ORIGINAL ARTICLE Evidence for a genetic etiology to ejaculatory dysfunction

(2009) 21, 62 67 & 2009 Nature Publishing Group All rights reserved 0955-9930/09 $32.00 www.nature.com/ijir ORIGINAL ARTICLE Evidence for a genetic etiology to ejaculatory dysfunction P Jern, P Santtila, A Johansson, M Varjonen, K Witting, B von der Pahlen and NK Sandnabba Department of Psychology, Center of Excellence in Behavior Genetics, A bo Akademi University, Turku, Finland A number of theoretical approaches to understanding the etiology of ejaculatory dysfunction have been proposed, but no study has yet found conclusive evidence that premature (PE) or delayed (DE) ejaculation is under genetic control. We conducted twin model fitting analyses on different indicator variables of ejaculatory function on a population-based sample of 3946 twins and their siblings (age 18 48; mean ¼ 29.9 years) to investigate genetic, shared environmental and unique environmental effects on PE and DE. A significant moderate genetic effect (28%) was found for PE. No clear-cut familial effect could be detected for DE. Significant associations between ejaculatory function and age were detected, but effects of age were generally very weak. The findings from the present study provide useful information regarding the etiology and understanding of ejaculatory dysfunction. (2009) 21, 62 67; doi:10.1038/ijir.2008.61; published online 11 December 2008 Keywords: premature ejaculation; genetic effects; twin study Introduction Premature ejaculation (PE) is one of the most commonly reported sexual dysfunctions in men. 1 Prevalence estimates of PE vary widely between studies, with reports ranging from around 3% 2 to around 30% 3 depending on definition and diagnostic criteria. For its etiology, psychological, environmental and contextual factors are all still being considered, 4,5 whereas recent research has focused on neurobiological, 6 endocrine 7,8 and genetic explanatory frameworks. 9,10 In recent years, researchers have tended to favor theoretical approaches that incorporate several different etiological factors to PE. For instance, Waldinger and Schweitzer 11 have proposed a system dividing PE into four subtypes as a function of the intravaginal ejaculation latency time (IELT) and their courses in life. The different subtypes are hypothesized to have different etiologies, with lifelong PE thought to be mostly neurobiological and/or genetic in nature, whereas the other types are thought to have a psychological or medical etiology. This approach, if proven to be empirically valid and Correspondence: P Jern, MPsych, Department of Psychology, Center of Excellence in Behavior Genetics, Åbo Akademi University, 20500 Turku, Finland. E-mail: pjern@abo.fi Received 30 June 2008; revised 9 October 2008; accepted 20 November 2008; published online 11 December 2008 clinically suitable, allows for the inclusion of different etiological frameworks. Delayed ejaculation (DE) is a less common sexual dysfunction, and few studies have looked into its etiology. 12 Approximate prevalence estimates of DE in sexually active men have been reported to be 1 4%, 12,13 whereas Nathan 14 found the prevalence to be as low as 0.15% in the general male population. Rosen 12 suggested that DE may be associated with a variety of surgical or medical conditions, and in most cases, the use of antiadrenergic or neuroleptic drugs. Jern et al. 10 found a moderate familial effect active in DE, but results were inconclusive regarding separation between shared environmental and genetic effects. Indeed, they concluded that the probability of a genetic etiology in DE is negligible. A genetic etiology to PE in humans has been suggested already in the 1940s, when Schapiro 15 noted (among other findings) that PE seems to be familial. This conclusion has since been supported by Waldinger et al. 9 They found that the odds of the problem co-occurring between family members were considerably higher than those expected solely on the basis of prevalence rates in the population. The authors concluded that this observation supports the role of genetics in affecting PE. Although plausible, familial resemblance in itself cannot prove genetic influence, as individuals who are genetically related also most often share their environment to some extent. Therefore, shared environmental factors could also explain a familial occurrence of PE. Jern et al., 10 in a population-based

study of Finnish men, found a highly significant familial effect in PE, which strong implications that this effect be genetic. However, they were not able to separate clearly the genetic effects from shared environmental effects, probably because of insufficient statistical power. Furthermore, animal studies have revealed that mice lacking the gene for endothelial nitric oxide synthase develop symptoms similar to PE, whereas mice lacking the gene for heme oxygenase-2 develop DE or anejaculation. 16 However, no study has to date, empirically or otherwise, proven a genetic effect on PE in humans, although results of these preliminary studies have been encouraging. Age has historically been thought to be an important causal factor of PE, but research conducted in recent years has indicated that this may not be the case. For example, Rosen 12 noted that the prevalence of PE is stable across age groups, and Waldinger et al. 17 found that median IELTs actually decreased with age. The aim of this study was to examine a number of different indicator variables of ejaculatory function and performance, and to what extent genetic and environmental effects are active in ejaculatory dysfunction. In continuation of the results of Jern et al., 10 PE was expected to be under moderate genetic influence, whereas no such effects were expected for DE. Instead, shared environmental factors were hypothesized to have some effect on DE. Finally, the effects of age on different PE indicator variables were investigated, with little or no effects of age expected. Materials and methods Participants The analyses presented in this study were based on 3946 male twins and their siblings who had provided information on their ejaculatory function. The mean age of the men was 29.92 (s.d. ¼ 6.91) years. The sample consisted of 933 monozygotic (MZ) twins, 1104 dizygotic (DZ) male twins, 921 male twins from opposite-sex DZ twin pairs, 136 male twins whose zygosity could not be determined and 852 male siblings of twins. Participants were a subset from the Genetics of Sex and Aggression (GSA) sample. The main GSA sample consists of two different data collections. The first data collection was carried out in 2005 and targeted 33- to 43-year-old twins. The second data collection was carried out in 2006 and targeted 18- to 33-year-old (there was no overlap between the data collections) twins and their over 18-year-old siblings. The individuals targeted for the second data collection were contacted by postal mail and asked if they would be interested in completing a questionnaire and to give saliva samples for DNA and hormone analyses (as the present study does not report on the DNA and hormone samples, no further information is given regarding them). The research plan for the first data collection consisting only of a questionnaire study was approved by the ethics committee of the Department of Psychology at Åbo Akademi University and for the second data collection also including DNA and hormone samples by the ethics committee of the Åbo Akademi University. For more information regarding the sample, please consult Santtila et al. 18 Instruments Ten questions were used to tap objective and subjective aspects of ejaculatory function. The questions were designed to take different definitional aspects into account (for example, ejaculation latency time, subjective experience, frequency of ejaculatory problems, number of thrusts). To avoid heterosexist bias and exclusion of both female male and male male anal intercourse, we used a gender-neutral definition of ejaculation latency time inclusive of anal intercourse. The questions were adapted from an unpublished questionnaire developed by Grenier and Byers 19 (see Jern et al. 10 for details). Factor analyses First, exploratory factor analyses (EFAs) were conducted on raw data. Variables not distributed normally were subjected to either logarithmic or square-root transformations. After excluding participants with more than 50% missing values, 3912 men were selected for the EFA, starting with a principal components analysis. Missing values were replaced with item-specific means. Second, based on the results of the EFA, confirmatory factor analyses were conducted. Given the relatively large sample size, and the fact that the individual variables were on the ordinal level, we opted for the asymptotically distribution-free estimation method. 20 To improve model fit, the error terms between two variables ( feeling of control and tried to slow down intercourse ) were allowed to correlate. The associations between ejaculatory function and age were measured with the Complex Samples General Linear Model application of SPSS 14.0. This procedure allows for correlated data and adjusts the estimates of standard errors, which was necessary for the present data consisting of responses by siblings and twins. We performed these analyses on the individual items as well as the composite variables. Genetic modeling The present sample yielded 228 complete MZ pairs (r g ¼ 1.0), 225 complete DZ pairs with an additional 63

64 349 complete twin sibling or sibling sibling pairs (all other types of brother pairings have a genetic resemblance of around 50%; thus r g ¼ 0.50 for all pairings except MZ twins). In 683 families, at least two family members had answered the questionnaire, with the highest number of participating male family members being six (one twin pair and four additional brothers). First, we tested for age effects on the composite variables. As small, but significant, age effects could be detected for two out of three composite variables, an age regression was performed on all variables before genetic model fitting. Next, we compared the different twin groups (monozygotes, dizygotes from an all-male twin pair, dizygotes from an opposite-sex twin pair and twins whose zygosity could not be determined) with one another to check for level differences between the groups. No such differences could be detected (all Ps40.941). Thus, dizygotes from both all-male and opposite-sex twin pairs were selected into the same group for the genetic analyses, resulting in one group of MZ twins (r g ¼ 1.0) and one group with all additional combinations of twins and siblings (r g ¼ 0.5). In addition, for those twin pairs whose zygosity could not be determined by standard criteria, one twin of each full nondetermined twin pair was randomly selected to this group in order not to lose information from the pairings of the siblings with one of the twins. Phenotypic correlations with adjacent 95% confidence intervals were then calculated for both factor models using the composite variables. Genetic effects were expected to be present if the correlation of the r g ¼ 1.0 group was substantially larger than (that is, around twice) the correlation of the r g ¼ 0.5 group (should the correlations of the r g ¼ 1.0 group be notably larger than twice those of the r g ¼ 0.5 group, there would be reason to expect dominance effects). The calculations were carried out with the Mx statistical package. 21 We also tested for differences in means and standard deviations between the groups. Next, the composite variables were used in a univariate model-fitting Mx script with full-information maximum likelihood analysis. Excluding epistatic interaction and dominance, we considered three sources of variance: additive genetic effects (A), shared environmental effects (C) and nonshared environmental effects (E). After examination of the phenotypic correlations, we decided to fit the ACE model for all variables. After the full model (ACE) was fitted to the data, it was followed by an AE model, a CE model and an E model containing only nonshared environmental effects. An AE model tests the fit of a model in which the component measuring shared environmental effects has been omitted against the full ACE model measuring all three components, such as additive genetic effects (A), shared environmental effects (C) and nonshared environmental effects (E), the latter also including measurement error. If the fit of the simpler or nested Table 1 Factor loadings for the one- and two-factor solutions from the exploratory factor analysis Variable model (that is, with omitted components; in this example the AE model where the C component has been dropped) does not decrease significantly from that of the full ACE model, the dropped component has very little or no measurable effect on the phenotype, and the nested model should be preferred. 21 The fit of the different models was compared by taking the fit function ( 2 log-likelihood of raw data) and the degrees of freedom of the ACE model, and subtracting it from the fit function and degrees of freedom of the AE, CE and E models. The subtraction gives a w 2 -value with associated degrees of freedom, which in turn can be tested for significance. Between models with equal fit according to the w 2 -test, the Akaike s Information Criterion (AIC) was used as an additional index of model fit. 22 Results One-factor model Two-factor model Factor 1 (PE) Factor 2 (DE) Subjective experience of 0.797 0.797 PE Worrying about PE 0.625 0.710 Feeling of control 0.524 0.624 Ejaculation latency time 0.648 0.615 Ejaculation before 0.408 0.414 intercourse Number of thrusts 0.446 0.414 Tried to slow down 0.480 0.391 ( 0.351) intercourse Tried to speed up 0.285 0.764 intercourse Later ejaculation than 0.256 0.761 desired Pretended to ejaculate 0.025 0.271 Abbreviations: DE, delayed ejaculation; PE, premature ejaculation. Note. Principal components analysis. The two-factor solution employs direct oblimin rotation. Exploratory factor analyses A principal components analysis revealed four factors with an eigenvalue exceeding 1, with the scree plot suggesting a two-factor solution to be the best suited. The resulting two-factor solution replicated the finding of Jern et al., 10 with one variable ( tried to slow down intercourse ) having a complex loading. Overall, test statistics indicated good factorability (KMO measure of sampling adequacy ¼ 0.693; Bartlett s test of sphericity Po0.001). As previous research has suggested a unidimensional construct, we also extracted a onefactor solution. Factor loadings for the one- and twofactor solutions can be seen in Table 1.

Table 2 Model estimation fit statistics for the confirmatory factor analysis 65 Model P GFI TLI RMSEA PCLOSE Hoelter s N (Po0.01) Two-factor model 0.000 0.976 0.862 0.043 0.984 766 One-factor model 0.000 0.968 0.813 0.050 0.461 580 Abbreviations: GFI, goodness-of-fit index; PCLOSE, null hypothesis test for the RMSEA; RMSEA, root mean square error of approximation; TLI, Tucker Lewis index. Note. Asymptotically distribution-free estimation. Two error terms were allowed to correlate for each model. Table 3 Age effects on the composite variables and individual PE indicator variables Variable Correlation with age F t R 2 PE composite 0.037* 4.269* 2.066* 0.001 DE composite 0.030 3.356 1.832 0.001 Number of thrusts 0.061** 10.954*** 3.310*** 0.004 Ejaculation latency time 0.122** 32.332*** 5.686*** 0.011 Feeling of control 0.082** 26.143*** 5.113*** 0.007 Worrying about PE 0.048** 9.877** 3.143** 0.002 Pretended to ejaculate 0.028 2.136 1.462 0.001 Ejaculation before intercourse 0.099** 27.773*** 5.270*** 0.010 Tried to speed up intercourse 0.030 4.170* 2.042* 0.001 Tried to slow down intercourse 0.059** 11.850*** 3.442*** 0.004 Subjective experience of PE 0.000 0.001 0.039 0.000 Later ejaculation than desired 0.013 0.644 0.802 0.000 Abbreviations: DE, delayed ejaculation; PE, premature ejaculation. ***, significant on the Po0.001 level. **, significant on the Po0.01 level. *, significant on the Po0.05 level. Table 4 Phenotypic correlations (with 95% confidence intervals) for the two-factor model of ejaculatory function Sibling pair type Two-factor model A C E Premature ejaculation Delayed ejaculation r g ¼ 1.0 0.30 (0.18 0.40) 0.03 ( 0.09 0.14) r g ¼ 0.50 0.13 (0.04 0.21) 0.09 (0.01 0.17) PE = 0.28 DE = 0.00 PE = 0.00 DE = 0.07 PE = 0.72 DE = 0.93 r g ¼ 1.0 refers to monozygotic twin pairs. r g ¼ 0.50 refers to any kind of sibling pairs excluding monozygotic twin pairs (i.e., siblings that share 50% of their genes). Confirmatory factor analyses We began this procedure by computing two different two-factor models. In the first of these, the variable tried to slow down intercourse, which had complex loadings (that is, reasonably strong loadings on each of the two factors in the EFA), was loaded on the PE factor. In the second of these, this variable was loaded on the DE factor. The fit of the first model was better (w 2 ¼ 394.095, d.f. ¼ 34) compared to the second model (w 2 ¼ 453.663, d.f. ¼ 34); thus the former model was selected for further analyses. Next, we computed a one-factor model to compare it to the selected two-factor model. The fit of the onefactor model was worse (w 2 ¼ 453.663, d.f. ¼ 35). A model fit comparison indicated that this difference Ejaculative function Figure 1 Additive genetic (A), shared environmental (C) and unique environmental (E) effects on ejaculative function. PE, premature ejaculation; DE, delayed ejaculation. was significant (Dw 2 ¼ 59.568, Dd.f. ¼ 1, Po0.001). The model fit statistics for solutions can be seen in Table 2. As the model fit results for the one-factor solution were rather inadequate in both factor analyses, we proceeded using only the two-factor solution selected for all subsequent analyses. Hence, two different composite variables (measuring PE and DE; respectively) were computed. Significant age effects were detected for some of the variables. For the composite scores, age had a significant but weak association with PE (r ¼ 0.037, Po0.05; F ¼ 4.269, P ¼ 0.039; R 2 ¼ 0.001); with problems related to PE increasing with age. No

66 Table 5 Results of the univariate model fitting analyses Test statistic Two-factor solution Premature ejaculation Delayed ejaculation Full model Nested models a Full model Nested models a ACE AE CE E ACE AE CE E 2 LL (Dw 2 ) 21 187.384 0.000 5.167 32.459 18 084.601 1.523 0.000 4.306 P 1.000 0.023 0.000 0.217 1.000 0.116 AIC (DAIC) 13 401.384 2.000 3.167 28.459 10 298.634 0.477 2.000 0.306 Abbreviations: AIC, Akaike s Information Criterion. 2 LL ¼ 2 times log-likelihood of data. a Note. Nested models use test statistics mentioned in brackets. significant age effects could be found for DE (r ¼ 0.030, NS; F ¼ 3.356, P ¼ 0.067). Generally, items from the PE factor tended to have significant but very weak associations (both in terms of correlations and effect sizes) with age, whereas items from the DE factor showed no such tendency. The associations between age and all variables can be seen in Table 3. Genetic modeling The phenotypic correlations for both factor solutions were rather modest. As expected, there were indications of significant genetic effects on PE, but not DE, in the two-factor solution. The phenotypic correlations are presented in Table 4. As the correlations of the r g ¼ 1.0 group were slightly larger than twice the correlations of the r g ¼ 0.5 group, models testing for genetic dominance were also fitted to the data. However, no significant effects of genetic dominance were detected. The full model fitting analyses revealed a significant familial effect (that is, evidence of either additive genetic or shared environmental influence). A summary of the model fitting analyses of the full models can be seen in Figure 1. As expected, moderate genetic effects (28%) were present for the PE factor in the two-factor model, but not for DE. The full model suggested no additive genetic effects for DE, instead a weak shared environmental effect could be distinguished. For both phenotypes, unique environmental effect accounted for most of the variance. When nested models were tested against the full model, a different picture emerged. The results from these analyses can be viewed in Table 5. For the twofactor solution, the A component cannot be dropped without a significant decrease in model fit in the case of PE, unlike the case of the C component, which could be omitted without jeopardizing the adequacy of the model s fit. For the DE factor, on the other hand, both familial components could be dropped without significantly reducing model fit; contrary to the finding of Jern et al., 10 where a both the A and C components could be dropped, but not simultaneously. However, the fact that the AE model is approaching significance for DE in a clearly more striking fashion than the CE model does suggest that the general tendency of these results is in line with the results of the previously mentioned study. Discussion The results of this study advocate the use of a twofactor solution, such as the one presented in Jern et al. 10 From this follows that PE and DE may not necessarily be the opposite extremes of a distribution as a function of latency time, but two separate conditions with distinct etiologies. From an evolutionary point of view, this makes sense in that PE could be (or have been) beneficial under some circumstances, thus making PE a genetic adaptation, 23 whereas DE would be completely maladaptive. Thus, there is reason to expect genetic effects for PE, but not for DE. A significant genetic effect was detected for PE, confirming the hypothesis of Waldinger et al. 9 in accordance with the findings of Jern et al. 10 This finding, together with other recent discoveries relating to the neurobiology of PE, 7 suggests that there is reason to be cautiously optimistic about identifying specific genes or polymorphisms that are associated with PE. These could be effective either directly, or indirectly, by mediation of hormones such as estrogen or testosterone. Endocrine factors have been recently associated with both PE and DE. For example, elevated testosterone levels were more prevalent in PE patients in 25- to 40-year-old Italian men, whereas decreased levels of testosterone were associated with DE in 55- to 70-year olds. 8

In addition, estrogen may have a mediating role in ejaculatory function. 7 Waldinger and Schweitzer 24 hypothesized that genetic effects would have a larger role in the development of lifelong PE, whereas other etiological factors would be the main causal factors of the other three PE subcategories. In other words, a strong genetic etiology should be more likely in men with IELT values of less than 1 1.5 min, which coincides with the 0.5 and 2.5 percentiles of the IELT in the general population. This would mean that genetic effects would be influential only in a rather small minority (2%) of cases of PE. The finding of some evidence for a genetic influence on PE is concordant with Waldinger s idea that only a small part of the men suffering from PE (that is, men with lifelong PE and IELTs o1 min) have a genetically determined dysfunction. The other men complaining of rapid ejaculation, but with IELTs of more than 1.5 min, probably do not suffer from a genetic form of PE but from one of the other subtypes of PE. The evidence for genetic effects in the present study applies to the whole range of variation in PE measured using a number of difference indicator variables. In future studies, it could indeed be worthwhile to investigate genetic effects separately for different indicators of PE and DE, in addition to comparing genotypes of patients suffering from lifelong PE to controls. Effects of age on PE have been recently dismissed in a number of studies. In his review article, Rosen 12 reported that prevalences for PE were stable across age levels, whereas Waldinger et al. 17 found that ejaculation latency time actually seemed to decrease with age. In this study, the results of Waldinger et al. 17 were replicated in that problems related to PE showed a tendency to increase slightly with age. However, although significant, the effects of age in the present study were very weak, and probably negligible in practice. In conclusion, the findings of the present study confirmed the existence of a moderate (28%) genetic effect on PE, with environmental effects unique to the individual accounting for most of the variance. No significant familial effect could be detected for DE. Age was significantly associated with PE, although effects of age were generally very weak. Acknowledgments This research was financed by grant no. 210298 from the Academy of Finland and a Centre of Excellence Grant from the Stiftelsen för Åbo Akademi Foundation. References 1 Jannini E, Lombardo F, Lenzi A. Correlation between ejaculatory and erectile dysfunction. Int J Androl 2005; 28: 40 46. 2 Waldinger M. Lifelong premature ejaculation: current debate on definition and treatment. J Men s Health Gend 2005; 2: 333 338. 3 Montorsi F. Prevalence of premature ejaculation: a global and regional perspective. J Sex Med 2005; 2: 96 102. 4 McMahon C. Premature ejaculation: past, present and future perspectives. J Sex Med 2005; 2(Suppl 2): 94 95. 5 Althof S. The psychology of premature ejaculation: therapies and consequences. J Sex Med 2006; 3(Suppl 4): 324 331. 6 Waldinger M. The neurobiological approach to premature ejaculation. J Urol 2002; 168: 2359 2367. 7 Donatucci C. Etiology of ejaculation and pathophysiology of premature ejaculation. J Sex Med 2006; 3(Suppl 4): 303 308. 8 Corona G, Jannini E, Mannucci E, Fisher A, Lotti F, Petrone L et al. Different testosterone levels are associated with ejaculatory dysfunction. J Sex Med 2008; published article online April 10, 2008. DOI: 10.1111/j.1743-6109.2008.00846.x. 9 Waldinger M, Rietschel M, Nöthen M, Hengeveld M, Olivier B. Familial occurrence of primary premature ejaculation. Psychiatr Genet 1998; 8: 37 40. 10 Jern P, Santtila P, Witting K, Alanko K, Harlaar N, Johansson A et al. Premature and delayed ejaculation: genetic and environmental effects in a population-based sample of Finnish twins. J Sex Med 2007; 4: 1739 1749. 11 Waldinger M, Schweitzer D. The use of old and recent DSM definitions of premature ejaculation in observational studies: a contribution to the present debate for a new classification of PE in the DSM-V. J Sex Med 2008; 5: 1079 1087. 12 Rosen R. Prevalence and risk factors of sexual dysfunction in men and women. Curr Psychiatry Rep 2000; 2: 189 195. 13 Jannini E, Simonelli C, Lenzi A. A sexological approach to ejaculatory dysfunction. Int J Androl 2002; 25: 317 323. 14 Nathan S. The epidemiology of the DSM-III psychosexual dysfunctions. J Sex Marital Ther 1986; 12: 267 281. 15 Schapiro B. Premature ejaculation: a review of 1130 cases. J Urol 1943; 50: 37 40. 16 Kriegsfeld L, Demas G, Huang P, Burnet A, Nelson R. Ejaculatory abnormalities in mice lacking the gene for endothelial nitric oxide synthase (enos / ) a critical review of the empirical literature. Physiol Behav 1999; 67: 374 379. 17 Waldinger M, Quinn P, Dilleen M, Mundayat R, Schweitzer D, Boolell M. A multinational population survey of intravaginal ejaculation latency time. J Sex Med 2005; 2: 492. 18 Santtila P, Sandnabba NK, Jern P, Varjonen M, Witting K, von der Pahlen B. Recreational use of erectile dysfunction medication may decrease trust in ability to gain and hold erections in young males. Int J Impot Res 2007; 19: 591 596. 19 Grenier G, Byers S. The relationships among ejaculatory control, ejaculatory latency and attempts to prolong heterosexual intercourse. Arch Sex Behav 1997; 26: 27 47. 20 Tabachnich B, Fidell L. Using Multivariate Statistics, 4th edn. HarperCollins: New York, 2001. 21 Neale M, Boker S, Xie G, Maes H. Mx: Statistical Modeling 6th edn. Department of Psychiatry, 2002. Available at: http://www.vcu.edu/mx/. (accessed 18 March 2006). 22 Akaike H. Factor analysis and AIC. Psychometrika 1987; 52: 317 332. 23 Hong LK. Survival of the fastest: on the origin of premature ejaculation. J Sex Res 1984; 20: 109 122. 24 Waldinger MD, Schweitzer DH. Changing paradigms from a historical DSM-III and DSM-IV view toward an evidencebased definition of premature ejaculation. Part II proposals for DSM-V and ICD-11. J Sex Med 2006; 3: 693 705. 67