a Departamento de Ortopedia, Santa Casa de São Paulo, São Paulo, SP, Brasil
b Serviço de Ortopedia e Traumatologia, Hospital Israelita Albert Einstein, Universidade Federal de São Paulo (Unifesp), São Paulo, SP,Brasil
c Instituto de Ortopedia, Hospital das Clínicas, Faculdade de Medicina, Universidade de São Paulo (USP), São Paulo, SP, Brasil
d Escola Paulista de Medicina, Universidade Federal de São Paulo (Unifesp), São Paulo, SP, Brasil
Introduction
Transtrochanteric fractures of the femur occur in the regionbetween the greater and lesser trochanters of the proximalfemur and are exclusively extracapsular.1In Brazil, in a sur-vey conducted by the Ministry of Health through the NationalHealth System (SUS), it was found that 90% of the funding des-tined for orthopedic diseases was consumed by nine diseases,which included transtrochanteric fractures.
Their consequences for society are alarming. Around one-third of the patients die during the first year after the injury,approximately 50% become incapable of walking unaided orgoing up stairs and 20% start to need continual home care.3
Several systems have been used to classifytranstrochanteric fractures and thus guide their treat-ment. The commonest of these are the Tronzo,4Evans5,6andAO classifications.
The Tronzo classification is widely used in Brazilian ortho-pedics and traumatology services.
Exact documentation of the fracture depends on thecapacity for radiographic evaluation and classification. Itsreproducibility depends on the surgeon's skill in interpretinga classification system. The position of the fractured limb, theradiographic technique and the surgeons' levels of experienceare factors that contribute toward the reproducibility of a clas-sification system
Ideally, a classification system should be easy to apply,reliable and helpful in making treatment decisions, and con-sequently should influence the final result. An ideal systemshould not have interobserver discrepancies.
Therefore, the aim of this study was to assess theinterobserver reproducibility of the Tronzo classification fortranstrochanteric fractures of the femur using the kappa coef-ficient of concordance ().
Materials and methods
A cross-sectional observational study was conducted, inwhich 20 radiographic images of hips with transtrochantericfractures of the femur, in anteroposterior view, were used. Allthe patients were over the age of 65 years and had sufferedlow-energy trauma. The following were used as exclusioncriteria for the images: pathological fractures caused by bone tumors previous surgery in the region of the proximal femurand images of transtrochanteric fractures from patients underthe age of 65 years.
The radiographs were classified in accordance with Tronzo,by 12 observers: four specialists who were members of thesociety for orthopedic trauma surgery, four third-year resi-dents and four first-year residents of an orthopedics andtraumatology service. All of these observers were given priorexplanations regarding the Tronzo classification, with graphicimages on spreadsheets.
The radiographs of the fractures were presented in theform of slides in the Power Point®software, in sequence, indi-vidually numbered from one to twenty, and each image wasanalyzed for a maximum of 45 s. During the presentation, aquestionnaire was filled out to gather data on all the optionsof the Tronzo classification (Anexo 1).
The series of radiographs were analyzed by means of thecomplete Tronzo classification (six types). Subsequently, thisclassification was subdivided for analysis into two simple sub-types: stable fractures (Tronzo I and II) and unstable fractures(Tronzo III, III variant, IV and V).
In 1974, Tronzo subdivided these fractures into five types.Types I and II were stable; type I was described as an incom-plete transtrochanteric fracture, while type II could presentfracturing of the lesser trochanter, but without posteromedialcomminution. Types III and IV presented posteromedial com-minution; in type III, the diaphysis was brought to a medialposition and proximal calcar was fitted to it. When fractur-ing of the greater trochanter was also present, the situationwas classified as III variant (Fig. 1). In type IV, the diaphysiswas brought to a lateral position, the fracture line was morevertical and the comminution was generally greater. Type Vhad an inverted line, from lateral to medial and from distalto proximal, which made the fracture unstable (Fig. 2). In pre-senting his classification, Tronzo described the osteosynthesistechniques proposed for the various types of fracture.1,4
The data gathered were analyzed statistically using thekappa concordance test. The software used comprised SPSSV16, Minitab 15 and Excel Office 2007.
Results
Taking all of the observers into account, we found the fol-lowing kappa indices: for images with stable fractures, 0.11;
for images with unstable fractures, 0.52; and for the completeclassification, 0.44 (Table 1).
When we divided the observers into groups consisting ofall the residents together versus the orthopedic trauma spe-cialists, we found the following results: for residents, 0.49 forthe complete classification, 0.18 for stable fractures and 0.55for unstable fractures; for orthopedic trauma specialists, 0.39for complete fractures, 0.20 for stable fractures and 0.52 forunstable fractures (Table 2).
Lastly, when we took into consideration the division ofthe observers into all of the groups, we obtained the follow-ing results for the group of first-year residents: 0.50 for thecomplete classification, 0.24 for stable fractures and 0.66 forunstable fractures. For the third-year residents, the resultswere 0.53 for the complete classification, 0.37 for stable frac-tures and 0.51 for unstable fractures. Meanwhile, the result fororthopedic trauma specialists was 0.39 for the complete classi-fication, 0.20 for stable fractures and 0.52 for unstable fractures(Table 3).
Discussion
The kappa concordance coefficient is often used when theintra- and interobserver reliability and reproducibility areevaluated. This coefficient provides paired proportions of con-cordance between observers, which may by chance be correct.Kappa values can range from -1 to +1; -1 indicates total dis-cordance, zero indicates a fortuitous level of concordance and+1 indicates total concordance. In general terms, kappa valuesbelow 0.5 are considered unsatisfactory, values between 0.5and 0.75 are considered adequate and satisfactory and valuesabove 0.75 are considered excellent.
Landis classified kappa values as follows: poor (below 0),slight (0-0.2), weak (0.2-0.41), moderate (0.41-0.6), substantial(0.61-0.8) and almost perfect (0.81-1).
The classification of a fracture is the basis for the choiceof treatment. Thus, it is important to assess the validityof a classification system. Gusmão et al.11evaluated theGarden classification system for femoral neck fractures andfound poorly reproducible concordance (0.32). Schwartsmannet al.7did the same in relation to the AO classification fortranstrochanteric fractures and found weak concordance forthe complete AO classification (0.34). Pervez et al.5found aweak concordance index (0.34) for the Evans/Jensen classifi-cation for transtrochanteric fractures.
In our searches in the Lilacs and Medline databases, we didnot find any studies that evaluated the concordance index ofthe Tronzo classification.
The greatest difficulty regarding the Tronzo classification4is in interpreting the stability of the fracture, since the author'sdescription states that the stability is in the comminutionof the posteromedial wall, and also that in type II frac-tures (which are considered stable), the lesser trochanter maybe fractured. This probably explains the insignificant repro-ducibility for stable fractures (0.11) that we found in our study,which gave rise to large bias in the observers' classificationsof type II and III fractures.
The unstable fractures presented a higher degree of con-cordance than shown by the stable fractures (moderate; 0.52),between the observers in general. This can be explained byTronzo's descriptions of the patterns of fracture types IV andV, given that they are specific and do not leave room for doubtregarding their pattern of instability.
Schipper et al.6studied 20 fractures using 15 observers tomake the complete AO classification. The interobserver kappavalue was 0.33. When they used the simplified AO classifi-cation, the kappa value was 0.67. They concluded that thesimplified AO classification had good reproducibility but thatthis was not repeated in the complete classification. In ourstudy, simplification of the Tronzo classification into stableand unstable fractures did not increase the concordance indexin relation to the complete classification.
Fung et al.9reported that, unexpectedly, in a study eval-uating the reproducibility of the Evans/Jensen classification,increasing experience among the observers correlated with adeclining concordance index and they inferred that this clas-sification system was unclear or that another system wasnecessary.
In another study that assessed the Evans/Jensen classifica-tion using four residents (two seniors and two juniors) and 52radiographs to evaluate the concordance, Gehrchen et al.12didnot find any differences and took increasing experience intoconsideration.
Schipper et al.6used the AO classification and did notobserve any appreciable difference in the concordance amongdifferent groups of professionals (residents, orthopedic sur-geons and radiologists).
We observed a decline in the kappa concordance indexwhen the increase in the observers' experience was takeninto consideration: 0.50 for the first-year residents, 0.53 forthe third-year residents and 0.39 for the orthopedic traumasurgeons. This difference was statistically insignificant whenthe groups of residents were taken into consideration. Thisresult was unexpected because greater training and greaterexperience generally ought to imply greater reliability.
Although the Tronzo classification is used in manyBrazilian orthopedic services, its value in practice is very ques-tionable, given that we found concordance of 0.44 for thecomplete classification, which is considered moderate accord-ing to Landis. Its subdivision into stable and unstable fracturespresented major bias, in that the stable fractures presentedslight to weak concordance (0.11) and the unstable fractures,moderate to substantial (0.52).
Conclusion
The Tronzo classification presented moderate concor-dance, with kappa of 0.44. However, simplification of the classification into stable and unstable fractures did notincrease the concordance indices.
Therefore, the Tronzo classification does not fit within thecriteria for an adequate classification for clinical practice.Thus, we suggest that another system should be used oranother one should be created for this type of fracture.
Conflicts of interest
The authors declare no conflicts of interest.
1. Köberle G. Fraturas transtrocanterianas. Rev Bras Ortop.2001;36(9):325-9.2. Dani WS, Azevedo E. Fraturas transtrocanterianas. Periódicona internet, 2008. Consultado em 22 de julho de 2011.Disponível em: http://www.moreirajr.com.br/revistas.asp?fase=r003&id materia=32453. Lima ALP, Azevedo AJ, Amaral NP, Franklin CE, Giordano V.Tratamento das fraturas intertrocanterianas com placa eparafuso deslizante. Rev Bras Ortop. 2003;38(5):271-9.4. Tronzo RG. Symposium on fractures of the hip. Specialconsiderations in management. Orthop Clin North Am.1974;5(3):571-83.5. Pervez H, Parker MJ, Pryor GA, Lutchman L, Chirodian N.Classification of trochanteric fracture of the proximal femur:a study of the reliability of current systems. Injury.2002;33(8):713-5.6. Schipper IB, Steyerberg EW, Castelein RM, van Vugt B.Realiability of the AO/ASIF classification for pertrochantericfemoral fractures. Acta Orthop Scand. 2001;72(1):36-41.7. Schwartsmann CR, Boschin LC, Moschen GM, Gonc¸alves RZ,Ramos ASN, Gusmão PDF, et al. Classificac¸ão das fraturastrocantéricas: avaliac¸ão da reprodutibilidade da classificac¸ãoAO*. Rev Bras Ortop. 2006;41(7):264-7.8. Dirchl DR, Cannada LK. Classification of fractures. In: BucholzRW, Heckman JD, Court-Brown CM, Tornetta P 3rd, McQueenMM, Ricci WM, editors. Rockwood & Grenn fractures inadults. 7th ed. Philadelphia: Lippincott Williams & Wilkins;2009. p. 45-51.9. Fung W, Jonsson A, Buhren V, Bhandari M. Classifyingintertrochanteric fractures of the proximal femur: doesexperience matter? Med Princ Pract. 2007;16(3):198-202.10. Landis JR, Koch GG. The measurement of observer agreementfor categorical data. Biometrics. 1977;33(1):159-74.11. Gusmão PDF, Mothues FC, Rubin LA, Gonc¸alves RZ, TelökenMA, Schwartsmann CR. Avaliac¸ão da reprodutibilidade daclassificac¸ão de Garden para fraturas do colo femoral. RevBras Ortop. 2002;37(9):381-6.12. Gerchen PM, Neilsen J, Olensen B. Poor reproducibility of theEvan's classification of the trochanteric fracture: assessmentof 4 observers in 52 cases. Acta Orthop Scand. 1993;64(1):71-2.
