score torsión testicular

SAVE OFFLINE

Development and Initial Validation of a Scoring System to Diagnose Testicular Torsion in Children João A. Barbosa,* Bruno Camargo Tiseo, Ghassan A. Barayan, Brian M. Rosman, Fábio César Miranda Torricelli, Carlo C. Passerotti, Miguel Srougi, Alan B. Retik and Hiep T. Nguyen From the Department of Urology, Boston Children’s Hospital, Boston, Massachusetts (JAB, BCT, GAB, BMR, ABR, HTN), and Department of Urology, University of São Paulo, São Paulo, Brazil (JAB, BCT, FCMT, CCP, MS)

Purpose: Testicular torsion is a surgical emergency requiring prompt intervention. Although clinical diagnosis is recommended, scrotal ultrasound is frequently ordered, delaying treatment. We created a scoring system to diagnose testicular torsion, decreasing the indication for ultrasound. Materials and Methods: We prospectively evaluated 338 patients with acute scrotum, of whom 51 had testicular torsion. Physical examination was performed by a urologist, and all patients underwent scrotal ultrasound. Univariate analysis and logistic regression were performed, and a scoring system for risk stratification of torsion was created. Retrospective validation was performed with 2 independent data sets. Results: The scoring system consisted of testicular swelling (2 points), hard testicle (2), absent cremasteric reflex (1), nausea/vomiting (1) and high riding testis (1). Cutoffs for low and high risk were 2 and 5 points, respectively. Ultrasound would be indicated only for the intermediate risk group. In the prospective data set 69% of patients had low, 19% intermediate and 11.5% high risk. Negative and positive predictive values were 100% for cutoffs of 2 and 5, respectively (specificity 81%, sensitivity 76%). Retrospective validation in 1 data set showed 66% of patients at low, 16% intermediate and 17% high risk. Negative and positive predictive values for cutoffs of 2 and 5 were 100% (specificity 97%, sensitivity 54%). The second retrospective data set included only torsion cases, none of which was misdiagnosed by the scoring system. Conclusions: This scoring system can potentially diagnose or rule out testicular torsion in 80% of cases, with high positive and negative predictive values for selected cutoffs. Ultrasound orders would be decreased to 20% of acute scrotum cases. Prospective validation of this scoring system is necessary.

Abbreviations and Acronyms DUS ⫽ Doppler ultrasound NPV ⫽ negative predictive value PPV ⫽ positive predictive value TT ⫽ testicular torsion US ⫽ ultrasound Accepted for publication October 16, 2012. Study received institutional review board approval. * Correspondence: Department of Urology, Hunnewell-353, Boston Children’s Hospital, 300 Longwood Ave., Boston, Massachusetts 02115 (telephone: 55-11-988807985; FAX: 55-11-38858852; e-mail: [email protected]).

Key Words: diagnosis, research design, scrotum, spermatic cord torsion, ultrasonography TESTICULAR torsion is a surgical emergency requiring prompt intervention to prevent testicular loss.1 The most frequent presenting symptom of testicular torsion is acute scrotal pain, which is common to several conditions, including torsion of the appendix testis, orchiepididymitis and scrotal trauma. Expedited

diagnosis of testicular torsion is critical to promptly indicate surgery.2–4 Several symptoms and findings on physical examination are associated with TT, including absent cremasteric reflex, horizontal lie and high riding testis.3–5 However, clinical overlap between the most common etiologies of

0022-5347/13/1895-1859/0 THE JOURNAL OF UROLOGY® © 2013 by AMERICAN UROLOGICAL ASSOCIATION EDUCATION

http://dx.doi.org/10.1016/j.juro.2012.10.056 Vol. 189, 1859-1864, May 2013 RESEARCH, INC. Printed in U.S.A.

AND

www.jurology.com

1859

1860

SCORING SYSTEM TO DIAGNOSE TESTICULAR TORSION

acute scrotum may preclude a conclusive diagnosis.6 Furthermore, examination of acute scrotum can be challenging due to swelling and uncooperative patients in distress, rendering the clinical diagnosis of TT dubious in daily practice.7–9 Imaging studies are frequently necessary for evaluation of acute scrotum. High resolution scrotal DUS is sensitive and specific for TT, and has become the standard imaging technique.10 –12 However, obtaining imaging studies delays diagnosis and treatment, which could prolong ischemia and decrease viability. Additionally, less than a third of acute scrotum cases are diagnosed as TT,13,14 which results in unnecessary imaging and cost for low risk groups. Error and delay associated with scrotal US motivate research into new techniques for diagnosis of TT, such as scintigraphy, magnetic resonance imaging and near infrared spectroscopy.15–17 To date, no systematic approach has been established for the objective evaluation of acute scrotum without complementary examinations. A diagnostic tool to help physicians in the clinical evaluation of acute scrotum could potentially decrease ischemic time in TT and reduce cost of treatment. We created a scoring system to diagnose testicular torsion in the pediatric population.

MATERIALS AND METHODS Study Design The TWIST (Testicular Workup for Ischemia and Suspected Torsion) study comprised a prospective evaluation and retrospective validation performed at 2 different institutions. A PubMed literature review (1965 to 2011) was performed to identify clinical variables associated with TT. Key words included testicular torsion, acute scrotum and testicular pain. The next phase consisted of a prospective evaluation of the identified variables in patients presenting with acute scrotal pain to the emergency department of a pediatric tertiary care hospital (Boston Children’s Hospital). A scoring system was created with the goal of predicting TT. Number of events of TT was estimated at 50 so as to allow 10 events per variable on 5-variable logistic regression.18 The second phase of the study consisted of retrospective validation of the scoring system performed with data from patients seen with acute scrotum at 2 tertiary centers (Boston Children’s Hospital and University of São Paulo).

sentation and physical examination.19 Patients older than 18 years were also excluded since the aim of the study was to evaluate TT in the pediatric population. Finally, patients with onset of symptoms longer than 1 week were excluded, as were those with known previous scrotal disease or history of surgery. A subset analysis was performed of patients presenting within 72 hours of onset of symptoms. Data from history and physical examination were first collected by the urologist or senior urology resident to evaluate the patient on a structured data sheet that included all variables selected from the literature search. Scrotal DUS, urinalysis, urinary culture and blood count were obtained for all patients. DUS was the gold standard for nonsurgical cases, and postoperative diagnosis was definitive for cases managed surgically. Univariate analysis was performed for all variables collected. Variables associated with TT were further analyzed in multivariate analysis. Chart review was performed to evaluate interobserver agreement of variables between urologists and nonspecialists. Variables with a kappa coefficient of less than 0.6 were excluded from analysis. Five-variable models were generated and tested for goodness of fit. Variables were selected and weighed to obtain the highest sensitivity for the low risk category stratification. Variable weight and cutoffs were selected based on ROC analysis, stratifying patients into low, medium and high risk categories.

Retrospective Validation The scoring system was retrospectively applied to distinct populations at 2 tertiary care centers. Data from patients seen at Boston Children’s Hospital from January 2007 to December 2008 were collected from charts based on ICD-9 codes for orchitis and epididymitis (604), and all categories for torsion of testis (608.2), including torsion of appendix testis and appendix epididymis. Patients with incomplete data or nonacute presentation were excluded. At the University of São Paulo data from patients diagnosed with TT between January 2007 and January 2011 were retrospectively reviewed and analyzed similarly.

Statistical Analysis Statistical analysis was performed using Stata®, version 11. Univariate analysis of categorical variables was performed using the chi-square test, time of symptoms using the Mann-Whitney U test and age using the Student t test. Multivariate analysis was performed with backward stepwise logistic regression. Kappa statistics were used to describe interobserver agreement of variables. HosmerLemeshow test and Nagelkerke R2 were used to evaluate goodness of fit of logistic regression. ROC curves and AUC were generated to construct the scoring system and determine its cutoffs.

Prospective Evaluation Institutional review board approval was obtained before beginning the investigation. Participating patients and parents completed a consent form before enrollment. Patients presenting with acute scrotal pain to Boston Children’s Hospital between January 2009 and January 2012 were enrolled in the first phase of the study. Patients younger than 3 months were excluded because the condition in this age group has significant differences in pre-

RESULTS Prospective Evaluation and Development of Scoring System For the prospective phase 338 patients with acute scrotum were enrolled, of which 51 were diagnosed with TT. Mean patient age was 11.6 years (12.0 for

SCORING SYSTEM TO DIAGNOSE TESTICULAR TORSION

1861

ROC curves for prospective and retrospective data sets. For prospective data set AUC was 0.983 (range 0.971 to 0.994). For cutoff of 2 NPV was 100% (range 98% to 100%), PPV 49% (39% to 59%), sensitivity 100% (91% to 100%) and specificity 82% (76% to 86%). For cutoff of 5 NPV was 96% (range 93% to 98%), PPV 100% (89% to 100%), sensitivity 76% (62% to 87%) and specificity 100% (98% to 100%). For retrospective data set AUC was 0.996 (range 0.988 to 1.000). For cutoff of 2 NPV was 100% (range 94% to 100%), PPV 95% (81% to 99%), sensitivity 100% (88% to 100%) and specificity 97% (90% to 99%). For cutoff of 5 NPV was 82% (range 73% to 89%), PPV 100% (80% to 100%), sensitivity 54% (37% to 70%) and specificity 100% (94% to 100%).

TT and 11.6 for other diagnoses, p ⫽ 0.59). Mean duration of symptoms before hospitalization was 67.2 hours (25.0 for TT, range 1 to 144, and 58.8 for other diagnoses, range 1 to 168). All patients with TT underwent surgery (orchiectomy in 15 and orchiopexy in 36). Subset analysis of patients presenting within 72 hours of onset of symptoms included 295 subjects, of which 49 had TT. Mean duration of symptoms was 18.3 hours (13.6 for TT and 19.0 for other diagnoses). Variables from history associated with TT on univariate analysis were testicular swelling (p ⬍0.001), nausea (p ⬍0.001), vomiting (p ⫽ 0.002) and severe pain (p ⫽ 0.04), considered as binary. Findings on physical examination associated with TT were transverse lie of testis (p ⬍0.001), testicular swelling (p ⬍0.001), scrotal erythema (p ⫽ 0.01), high riding testis (p ⬍0.001), localized pain (p ⫽ 0.01), hard testicle (p ⬍0.001), thickened spermatic cord (p ⬍0.001), inguinal fullness (p ⫽ 0.04), epididymis moved from posterior position (p ⬍0.001), fixed scrotal skin to testis (p ⬍0.001) and absent cremasteric reflex (p ⬍0.001). All variables on examination were binary. For logistic regression presence of nausea or vomiting was combined into 1 variable. Testicular swelling based on history or physical examination was similarly combined. Independent predictors of TT were nausea/vomiting, testicular swelling, high riding testis, transverse lie, hard testicle, thick spermatic cord, absent cremasteric reflex and fixed scrotal skin to testis (p ⬍0.05). Thick spermatic cord and fixed scrotal skin were excluded due to low interobserver agreement. The prediction model of choice

featured testicular swelling (2 points), hard testis on palpation (2), nausea or vomiting (1), high riding testis (1) and absent cremasteric reflex (1). HosmerLemeshow test p value for goodness of fit of logistic regression was 0.972, Nagelkerke R2 was 0.885 and AUC for the scoring system was 0.983 (range 0.971 to 0.994, see figure). A score of 2 or less corresponded to low risk of TT with no recommendation for US. A score of 3 or 4 points determined intermediate risk, for which US would be indicated. A score of 5 or more indicated high risk with recommendation of surgical exploration. Of 338 patients evaluated 234 (69.2%) were at low risk, none of whom had TT (table 1). For a cutoff of Table 1. Score distribution and incidence of torsion in prospective data set Score

No. TT Present (%)

No. TT Absent (%)

0 1 2 3 4 5 6 7

0 (0) 0 (0) 0 (0) 3 (7) 9 (40.9) 17 (100) 9 (100) 13 (100)

95 (100) 50 (100) 89 (100) 40 (93) 13 (59.1) 0 (0) 0 (0) 0 (0)

51 (15)

287 (85)

Totals/av

Distribution of acute scrotum cases according to scoring system in prospective evaluation. Risk stratification of patients at low (0 to 2 points), intermediate (3 to 4) and high risk (5 to 7) for testicular torsion shows no torsions in low risk category and only torsions in high risk category. Low and high risk categories combined comprise 80.7% of all cases evaluated.

1862

SCORING SYSTEM TO DIAGNOSE TESTICULAR TORSION

2 NPV was 100% (95% CI 98 –100) and sensitivity was 100% (95% CI 91–100). The intermediate risk group included 65 patients (19.2%). A total of 39 patients (11.5%) were at high risk, all of whom had TT. PPV was 100% (95% CI 88 –100) for a cutoff of 5 and specificity was 100% (95% CI 98 –100). Such individuals comprised 76% of all TT cases. Of 287 patients who did not have TT 234 were identified by score as being at low risk. Subset analysis of patients with shorter duration of symptoms revealed similar results, with 68.1% (201 of 295) in low risk categories (none with TT), 19.3% (57) at intermediate risk (12 with TT) and 12.5% (37) at high risk (all with torsion). Retrospective Validation Retrospective analysis of 116 acute scrotum cases at the same institution yielded similar results, with an AUC of 0.996 (range 0.988 to 1.000, see figure and table 2). Individuals at low risk for TT represented 66.3% of all patients (77 of 116), and no patient in this category had a diagnosis of torsion (NPV 100%, 95% CI 94 –100). This group represented nearly all patients (77 of 79) with nonischemic acute scrotum. Conversely of the 37 patients diagnosed with TT 20 were at high risk for torsion. All patients in the high risk group had a confirmed diagnosis of TT, and PPV was 100% (95% CI 80 –100). The intermediate risk group represented 16.3% of all patients analyzed (19 of 116). Retrospective analysis of data from the University of São Paulo included 59 cases of testicular torsion confirmed after either orchiopexy or orchiectomy. No patient was classified in the low risk category, while 45 were included in the intermediate risk category and 14 were identified as being at high risk for TT. Measures of performance were not attempted for this data set as this sample did not include individuals without TT.

Table 2. Score distribution and incidence of torsion in retrospective data set Score

No. TT Present (%)

No. TT Absent (%)

0 1 2 3 4 5 6 7

0 (0) 0 (0) 0 (0) 3 (75) 14 (93) 11 (100) 8 (100) 1 (100)

48 (100) 10 (100) 19 (100) 1 (25) 1 (7) 0 (0) 0 (0) 0 (0)

37 (32)

79 (68)

Totals/av

Distribution of acute scrotum cases according to scoring system in retrospective analysis. Risk stratification of patients at low (0 to 2 points), intermediate (3 to 4) and high risk (5 to 7) for testicular torsion reveals no torsions in low risk category and only torsions in high risk category. Low and high risk categories combined comprise 83.6% of all cases evaluated.

DISCUSSION We present a scoring system for prediction of testicular torsion. Although signs and symptoms associated with TT are well defined,6,14 DUS has become a ubiquitous tool for differential diagnosis of acute scrotum.12 Despite dependence on US in practice, several studies have attempted to provide physicians with substitutes for clinical diagnosis of TT.5,20,21 The present study synthesizes such efforts into an objective evaluation of risk in the acute scrotum, delivering a diagnostic tool that can help confirm or rule out TT. This scoring system could potentially decrease the number of ultrasounds for acute scrotum by 80%. Our population is fairly comparable to existing reports regarding age, duration of symptoms and percentage of TT as the etiology of acute scrotum.5,14,22,23 Variables associated with testicular torsion in this cohort consistently replicate previous findings in the literature.6,8,14 A prospective evaluation of acute scrotum cases by urologists allows for reliable data collection and accurate physical examination, minimizing selection bias and error. Furthermore, 2 independent data sets for validation contribute to the reliability of our results. Debate continues regarding whether US should be mandatory for all cases of acute scrotum.12,21 While most studies advocating clinical diagnosis fail to establish a consistent rationale, we present an objective instrument for evaluation and risk assessment to safely decrease the number of imaging studies ordered. Furthermore, we opted for the most easily reproducible features for the scoring system to allow for accurate application of our instrument. Thus, we avoided specific but less reproducible findings such as thickened spermatic cord and fixed skin to testis. Determining 3 categories of risk for TT, we propose to eliminate the indication of US in low and high risk groups. NPV for scores of 2 or less was 100% in both data sets tested, which encourages safe elimination of US in such cases. Similarly for scores of 5 points or more PPV was 100% for both data sets, which provides enough evidence to indicate surgical exploration in this group, as this procedure is justified by high suspicion of torsion in the absence of a conclusive diagnosis.24 Thus, the TWIST score could be used as a screening method for indication of US. While low and high risk groups could be safely treated according to the score classification, patients at intermediate risk for torsion would benefit from complementary imaging studies, as this group had variable incidences of torsion. These individuals comprised approximately 20% of all acute scrotum cases, suggesting that US orders could be decreased by 80%. The fact that 54% to 76% of torsion cases were identified in the high risk group suggests that half of

SCORING SYSTEM TO DIAGNOSE TESTICULAR TORSION

torsion cases could have surgery indicated before US in an effort to reduce ischemia time. It is noteworthy that sensitivity and specificity rates for this scoring system are not to be compared with those of US, as the aim of this instrument is not to substitute imaging studies, but to better delineate their indication. Rather, NPV and sensitivity for low risk, and specificity and PPV for high risk cutoffs are the indicators to be considered when evaluating the safety of dispensing US in these groups. In our samples high values of such performance measures suggest that the use of this scoring system could decrease the number of ordered ultrasounds without adding to the risk of misdiagnosis. It is noteworthy that no diagnostic method is free of error, and even US may have sensitivity as low as 91%.25 In the event of high suspicion for torsion despite contrasting evidence from diagnostic methods surgical exploration is indicated at the discretion of the attending physician. Our study has some limitations. Although retrospective validation has been performed with 2 independent cohorts, data from these cohorts are limited and prospective validation in larger populations is critical to ascertain the safety of this instrument in clinical practice. All of the data sets included patients from tertiary hospitals, which could lead to selection bias. Another source of bias could arise from relatively high values and range of duration of symptoms in our prospective sample, which could influence the incidence of testicular loss as well as include a number of less acute conditions. Additionally, early torsion may present with less prominent signs. However, this scoring system was demonstrated to work equally for shorter and longer

1863

ranges of duration of symptoms, particularly in those most frequently found at our emergency department. Furthermore, testicular salvage rate in our group was fairly comparable to others found in the literature.23 Another limitation is the fact that examination was performed by urologists only. Performance measures should be reevaluated when this instrument is tested and validated by nonexperts, as some of the variables, such as swelling and hard consistency, are subject to interobserver variability. It is also noteworthy that a scoring system for the entire pediatric population does not consider characteristics of younger children, such as physiological absence of the cremasteric reflex. Additionally this scoring system does not clarify the differential diagnosis of acute scrotum etiologies other than testicular torsion, for which it provides less information than scrotal ultrasound. Also this instrument is not a substitute from a medicolegal standpoint for ultrasound documentation. Finally, we adopted DUS as a gold standard, which could lead to misdiagnosis in a percentage of cases, as this technique is not free of error.

CONCLUSIONS Our scoring system could be a valuable tool for clinical diagnosis of testicular torsion. Risk stratification could help decrease orders for ultrasound in up to 80% of TT cases, and more than 50% of cases could have ischemia time abbreviated. Despite encouraging results with high positive and negative predictive values, further validation of this scoring system is necessary.

REFERENCES 1. Barada JH, Weingarten JL and Cromie WJ: Testicular salvage and age-related delay in the presentation of testicular torsion. J Urol 1989; 142: 746. 2. Nöske HD, Kraus SW, Altinkilic BM et al: Historical milestones regarding torsion of the scrotal organs. J Urol 1998; 159: 13. 3. Davenport M: ABC of general surgery in children. Acute problems of the scrotum. BMJ 1996; 312: 435. 4. Lewis AG, Bukowski TP, Jarvis PD et al: Evaluation of acute scrotum in the emergency department. J Pediatr Surg 1995; 30: 277. 5. Beni-Israel T, Goldman M, Bar Chaim S et al: Clinical predictors for testicular torsion as seen in the pediatric ED. Am J Emerg Med 2010; 28: 786. 6. Nelson CP, Williams JF and Bloom DA: The cremasteric reflex: a useful but imperfect sign in testicular torsion. J Pediatr Surg 2003; 38: 1248.

7. McAndrew HF, Pemberton R, Kikiros CS et al: The incidence and investigation of acute scrotal problems in children. Pediatr Surg Int 2002; 18: 435. 8. Schmitz D and Safranek S: Clinical inquiries. How useful is a physical exam in diagnosing testicular torsion? J Fam Pract 2009; 58: 433. 9. Livne PM, Sivan B, Karmazyn B et al: Testicular torsion in the pediatric age group: diagnosis and treatment. Pediatr Endocrinol Rev 2003; 1: 128. 10. Yazbeck S and Patriquin H: Accuracy of Doppler sonography in the evaluation of acute conditions of the scrotum in children. J Pediatr Surg 1994; 29: 1270. 11. al Mufti RA, Ogedegbe AK and Lafferty K: The use of Doppler ultrasound in the clinical management of acute testicular pain. Br J Urol 1995; 76: 625. 12. Liguori G, Bucci S, Zordani A et al: Role of US in acute scrotal pain. World J Urol 2011; 29: 639.

13. Kass EJ and Lundak B: The acute scrotum. Pediatr Clin North Am 1997; 44: 1251. 14. Kadish HA and Bolte RG: A retrospective review of pediatric patients with epididymitis, testicular torsion, and torsion of testicular appendages. Pediatrics 1998; 102: 73. 15. Nussbaum Blask AR, Bulas D, Shalaby-Rana E et al: Color Doppler sonography and scintigraphy of the testis: a prospective, comparative analysis in children with acute scrotal pain. Pediatr Emerg Care 2002; 18: 67. 16. Mäkelä E, Lahdes-Vasama T, Ryymin P et al: Magnetic resonance imaging of acute scrotum. Scand J Surg 2011; 100: 196. 17. Aydogdu O, Burgu B, Gocun PU et al: Near infrared spectroscopy to diagnose experimental testicular torsion: comparison with Doppler ultrasound and immunohistochemical correlation of tissue oxygenation and viability. J Urol 2012; 187: 744.

1864

SCORING SYSTEM TO DIAGNOSE TESTICULAR TORSION

18. Peduzzi P, Concato J, Kemper E et al: A simulation study of the number of events per variable in logistic regression analysis. J Clin Epidemiol 1996; 49: 1373. 19. Chiang MC, Chen HW, Fu RH et al: Clinical features of testicular torsion and epididymo-orchitis in infants younger than 3 months. J Pediatr Surg 2007; 42: 1574. 20. Srinivasan A, Cinman N, Feber KM et al: History and physical examination findings predictive of testicular torsion: an attempt to promote clinical

diagnosis by house staff. J Pediatr Urol 2011; 7: 470. 21. Ciftci AO, Senocak ME, Tanyel FC et al: Clinical predictors for differential diagnosis of acute scrotum. Eur J Pediatr Surg 2004; 14: 333. 22. Mushtaq I, Fung M and Glasson MJ: Retrospective review of paediatric patients with acute scrotum. Aust N Z J Surg 2003; 73: 55. 23. Zhao LC, Lautz TB, Meeks JJ et al: Pediatric testicular torsion epidemiology using a national

database: incidence, risk of orchiectomy and possible measures toward improving the quality of care. J Urol 2011; 186: 2009. 24. Molokwu CN, Somani BK and Goodman CM: Outcomes of scrotal exploration for acute scrotal pain suspicious of testicular torsion: a consecutive case series of 173 patients. BJU Int 2011; 107: 990. 25. Yang C Jr, Song B, Liu X et al: Acute scrotum in children: an 18-year retrospective study. Pediatr Emerg Care 2011; 27: 270.