Immediate versus delayed intramedullary nailing for open fractures of the tibial shaft: A multivariate analysis of factors affecting deep infection and fracture healing

Background: The purpose of this study was to evaluate contributing factors affecting deep infection and fracture healing of open tibia fractures treated with locked intramedullary nailing (IMN) by multivariate analysis. Materials and Methods: We examined 99 open tibial fractures (98 patients) treated with immediate or delayed locked IMN in static fashion from 1991 to 2002. Multivariate analyses following univariate analyses were derived to determine predictors of deep infection, nonunion, and healing time to union. The following predictive variables of deep infection were selected for analysis: age, sex, Gustilo type, fracture grade by AO type, fracture location, timing or method of IMN, reamed or unreamed nailing, debridement time (≤6 h or >6 h), method of soft-tissue management, skin closure time (≤1 week or >1 week), existence of polytrauma (ISS< 18 or ISS≥18), existence of floating knee injury, and existence of superficial/pin site infection. The predictive variables of nonunion selected for analysis was the same as those for deep infection, with the addition of deep infection for exchange of pin site infection. The predictive variables of union time selected for analysis was the same as those for nonunion, excluding of location, debridement time, and existence of floating knee and superficial infection. Results: Six (6.1%; type II Gustilo n=1, type IIIB Gustilo n=5) of the 99 open tibial fractures developed deep infections. Multivariate analysis revealed that timing or method of IMN, debridement time, method of soft-tissue management, and existence of superficial or pin site infection significantly correlated with the occurrence of deep infection (P< 0.0001). In the immediate nailing group alone, the deep infection rate in type IIIB + IIIC was significantly higher than those in type I + II and IIIA (P = 0.016). Nonunion occurred in 17 fractures (20.3%, 17/84). Multivariate analysis revealed that Gustilo type, skin closure time, and existence of deep infection significantly correlated with occurrence of nonunion (P < 0.05). Gustilo type and existence of deep infection were significantly correlated with healing time to union on multivariate analysis (r2 = 0.263, P = 0.0001). Conclusion: Multivariate analyses for open tibial fractures treated with IMN showed that IMN after EF (especially in existence of pin site infection) was at high risk of deep infection, and that debridement within 6 h and appropriate soft-tissue managements were also important factor in preventing deep infections. These analyses postulated that both the Gustilo type and the existence of deep infection is related with fracture healing in open fractures treated with IMN. In addition, immediate IMN for type IIIB and IIIC is potentially risky, and canal reaming did not increase the risk of complication for open tibial fractures treated with IMN.


INTRODUCTION
The treatment of open fractures of the tibial shaft remains controversial. The precarious blood supply and the lack of soft-tissue cover of the shaft of the tibia make these fractures vulnerable to nonunion and infection. Basic concepts of the current strategy for open tibial fractures reducing these complications are as follows: (1) immediate intravenous antibiotics; (2) urgent and repeated surgical debridement; (3) immediate rigid skeletal stabilization; (4) early, appropriate soft-tissue coverage, and (5) subsequent early bone grafting beneath a stable soft-tissue cover. [1][2][3] Moreover, there have been still various questions or problems in the treatment as follows: What kind of stabilization, such as external fixation (EF), intramedullary nailing (IMN), plating, or Ender nailing, is appropriate for immediate stabilization in open tibial fractures? 4-8 Whether reamed or unreamed nailing is appropriate in the use of IMN? 7,9 What kind of soft-tissue injury grade is the limit in immediate IMN? 5,7,9,10 When should soft-tissue cover be performed? 2,11 When is appropriate conversion to IMN from EF? 12,13 In the point of view of evidence-based medicine (EBM) for the resolution of the above problems, retrospective case-control cohort study, randomized controlled trial (RCT) study, multicenter study, and meta-analysis are more effective. However, it is difficult to perform the above trials because of its cost or ethics problem in randomization. Multivariate analysis in retrospective study could reduce several biases in clinical setting. The objective of the present study is to evaluate factors affecting deep infections and fracture healing of an open fracture of the tibial shaft treated with immediate or delayed locked IMN, using multivariate analysis in retrospective setting.

Clinical materials
Our institutional review board approved this study. This study was a retrospective chart and radiographic review. Eighty-four patients were male, and 14 patients were female. The mean age of the patients at the time of injury was 34.6 years (range, 15-86 years). Patients were divided into three groups on the basis of age as follows: 70 fractures belonged to group 1 aged 45 years or younger, 20 fractures to group 2 aged 46-59 years, and nine fractures to group 3 aged 60 or older.
Ninety patients were injured in motor vehicle accidents (53 were injured in motorcycle accidents, 21 were passengers or drivers in cars, 12 were pedestrians struck by automobiles, and four were injured bicycle drivers struck by automobiles), four patients were injured in labor accidents, and the remaining four patients were injured in other causes.
The mean Injury Severity Score (ISS) 14 of the 98 patients was 12.8 (range, 9-27). Patients were divided into two categories according to the criteria of Bone et al.,15 to distinguish between multiple injured patients and those with multiple skeletal injuries alone (additional fracture of the femur, tibia, humerus, forearm, simple pelvis or spine, without cord injury): 74 fractures had an ISS < 18, and 25 had an ISS ≥ 18. All patients with multiple skeletal injuries alone had an ISS < 18.
The 99 open tibial fractures were classified according to the criteria of Gustilo et al. 16,17 : type I, 22 fractures; type II, 42 fractures; type IIIA, 13 fractures; type IIIB, 20 fractures, and type IIIC, two fractures. We used Hannover Fracture Scale '98 (HFS) 18 to distinguish another grading of soft-tissue injury, although HFS was devised for predictive indices of limb salvage or amputation using a clear point system. The mean HFS scores in types I, II, IIIA, IIIB, and IIIC were 1.1, 1.5, 2.9, 6.9, and 9.0, respectively. The order of the mean HFS in respective Gustilo type was as follows: type 1 = type II < type IIIA < type IIIB = type IIIC (P < 0.01 in one-way analysis of variance (ANOVA) and post hoc Tukey's test).
Accordingly, the open tibia fractures were divided into thee groups as following: type I + II, IIIA, and IIIB + IIIC.
Severity of fractures was classified according to AO/ASIF classification 19 : type A, 49 fractures; type B, 39; type C, 11. One fracture was in the proximal third of the tibial shaft; 91 were in the middle third; four were in the distal third, and three were segmental fractures.

Treatment and group
Patients were divided into the following three treatment groups according to the timing and circumstances of IMN: group I, immediate IMN at the time of initial debridement (n = 62); group D1, delayed IMN following nonoperative treatment such as skeletal traction or splint (n = 16); group D2, delayed IMN following external fixation (n = 21). Group D2 was subdivided into group D2a: planned conversion (n = 17); and group D2b: nailing for established nonunion treated with EF (n = 4). Conversion from external fixation to IMN was performed, if no bacteria are found in smears from the open wound and the entire pin site area immediately before the second operation. The mean stabilization times in groups D1 and D2 were 18.3 ± 3.0 days (mean ± SE) and 52.1 ± 16.0 days, respectively. Unreamed procedures in these IMNs were performed for 63 fractures (63.6%). The distribution of Gustilo type in each treatment group was shown in Figure 1. There were many cases of type II in group I. There were many cases of type I and II in group D1. And there were many cases of type IIIB in group D2. In addition, the mean HFS scores in groups I, D1, and D2 were 2.2 ± 0.27, 1.9 ± 0.44, and 5.3 ± 0.73, respectively. Although there was no significant difference between the mean HFS scores in group I and that in group D1, the mean HFS score in group D2 was significantly higher than those in groups D1 and I (P < 0.001 in one-way ANOVA and post hoc Tukey's test).
No protocol was followed for deciding between immediate and delayed IMN, or for deciding between reamed and unreamed IMN. Decisions between immediate and delayed IMN and between reamed and unreamed IMN were based on the experience and discretion of the attending orthopedic staff, the interval since injury, the degree of contamination of the wound, the extent of injury to the soft tissues, and the degree of associated vital organ injuries. Kitasato Cylinder Nails (Mizuho Inc., Tokyo, Japan) manufactured at our institution in 1979 20 and AO/ASIF unreamed Tibial Nails (Synthes Inc./Mathys Medical Ltd., Japan, Tokyo, Japan) were used for reamed and unreamed IMN, respectively. Reamed IMN was performed with as limited reaming technique. 21 All fractures were initially stabilized by static locking.
Intravenous antibiotic treatment with a first degree cephalosporin for Gustilo type I and II fractures with the addition of an aminoglycoside (usually gentamicin) for type-III fractures was begun in the emergency room, and continued for 72-96 hours after the initial procedure. After the patient was resuscitated and all required emergency surgical procedures were completed, the open wound was irrigated and debrided. Irrigations were performed by using low-pressure bulb syringes to type I and II fractures, and performed by using high-pressure pulsating water jet devices to type III fractures. Debridement was repeated at 48-hour intervals until the wound was clean and all devitalized tissue had been resected.
The soft-tissue managements in the present series were as follows: primary suture (n = 66; 66.7%); delayed primary suture (n = 6; 6.1%); secondary split-thickness skin grafting (STSG) (n = 4; 4.0%); local flap (n = 18; 18.2%); and free flap (n = 5; 5.1%). The distributions of soft-tissue management in each Gustilo type and each IMN group are shown in Figures 2 and 3. The primary sutures were performed in many cases of types I and II. The soft-tissue management in type IIIA consisted of primary sutures or delayed primary sutures, and that in type IIIB consisted of local or free flaps. The soft-tissue management in treatment group I and D1 consisted of primary sutures or local flaps and in group D2 consisted of primary sutures and local or free flaps.

Evaluations and statistical analysis
Follow-up evaluations lasted from 1.6 to 10 years (mean, 3.1    years) after the original injury. Medical records of all patients were reviewed in detail. Roentgenograms were available for 84 fractures. The union time could be calculated in 78 fractures (79%). Union time in the remaining six cases could not be clearly obtained, because the status of bony union had already been consolidated at the visit to our outpatient clinic including taking plain radiographic films. Several senior staff doctors, who were double-blinded with respect to the detail of respective cases, performed the evaluations of the fracture healing time.
The deep, superficial, and pin site infection rates (in D2 group alone), nonunion rate, and time to union were assessed on the basis of clinical charts and radiographs. In addition, each case belonging to group D2b was excluded in the fracture healing analysis. Superficial and deep infections were defined according to Dellinger et al. 22 A superficial wound infection was located entirely above fascia with erythema and tenderness that required antibiotic therapy and opening the wound. A deep infection involving bone was defined as infection involving tissue below the muscular fascia. Pin tract infection was defined as any persistent drainage from a pin site requiring intervention or positive bacterial cultures from the pin site. 1,10 Bony union was defined as follows: clinically, there was no pain or tenderness, and the patient walked without aids; radiographically, solid bridging callus had connected the fracture fragment on both anteroposterior view and lateral view. 23 Nonunion was defined as a lack of clinical or radiographic evidence of healing 12 months after the injury, requiring a secondary procedure.
The three measures of final outcome were deep infection, nonunion, and healing time to union. Categorical regression analyses were performed if P values in chi-square test or Fisher's exact test showed less than 0.05 to determine predictors of the former two outcomes. Multiple regression analysis (stepwise method) after unpaired t-test or one-way ANOVA combined with Tukey's post hoc test was derived to determine predictors of the healing time to union.
For the analysis of predictive variables of nonunion, we selected the factors used for analysis of deep infection, minus pin site infection, plus the following factors: treatment type 2 (I group, D1 group, or D2a group) and existence of deep infection (+ or -). We excluded D2b group because of cases in need of nailing due to the existence of clear nonunions.
For the analysis of predictive healing time to union, we selected age, sex, Gustilo type, fracture grade (AO type), treatment type (I group, D1 group, or D2 group), existence of reaming, method of soft-tissue management, skin closure time, ISS, and existence of deep infection, similar to predictive analysis of nonunion.
The above categorical and numerical data were used for several univariate analyses, categorical regression analysis, or multiple regression analysis, which was performed using SPSS 11.0 for Windows (SPSS, Chicago, IL, USA) and a personal computer. The regression coefficient for individual predictors of the three outcomes were calculated, and a P value of <0.05 was considered to indicate significance.

Deep infections
Six (6.1%) of the 99 open tibial fractures developed deep infections. Three of these infections were caused by methicillin-resistant Staphylococcus aureus (MRSA); two were caused by Pseudomonas aeruginosa, and one was caused by both . The deep infections occurred in one Gustilo type II (1.0%, 1/99) and five type IIIB (5.1%, 5/99).
The relationship between deep infection and the following factors was evaluated: Gustilo type, treatment type 1, debridement time, method of soft-tissue management, skin closure time, existence of superficial infection, and existence of pin tract infection showed P < 0.05 on univariate analysis, among examined factors. Then, the relationship between the occurrence of deep infection and the above significant predictive factors was evaluated by means of categorical regression analysis. Finally, the above five factors excluding skin closure time were significant factors affecting the occurrence of deep infection on multivariate analysis (P = 0.0001). The orders of significant factors were as follows: (1) treatment type 1 and method of soft-tissue management,

Fracture healing
Seventeen (20.3%) of the 84 open tibial fractures developed nonunions The relationship between nonunion rate and the following factors: Gustilo type, method of soft-tissue management, skin closure time, existence of superficial infection, and existence of deep infection showed P < 0.05 on univariate analyses, among examined factors. Then, the relationship between the occurrence of nonunion and the above significant predictive factors were evaluated by means of categorical regression analysis. Finally, the above three factors, such as Gustilo type, skin closure time, and existence of deep infection, were significant factors affecting the occurrence of nonunion on multivariate analysis (P < 0.05). The orders of significant factors were as follows: (1) Gustilo type and existence of deep infections and (2) skin closure time. The statistical results about nonunion analysis are summarized in Table 2.
The relationship between fracture healing time and the following factors: Gustilo type, skin closure time, and existence of deep infection showed P < 0.05 on univariate analyses, among examined factors. Moreover, the relationship between the healing time and all the above evaluated factors was evaluated by means of multiple    Moreover, the debate has focused on whether IMN should be performed with or without reaming. Several authors 9,10,24 have failed to clearly identify any differences with regard to rates of infection or nonunion or functional outcome between reamed IMN regimen and unreamed IMN regimen. Bhandari et al. 25 also failed to demonstrate any significant differences with regard to infection, nonunion, or reoperations. Currently, it is not possible to make a recommendation for or against reaming in the fixation of open tibial fractures, according to a most recent review. 26 In our study also, canal reaming did not increase the risk of complication for open tibial fractures treated with IMN.
The present multivariate analyses about infections in open tibial fractures treated with immediate and delayed IMN showed that IMN after EF (especially in existence of pin site infection) was at high risk of deep infection as a first conclusion, and that debridement within 6 h and appropriate soft-tissue management were also important factors in preventing deep infections as a second conclusion.
As a third conclusion, it was shown that immediate IMN for type IIIB and IIIC was potentially risky.
In managing severe open tibial fractures initially treated with external fixations, several surgeons, including us, 13,26 performed conversion IMN from external fixators to decrease various complications, such as delayed unions, nonunions, malunions, and ankle joint stiffness. However, some authors 27,28 reported a high incidence of deep infection in secondary or delayed IMN after external fixation. Bhandari et al. 29 concluded in a meta-analysis study about this conversion method for tibial fractures that the existence of pin site infections was the most important factor in the expansion of infections, which is the same result as in our study.  10.6 ± 0.9 (n=59) 14.3 ± 2.5 (n=13) 12.2 ± 1.82 (n=6) 11.7 ± 1.0 (n=65) 9.8 ± 1.4 (n=13) 9.6 ± 0.7 (n=49) 8.6 ± 1.2 (n=9) 16.7 ± 2.4 (n=20) 10.1 ± 0.8 (n=38) 12.0 ± 1.7 (n=30) 14.3 ± 2.5 (n=10) 10.7 ± 0.9 (n=47) 9.4 ± 1.5 (n=13) 14.4 ± 2.6 (n=18) 11.6 ± 1.1 (n=50) 11.2 ± 1.4 (n=28) 10.2 ± 1.0 (n=49) 11.6 ± 3.8 (n=5) 6.2 ± 1.0 (n=4) 13.1 ± 1.6 (n=17) 21.3 ± 7.8 (n=3) 10.7 ± 0.8 (n=70) 17.4 ± 3.3 (n=8) 11.4 ± 1.0 (n=61) 11.2 ± 1.6 (n=17) 10.5 ± 0.8 (n=73) 24.0 ± 4.6 (n=5) Although skin closure time (less than or more than 7 days) is significant in the present univariate analysis, this factor is not significant in this multivariate analysis. Several authors 2,11,34-36 have documented significantly better outcomes with early closure (within 7 days). The timing of soft-tissue coverage in open fractures remains controversial. 3,28,[34][35][36][37][38] reported that wound closure should be obtained within 7 days to prevent infectious complications in severe open fractures managed with the antibiotic bead pouch technique and stabilization by external fixators. Recently, a radical immediate or very early fix and flap (within 24-72 h) protocol has been developed, based on a close collaboration between the orthopedic and microsurgical teams by Gopal et al. 39 Their report on such management showed a rate of union of 100% although with delayed union in 62% and a rate of infection of 9.5%. On the other hand, Hertel et al. 37 concluded that immediate softtissue reconstruction in association with immediate radical debridement and early definitive skeletal stabilization was the timing of choice for soft-tissue coverage in severe lower leg fractures if the general condition of the patient permitting. Anyway, the issue of infections related to softtissue coverage timing can depend on various factors, such as, coverage method, stabilization method, soft-tissue injury itself, and other organ injuries. In other words, it is difficult to conclude these problems briefly and simply. Our analyses postulated that both the Gustilo type and the existence of deep infection were associated with fracture healing in open fractures treated with IMN. Henley et al. 5 described, as previously mentioned in the initial section of discussion, that the severity of the soft-tissue injury rather than the choice of implant appeared to be the predominant factor influencing the rapidity of bone healing. Their results were similar to our results. In most of the reports about fracture healing in IMN for open tibial fractures, the use of unreamed, small-diameter nail resulted in delayed union or implant failure associated with delayed fracture healing, when compared with the use of reamed nailing. 9,10,25 However, reamed or unreamed procedures for open tibial fractures did not affect fracture healing in our multivariate analyses.
Lastly, we must mention an important issue in discussing the problem of treatment for open fractures. The Gustilo and Anderson classification is based on subjective description and is not an objective criterion based on scoring system. Two authors 43,44 have found this classification system to be associated with low interobserver agreement. According to Brumback et al.,43 interobserver agreement in the classification of open fractures of the tibia was about 60% based on 9 results of a survey of 245 orthopedic surgeons. Thus, the reliability of this classification system has been problematic. A recent classification based on an objective scoring system composed of the Ganga Hospital Injury Severity Score is thought to be useful for predicting the occurrence of deep infection and functional outcome. 45 Although HFS 18 has been devised for predictive indices of limb salvage or amputation, this scale is composed of eight objective items ( 46 Thus, we reclassified our cases according to HFS, as another classification tool. In addition, we should evaluate the validity and reliability of a treatment and regimen for open fractures under the consideration of the above. In conclusion, multivariate analyses for open tibial fractures treated with IMN showed that IMN after EF (especially in existence of pin site infection) was at high risk of deep infection, and that debridement within 6 h and appropriate soft-tissue management were also important factors in preventing deep infections. These analyses postulated that both the Gustilo type and the existence of deep infection were associated with fracture healing in open fractures treated with IMN. Moreover, immediate IMN for type IIIB and IIIC is potentially risky, and canal reaming did not increase the risk of complication for open tibial fractures treated with IMN.