SEARCH

SEARCH BY CITATION

Keywords:

  • Impotence;
  • Factor Analysis;
  • Psychometrics;
  • Outcome Assessment

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Aims
  5. Methods
  6. Main Outcome Measures
  7. Results
  8. Discussion
  9. Conclusions
  10. References

Introduction.  The International Index of Erectile Function (IIEF) is the most widely used instrument to assess erectile function in research and clinical practice. However, there are heterogeneous results concerning the factor structure of this questionnaire. The original model assumes five factors (erectile function, intercourse satisfaction, orgasmic function, sexual desire, and overall satisfaction). Others suggested four factors (composite domain of erectile dysfunction and intercourse satisfaction, orgasmic function, sexual desire, and overall satisfaction) or only two factors (sexual function and sexual desire). Because of the high intercorrelation between the different domains, a one-factor model is also plausible.

Aims.  This study assesses and compares these four models of the German version of the IIEF.

Methods.  It was examined which of the models fit best our data from 261 German men in cardiovascular rehabilitation participating in the SPARK study (Sexuality of Patients in the Rehabilitation of Cardiovascular Diseases). Contrary to the former exploratory studies, we used confirmatory factor analysis.

Main Outcome Measures.  Local and global goodness-of-fit measures were calculated.

Results.  The results show that two items (ability to maintain erection and intercourse frequency) could not be represented sufficiently through any of the four models. Based on the global goodness-of-fit indexes, our data proved to be fairly congruent with the original five-factor model and were acceptably represented by a four-factor model as well.

Conclusions.  The original five-factor structure could be confirmed. Due to high intercorrelations, the different domains cannot optimally be discriminated and should be interpreted with caution. Further research is needed to clarify the association between the domains of male sexual function. Kriston L, Günzler C, Harms A, and Berner M. Confirmatory factor analysis of the German version of the International Index of Erectile Function (IIEF): A comparison of four models. J Sex Med 2008;5:92–99.


Introduction

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Aims
  5. Methods
  6. Main Outcome Measures
  7. Results
  8. Discussion
  9. Conclusions
  10. References

Sexuality is a central issue of human life. Therefore, establishing valid and reliable measures of sexual function is indispensable for both clinical work and research. The most frequently used self-report questionnaire for the evaluation of male sexual function is the International Index of Erectile Function (IIEF) [1,2]. This instrument was developed for efficacy assessment of clinical trials and has been translated into several languages [3]. Its use is also widespread in epidemiological surveys [3–5]. Both in clinical trials and in epidemiologolical studies, it became a standard instrument and its application is recommended by several experts [6,7]. The IIEF meets psychometric criteria for validity, reliability, and sensitivity-of-change [3]. Brief versions of the instrument were developed, and validated cutoff values to assess erectile dysfunction (ED) are available [1,3,8–10].

The IIEF is a 15-item multidimensional scale that measures male sexual function along fivedomains: erectile function (six items), intercourse satisfaction (three items), orgasmic function (two items), sexual desire (two items), and overall satisfaction (two items) [1]. Apart from the original article describing the test development, the factorial structure of the IIEF was examined in two published studies. A research group validating the Malay version of the questionnaire could not replicate the original factor structure. Although the orgasmic function, sexual desire, and overall satisfaction domains proved to be reproducible, the erectile function and intercourse satisfaction domains were different from that proposed by the original version [11]. A validation study in Germany revealed a two-factor structure with a sexual function and a sexual desire domain [12]. A common finding in these studies is the high intercorrelation of the domain scores of the IIEF. However, to date all studies examining the factor structure of the questionnaire used exploratory methods, mostly principal component analysis with varimax rotation. This approach is largely data-driven and aims rather to replicate the findings of the test developers, than to confirm them. In general, this exploratory method defines the theoretical context a posteriori, after the factor structure that best fits the data was derivated.

A theory-driven procedure of the examination of factor structures is the confirmatory factor analysis (CFA). This method rests largely on structural equation modeling (SEM) and defines a priori several attributes of the tested factor structure, e.g., the number of factors, item loadings on the factors, and the intercorrelation between factors. In CFA, first the assumed factor structure (model) is defined, then the congruence of the collected data with this model is tested [13]. For further methodological and statistical details for nonexperts, see Table 1.

Table 1.  Methodological and statistical definitions
TermDefinition
Confirmatory factor analysis (CFA)Factor analysis is a statistical technique used to examine the interrelations among a set of variables, or items, in order to identify an underlying structure to those items. This process can be confirmatory, which means that an underlying causal structure (→model) is hypothesized. CFA is mostly performed using →structural equation modeling.
ConstructA theoretical concept or idea. Generally measured indirectly by →indicators. In →confirmatory factor analysis, constructs are usually test →scales or domains.
Goodness-of-fit indexesMeasures of the extent to which data are consistent with theoretical →model. Local indexes concern items (→indicators) and scales (→constructs), global indexes refer to the whole →model.
IndicatorAn indirect measure of a broad concept (→construct) which cannot be measured directly. In →confirmatory factor analysis, indicators are usually test →items.
ItemA question or a preference judgement in a questionnaire or test. In →confirmatory factor analysis, items are usually used as →indicators.
ModelA set of hypotheses about (causal) associations between numerous variables. In →confirmatory factor analysis, the model is the theoretically assumed factor structure.
ReliabilityThe degree to which an assessment or instrument consistently measures an attribute or a →construct.
ScaleAn aggregation (mostly summation) of test →items in order to measure a →construct. Also called domain or dimension. In →confirmatory factor analysis, scales correspond to factors.
Structural equation modeling (SEM)A method for determining the extent to which data on a set of variables are consistent with hypotheses about causal association among the variables. Among other applications, SEM is appropriate for →confirmatory factor analysis.
ValidityThe extent to which a measure (e.g., a →scale) is measuring the →construct, which it is supposed to measure.

Aims

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Aims
  5. Methods
  6. Main Outcome Measures
  7. Results
  8. Discussion
  9. Conclusions
  10. References

The objective of the present study is to test the compatibility of the four proposed models of the factorial structure of the IIEF with data obtained in German rehabilitation centers for patients with cardiovascular diseases. The models that are aimed to be confirmed were extracted from published validation studies (five, four, and two factors) [1,11,12] or rested on theoretical considerations (one factor). The principal aim was to decide which of the competing models shows the best fit with the data. Findings are reported under consideration of publication standards for SEM [14].

Methods

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Aims
  5. Methods
  6. Main Outcome Measures
  7. Results
  8. Discussion
  9. Conclusions
  10. References

Patients

In the context of the project SPARK (Sexuality of Patients in the Rehabilitation of Cardiovascular Diseases), all patients with cardiovascular diseases in five voluntarily participating German rehabilitation centers were administered a questionnaire concerning sociodemographic, cardiovascular disease-specific, sexuality-related, and quality-of-life issues. The questionnaire contained the German version of the IIEF. In total, 1,475 men were asked to participate in the study. A total of 395 (26.8%) agreed on participating in the study and provided analyzable data. Of the 395 men, 261 attempted at least one sexual intercourse over the past 4 weeks and could, therefore, be included in the analysis. The mean age of the analyzed patients was 54.0 years (standard deviation 9.7 years, range 27–78 years). Around a third (37.9%) had a higher education (at least secondary graduation), and the majority (62.6%) was employed full time. A total of 81.5% were married. Most patients (83.5%) were treated for a long-term heart disease or due to (first or repeated) myocardial infarction.

The study was approved by the local ethics committee of the Albert-Ludwigs-University of Freiburg.

Statistical Analysis

A CFA was performed to test four models (factor structure assumptions) of the IIEF: (i) the original five-factor structure as defined by the developers [1]; (ii) a four-factor solution derived from the validation of the Malaysian version of the IIEF [11]; (iii) a two-factor model obtained from the validation of the German version of the IIEF [12]; and (iv) a one-factor solution based on the high intercorrelation of the factors found in these validation studies [1,11,12]. The model published in the Malaysian study originally consisted of five factors (number was fitted) [11]; however, preliminary plausibility analyses with this model led to improper (offending) estimates and, therewith, to uninterpretable results. The reason for this problem was identified in the extremely high intercorrelation of the erectile function and intercourse satisfaction domains; therefore, these two domains were combined in one, resulting in a four-factor model for further analyses. Table 2 shows to which domains the items of the IIEF were assigned in the different models.

Table 2.  Assignment of items to domains in the tested models
Item (indicator)Model
Original five-factorMalay four-factorGerman two-factorGeneral one-factor
  • *

    Loading was fixed to 1.

  • EF = erectile function; IS = intercourse satisfaction; OF = orgasmic function; SD = sexual desire; OS = overall satisfaction; EF-IS = composite domain of erectile function and intercourse satisfaction; SF = sexual function; GSF = general sexual function; SDm = sexual desire (modified).

 1. Erection frequencyEF*EF-IS*SF*GSF*
 2. Erection firmnessEFEF-ISSFGSF
 3. Penetration abilityEFEF-ISSFGSF
 4. Maintenance frequencyEFEF-ISSFGSF
 5. Maintenance abilityEFEF-ISSFGSF
 6. Intercourse frequencyIS*EF-ISSDmGSF
 7. Intercourse satisfactionISEF-ISSFGSF
 8. Intercourse enjoymentISEF-ISSFGSF
 9. Ejaculation frequencyOF*OF*SFGSF
10. Orgasm frequencyOFOFSFGSF
11. Desire frequencySD*SD*SDm*GSF
12. Desire levelSDSDSDmGSF
13. Overall satisfactionOSOS*SFGSF
14. Relationship satisfactionOSOSSFGSF
15. Erection confidenceEFEF-ISSFGSF

The analysis was proceeded by an examination of the assumptions of SEM. CFA was carried out by means of maximum likelihood estimation analyzing the covariance matrix of the items. Constructs were scaled indirectly by fixation of loadings of specific indicators. In all models, intercorrelation of all factors was allowed.

A total of 8.8% (23 of 261) of the cases had at least one missing value through the IIEF, with most of them presenting one missing value. To make an estimation of all indexes in SEM possible, missing values were imputed by multiple imputation. In this way, 1.1% (45 of 3,915) of all data points were estimated.

CFA was performed with AMOS 4.0 (SPSS Inc., Chicago, IL, USA). Missing values were imputed by LISREL 8.51 (Scientific Software International, Inc., Lincolnwood, IL, USA). Additional calculations were made using SPSS 13.0 and Microsoft Excel 2002 (Microsoft Corporation, Redmond, WA, USA).

Main Outcome Measures

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Aims
  5. Methods
  6. Main Outcome Measures
  7. Results
  8. Discussion
  9. Conclusions
  10. References

To evaluate and compare models, both local and global goodness-of-fit measures were calculated. Local goodness-of-fit indexes evaluate the so-called measurement model, i.e., how good the defined constructs were measured. Global goodness-of-fit indexes on the other hand, evaluate the whole (both measurement and factorial) model and allow comparisons between competing models with varying complexity. Table 3 gives an overview of the central measures used with corresponding definitions and assessment guidelines from experts [13,15–18].

Table 3.  Goodness-of-fit indexes with definitions and recommended values
MeasureDefinitionRecommendation for good fit*
  • *

    Multiple values indicate diverse recommendations with the most strict recommendation presented always first.

Local goodness-of-fit
 Indicator reliability (communality)Amount of variance in a given indicator explained by all the constructs jointly≥0.50, ≥0.40
 Critical ratio (CR)Test statistic of significance of correlation between indicator and construct≥1.96
 Construct reliabilityConsistency of the construct≥0.70, ≥0.60
 Variance extractedOverall amount of variance in the indicators accounted for by the construct≥0.50
 Fornell-Larcker ratioRatio of the squared highest intercorrelation of a construct with another construct and the variance extracted<1.0
 χ2-difference testDifference between the tested model and a model in which all construct intercorrelations are set to 1Significant
Global goodness-of-fit
 Discrepancy χ2 testTest of discrepancy between theoretical and observed relationsNot significant by 300 > N > 100
 Goodness-of-fit index (GFI)Relative amount of variance and covariance in observed relations that is explained by the model≥0.90, ≥0.80
 Standardized root mean residual (SRMR)Standardized average residual value derived from the fitting of the theoretical relations to the observed relations<0.05, <0.08
 Root mean square error of approximation (RMSEA)Hypothetical discrepancy between observed and population relations assuming they were available<0.05, <0.06, <0.08
 Tucker-Lewis Index (TLI)Complexity-adjusted comparison of the model with a model defining independent indicators≥0.95, ≥0.90, ≥0.80
 Comparative fit index (CFI)Sample-size adjusted comparison of the model with a model defining independent indicators≥0.95, ≥0.90, ≥0.80
 Normed χ2Degrees of freedom adjusted discrepancy between theoretical and observed relations<1.0, <1.5, <2.0, <3.0, <5.0
 Consistent Akaike Information Criterion (CAIC)Sample-size adjusted extent to which parameter estimates from the original sample will cross-validate in future samplesSmaller values (for model comparisons)

Results

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Aims
  5. Methods
  6. Main Outcome Measures
  7. Results
  8. Discussion
  9. Conclusions
  10. References

Assumptions

Descriptive statistics of items and their intercorrelations are presented in Table 4. All Pearson product–moment correlations were positive, with a median of 0.55. A notable proportion of the correlations (18.1%) were over 0.70. Skewness and kurtosis values ranged from −1.96 to 2.79.

Table 4.  Pearson product–moment correlation matrix and univariate descriptive statistics of items
 Item 1Item 2Item 3Item 4Item 5Item 6Item 7Item 8Item 9Item 10Item 11Item 12Item 13Item 14Item 15
  1. SD = standard deviation.

Item 11.00              
Item 20.7891.00             
Item 30.7790.7721.00            
Item 40.7790.7830.8641.00           
Item 50.5150.4800.5750.6071.00          
Item 60.3500.3390.4100.3310.2211.00         
Item 70.7450.7620.8450.7800.5600.3961.00        
Item 80.7010.6580.7870.7190.5220.4170.8361.00       
Item 90.7170.6590.6790.6770.5100.2650.7080.7011.00      
Item 100.6900.6480.6220.6160.4600.2110.6740.6640.8231.00     
Item 110.3460.3430.4510.3720.2750.3780.4060.4640.4010.3531.00    
Item 120.3920.3560.4320.3270.2260.3890.4350.4790.3940.3380.7001.00   
Item 130.6060.5880.6300.5660.4120.4810.7210.7050.5300.5470.4260.5151.00  
Item 140.5620.5210.5910.5510.3440.3450.6870.6440.4290.4530.3340.3690.7131.00 
Item 150.6670.6860.6530.6850.3920.2960.6880.6550.5590.5270.3920.4840.6440.5531.00
Mean4.053.983.993.743.342.743.933.424.284.133.573.273.023.753.52
SD1.431.401.471.531.651.361.441.221.341.401.000.830.951.141.10
Skewness−1.41−1.24−1.40−1.02−0.550.27−1.15−1.04−1.96−1.56−0.48−0.21−0.48−0.78−0.34
Kurtosis0.860.480.89−0.16−1.12−1.160.100.792.791.31−0.080.26−0.14−0.06−0.64

Local Goodness-of-fit

Indicator reliability exceeded 0.40 in most cases, with the exception of items 5 and 6 in all tested models, as well as items 11 and 12 in the general one-factor model (see Table 5). Critical ratios were significant (≥1.96) by all items in all models.

Table 5.  Local goodness-of-fit indexes for indicators
Item (indicator)Model
Original five-factorMalay four-factorGerman two-factorGeneral one-factor
IRCRIRCRIRCRIRCR
  • *

    P < 0.001.

  • IR = indicator reliability; CR = critical ratio; ref. = reference.

 1. Erection frequency0.750ref.0.733ref.0.734ref.0.731ref.
 2. Erection firmness0.73818.9*0.71418.0*0.71018.0*0.70617.8*
 3. Penetration ability0.85221.9*0.83821.1*0.82620.1*0.82620.7*
 4. Maintenance frequency0.83021.30.78319.7*0.77119.4*0.76619.2*
 5. Maintenance ability0.38211.3*0.37411.1*0.37111.0*0.36911.0*
 6. Intercourse frequency0.204ref.0.1847.2*0.2387.4*0.1887.3*
 7. Intercourse satisfaction0.8767.8*0.84021.1*0.84221.2*0.84121.1*
 8. Intercourse enjoyment0.7937.7*0.74418.80.75118.9*0.75519.0*
 9. Ejaculation frequency0.888ref.0.886ref.0.62016.0*0.61815.9*
10. Orgasm frequency0.76319.9*0.76419.9*0.56214.8*0.55814.7*
11. Desire frequency0.640ref.0.640ref.0.674ref.0.2338.3*
12. Desire level0.76610.2*0.76510.3*0.70511.4*0.2478.6*
13. Overall satisfaction0.777ref.0.790ref.0.55214.6*0.56014.7*
14. Relationship satisfaction0.65515.5*0.64414.9*0.46612.9*0.46912.9*
15. Erection confidence0.56415.0*0.57515.1*0.57815.1*0.57915.1*

Construct reliability exceeded 0.70 and the variance extracted was above 0.50 by all domains of the tested models (see Table 6). Also, all models were significantly better than their alterations, with maximal intercorrelation (1.00) between the domains. The Fornell-Larcker ratio exceeded the critical value of 1.00 by the erectile function, intercourse satisfaction, and overall satisfaction domains of the original five-factor model, as well as by the composite domain of erectile function and intercourse satisfaction in the Malay four-factor model (Table 6).

Table 6.  Local goodness-of-fit indexes for constructs
Model/constructConstruct reliabilityVariance extractedFornell-Larcker ratioχ2-difference test (χ2/d.f.)
  • *

    P < 0.001.

  • d.f. = degrees of freedom; n.c. = not computable.

Original five-factor
 Erectile function0.9280.6861.27139.13/10*
 Intercourse satisfaction0.8220.6241.40 
 Orgasmic Function0.9040.8260.80 
 Sexual desire0.8250.7030.51 
 Overall satisfaction0.8340.7151.12 
Malay four-factor
 Erectile function–intercourse satisfaction0.9400.6431.08125.56/6*
 Orgasmic function0.9040.8250.83 
 Sexual desire0.8250.7030.52 
 Overall satisfaction0.8350.7170.97 
German two-factor
 Sexual function0.9650.6480.5414.27/1*
 Sexual desire (modified)0.7700.5390.65 
General one-factor
 General sexual function0.9490.563n.c.n.c.

In the original five-factor model, intercorrelation between domains ranged from 0.49 to 0.94, with three intercorrelations exceeding 0.80 (between erectile function and intercourse satisfaction r = 0.94; between intercourse satisfaction and overall satisfaction r = 0.90; between erectile function and orgasmic function r = 0.81). In the Malay four-factor model, correlations ranged from 0.49 to 0.83. Two intercorrelations were above 0.80 (the composite domain of erectile function and intercourse satisfaction correlated both with overall satisfaction and orgasmic function at an r = 0.83). In the German two-factor model, a moderate correlation (= 0.59) was found between the sexual function domain and the modified sexual desire domain.

Global Goodness-of-fit

Global goodness-of-fit indexes are displayed in Table 7. The discrepancy χ2 test proved to be statistically significant by all models, indicating a difference between theoretical and observed relations. The GFI was acceptably high and the standardized root mean residual (SRMR) was desirably low in the original five-factor model (GFI = 0.889; SRMR = 0.045) and in the Malay four-factor model (GFI = 0.849; SRMR = 0.049). the root mean square error of approximation exceeded the critical value of 0.80 for an acceptable fit in all tested models. Both the Tucker-Lewis Index and the comparative fit index showed a good fit (≥0.90) in the original five-factor and the Malay four-factor model, and were acceptable (≥0.80) in the German two-factor and the general one-factor model. The normed χ2 values of 3.13 and 3.95 for the original five-factor and the Malay four-factor model proved to be acceptable, however by far not optimal. The Consistent Akaike Criterion favored the original five-factor model with a value of 512.68.

Table 7.  Global goodness-of-fit measures in the tested models
Goodness-of-fit indexModel
Original five-factorMalay four-factorGerman two-factorGeneral one-factor
  • *

    P < 0.001.

Discrepancy χ2 test (χ2/d.f.)250.10/80*332.15/84*503.95/89*630.05/90*
Goodness-of-fit index0.8890.8490.7830.743
Standardized root mean residual0.0450.0490.0640.072
Root mean square error of approximation0.0900.1070.1340.152
Tucker-Lewis Index0.9330.9080.8540.812
Comparative fit index0.9490.9260.8760.839
Normed χ23.133.955.667.00
Consistent Akaike Information Criterion512.68568.47707.45826.99

Discussion

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Aims
  5. Methods
  6. Main Outcome Measures
  7. Results
  8. Discussion
  9. Conclusions
  10. References

A CFA of the IIEF with competing models was performed in patients with cardiovascular diseases who were recruited from German rehabilitation centers.

Central conditions of SEM were met: (i) the sample size exceeded N = 200; (ii) skewness and kurtosis values were below 3 and 10, respectively; and (iii) sufficient correlations between items were present to enable a structure definition [15]. However, some item intercorrelations were very high and suggested possible redundance.

The analysis of local goodness-of-fit indexes revealed that the variance in item 5 (maintenance ability) and item 6 (intercourse frequency) could not be sufficiently explained by any of the models. Erection maintenance ability seems to be only moderately associated with the construct of erectile function in the questionnaire. This is interesting insofar as the maintenance forms the central condition in the clinical definition of ED. Similarly, intercourse frequency can only marginally be claimed as an indicator of intercourse satisfaction. A possible explanation for this may be that intercourse frequency depends on several factors (partner desire, living conditions, etc.) and can be only partly explained by intercourse satisfaction. The item-related indexes of the general one-factor model suggest that all items might be considered as indicators of general sexual function, with exception of the mentioned items 5 and 6, as well as the sexual desire domain (items 11 and 12). However, combining the erectile function and intercourse satisfaction domains in one factor (as in the Malay four-factor model), or differentiating between only general sexual function and sexual desire (with intercourse frequency as an indicator of the latter, as in the German two-factor model), did not lead to improved reliability of items 5 and 6.

Considering construct reliability and extracted variance, all models proved to be acceptable. However, Fornell-Larcker ratios indicate difficulties in differentiating between constructs in some cases (sometimes termed as discriminant validity). In this way it seems indeed questionable, if the erectile function and intercourse satisfaction domains of the original five-factor model are distinct constructs. Even if they are considered as one, as in the Malay four-factor model, a separation from other domains, mainly from overall satisfaction, remains still problematic. A sufficient discrimination between constructs is first achieved in the German two-factor model (sexual function and sexual desire).

The examination of global goodness-of-fit indexes suggests that the original five-factor model provides the best fit with observed data. Although not optimal, an acceptable congruence between model and observed data was achieved. Therefore, the factor structure of the IIEF, as proposed and disseminated by Rosen et al. [1], could be confirmed and may count as cross-validated. Therefore, the wide use of the German scale with the proposed factors seems warranted. Also, the four-factor structure of the Malay version, with combining erectile function and intercourse satisfaction in one domain, fits the data sufficiently. Based on the global goodness-of-fit indexes, the German two-factor solution must be rejected, and also, the consideration of all items as indicators of a general sexual function is inappropriate.

The relatively weak results for item 5 and item 6 are also to find in previous validation studies. So erection maintenance ability (item 5) shows the weakest association with the erectile function domain also in the test development sample [1], and remains largely unexplained in the validation study of the Malay version [11]. Similarly, intercourse satisfaction (item 6) shows only moderate correlations with domains and items in both the German and the Malay validation studies [11,12]. However, these item-related limitations alone would not justify a complete revision of the instrument.

Differences in the results obtained in different studies could be due to either differences in the populations or linguistic-cultural differences. All of the subjects in this study had cardiovascular disease, which may have affected the pattern or distribution of the scores. For example, it may be expected that these patients, in general, practice less frequent sexual intercourse because of health complaints or anxiety than healthy persons. Furthermore, subtle differences in the meaning of certain terms from the English to the German version may be present. Although linguistic validation procedures were used in translating the questionnaire, there may still be differences in the meaning of the terms, which could influence the factor structure. For example, the term “maintain erection” may have different semantic characteristics in the United States, in Germany, and in Malaysia. A culture-based subjective assessment is inevitable in these cases, and may lead to discrepancies between perceptions of the national questionnaire versions and study results from different countries. Nevertheless, as long as these discrepancies do not exceed a certain level and, statistically, the item correlations and factor structures are similar to a sufficient degree, they support external (implicating generalizability) rather than limit internal validity of the instrument.

A central issue of using the IIEF is the consistently high intercorrelation between the domains, mainly erectile function, intercourse satisfaction, and overall satisfaction. Although some association may be expected, such strong interrelations (in this and other validation studies, correlations above 0.70) raise the question of whether these constructs can reliably be discriminated from each other. The answer probably depends on the point of view. Psychometrically, these constructs cannot be optimally differentiated, but an examination of the content reveals a clear face validity. The use of short versions of the IIEF, such as the erectile function domain alone (IIEF-6) or the Sexual Health Inventory for Men (IIEF-5), is not affected by these findings, because their reliability and validity were demonstrated in several studies [3,8–10].

Conclusions

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Aims
  5. Methods
  6. Main Outcome Measures
  7. Results
  8. Discussion
  9. Conclusions
  10. References

In general, we encourage the use of the full version of the IIEF but advise caution by interpreting and comparing the results obtained by the different domains of the instrument. This must be especially taken into account when using the instrument for diagnostic purposes. Although the full version seems to be more appropriate to follow research goals, treatment efficacy should be assessed rather by the IIEF-5 or IIEF-6. Further research may focus on the structure of male sexuality by establishing models that adequately fit to observed data. Findings regarding the IIEF suggest that the constructs sexual desire, erectile function, and, to a lesser extent, also orgasmic function as aspects of male sexual function are reproducible corresponding to the classification of sexual appetence, arousal, and orgasm disorders in the Diagnostic and Statistical Manual of Mental Disorder, Fourth Edition [19]. Furthermore, procedural (if possible, causal) associations between these dimensions should be a main focus of future research interest.

Conflict of Interest: Dr. Berner currently holds research grants from Pfizer Inc. He received tuition fees from Bayer Inc., Lilly Icos Inc., and Pfizer Inc. He also received travel expenses from the European Sexual Dysfunction Alliance (ESDA). Mr. Kriston previously received travel expenses reimbursement from Pfizer Inc.

References

  1. Top of page
  2. ABSTRACT
  3. Introduction
  4. Aims
  5. Methods
  6. Main Outcome Measures
  7. Results
  8. Discussion
  9. Conclusions
  10. References
  • 1
    Rosen RC, Riley A, Wagner G. The International Index of Erectile Function (IIEF): A multidimensional scale for assessment of erectile dysfunction. Urology 1997;49:82230.
  • 2
    Corona G, Jannini EA, Maggi M. Inventories for male and female sexual dysfunctions. Int J Impot Res 2005;18:23650.
  • 3
    Rosen RC, Cappelleri JC, Gendrano N III.The International Index of Erectile Function (IIEF): A state-of-the-science review. Int J Impot Res 2002;14:22644.
  • 4
    De Boer BJ, Bots ML, Nijeholt AA, Moors JP, Pieters HM, Verheij TJ. Impact of various questionnaires on the prevalence of erectile dysfunction. The ENIGMA Study. Int J Impot Res 2004;16:2149.
  • 5
    Dean J. Characterisation, prevalence, and consultation rates of erectile dysfunction. Clin Cornerstone 2005;7:511.
  • 6
    Hirsch M, Donatucci C, Glina S, Montague D, Montorsi F, Wyllie M. Standards for clinical trials in male sexual dysfunction: Erectile dysfunction and rapid ejaculation. J Sex Med 2004;1:8791.
  • 7
    Lewis RW, Fugl-Meyer KS, Bosch R, Fugl-Meyer AR, Laumann EO, Lizza E, Martin-Morales A. Epidemiology/risk factors of sexual dysfunction. J Sex Med 2004;1:359.
  • 8
    Cappelleri JC, Rosen RC, Smith MD, Mishra A, Osterloh IH. Diagnostic evaluation of the erectile function domain of the International Index of Erectile Function. Urology 1999;54:34651.
  • 9
    Cappelleri JC, Rosen RC, Smith MD, Quirk F, Maytom MC, Mishra A, Osterloh IH. Some developments on the International Index of Erectile Function (IIEF). Drug Inf J 1999;33:17990.
  • 10
    Rosen RC, Cappelleri JC, Smith MD, Lipsky J, Pena BM. Development and evaluation of an abridged, 5-item version of the International Index of Erectile Function (IIEF-5) as a diagnostic tool for erectile dysfunction. Int J Impot Res 1999;11:31926.
  • 11
    Lim TO, Das A, Rampal S, Zaki M, Sahabudin RM, Rohan MJ, Isaacs S. Cross-cultural adaptation and validation of the English version of the International Index of Erectile Function (IIEF) for use in Malaysia. Int J Impot Res 2003;15:32936.
  • 12
    Wiltink J, Hauck EW, Phadayanon M, Weidner W, Beutel ME. Validation of the German version of the International Index of Erectile Function (IIEF) in patients with erectile dysfunction, Peyronie's disease and controls. Int J Impot Res 2003;15:1927.
  • 13
    Hair JF, Anderson RE, Tatham RL, Black WC. Structural equation modeling. In: HairJF, AndersonRE, TathamRL, BlackWC, eds. Multivariate data analysis. Upper Saddle River, NJ: Prentice Hall; 1998:577667.
  • 14
    Boomsma A. Reporting analyses of covariance structures. Struct Equation Model 2000;7:46183.
  • 15
    Kline RB. Principles and practice of structural equation modeling. New York, NY: Guilford Press; 1998.
  • 16
    Byrne BM. Structural equation modeling with AMOS. Basic concepts, applications, and programming. Hillsdale, NJ: Lawrence Erlbaum; 2001.
  • 17
    Arbuckle JL, Wothke W. AMOS 4.0 user's guide. Chicago, IL: Small Waters Corporation; 1999.
  • 18
    Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct Equation Model 1999;6:155.
  • 19
    American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th edition. Washington, DC: American Psychiatric Association; 1994.