Vinícius Alexandre de Souza Almeida a,*, Carlos Henrique Fernandes a,
João Baptista Gomes dos Santos a, Francisco Alberto Schwarz-Fernandes b,
Flavio Faloppa a, Walter Manna Albertoni a
Introduction
Lesions of the extensor mechanism of the fingers are among the most prevalent in the orthopedic practice. The terminal extensor tendon, formed by the union of two lateral slips, is inserted into the dorsal surface at the base of the distal phalanx. Injury of this tendon, or intra-articular fractures at the base of the distal phalanx, lead to a flexion deformity of the distal interphalangeal joint (DIPJ) known as mallet finger.1 This lesion mainly affects the young population; it is common in sporting practices and may lead to a significant functional deficit if not treated properly.
Several clinical classifications have been described, aiming to categorize this condition. In 1957, Pratt et al.2 classified mallet finger based on the etiology: laceration, crushing, and indirect trauma. In 1984, Wehbé and Schneider described a system that categorized these lesions into three types.3 Doyle et al.4 have also described another system widely used in the literature. In Brazil, Albertoni's.5 clinical-radiological classification, described in 1986, is widely used. A good quality classification should primarily be written in simple language and provide reliable guidelines to aid in treatment, prognosis, and reducing the possibility of complications. Moreover, it must be feasible, reliable, and reproducible; the latter characteristic is measured by interobserver agreement.1,6 A classification is reproducible when several individuals are able to reproduce the same result at any time, anywhere.1 Thus, it becomes possible to compare the results of different centers with different patients and the respective outcomes for each type of treatment.
Reproducibility studies are classic in the literature when measuring the quality of classification systems, especially in orthopedics. These studies usually include few observers, due to the difficulty in maintaining a reliable assessment. Any classification system worsens its agreement as the number of observers and categories increase. The low experience of observers in the assessed condition and multicenter studies also tend to decrease agreement.
No studies on the reproducibility of the Albertoni classification were retrieved in the literature, nor any study on the reproducibility of any mallet finger classification. The authors conjectured that this classification has good interobserver agreement. This study is aimed at evaluating the interobserver agreement of the Albertoni classification for mallet finger, and to quantify its reproducibility in the management of this condition. Materials
This study was approved by the Research Ethics Committee of the institution where it was conducted (under CAAE No. 49960815.8.0000.5505).
A questionnaire survey was carried out in which 43 photographs of DIPJ radiographs in lateral view of hands with mallet finger injury were assessed. All radiographs were considered by the researchers to be of good quality.
The Albertoni classification was presented at the beginning of the questionnaire. It divides the lesions according to findings on a DIPJ radiograph in lateral view, categorizing them into four types: (A), pure tendon lesion without fracture; (B), bone avulsion lesion; (C), lesion associated with fracture of the dorsal region of the base of the distal phalanx, comprising one-third or more of the articular surface; and (D), epiphyseal detachment in children. Each type is divided into two Lesions of the extensor mechanism of the fingers are among the most prevalent in the orthopedic practice. The terminal extensor tendon, formed by the union of two lateral slips, is inserted into the dorsal surface at the base of the distal phalanx. Injury of this tendon, or intra-articular fractures at the base of the distal phalanx, lead to a flexion deformity of the distal interphalangeal joint (DIPJ) known as mallet finger.1 This lesion mainly affects the young population; it is common in sporting practices and may lead to a significant functional deficit if not treated properly.
Several clinical classifications have been described, aiming to categorize this condition. In 1957, Pratt et al.2 classified mallet finger based on the etiology: laceration, crushing, and indirect trauma. In 1984, Wehbé and Schneider described a system that categorized these lesions into three types.3 Doyle et al.4 have also described another system widely used in the literature. In Brazil, Albertoni's.5 clinical-radiological classification, described in 1986, is widely used. A good quality classification should primarily be written in simple language and provide reliable guidelines to aid in treatment, prognosis, and reducing the possibility of complications. Moreover, it must be feasible, reliable, and reproducible; the latter characteristic is measured by interobserver agreement.1,6 A classification is reproducible when several individuals are able to reproduce the same result at any time, anywhere.1 Thus, it becomes possible to compare the results of different centers with different patients and the respective outcomes for each type of treatment.
Reproducibility studies are classic in the literature when measuring the quality of classification systems, especially in orthopedics. These studies usually include few observers, due to the difficulty in maintaining a reliable assessment. Any classification system worsens its agreement as the number of observers and categories increase. The low experience of observers in the assessed condition and multicenter studies also tend to decrease agreement.
No studies on the reproducibility of the Albertoni classification were retrieved in the literature, nor any study on the reproducibility of any mallet finger classification. The authors conjectured that this classification has good interobserver agreement. This study is aimed at evaluating the interobserver agreement of the Albertoni classification for mallet finger, and to quantify its reproducibility in the management of this condition. Materials
This study was approved by the Research Ethics Committee of the institution where it was conducted (under CAAE No. 49960815.8.0000.5505). A questionnaire survey was carried out in which 43 photographs of DIPJ radiographs in lateral view of hands with mallet finger injury were assessed. All radiographs were considered by the researchers to be of good quality.
The Albertoni classification was presented at the beginning of the questionnaire. It divides the lesions according to findings on a DIPJ radiograph in lateral view, categorizing them into four types: (A), pure tendon lesion without fracture; (B), bone avulsion lesion; (C), lesion associated with fracture of the dorsal region of the base of the distal phalanx, comprising one-third or more of the articular surface; and (D), epiphyseal detachment in children. Each type is divided into two subtypes. In types A and B, subtype 1 is characterized by a flexion deformity of less than 30? and subtype 2, by a flexion deformity greater than or equal to 30?. Deformities greater than 30? indicate injury to the retinacular ligaments and capsular structures in types A2 and B2. Type C is subdivided into C1, congruent joint (stable), and C2, sub-dislocated or dislocated joint (unstable). Type D is subdivided into D1, epiphyseal detachment (Salter and Harris lesion type 1) and D2, fracturedetachment (Salter and Harris type 3).7,8 Below each photograph of a radiograph, the options A1, A2, B1, B2, C1, C2, and D were presented, so that the observer could choose only one of them. Given the rarity of type D, it was not subdivided into D1 and D2. A goniometer and a pen were provided to the evaluator for the necessary measurements, so as to determine the subgroups 1 or 2 in type A and B, It was considered that all observers were able to measure the angles, as all have specialty degrees in orthopedics and traumatology.
The questionnaire was applied to 19 observers, all from the same institution, divided into 12 hand surgeons and seven hand surgery residents. Each observer answered the questionnaire separately, with no debate among them. Statistical methods The interobserver agreement was calculated using the Fleissk agreement coefficient, a generalization for more than two evaluators based on Scott's agreement measure.9,10 The standard error and consequently the confidence intervals were calculated according to Fleiss' algorithm.11 The agreements of the seven possible classification options were compared. The study also compared the agreement when the classification was grouped into four major categories: A1 and A2, in A; B1 and B2, in B; C1 and C2, in C; and D. In this way, it was possible to assess the difficulty of differentiating the groups (A, B, C, D) or the subgroups (1 and 2). In all evaluations, in addition to the results from all 19 professionals, the results by the residents and hand surgeons were also compared separately. This comparison was made to assess the presence of differences of agreement between different levels of professional experience. For all inferential tests, the alpha error value was set at 0.05. The k-value ranges from -1 to 1, where 1 means total agreement, -1 means total disagreement, and zero means that the evaluators classified the items at random. The
agreement classification scale of Landis and Koch12 was adopted (Table 1). Results Table 2 presents the responses of the observers. Regarding the distribution of types per observer group, the most recurrent type was C1 in the groups of surgeons and residents; the least recurrent type was D for surgeons and B2 for residents (Table 3). When the Albertoni classification was grouped into larger categories for types A, B, C and D, the most prevalent type
was C, followed by A, B, and D, both in the hand surgeons and residents groups (Table 4).
Table 5 presents the results for the agreement in the Albertoni classification according to the Fleiss agreement coefficient. Among the hand surgeons, classifications A1 (k = 0.75 [0.71-0.78]) and A2 (k = 0.84 [0.80-0.87]) presented higher agreement than the other groups, having good and excellent agreements, respectively. In types B1 (k = 0.34 [0.30-0.37]) and B2 (k = 0.19 [0.15-0.23]), the agreement was poor and very poor, respectively. Types C1 (k = 0.51 [0.48-0.55]) and C2 (k = 0.44 [0.41-0.48]) had moderate agreement, whereas in type D (k = 0.10 [0.06-0.14]), the agreement was very poor. The general agreement presented k = 0.56 (0.54-0.58), was considered moderate.
Among the residents, the types A1 (k = 0.82 [0.76-0.89]) and A2 (k = 0.90 [0.83-0.96]) also presented better agreement than the others, and were considered to be excellent. The groups B1 (k = 0.46 [0.39-0.52]), B2 (k = 0.49 [0.43-0.56]), and C1 (k = 0.49 [0.43-0.56]) presented moderate agreement, while groups C2 (k = 0.37 ([0.30-0.43]) and D (k = 0.24 [0.18-0.31]), presented poor agreement. However, taking into account the confidence interval, it can be considered that groups B1 and B2 and C1 and C2 had the same agreement. The general agreement was k = 0.59 (0.55-0.62), considered moderate.
When hand surgeons and residents were analyzed together, the results were similar. In A1 (k = 0.77 [0.74-0.79])
and A2 (k = 0.86 [0.84-0.88]), the agreement was considered good and excellent, respectively. In turn, in B1 (k = 0.38 [0.36-0.40]) and B2 (k = 0.28 [0.26-0.30]) the agreements were considered poor. Types C1 (k = 0.52 [0.50-0.54]) and C2 (k = 0.42 [0.39-0.44]) presented moderate agreement and type D (k = 0.16 [0.14-0.19]), very poor agreement. The general agreement was k = 0.57 (0.56-0.58), considered moderate.
When assessing the classification into large groups (A, B, C, D), an improvement was observed for all groups, and the most significant improvement was observed in group C. Taking into account the 95% confidence interval, in all analyses combining subtypes 1 and 2 improved the agreement.
Among the surgeons, the agreement in A (k = 0.95
[0.91-0.99]) was excellent, in B it was poor (k = 0.34 [0.31-0.38]),
in C (k = 0.71 [0.67-0.75]) it was good and, in D it remained very
poor. The overall agreement was good, with k = 0.72 (0.69-0.74).
Among the residents, the agreement in A (k = 0.96
[0.89-1.00]) was excellent, in B (k = 0.55 [0.48-0.61]), moderate,
in C (k = 0.77 [0.70-0.83]), good, and in D it remained poor. The
overall agreement was good, with k = 0.76 (0.72-0.81).
When combining hand surgeons and residents, the agreement in A (k = 0.95 [0.93-0.97]) was excellent, in B (k = 0.42
0.39-0.44]), moderate, in C (k = 0.72 [0.70-0.74]), good, and in D it remained poor. The general agreement in this second evaluation was good, with k = 0.73 (0.71-0.74). The data plotted in the graphs present the respective confidence intervals (Figs. 1 and 2). Discussion Mallet finger is a very prevalent condition, mainly affecting an economically active age population. An efficient and reproducible classification is able to guide the professional to a more efficient treatment.
Among the previously described classifications for this type of deformity, that of Pratt et al.2 only divides the etiology and does not lead to treatment or prognosis. The classification of Wehbé and Schneider,3 which uses the extent of joint surface involvement, has the same limitation. It stratifies each of the three types into three subtypes (A, less than one-third of the affected joint surface, B between one-third and twothirds of the articular surface, and C, more than two-thirds of the affected surface). However, this does not provide guidance
regarding treatment or prognosis. Doyle's classification provides a stratification based on clinical parameters that remain uncontemplated in the other systems, but there is no categorization of the radiographic patterns for all types.4 The latter is the most widely used in literature worldwide.1
Compared with other classifications, the Albertoni classification defines a therapeutic schedule, which changes according to the lesion in question. Each type has a specific treatment: type A1 and B1 injuries are classically treated with immobilization. Types C1 and D are treated with non-surgical reduction and immobilization with a metal splint. Type A2, B2, and C2 lesions usually require surgical treatment.5,8 This classification is widely used among Brazilian orthopedic surgeons, hence the importance of assessing its reproducibility.
In the literature, no study on the reproducibility of the Albertoni classification was retrieved, nor was any study on the reproducibility of mallet finger classifications. This demonstrates the originality of the present study.
The Albertoni classification is based on radiographs. In the literature, several interobserver agreement studies have evaluated radiographic classifications. Audigé et al.13 evaluated 44 reproducibility studies on orthopedic classifications with the use of imaging criteria and found little uniformity in the methodology, which hinders the comparison of reproducibility. Belloti et al.6 assessed the reproducibility of distal radius fracture classifications, while Utino et al.14 studied the agreement in the AO classification for long bones in the pediatric population.
The number of radiographs under evaluation is an important factor for assessing agreement. Both too few and too many evaluations tend to worsen agreement.15 In Audigé's systematic review of orthopedic reproducibility studies, a large variation in the number of radiographs per study (from 14 to 200 evaluations) was observed.13 Berger et al.,16 in a systematic review on the reproducibility of the Eaton classification for rhizarthrosis, assessed four studies. In these studies, the number of radiographs ranged from 40 to 43. Based on these studies, the authors consider that the 43 radiographs used in the present study were sufficient to evaluate the interobserver agreement in the Albertoni classification.
The number of observers is another factor that interferes with the coefficient of agreement. The greater the quantity, the lower the probability of agreement. The literature also did not present uniformity regarding the number of observers. Thomsen et al.17 used only four observers, while Randsborg and Sivertsen18 worked with 12 observers and Audigé et al.13 analyzed works with 2-36 observers (median = 5). The present study included 19 observers in subgroups of residents and hand surgeons. The number of radiographs and observers in this study is therefore within the norms found in the literature.6,13,14,16-24
The higher the number of categories in a classification, the worse the agreement.6,18 Albertoni's classification, with seven possible options, would tend to have a worse agreement when compared with other classifications. This was demonstrated when grouping types A1 and A2 into A, B1 and B2 into B, and C1 and C2 into C; an increase of agreement was observed in all the analyses, since the number of categories had decreased to four.
There is no consensus in the literature regarding which is the cut-off value of k to consider a classification as reproducible.13,15 These values are arbitrarily defined by the authors.15 Fleiss9 consider k-values between 0.40 and 0.75 to present moderate to good agreement. Svanholm et al.25 only consider as good values of k greater than 0.75. In turn, Brage et al.22 consider k-values above 0.50 as reproducible. Landis and Koch,12 the parameter used in the present study (Table 1) and the most used today,13 considered moderate agreement as those in the range of 0.4-0.6, and good agreement, above 0.6.
When assessed as a single group, Albertoni types A1 and A2 presented k = 0.95 (0.93-0.97), which indicates an excellent agreement. When stratified into A1 (k = 0.77 [0.74-0.79]) and A2 (k = 0.84 [0.80-0.87]), the coefficient of agreement decreased slightly, but remained good. The k coefficient presented a decrease of 16%, i.e., there is little change in the agreement when combining the categories and removing the parameter that differentiates them. This shows that an angle of <30? (A1) or >30? (A2) can be considered a reproducible parameter.
Albertoni types B1 and B2, when evaluated as a single group, presented a moderate agreement coefficient, with (k = 0.42 [0.39-0.44]). When stratified into B1 (k = 0.38 [0.36-0.40]) and B2 (k = 0.28 [0.26-0.30]), the coefficient of agreement presented a 21% decrease, being classified as poor. Similarly to the findings for type A, there is no difficulty in differentiating between B1 and B2, as well as between A1 and A2, which corroborates the fact that the angle parameter is reproducible. Regarding the poor agreement, the authors believe that the main reason for this result is the low prevalence of type B in the questionnaire applied (14.6%), as seen in Table 4.
When evaluated as a single group, Albertoni types C1 and C2 presented a good agreement coefficient, with k = 0.77 (0.70-0.83). When stratified into C1 (k = 0.52 [0.50-0.54]) and C2 (k = 0.42 [0.39-0.44]), the coefficient of agreement presented a pronounced decrease of 41%, which became moderate for C1 and poor for C2. This leads to the assumption that the joint congruence of the DIPJ is difficult to define. That is, in relation to joint congruence, agreement decreased considerably. The authors believe that one of the ways to improve agreement in type C would be to better define the criterion for joint congruence. This may guide future modifications to this classification.
Another relevant factor is that out of the 43 radiographs, in 22 the observers presented doubt in the choice between B and C (Table 2). In contrast, when type A (A1 or A2) was chosen, it was always concordant, except for radiographs 4 and 38 (Table 2). This was evidenced by the lower agreement in types B and C when compared with type A. The authors believe that the parameter to distinguish between bone avulsion (type B) and fracture of the dorsal region of the base of the distal phalanx (type C) is not well understood by the observers.
Due to the low incidence of this type of lesion, type D was not separated in D1 and D2. A very poor agreement was observed, with k = 0.16 (0.14-0.19). The low prevalence of this type (1.8%), in only four cases, also justifies the low agreement.23
The evaluators' experience in the assessed condition tends to change agreement.18 Mattos et al.24 concluded that the lack of experience of the observers decreased agreement. However, in the present study, among the residents, the general agreement was k = 0.76 (0.72-0.81), while among hand surgeons it was k = 0.72 (0.69-0.74). Although the group of residents presented k-values greater than the group of hand surgeons, considering the 95% confidence interval (Figs. 1 and 2), it cannot be stated that the agreement is higher in that group than in the group of surgeons. This is an advantage of the Albertoni classification, in which the agreement is not altered with the experience of the observers. The authors believe that there is a tendency for greater agreement among residents because they have similar levels of knowledge and experience, and because they are in training in the same center, which leads to uniformity. However, this was not demonstrated in the present studyAudigé mentions that in the evaluation of 44 studies on the reproducibility of orthopedic classifications, of the 86 coefficients of agreement calculated, only four were excellent (k > 0.80), 17 were good (between 0.60 and 0.80), 32 were moderate (0.40-0.60), and 33 were fair or poor (<0.40).13 The overall agreement in the Albertoni classification was moderate, with k = 0.57 (0.56-0.58), based on the classification by Landis and Koch.12 Despite the moderate agreement, when compared with the literature and taking into account all the factors discussed, the authors consider the Albertoni classification to be reproducible. A selection bias has to be considered, as the radiographs in the present study were not randomized, having been chosen by the researchers for their quality. Another relevant feature is that it cannot be guaranteed that measurements were made using the appropriate methods, even though observers were correctly instructed and provided correct measurement material. This was a single-center study, which tends to homogenize responses and improve agreement. Conclusion
The Albertoni classification presented good or excellent interobserver agreement for types A1 and A2, moderate for types C1 and C2, and poor for types B1, B2, and D. Following the statistical methods employed, and compared with research in the literature, the authors consider the Albertoni classification to be reproducible. The authors believe that a better definition of the criteria for joint congruence would substantially improve agreement.
Conflicts of interest
The authors declare no conflicts of interest
1. Nieves JW, Bilezikian JP, Lane JM, Einhorn TA, Wang Y, Steinbuch M, et al. Fragility fractures of the hip and femur: incidence and patient characteristics. Osteoporos Int. 2010;21(3):399-408. 2. Dobbs RE, Parvizi J, Lewallen DG. Perioperative morbidity and 30-day mortality after intertrochanteric hip fractures treated by internal fixation or arthroplasty. J Arthroplasty. 2005;20(8):963-6. 3. Wolfgang GL, Bryant MH, O'Neill JP. Treatment of intertrochanteric fracture of the femur using sliding screw plate fixation. Clin Orthop Relat Res. 1982;(163): 148-58. 4. Kokoroghiannis C, Aktselis I, Deligeorgis A, Fragkomichalos E, Papadimas D, Pappadas I. Evolving concepts of stability and intramedullary fixation of intertrochanteric fractures - a review. Injury. 2012;43(6):686-93. 5. Boldin C, Seibert FJ, Fankhauser F, Peicha G, Grechenig W, Szyszkowitz R. The proximal femoral nail (PFN) - a minimal invasive treatment of unstable proximal femoral fractures: a prospective study of 55 patients with a follow-up of 15 months. Acta Orthop Scand. 2003;74(1):53-8. 6. Stern R. Are there advances in the treatment of extracapsular hip fractures in the elderly? Injury. 2007;38 Suppl. 3:S77-87. Review. Erratum in: Injury 2008;39(1):140. 7. Haidukewych GJ. Intertrochanteric fractures: ten tips to improve results. J Bone Joint Surg Am. 2009;91(3):712-9. 8. Gotfried Y. The lateral trochanteric wall: a key element in the reconstruction of unstable pertrochanteric hip fractures. Clin Orthop Relat Res. 2004;(425):82-6. 9. Palm H, Jacobsen S, Sonne-Holm S, Gebuhr P, Hip Fracture Study Group. Integrity of the lateral femoral wall in intertrochanteric hip fractures: an important predictor of a reoperation. J Bone Joint Surg Am. 2007;89(3):470-5. 10. Kinast C, Bolhofner BR, Mast JW, Ganz R. Subtrochanteric fractures of the femur. Results of treatment with the 95 degrees condylar blade-plate. Clin Orthop Relat Res. 1989;(238):122-30. 11. Zha GC, Chen ZL, Qi XB, Sun JY. Treatment of pertrochanteric fractures with a proximal femur locking compression plate. Injury. 2011;42(11):1294-9. 12. Hasenboehler EA, Agudelo JF, Morgan SJ, Smith WR, Hak DJ, Stahel PF. Treatment of complex proximal femoral fractures with the proximal femur locking compression plate. Orthopedics. 2007;30(8):618-23. 13. Egol KA, Kubiak EN, Fulkerson E, Kummer FJ, Koval KJ. Biomechanics of locked plates and screws. J Orthop Trauma. 2004;18(8):488-93. 14. Jensen JS. Classification of trochanteric fractures. Acta Orthop Scand. 1980;51(5):803-10. 15. Nilsdotter A, Bremander A. Measures of hip function and symptoms: Harris Hip Score (HHS), Hip Disability and Osteoarthritis Outcome Score (HOOS), Oxford Hip Score (OHS), Lequesne Index of Severity for Osteoarthritis of the Hip (LISOH), and American Academy of Orthopedic Surgeons (AAOS) Hip and Knee Questionnaire. Arthritis Care Res (Hoboken). 2011;63 Suppl. 11:S200-7. 16. Jensen JS, Sonne-Holm S, Tøndevold E. Unstable trochanteric fractures. A comparative analysis of four methods of internal fixation. Acta Orthop Scand. 1980;51(6):949-62. 17. Saarenpää I, Heikkinen T, Ristiniemi J, Hyvönen P, Leppilahti J, Jalovaara P. Functional comparison of the dynamic hip screw and the Gamma locking nail in trochanteric hip fractures: a matched-pair study of 268 patients. Int Orthop. 2009;33(1):255-60. 18. Sehat K, Baker RP, Pattison G, Price R, Harries WJ, Chesser TJ. The use of the long gamma nail in proximal femoral fractures. Injury. 2005;36(11):1350-4. 19. Finsen V, Børset M, Buvik GE, Hauke I. Preoperative traction in patients with hip fractures. Injury. 1992;23(4):242-4. 20. Resch S, Thorngren KG. Preoperative traction for hip fracture: a randomized comparison between skin and skeletal traction in 78 patients. Acta Orthop Scand. 1998;69(3):277-9. 21. Fogagnolo F, Kfuri M Jr, Paccola CA. Intramedullary fixation of pertrochanteric hip fractures with the short AO-ASIF proximal femoral nail. Arch Orthop Trauma Surg. 2004;124(1):31-7. 22. Uzun M, Ertürer E, Oztürk I, Akman S, Sec¸kin F, Ozc¸elik IB. Long-term radiographic complications following treatment of unstable intertrochanteric femoral fractures with the proximal femoral nail and effects on functional results. Acta Orthop Traumatol Turc. 2009;43(6):457-63. 23. Ekström W, Karlsson-Thur C, Larsson S, Ragnarsson B, Alberts KA. Functional outcome in treatment of unstable trochanteric and subtrochanteric fractures with the proximal femoral nail and the Medoff sliding plate. J Orthop Trauma. 2007;21(1):18-25. 24. Suckel AA, Dietz K, Wuelker N, Helwig P. Evaluation of complications of three different types of proximal extra-articular femur fractures: differences in complications, age, sex and surviving rates. Int Orthop. 2007;31(5): 689-95. 25. Lee WT, Murphy D, Kagda FH, Thambiah J. Proximal femoral locking compression plate for proximal femoral fractures. J Orthop Surg (Hong Kong). 2014;22(3):287-93.
