate socio-demographic, health, and tinnitus variables that could easily be assessed by clinicians in the medical office or by audiometry. Of the 28 variables previously described to possibly be related to tinnitus severity 18 variables were included in this study because they fulfilled these criteria. An additional ten variables were included in the study because they could theoretically be associated with tinnitus severity, were only once or not investigated before or because they form important tinnitus characterizations. This resulted in a total of 28 variables. Table 2 shows an overview of the included variables, if applicable a short explanation, and the assessment method for each variable. For four variables (hyperacusis, masking, loudness, and pitch) multiple measurements were included because they can be measured in different ways and have been measured differently in earlier studies. Pure-tone audiometry and tinnitus analysis (pitch and loudness matching, measurement of minimal masking level, and complete and/or partial residual inhibition) were performed by one of two audiology assistants trained in tinnitus analysis. For the tinnitus analyses tones were used in the analysis for tonal tinnitus and a narrow-band noise was used for noise-like tinnitus. Potential octave ambiguity was tested as part of tinnitus pitch matching. Tinnitus loudness was matched at the pitch-matched frequency. When a pitch match could not be obtained, the loudness was matched with a 1000-Hz pure tone. Minimal masking was tested at the tinnitus frequency. Residual inhibition was tested 10 dB above the minimal masking level. Uncomfortable loudness was measured at 0.5, 1, 2, and 4 kHz by increasing the sound level by 5-dB each time until the patient reported that the loudness was uncomfortable. Testing was done in a soundproof cabin with Telephonics TDH 39 earphones. Pure-tone audiometry was performed according to international standards (International Standards Organization 8253-1) on a Decos Audiology audiometer (Decos Technology Group, Noordwijk, The Netherlands), which is compliant with International Standards Organization 389 standards.
For a continuous outcome measure as used in this study at least 10 participants are needed per variable in the model for the analyses to be reliable (Peduzzi et al. 1996). Thus, using 28 variables, at least 280 participants are needed.
RESULTS Patient Characteristics A total of 321 patients were consecutively seen at the Tinnitus Care Group of the University Medical Center, Utrecht, between June 2007 and November 2012. Twelve patients had to be excluded because they had not returned their questionnaire booklet, and thus, no TQ or THI score was available. Therefore, a total of 309 patients were included in this study, consisting of 208 male and 101 female patients, with a mean age of 51 years (range: 17–82 years). Mean tinnitus duration was 7 years (range: 2 months to 48 years), mean matched tinnitus pitch was 6 kHz (range: 100–16,000 Hz), and mean matched tinnitus loudness was 51 dB HL (range: 4–115 dB HL). Overall completeness of data was high; in total 3% of data were missing. The lowest response rate was on level of education (71%). Response varied between 90 and 94% on four variables (onset, pitch on a VAS, variability on VAS, and presence on a VAS), between 95 and 99% on seventeen variables, and all data were available for 14 variables. Total numbers can be found in Tables 3–5. Tinnitus analysis was performed in all 309 patients, but audiometric pitch and loudness matches were not always measurable (in 270 and 264 patients respectively). Because the VAS scores for pitch, loudness, presence, and variability in loudness and/or pitch were included from December 2007, the maximum number of possible respondents for these variables was 281 patients.
548
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
Table 2. Description of variables investigated for a possible relationship with tinnitus severity Included Variables
Explanation (If Applicable)
Assessment Method
Previously Described (n = 18) Additional (n = 10) 1 2 3 4 5 6 7 8 9 10 11 12
Age Gender Educational level Employment status Age at onset Duration Number of sounds Location Type Onset Change in perception over time Awareness during the day
13
Presence
14
Pitch
14a Subjective patient view
15
Loudness
14b Measurement 15a Subjective patient view 15b Measurement
16 17
18 19 20 21
Variability in loudness and/or pitch Masking
Residual inhibition Self-reported depression and/or anxiety Pain complaints Somatic complaints
22 23 24 25
Vertigo Otalgia Ear fullness Hyperacusis
26 27
Distortion of sound Somatosensory modulation
28
Hearing loss
patient rated awareness of tinnitus in percentage during the day †
His otorhinolaryngologist His otorhinolaryngologist His otorhinolaryngologist His otorhinolaryngologist His audiologist His audiologist His audiologist His audiologist His audiologist His audiologist His otorhinolaryngologist His audiologist VAS - continuously absent to continuously present VAS - extremely low to extremely high pitch Audiometrically VAS - inaudible to extremely loud Audiometrically VAS - no variability to extreme variability
17a Masking by external sounds (patient view, yes/no) 17b Audiometric maskability (yes/no) 17c Minimal masking level
Possible range 0–46 (9 questions concerning auto-immune disease, 11 neurologic disease complaints, 4 infectious disease complaints, 2 temporomandibular dysfunction or dental complaints, 12 metabolic or endocrine disease complaints, 8 cardiovascular disease complaints, and 2 on pain).
25a Subjective patient view (do you experience oversensitivity/pain to sound?) 25b Uncomfortable loudness level 23 maneuvers (six jaw movements, four jaw movements against resistance, five pressure points on the head, four head movements against resistance, four eye movements) Averaged (1, 2, 4kHz) pure-tone-threshold (dB HL)
His audiologist Audiometrically Audiometrically Audiometrically His otorhinolaryngologist His otorhinolaryngologist His otorhinolaryngologist, assessed through a standard list
His otorhinolaryngologist His otorhinolaryngologist His otorhinolaryngologist Likert scale (“never”, “hardly ever”, “sometimes”, “often”, “daily”) Audiometrically His audiologist Phys ex otorhinolaryngologist
Audiometrically
VAS were assessed through the earlier mentioned questionnaire booklet on a 10 cm line ranging from 0 to 10 *First proposed and reported on by (Stouffer and Tyler 1990). his, history taking; phys. ex., physical examination; VAS, Visual Analogue Scale.
Tinnitus Severity Mean TQ score was 40 (SD: 17, range: 3–80) and mean THI score was 45 (SD: 23, range: 2–100) (Figs. 1 and 2). Both outcomes were distributed normally and related significantly with each other (r = 0.85, strong relationship).
Univariate Relationships Between Demographic Characteristics and Tinnitus Severity Table 3 shows the univariate effects of demographic characteristics on tinnitus severity. Age and sex were not significantly associated with tinnitus severity. A significant effect
549
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
Table 3. Patient characteristics, descriptives on and univariate relationships with tinnitus severity
Characteristics 1) Age 2) Sex Male Female 3) Level of education High education Middle education Low education 4) Employment status Employed Student/housewife Not employed due to tinnitus Not employed other reason Unemployed Retired
Number Included in Number (%) Analyses or Mean ± SD 309 309
TQ (Mean ± SD)
51 ± 12 208 (67) 101 (33)
39 ± 17 42 ± 17
29 (13) 77 (35) 114 (52)
54 ± 15 41 ± 16 33 ± 14
194 (63) 9 (3) 12 (4) 34 (11) 13 (4) 47 (15)
36 ± 15 40 ± 17 49 ± 15 52 ± 16 45 ± 19 43 ± 17
220
Univariate Test Test Statistic
p
r = 0.1 t = −1.4
ns ns
Univariate Test Test Statistic
p
r = −0.01 t = −1.3
ns ns
44 ± 23 47 ± 22 F = −22.5
309
THI (Mean ± SD)
≤0.0005
F = 11.5
≤0.0005
F = 6.5
≤0.0005
60 ± 26 48 ± 22 39 ± 21 F = 7.9
≤0.0005 41 ± 21 44 ± 22 54 ± 15 62 ± 24 55 ± 27 46 ± 25
Bonferroni post-hoc test educational level TQ/THI: patients with low education reported significantly lower TQ and THI scores than patients with middle education. Both these groups reported lower TQ and THI scores than high educated patients. Bonferroni post-hoc test employment status TQ: employed patients reported significantly lower TQ scores than patients who are not employed due to tinnitus or another reason. Bonferroni post-hoc test employment status THI: employed patients and pensioned patients reported significantly lower TQ scores than patients who are not employed due to another reason. ns, non significant; THI, Tinnitus Handicap Inventory; TQ, Tinnitus Questionnaire.
was found for level of education and employment status on tinnitus severity. Consequent Bonferroni tests on educational level showed that all three groups (high, middle, and low education) differed significantly from each other on both questionnaires. The largest difference was found between patients with low and high level of education: patients with a low level of education reported significantly lower TQ and THI scores than patients with a high level of education (large effect sizes: 1.40 and 0.81, respectively). Bonferroni post hoc test on employment status showed that employed patients reported significantly lower TQ scores than patients who are not employed because of tinnitus or because of a reason other than tinnitus (large effect sizes: 0.81 and 1.0, respectively). Bonferroni post hoc tests on the THI showed that employed patients and retired patients reported significantly lower THI scores than patients who are not employed because of a reason other than tinnitus (effect sizes: 0.88 [large] and 0.64 [moderate], respectively).
Univariate Relationships Between Tinnitus Factors and Tinnitus Severity Nine tinnitus variables related significantly to tinnitus severity if measured using the TQ (Table 4). Four of these variables related (moderately) strongly to severity: longer tinnitus duration, number of sounds, higher percentage of awareness during the day, and louder tinnitus on a VAS. Weak correlations or small effect sizes were found with increased perception of tinnitus over time (effect size: 0.29), higher tinnitus pitch on a VAS, more variable tinnitus on a VAS, higher minimal masking level, and no masking through background sounds (effect size: 0.44). When measuring severity using the THI, relationships with the same variables were found, except for tinnitus duration and minimal masking level. Effect sizes and relationship strengths were similar for the seven variables that related both to the TQ and the THI.
Univariate Relationships Between Health Factors and Tinnitus Severity Table 5 shows relationships between health variables and tinnitus severity. Nine health variables related significantly to tinnitus severity if measured on the TQ. A large effect size was found for distortion of sounds (1.12 difference between patients who never had distortion of sounds and those who often experienced distortion of sounds). Moderately strong relationships or moderately large effect sizes were found for a number of somatic complaints, hearing loss, self-reported depression and/ or anxiety (effect size: 0.75), and hyperacusis (0.71 difference between patients who never had hyperacusis and those who often had hyperacusis, 0.59 difference between patients who sometimes had hyperacusis and those who often had hyperacusis). Small effect sizes were found for pain (effect size: 0.35) and vertigo (effect size: 0.44). When severity was measured using the THI, the same variables related univariately significantly with increased severity except for distortion of sounds. Effect sizes and relationship strengths were comparable for the seven variables that related both to the TQ and the THI. In addition when severity was measured on the THI, a significant relationship was found with presence of ear fullness and otalgia (small effect sizes: 0.39 and 0.28, respectively).
Multivariate Model Predicting Tinnitus Severity The factors that significantly univariately related to severity were entered in two separate stepwise multivariate models. Eighteen variables were found in the univariate analyses to significantly relate to the TQ, three with a strong effect/relationship, six with a moderate effect/relationship, and 11 with a small effect or weak relationship. Sixteen variables were found in the univariate analyses to significantly relate to the THI, three with a strong effect/relationship, five with a moderate effect/relationship, and nine with a small effect or weak relationship. Fourteen variables were found to relate to both the THI and the Tinnitus
550
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
Table 4. Tinnitus characteristics, descriptives on and univariate relationships with tinnitus severity
Characteristics 5) Age at onset 6) Duration (mo) 7) Number of sounds 8) Location Right Left 2× unilateral Bilateral In the head Varying locations 9) Type Tonal Noise 10) Onset Acute Gradual Unknown 11) Change in perception over time Increased Decreased Unchanged 12) Awareness during the day 13) Presence (VAS) 14) Pitch Pitch (VAS) Pitch (kHz)* (audiometric analysis) 15) Loudness Loudness (VAS) Loudness (dB HL)† (audiometric analysis) 16) Variability of sound (VAS) 17) Masking Maskability (by external sounds) Yes No Maskability (audiometric) Yes No Minimal masking level (dB) 18) Residual inhibition Yes No Duration residual inhibition (sec)
Number Included in Number (%) Analyses or Mean ± SD 301 301 308 308
TQ (Mean ± SD)
44 ± 12 85 ± 97 2 ± 1 40 (13) 54 (18) 40 (12) 77 (25) 73 (24) 24 (8)
40 ± 16 38 ± 18 42 ± 16 39 ± 17 41 ± 17 34 ± 12
180 (60) 119 (40)
38 ± 16 42 ± 17
136 (49) 131 (47) 11 (4)
41 ± 16 38 ± 17 41 ± 18 42 ± 17 36 ± 16 37 ± 17
298 264
145 (50) 21 (7) 127 (43) 76 ± 31 8 ± 2
262 270
299
Univariate Test Test Statistic
p
r = 0.03 r = 0.1 r = 0.1 F = 1.0
ns 0.045 0.009 Ns
Test Statistic
p
r = −0.07 r = 0.1 r = 0.2 F = 0.8
ns ns 0.005 ns
ns
t = −1.3
ns
F = 0.4
ns
F = 3.6
0.03
r = 0.3 r = 0.1
≤0.0005 Ns
43 ± 22 47 ± 24 F = 0.7
293
Univariate Test
41 ± 21 41 ± 24 47 ± 24 47 ± 22 48 ± 24 44 ± 19 t = −1.9
278
THI (Mean ± SD)
ns 46 ± 22 44 ± 24 48 ± 24
F = 4.5
0.012 49 ± 23 37 ± 18 43 ± 24
r = 0.4 r = 0.1
≤0.0005 ns
6 ± 2 6.0 ± 4.6
r = 0.2 r = −0.1
0.001 ns
r = 0.2 r =-0.06
0.014 ns
272 264
6 ± 2 51 ± 27
r = 0.4 r = 0.1
≤0.0005 ns
r = 0.3 r = 0.006
≤0.0005 ns
262
3 ± 3
r = 0.1
0.045
t = −3.5
0.001
308 182 (59) 126 (41)
37 ± 16 44 ± 16
186 121 59 ± 26
39 ± 16 41 ± 17
93 (31) 211 (69) 64 ± 59
39 ± 16 40 ± 17
307
0.001
t = −2.7
0.007
t = −1.1
ns
r = 0.06 t = −1.1
ns ns
r = −0.02
ns
42 ± 22 49 ± 23 t = −0.8
ns 44 ± 23 47 ± 23
r = 0.2 t = −0.6
304
r = 0.2
0.011 ns 43 ± 22 46 ± 23
r = −0.1
ns
Bonferroni post-hoc test change in tinnitus perception over time TQ: patients reporting unchanged tinnitus showed significantly lower TQ scores than patients reporting increased tinnitus. Bonferroni post-hoc test change in tinnitus perception over time THI: no significant differences between two of the groups. *Three hundred nine patients pitch matching in performed, no pitch measurable in 39 patients. †Three hundred nine patients loudness matching in performed, no loudness measurable in 45 patients. ns, non significant; THI, Tinnitus Handicap Inventory; TQ, Tinnitus Questionnaire; VAS, Visual Analogue Scale.
Handicap Questionnaire. Thus, 18 variables were included in the model using the TQ, and 16 in the model using the THI. Employment status was dichotomized according to the outcomes of the Bonferroni post hoc tests into “not/unemployed” versus “working/pension/students/housewives.” Stepwise multiple regression analysis on the TQ showed that four variables had a significant unique predictive effect on tinnitus severity, explaining 53% of the variance (Table 7). Analysis using the THI as dependent variable showed that five variables had a significant unique predictive effect explaining 40% of the variance (Table 6). Three variables
contributed significantly independently to the explained variance in both models: percentage of awareness during the day, selfreported depression and/or anxiety, and loudness measured on a VAS. The model based on the TQ included level of education as well. This was the first excluded variable in the THI model (p = 0.07). The model based on the THI also included somatic complaints and variability measured on a VAS. These were the second and sixth excluded variables in the TQ model (p = 0.12, p = 0.26, respectively). Percentage awareness of tinnitus during the day was the strongest contributing variable for tinnitus
551
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
Table 5. Health characteristics, descriptives on and univariate relationships with tinnitus severity
Characteristics 19) Self-reported depression and/or anxiety Yes No 20) Pain complaints Yes No 21) Somatic complaints 22) Vertigo Yes No 23) Otalgia Yes No 24) Ear fullness Yes No 25) Hyperacusis Subjective hyperacusis Never Hardly ever Sometimes Often Daily Uncomfortable loudness level 26) Distortion of sound Never Hardly ever Sometimes Often Daily 27) Somatosensory modulation 28) Averaged (1, 2, 4 kHz) pure-tone-hearing loss (dB HL) right and left ear
Number Univariate Test Included in Number (%) TQ Analyses or Mean ± SD (Mean ± SD) Test Statistic p 304
t = 6.1 74 (24) 230 (76)
49 ± 16 37 ± 16
143 (47) 163 (53) 4 ± 4 (306)
43 ± 17 37 ± 16
81 (27) 225 (74)
45 ± 16 38 ± 16
47 (15) 258 (85)
44 ± 17 39 ± 16
100 (33) 204 (67)
41 ± 18 39 ± 16
110 (36) 29 (10) 72 (24) 46 (15) 49 (16) 103 ± 12
36 ± 16 41 ± 15 38 ± 16 48 ± 17 42 ± 17
240 (78) 2 (1) 30 (10) 7 (2) 27 (9) 3 ± 6 29 ± 23
39 ± 17 44 ± 31 43 ± 17 58 ± 12 42 ± 12
306
306 306
t = 7.2
INTRODUCTION Tinnitus is a phantom auditory perception of meaningless sound in the absence of an external or internal acoustic stimulus. It is a common problem that affects 7 to 19% of the adult population (Chung et al. 1984; Coles 1984; Axelsson & Ringdahl 1989; Davis 1989; Nondahl et al. 2002). The majority of these individuals do not experience the tinnitus as problematic, but in up to 5% of the adult population it does lead to annoyance or even interferes with the ability to lead a normal life (Coles 1984; Davis & Rafaie 2000; Nondahl et al. 2002). Many recommendations have been made to recognize these distressed patients or to assess which patients are in need of further counseling (Tyler & Erlandsson 2003; Henry et al. 2005; Cima et al. 2012). It is difficult though, to identify beforehand individuals who may develop problems due to their tinnitus to such a degree that it may negatively affect their quality of life (clinically significant tinnitus). Knowledge about factors that are associated with tinnitus severity could be useful to identify these individuals who may develop problems. If these factors are present they could support an indication for follow-up of these patients to monitor them for development of potential distress. In addition, the knowledge about factors associated with tinnitus severity might provide a useful tool in counseling or psychological therapies and, therefore, in the effectiveness of the care provided. As a next step, the question would arise whether such factors associated with tinnitus severity would also be related to treatment outcomes. It seems likely that the severity of tinnitus can be influenced by different factors, such as socio-demographic, personality, psychological or tinnitus characteristics, or additional health complaints. Many studies have reported variables related to tinnitus severity (Table 1). Among these studies many relationships have only been examined once, contradicting results are reported across studies, and reliable multivariate analyses are scarce. Combined, the 18 studies shown in Table 1 identified 28 variables significantly related to tinnitus severity and 35 variables without such a relationship. Of these 28 variables, only three variables (continuous tinnitus, sleep problems, and vertigo) have been shown to significantly relate to severity in more than one study. Thirteen variables are described to both having and not having a significant relationship to severity (age, anxiety, depression, etiology, sex, hearing loss, location, loudness, loudness variability, maskability, pitch, [psycho]somatic complaints, and tinnitus type). The remaining 12 variables were reportedly significantly related to tinnitus in a single study only, namely, age at onset, avoidance, awareness, change since onset, chronic pain, controllability, hyperacusis, influence of hearing aid, internal locus of control, personality, tolerance, and various questions on quality of life. An important drawback of the studies performed so far is that most do not include a multivariate analysis. Multivariate
Design: This is a retrospective cohort study performed at the Tinnitus Care Group of the University Medical Center, Utrecht, in 309 consecutively seen chronic tinnitus patients. At this care group, patients are examined according to a structured diagnostic protocol, including history-taking by an otorhinolaryngologist and audiologist, physical examination, and audiometry. Based on results from previous research and theoretical considerations, a subset of data acquired through this diagnostic protocol were selected and used in this study. Univariate and multivariate correlations with tinnitus severity were investigated for 28 socio-demographic, health, and tinnitus variables. Tinnitus severity was measured with the Tinnitus Questionnaire (TQ) and the Tinnitus Handicap Inventory (THI). Results: Eighteen variables related univariately with the TQ and 16 variables related univariately with the THI. Among these, 14 variables related univariately with both the TQ and the THI. Multivariate analyses showed three variables with an independent significant effect on both the TQ and the THI: percentage of tinnitus awareness during the day, self-reported depression and/or anxiety, and loudness on a Visual Analogue Scale. Three additional variables contributed independently significantly to the explained variance in either the TQ or the THI: level of education, somatic complaints, and tinnitus variability in loudness and/or pitch on a Visual Analogue Scale. These variables were among the first not to reach significance on the other questionnaire. Conclusions: Tinnitus severity is shown to be strongly related to percentage of tinnitus awareness during the day, self-reported depression and/or anxiety, subjectively experienced loudness, level of education, existence of additional somatic complaints, and subjectively experienced variability in loudness and/or pitch. Further research is needed to investigate the causal relationship between these variables and tinnitus severity. This knowledge may contribute to a better recognition, follow-up, and/or counseling of more vulnerable patients. Key words: Chronic tinnitus, Severity. (Ear & Hearing 2014;35;544–554)
Department of Otorhinolaryngology and Head & Neck Surgery, University Medical Center Utrecht, Utrecht, The Netherlands; 2Rudolf Magnus Institute of Neuroscience, University Medical Center Utrecht, Utrecht, The Netherlands; and 3Faculty of Medicine, University Medical Center Utrecht, Utrecht, The Netherlands. 1
0196/0202/2014/355-0544/0 • Ear & Hearing • Copyright © 2014 by Lippincott Williams & Wilkins • Printed in the U.S.A. 544
5 5
11
112
72
166
6
71
136 68 755
Ooms et al. 2011 Pinto et al. 2010 Schlee et al. 2011
3 3 1
3 5 3
531 281 100
Mazurek et al. 2010 Meric et al.1998 Mondelli & da Rocha 2011 Ooms et al. 2012
13*
127
10
4
2378
4995
3
19
146
39
Holgers et al. 2005
Hiller & Goebel 2006
Andersson et al. 2001 Axelsson & Ringdahl 1989 Halford & Anderson 1991 Henry & Wilson 1995 Hiller & Goebel 1999
Andersson et al. 1999
No. Patients Included
No. Variables Analyzed
THI THI Mini-TQ
THI
TQ TRQ, THQ STSS THI
Own criteria for “tinnitus sufferer”
Mini-TQ
STI global score
TRQ
STSS
Likert scale
TRQ
K&L grading system
Outcome Parameter for Tinnitus Severity
Pitch (A), loudness (A), BDI-II Sex, age, hearing loss
Pitch (A), loudness (A)
Sex, age, duration, hearing loss, BMI, marital status, educational level, income level, professional status, alcohol, tobacco Sex Sex, age, duration Sex, age, hearing loss
Duration, familial status
Pitch (A), minimal masking level, ATQ Location, tinnitus type, pitch (S), pulsatile character, onset, nine types of etiology
Sex, age, variability
Sex
Hearing loss, loudness (A), duration, continuous loudness, influence of background sound, somatic problems, influence of substance or treatment, possibility to do something to ease, situations when less problematic, irritation, psychological factors, anxiety, depression, concentration difficulties Hearing loss, loudness (A)
Variables Not Related to Severity
STAI-state, STAI-trait, somatic anxiety — — Age at onset
Hearing loss, location Hearing loss, MMPI —
Hearing loss, continuous tinnitus, maskability, progressive loudness, SCL90-R scales, two types of etiology (sudden hearing loss, craniomandibular dysfunction) Sex, age, hearing loss, progressive onset, location, continuous tinnitus, vertigo, hyperacusis Various questions on NHP, mental disorders (mostly anxiety)
Loudness (A), BDI
Advancing age (females), hearing loss, sleep problems STAI-trait, DTQ
Maskability
Pitch (A), minimal masking level, noise sensitivity, change since onset, avoidance, tolerance
Variables Related to Severity
Table 1. Characteristics of studies assessing variables related to tinnitus severity (excluding various tinnitus questionnaires as variables)
Not performed Not performed Not performed
Not performed
Not performed Not performed Not performed
(Continued )
Three questions on NHP (find it hard to reach for things, finding live not worth living, bad sleep)
Not performed
Not performed
BDI
Not performed
Not performed
Not performed
Audiometric pitch, minimal masking level, avoidance, tolerance
Variables Related to Severity in Multiple Regression Analyses
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554 545
546
* Thirteen variables were analyzed, of which of four variables the individual components (different questions within a questionnaire or interview or different average hearing levels) were also analyzed separately. (A), audiometric; ATQ, Automatic Thoughts Questionnaire; BDI, Beck Depression Index; BMI, body mass index; DTQ, Depressive Tendency Questionnaire; K&L grading system, Klockhoff and Lindblom grading system; MMPI, Minnesota Multiphasic Personality Inventory; NHP, Nottingham Health Profile; (S), subjective; SCL-90-R, Symptom Checklist-90-Revised; STAI, State-Trait Anxiety Inventory; STI, Structured Tinnitus Interview; STSS, Subjective Tinnitus Severity Scale; THI, Tinnitus Handicap Inventory; THQ, Tinnitus Handicap Questionnaire; TQ, Tinnitus Questionnaire; TRQ, Tinnitus Reaction Questionnaire.
Binaural/central tinnitus, influence of hearing aid, sleep problems, chronic pain, depression, anxiety, somatic symptom severity 4705
15
Mini-TQ
Pitch (S), duration, perception as illness, comorbidity Sex, age, hearing loss, duration, onset, somatic comorbidity Unterrainer et al. 2003 WallhäusserFranke et al. 2012
149
6
THI
Depression, loudness (S), internal locus of control Awareness, location, influence of hearing aid, vertigo, chronic pain, sleep problems, depression, anxiety, somatic symptom severity
Tinnitus type, four psychosomatic complaints (depression, lack of concentration, insomnia, balance disturbance), maskability, presence, controllability, loudness variation Not performed Tinnitus type, nine psychosomatic complaints, maskability, presence, controllability, loudness variation Duration, onset, location, hearing loss Likert scale 18 3372 Scott et al. 1990
Table 1. Continued.
No. Patients Included
No. Variables Analyzed
Outcome Parameter for Tinnitus Severity
Variables Not Related to Severity
Variables Related to Severity
Variables Related to Severity in Multiple Regression Analyses
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
analyses are essential in finding uniquely related variables because these take into account the interdependencies between different variables. Variables that have been found in studies only performing univariate analyses could be based on unknown confounding effects. Eleven of these studies evaluated only a small number of potential variables and were thus not able to perform a contributory multivariate analysis (Axelsson & Ringdahl 1989; Halford & Anderson 1991; Henry & Wilson 1995; Meric et al. 1998; Andersson et al. 2001; Mazurek et al. 2010; Pinto et al. 2010; Mondelli & da Rocha 2011; Schlee et al. 2011; Ooms et al. 2011; Ooms et al. 2012). Three studies that could have performed a multivariate analysis did not include this analysis (Hiller & Goebel 1999; Unterrainer et al. 2003; Hiller & Goebel 2006). Another drawback of previous studies is that the included number of participants is not always in relation to the examined number of variables. As a rule of thumb at least 10 participants are needed per variable in a model for the analysis to be reliable (Peduzzi et al. 1996). Unfortunately, of the five studies including a multivariate analysis, two studies did not reach the 1:10 ratio (Andersson et al. 1999; Holgers et al. 2005). The three studies including a reliable multivariate analysis show 13 variables related to tinnitus severity, of which only depression, (psycho) somatic complaints, and sleep problems were shown in two studies to relate to severity independently (Scott et al. 1990; Henry & Wilson 1995; Wallhäusser-Franke et al. 2012). In sum, although much has been published about variables related to severity, it remains unclear from the current literature as to what are the most important and independent variables related to tinnitus severity. Our study addresses this problem by investigating variables previously described in the literature as well as additional variables, through univariate and multivariate analyses in a large enough sample size for the results to be considered reliable. The aim of this study is to find socio-demographic, health, and tinnitus variables that independently affect tinnitus severity. This study focuses on these three groups of variables because these variables are commonly evaluated in a standard outpatient visit. The results of this study may give clinicians a convenient number of risk variables for (chance for) more severe tinnitus that can easily and quickly be evaluated in the medical office or through audiometry. In addition to knowing risk variables for more severe tinnitus, clinicians will also have knowledge of which variables that are of less interest.
MATERIALS AND METHODS Study Design and Participants This study was performed at the Tinnitus Care Group of the Otorhinolaryngology department of the University Medical Center, Utrecht, as a retrospective cohort study. Patients with chronic tinnitus, defined as more than 2 months in duration, were examined according to a structured diagnostic protocol. Through this diagnostic protocol patients were evaluated by an otorhinolaryngologist, an audiologist, and a psychologist. Audiometry was performed and patients filled out various tinnitus questionnaires, Visual Analogue Scales (VASs) and psychological questionnaires (including the Symptom Checklist-90-Revised [SCL-90-R]). All patient data collected through this protocol were anonymized and entered in an Access database. For this retrospective study, a selection of these routinely collected data (see Independent Variables) from all consecutively
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
547
seen chronic tinnitus patients between June 2007 and November 2012 was included. The study was performed in accordance with the Declaration of Helsinki. Because this is a retrospective study with anonymized data, exemption for a full review from the Local Research Ethics Committee was obtained (12–611/C).
Sample Size
Dependent Variable (Outcome Measure)
Data Analysis
As part of their evaluation by the Tinnitus Care Group patients completed two questionnaires to measure tinnitus severity: the Tinnitus Questionnaire (TQ) and the Tinnitus Handicap Inventory (THI). The TQ is a 52 item self-response questionnaire with three answer possibilities (not true, partly true, true) and a total score ranging from 0 to 84 (Hallam et al. 1988). The THI is a 25-item self-response questionnaire with three answer possibilities (no, sometimes, yes) and a score range of 0 to 100 (Newman et al. 1996). For both the TQ and the THI higher scores account for more severe tinnitus. Since there is no standard definition or measurement for tinnitus severity, both questionnaires are included in this study. This allows to find potential similarities or differences in the results that might arise from using different ways to measure severity.
Statistical analyses were conducted with SPSS 20.0 (IBM, Armonk, NY). Descriptive analyses were calculated for the socio-demographic, health, and tinnitus variables. Univariate effects on tinnitus severity (TQ and THI) were explored through independent-sample t tests for dichotomous variables and through analysis of variance for ordinal variables. Effect sizes were computed using Cohen’s d to examine clinical significance of differences found between groups. Effect sizes ≥0.80 reflected a large clinically relevant difference, those between 0.50 and 0.80 were considered moderately large, and those between 0.20 and 0.49 were considered small (Cohen 1988). Bonferroni post hoc tests were executed as part of the analysis of variance to assess which groups differed significantly. For continuous variables, bivariate correlations were calculated (Pearson correlation for normally distributed data and Spearman correlations for abnormally distributed data). Correlation coefficients 0.50 strong (Cohen et al. 2003). Socio-demographic, health, and tinnitus variables showing a univariate significant relationship with tinnitus severity were entered into two separate stepwise multiple regression analyses to examine unique predictive effects. The first analysis used the TQ as dependent variable, the second analysis used the THI as dependent variable, and p values ≤0.05 were defined as statistically significant.
Independent Variables This study had the aim to investigate socio-demographic, health, and tinnitus variables that could easily be assessed by clinicians in the medical office or by audiometry. Of the 28 variables previously described to possibly be related to tinnitus severity 18 variables were included in this study because they fulfilled these criteria. An additional ten variables were included in the study because they could theoretically be associated with tinnitus severity, were only once or not investigated before or because they form important tinnitus characterizations. This resulted in a total of 28 variables. Table 2 shows an overview of the included variables, if applicable a short explanation, and the assessment method for each variable. For four variables (hyperacusis, masking, loudness, and pitch) multiple measurements were included because they can be measured in different ways and have been measured differently in earlier studies. Pure-tone audiometry and tinnitus analysis (pitch and loudness matching, measurement of minimal masking level, and complete and/or partial residual inhibition) were performed by one of two audiology assistants trained in tinnitus analysis. For the tinnitus analyses tones were used in the analysis for tonal tinnitus and a narrow-band noise was used for noise-like tinnitus. Potential octave ambiguity was tested as part of tinnitus pitch matching. Tinnitus loudness was matched at the pitch-matched frequency. When a pitch match could not be obtained, the loudness was matched with a 1000-Hz pure tone. Minimal masking was tested at the tinnitus frequency. Residual inhibition was tested 10 dB above the minimal masking level. Uncomfortable loudness was measured at 0.5, 1, 2, and 4 kHz by increasing the sound level by 5-dB each time until the patient reported that the loudness was uncomfortable. Testing was done in a soundproof cabin with Telephonics TDH 39 earphones. Pure-tone audiometry was performed according to international standards (International Standards Organization 8253-1) on a Decos Audiology audiometer (Decos Technology Group, Noordwijk, The Netherlands), which is compliant with International Standards Organization 389 standards.
For a continuous outcome measure as used in this study at least 10 participants are needed per variable in the model for the analyses to be reliable (Peduzzi et al. 1996). Thus, using 28 variables, at least 280 participants are needed.
RESULTS Patient Characteristics A total of 321 patients were consecutively seen at the Tinnitus Care Group of the University Medical Center, Utrecht, between June 2007 and November 2012. Twelve patients had to be excluded because they had not returned their questionnaire booklet, and thus, no TQ or THI score was available. Therefore, a total of 309 patients were included in this study, consisting of 208 male and 101 female patients, with a mean age of 51 years (range: 17–82 years). Mean tinnitus duration was 7 years (range: 2 months to 48 years), mean matched tinnitus pitch was 6 kHz (range: 100–16,000 Hz), and mean matched tinnitus loudness was 51 dB HL (range: 4–115 dB HL). Overall completeness of data was high; in total 3% of data were missing. The lowest response rate was on level of education (71%). Response varied between 90 and 94% on four variables (onset, pitch on a VAS, variability on VAS, and presence on a VAS), between 95 and 99% on seventeen variables, and all data were available for 14 variables. Total numbers can be found in Tables 3–5. Tinnitus analysis was performed in all 309 patients, but audiometric pitch and loudness matches were not always measurable (in 270 and 264 patients respectively). Because the VAS scores for pitch, loudness, presence, and variability in loudness and/or pitch were included from December 2007, the maximum number of possible respondents for these variables was 281 patients.
548
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
Table 2. Description of variables investigated for a possible relationship with tinnitus severity Included Variables
Explanation (If Applicable)
Assessment Method
Previously Described (n = 18) Additional (n = 10) 1 2 3 4 5 6 7 8 9 10 11 12
Age Gender Educational level Employment status Age at onset Duration Number of sounds Location Type Onset Change in perception over time Awareness during the day
13
Presence
14
Pitch
14a Subjective patient view
15
Loudness
14b Measurement 15a Subjective patient view 15b Measurement
16 17
18 19 20 21
Variability in loudness and/or pitch Masking
Residual inhibition Self-reported depression and/or anxiety Pain complaints Somatic complaints
22 23 24 25
Vertigo Otalgia Ear fullness Hyperacusis
26 27
Distortion of sound Somatosensory modulation
28
Hearing loss
patient rated awareness of tinnitus in percentage during the day †
His otorhinolaryngologist His otorhinolaryngologist His otorhinolaryngologist His otorhinolaryngologist His audiologist His audiologist His audiologist His audiologist His audiologist His audiologist His otorhinolaryngologist His audiologist VAS - continuously absent to continuously present VAS - extremely low to extremely high pitch Audiometrically VAS - inaudible to extremely loud Audiometrically VAS - no variability to extreme variability
17a Masking by external sounds (patient view, yes/no) 17b Audiometric maskability (yes/no) 17c Minimal masking level
Possible range 0–46 (9 questions concerning auto-immune disease, 11 neurologic disease complaints, 4 infectious disease complaints, 2 temporomandibular dysfunction or dental complaints, 12 metabolic or endocrine disease complaints, 8 cardiovascular disease complaints, and 2 on pain).
25a Subjective patient view (do you experience oversensitivity/pain to sound?) 25b Uncomfortable loudness level 23 maneuvers (six jaw movements, four jaw movements against resistance, five pressure points on the head, four head movements against resistance, four eye movements) Averaged (1, 2, 4kHz) pure-tone-threshold (dB HL)
His audiologist Audiometrically Audiometrically Audiometrically His otorhinolaryngologist His otorhinolaryngologist His otorhinolaryngologist, assessed through a standard list
His otorhinolaryngologist His otorhinolaryngologist His otorhinolaryngologist Likert scale (“never”, “hardly ever”, “sometimes”, “often”, “daily”) Audiometrically His audiologist Phys ex otorhinolaryngologist
Audiometrically
VAS were assessed through the earlier mentioned questionnaire booklet on a 10 cm line ranging from 0 to 10 *First proposed and reported on by (Stouffer and Tyler 1990). his, history taking; phys. ex., physical examination; VAS, Visual Analogue Scale.
Tinnitus Severity Mean TQ score was 40 (SD: 17, range: 3–80) and mean THI score was 45 (SD: 23, range: 2–100) (Figs. 1 and 2). Both outcomes were distributed normally and related significantly with each other (r = 0.85, strong relationship).
Univariate Relationships Between Demographic Characteristics and Tinnitus Severity Table 3 shows the univariate effects of demographic characteristics on tinnitus severity. Age and sex were not significantly associated with tinnitus severity. A significant effect
549
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
Table 3. Patient characteristics, descriptives on and univariate relationships with tinnitus severity
Characteristics 1) Age 2) Sex Male Female 3) Level of education High education Middle education Low education 4) Employment status Employed Student/housewife Not employed due to tinnitus Not employed other reason Unemployed Retired
Number Included in Number (%) Analyses or Mean ± SD 309 309
TQ (Mean ± SD)
51 ± 12 208 (67) 101 (33)
39 ± 17 42 ± 17
29 (13) 77 (35) 114 (52)
54 ± 15 41 ± 16 33 ± 14
194 (63) 9 (3) 12 (4) 34 (11) 13 (4) 47 (15)
36 ± 15 40 ± 17 49 ± 15 52 ± 16 45 ± 19 43 ± 17
220
Univariate Test Test Statistic
p
r = 0.1 t = −1.4
ns ns
Univariate Test Test Statistic
p
r = −0.01 t = −1.3
ns ns
44 ± 23 47 ± 22 F = −22.5
309
THI (Mean ± SD)
≤0.0005
F = 11.5
≤0.0005
F = 6.5
≤0.0005
60 ± 26 48 ± 22 39 ± 21 F = 7.9
≤0.0005 41 ± 21 44 ± 22 54 ± 15 62 ± 24 55 ± 27 46 ± 25
Bonferroni post-hoc test educational level TQ/THI: patients with low education reported significantly lower TQ and THI scores than patients with middle education. Both these groups reported lower TQ and THI scores than high educated patients. Bonferroni post-hoc test employment status TQ: employed patients reported significantly lower TQ scores than patients who are not employed due to tinnitus or another reason. Bonferroni post-hoc test employment status THI: employed patients and pensioned patients reported significantly lower TQ scores than patients who are not employed due to another reason. ns, non significant; THI, Tinnitus Handicap Inventory; TQ, Tinnitus Questionnaire.
was found for level of education and employment status on tinnitus severity. Consequent Bonferroni tests on educational level showed that all three groups (high, middle, and low education) differed significantly from each other on both questionnaires. The largest difference was found between patients with low and high level of education: patients with a low level of education reported significantly lower TQ and THI scores than patients with a high level of education (large effect sizes: 1.40 and 0.81, respectively). Bonferroni post hoc test on employment status showed that employed patients reported significantly lower TQ scores than patients who are not employed because of tinnitus or because of a reason other than tinnitus (large effect sizes: 0.81 and 1.0, respectively). Bonferroni post hoc tests on the THI showed that employed patients and retired patients reported significantly lower THI scores than patients who are not employed because of a reason other than tinnitus (effect sizes: 0.88 [large] and 0.64 [moderate], respectively).
Univariate Relationships Between Tinnitus Factors and Tinnitus Severity Nine tinnitus variables related significantly to tinnitus severity if measured using the TQ (Table 4). Four of these variables related (moderately) strongly to severity: longer tinnitus duration, number of sounds, higher percentage of awareness during the day, and louder tinnitus on a VAS. Weak correlations or small effect sizes were found with increased perception of tinnitus over time (effect size: 0.29), higher tinnitus pitch on a VAS, more variable tinnitus on a VAS, higher minimal masking level, and no masking through background sounds (effect size: 0.44). When measuring severity using the THI, relationships with the same variables were found, except for tinnitus duration and minimal masking level. Effect sizes and relationship strengths were similar for the seven variables that related both to the TQ and the THI.
Univariate Relationships Between Health Factors and Tinnitus Severity Table 5 shows relationships between health variables and tinnitus severity. Nine health variables related significantly to tinnitus severity if measured on the TQ. A large effect size was found for distortion of sounds (1.12 difference between patients who never had distortion of sounds and those who often experienced distortion of sounds). Moderately strong relationships or moderately large effect sizes were found for a number of somatic complaints, hearing loss, self-reported depression and/ or anxiety (effect size: 0.75), and hyperacusis (0.71 difference between patients who never had hyperacusis and those who often had hyperacusis, 0.59 difference between patients who sometimes had hyperacusis and those who often had hyperacusis). Small effect sizes were found for pain (effect size: 0.35) and vertigo (effect size: 0.44). When severity was measured using the THI, the same variables related univariately significantly with increased severity except for distortion of sounds. Effect sizes and relationship strengths were comparable for the seven variables that related both to the TQ and the THI. In addition when severity was measured on the THI, a significant relationship was found with presence of ear fullness and otalgia (small effect sizes: 0.39 and 0.28, respectively).
Multivariate Model Predicting Tinnitus Severity The factors that significantly univariately related to severity were entered in two separate stepwise multivariate models. Eighteen variables were found in the univariate analyses to significantly relate to the TQ, three with a strong effect/relationship, six with a moderate effect/relationship, and 11 with a small effect or weak relationship. Sixteen variables were found in the univariate analyses to significantly relate to the THI, three with a strong effect/relationship, five with a moderate effect/relationship, and nine with a small effect or weak relationship. Fourteen variables were found to relate to both the THI and the Tinnitus
550
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
Table 4. Tinnitus characteristics, descriptives on and univariate relationships with tinnitus severity
Characteristics 5) Age at onset 6) Duration (mo) 7) Number of sounds 8) Location Right Left 2× unilateral Bilateral In the head Varying locations 9) Type Tonal Noise 10) Onset Acute Gradual Unknown 11) Change in perception over time Increased Decreased Unchanged 12) Awareness during the day 13) Presence (VAS) 14) Pitch Pitch (VAS) Pitch (kHz)* (audiometric analysis) 15) Loudness Loudness (VAS) Loudness (dB HL)† (audiometric analysis) 16) Variability of sound (VAS) 17) Masking Maskability (by external sounds) Yes No Maskability (audiometric) Yes No Minimal masking level (dB) 18) Residual inhibition Yes No Duration residual inhibition (sec)
Number Included in Number (%) Analyses or Mean ± SD 301 301 308 308
TQ (Mean ± SD)
44 ± 12 85 ± 97 2 ± 1 40 (13) 54 (18) 40 (12) 77 (25) 73 (24) 24 (8)
40 ± 16 38 ± 18 42 ± 16 39 ± 17 41 ± 17 34 ± 12
180 (60) 119 (40)
38 ± 16 42 ± 17
136 (49) 131 (47) 11 (4)
41 ± 16 38 ± 17 41 ± 18 42 ± 17 36 ± 16 37 ± 17
298 264
145 (50) 21 (7) 127 (43) 76 ± 31 8 ± 2
262 270
299
Univariate Test Test Statistic
p
r = 0.03 r = 0.1 r = 0.1 F = 1.0
ns 0.045 0.009 Ns
Test Statistic
p
r = −0.07 r = 0.1 r = 0.2 F = 0.8
ns ns 0.005 ns
ns
t = −1.3
ns
F = 0.4
ns
F = 3.6
0.03
r = 0.3 r = 0.1
≤0.0005 Ns
43 ± 22 47 ± 24 F = 0.7
293
Univariate Test
41 ± 21 41 ± 24 47 ± 24 47 ± 22 48 ± 24 44 ± 19 t = −1.9
278
THI (Mean ± SD)
ns 46 ± 22 44 ± 24 48 ± 24
F = 4.5
0.012 49 ± 23 37 ± 18 43 ± 24
r = 0.4 r = 0.1
≤0.0005 ns
6 ± 2 6.0 ± 4.6
r = 0.2 r = −0.1
0.001 ns
r = 0.2 r =-0.06
0.014 ns
272 264
6 ± 2 51 ± 27
r = 0.4 r = 0.1
≤0.0005 ns
r = 0.3 r = 0.006
≤0.0005 ns
262
3 ± 3
r = 0.1
0.045
t = −3.5
0.001
308 182 (59) 126 (41)
37 ± 16 44 ± 16
186 121 59 ± 26
39 ± 16 41 ± 17
93 (31) 211 (69) 64 ± 59
39 ± 16 40 ± 17
307
0.001
t = −2.7
0.007
t = −1.1
ns
r = 0.06 t = −1.1
ns ns
r = −0.02
ns
42 ± 22 49 ± 23 t = −0.8
ns 44 ± 23 47 ± 23
r = 0.2 t = −0.6
304
r = 0.2
0.011 ns 43 ± 22 46 ± 23
r = −0.1
ns
Bonferroni post-hoc test change in tinnitus perception over time TQ: patients reporting unchanged tinnitus showed significantly lower TQ scores than patients reporting increased tinnitus. Bonferroni post-hoc test change in tinnitus perception over time THI: no significant differences between two of the groups. *Three hundred nine patients pitch matching in performed, no pitch measurable in 39 patients. †Three hundred nine patients loudness matching in performed, no loudness measurable in 45 patients. ns, non significant; THI, Tinnitus Handicap Inventory; TQ, Tinnitus Questionnaire; VAS, Visual Analogue Scale.
Handicap Questionnaire. Thus, 18 variables were included in the model using the TQ, and 16 in the model using the THI. Employment status was dichotomized according to the outcomes of the Bonferroni post hoc tests into “not/unemployed” versus “working/pension/students/housewives.” Stepwise multiple regression analysis on the TQ showed that four variables had a significant unique predictive effect on tinnitus severity, explaining 53% of the variance (Table 7). Analysis using the THI as dependent variable showed that five variables had a significant unique predictive effect explaining 40% of the variance (Table 6). Three variables
contributed significantly independently to the explained variance in both models: percentage of awareness during the day, selfreported depression and/or anxiety, and loudness measured on a VAS. The model based on the TQ included level of education as well. This was the first excluded variable in the THI model (p = 0.07). The model based on the THI also included somatic complaints and variability measured on a VAS. These were the second and sixth excluded variables in the TQ model (p = 0.12, p = 0.26, respectively). Percentage awareness of tinnitus during the day was the strongest contributing variable for tinnitus
551
HOEKSTRA ET AL. / EAR & HEARING, VOL. 35, NO. 5, 544–554
Table 5. Health characteristics, descriptives on and univariate relationships with tinnitus severity
Characteristics 19) Self-reported depression and/or anxiety Yes No 20) Pain complaints Yes No 21) Somatic complaints 22) Vertigo Yes No 23) Otalgia Yes No 24) Ear fullness Yes No 25) Hyperacusis Subjective hyperacusis Never Hardly ever Sometimes Often Daily Uncomfortable loudness level 26) Distortion of sound Never Hardly ever Sometimes Often Daily 27) Somatosensory modulation 28) Averaged (1, 2, 4 kHz) pure-tone-hearing loss (dB HL) right and left ear
Number Univariate Test Included in Number (%) TQ Analyses or Mean ± SD (Mean ± SD) Test Statistic p 304
t = 6.1 74 (24) 230 (76)
49 ± 16 37 ± 16
143 (47) 163 (53) 4 ± 4 (306)
43 ± 17 37 ± 16
81 (27) 225 (74)
45 ± 16 38 ± 16
47 (15) 258 (85)
44 ± 17 39 ± 16
100 (33) 204 (67)
41 ± 18 39 ± 16
110 (36) 29 (10) 72 (24) 46 (15) 49 (16) 103 ± 12
36 ± 16 41 ± 15 38 ± 16 48 ± 17 42 ± 17
240 (78) 2 (1) 30 (10) 7 (2) 27 (9) 3 ± 6 29 ± 23
39 ± 17 44 ± 31 43 ± 17 58 ± 12 42 ± 12
306
306 306
t = 7.2
