Abstract
We reviewed the literature to determine the clinical outcomes of the treatment of closed fractures of the tibial shaft with immobilization in a cast, open reduction with internal fixation, or fixation with an intramedullary rod. We reviewed 2372 reports of comparative trials and uncontrolled studies of series of patients published between 1966 and 1993. Nineteen reports, involving six controlled trials and twenty-seven groups of patients, met our inclusion criteria. A structured questionnaire was used to assess the quality of the literature in terms of the experimental design and the method of assessment of outcome. Outcomes from controlled trials were summarized with odds ratios and risk differences, and outcomes from case series were summarized by the medians of the reported results.The studies that were reviewed generally had few subjects and were poorly designed. The comparative trials showed treatment with a cast to be associated with a lower rate of superficial infection than open reduction and internal fixation (mean difference, -5.81 per cent; p = 0.02) and open reduction and internal fixation to be associated with a higher rate of union by twenty weeks than treatment with a cast (mean difference, -18.07 per cent; p = 0.008). There were no other significant associations.There were insufficient data for us to evaluate any aspect of functional status, level of pain, or other patient-reported outcomes of any of the methods of treatment. The results of the present review suggest that the data from the published literature are inadequate for decision-making with regard to the treatment of closed fractures of the tibia.
The most frequently used methods of treatment for closed fractures of the tibial shaft are immobilization in a cast, open reduction and internal fixation, and fixation with an intramedullary rod. However, it is not clear which method has the best outcome. Some surgeons prefer closed reduction and placement of an intramedullary rod because of the decreased duration of external immobilization of the extremity, the maintenance of the alignment of the fracture, and the possibility of a lower rate of infection compared with that associated with open reduction and internal fixation. However, there is a lack of substantive data to support the concept that fixation with an intramedullary rod or open reduction and internal fixation leads to a better outcome than immobilization in a cast or brace.
The lack of adequate randomized, controlled trials comparing the efficacy of the three methods of treatment for closed fractures of the tibial shaft led us to perform a structured and systematic review of the literature. Before we began the study, we established a rigorous procedure for this review in keeping with the methodology used to conduct a meta-analysis. Our goal was to determine if the outcomes of the three therapeutic options were comparable or whether any one method was better.
Meta-analysis is a method for pooling data to increase statistical power for defined end points and subgroups and a way of systematically analyzing conflicting reports. The technique may also answer new questions that were not considered in the original reports18.
*No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. Funds were received in total or partial support of the research or clinical study presented in this article. The funding sources were The American Academy of Orthopaedic Surgeons and the American College of Physicians (B.L.).
†Division of General Medical Sciences, Washington University School of Medicine, 660 South Euclid Avenue, Campus Box 8005, St. Louis, Missouri 63110. The e-mail address for Dr. Littenberg is bl@osler.wustl.edu.
‡Department of Orthopedics, Mayo Clinic, 200 First Street S.W., Rochester, Minnesota 55905.
§Center for Health Services Research, M/C 922, University of Illinois, 2121 West Taylor Street, Chicago, Illinois 60612.
#Dana Biomedical Library, 6168 Dartmouth College, Hanover, New Hampshire 03755.
**Department of Orthopedics, University of Minnesota, 420 Delaware Street, Box 492, Minneapolis, Minnesota 55455.
††Department of Orthopaedics, New England Medical Center, 750 Washington Street, P.O. Box 306, Boston, Massachusetts 02111.
‡‡Indiana University, 541 Clinical Drive, Suite 600, Indianapolis, Indiana 46202.
The general procedure for conducting a meta-analysis is to ask a specific question, define inclusion and exclusion criteria for eligible reports, perform an exhaustive search, and summarize the eligible published literature10,18. An effort is made to include unpublished studies if they can be identified from published abstracts or from a dialogue with authors of publications on the subject of analysis. In order to reduce the impact of bias and confounding, most meta-analyses are limited to randomized, controlled trials. However, because of the paucity of such studies on the treatment of closed fractures of the tibial shaft, we broadened our inclusion criteria and analyzed non-randomized, comparative studies and uncontrolled case series. In order to limit bias, we established a written protocol for the review of the literature and developed a quality-scoring sheet before we began the search.
Search
The minimum eligibility criteria for both the comparative studies and the case series were clearly defined before we conducted a Medline search of the literature on fractures of the tibial shaft published between January 1966 and August 1993. We also searched the citation lists of eligible reports for earlier eligible publications. A report was considered eligible when the study had included a minimum of twenty patients; when there was a clear description of an isolated, closed fracture of the tibial shaft treated within twenty-one days after the injury; when one of the three treatments of interest had been used; and when at least one of the outcomes of interest was described. There were three treatment groups: immobilization in a cast or brace, open reduction and internal fixation (with screws alone or with screws and a plate), and fixation with an intramedullary rod (including open nailing). Comparative studies, including the controlled trials, were separated into individual series of patients on the basis of the method of treatment. We analyzed only reports written in English because of the expense and uncertainty of translation. We excluded a report if the study had involved animal models, if the series had included patients with polytrauma or injuries involving a joint, if original outcome data were not presented, if the series had included patients who had an open fracture of the tibial shaft and the outcomes for those who had a closed fracture were not defined separately, or if the study had included children.
Each citation found by the Medline search was reviewed by two of us (an epidemiologist [M. McC.] and an orthopaedic surgeon [M. F. S.]). It was often possible to exclude a citation immediately on the basis of information in the title and abstract. If the citation could not be excluded immediately and without question, the complete report was obtained and was reviewed by the two reviewers. If there was a persistent difference of opinion, the complete report was reviewed by three orthopaedic surgeons (S. A. R., D. H., and M. F. S.). The names of the institution, authors, and journal were blocked out on all manuscripts to minimize bias during the review.
Quality Score
We used a 15-point quality score to evaluate the follow-up measures and the research design for each study. The follow-up measures were assessed on the basis of the answers to four specific questions:
1. Were the reviewers of outcomes blinded to the treatment?
2. Were more than 85 per cent of the patients in each treatment group followed?
3. Were any subjective (patient-reported) outcomes described?
4. Was follow-up active (meaning that patients were checked at prespecified intervals regardless of whether they had any complaints) rather than passive (meaning that a complaint triggered an assessment or the follow-up was through a review of the chart)?
A response of clearly yes received 3 points, probably yes (not explicitly described in the Materials and Methods section of the manuscript, but the Discussion suggested that this procedure was followed) received 2 points, probably no received 1 point, and clearly no received 0 points. Research design was the final component of the quality score: a randomized, comparative study was given 3 points; a non-randomized, comparative study was given 2 points; and a case-series report was given 0 points.
Data
Descriptive data obtained from the eligible reports included the size of the sample, the ages of the patients, the mechanisms of injury (high or low velocity), the locations of the fractures (the proximal, middle, or distal part of the shaft), the fracture patterns (simple or comminuted), whether antibiotics had been used prophylactically, and the year that the report was published.
Continuous data: Continuous data (such as age or time) are measured on a scale that allows many possible responses. Continuous outcome measures of treatment included time to union of the fracture, duration of hospitalization, and time to return to work.
Dichotomous data: Dichotomous data are those for which only two responses are possible, such as present or absent. Such data provide information on the rates of superficial and deep infection, pulmonary embolism, deep-vein thrombosis, compartment syndrome, a reoperation after internal fixation, an operative procedure after treatment with a cast, union by twenty weeks, delayed union, non-union, slow union, amputation, refracture, and death. The patient was assumed to have had a superficial infection when terms such as cellulitis, soft-tissue infection, local infection, and skin necrosis were used in the report. Deep infection was assumed when terms such as fistula, infected non-union, osteomyelitis, osteitis, deep abscess, and sequestration were used. Unspecified infections were also considered to be deep infections. Delayed union was defined as union twenty-four to thirty-six weeks after the injury, and non-union was defined as incomplete healing more than thirty-six weeks after the injury. Some authors used another, overlapping category of union between twenty and thirty-six weeks that we designated as slow union.
Categorical data: Categorical data are data for which more than two discrete responses are possible, such as worse, unchanged, or better. These were extracted to quantify varus or valgus angulation; anterior or posterior angulation; shortening; rotation; and pain, stiffness, function, and range of motion at the knee and ankle joints as evaluated by the physician and the patient at six months. Other data of interest, such as the status of Workers' Compensation claims, occupation, and comorbidity, were not extracted because they were reported infrequently.
Analysis
The analysis was divided into two phases. In the first phase, we treated each group of patients (whether from a case series or from one arm of a comparative study) as a single eligible cohort. Each cohort was then assigned to one of three groups on the basis of the treatment. The median of the published rates of a given outcome after each treatment was determined, and then the medians across groups were compared with the Kruskal-Wallis test (a ranked non-parametric analog to analysis of variance) to calculate p values.
In the second phase, we analyzed the comparative trials, which represent only a subset of the data analyzed and which provide only a subset of the various outcome measures, to detect any differences among the three treatment groups. This approach is more robust in the face of differences among studies with regard to patient selection or technique, but it cannot be used for data from case series.
We used odds ratios as estimates of relative risk as well as the difference in absolute risks between treatment groups15. The odds ratio estimates the impact of the treatment on the chances of the outcome. For instance, a mortality odds ratio of 1.46 indicates that open reduction and internal fixation is associated with a 46 per cent greater chance of death than treatment with a cast (Table I). The risk difference is the rate of an outcome in one group subtracted from the rate in the other group. If the death rate is 25 per cent with operative treatment and 20 per cent with non-operative treatment, the risk difference is 5 per cent.
We used a random effects model to estimate confidence limits and p values for both odds ratios and risk differences7. Both non-randomized comparisons and randomized clinical trials were included in this phase of the analysis.
Search of the Literature
The Medline search of the literature on tibial fractures revealed 2372 citations. Nineteen reports met the inclusion criteria for the present study. The most frequent reason for exclusion was failure of the report to address the clinical problem of interest. Some studies were excluded because they did not involve human subjects, they did not address any of the three eligible treatment methods, or they included only a few patients. Several articles fulfilled multiple exclusion criteria. Twenty-seven independent groups were extracted from the nineteen eligible studies. Sixteen groups (2005 patients) had been treated with a cast or brace, eight groups (474 patients) had had open reduction and internal fixation, and three groups (407 patients) had had fixation with an intramedullary rod. Two of five reports that included comparisons among thirteen cohorts were randomized clinical trials.
Quality Scores
The quality score for the nineteen eligible reports ranged from 3 to 11 points, with a median of 6 points and a mean of 6.4 points. Linear regression analysis indicated that these scores improved with time, by about 0.11 point per year (p = 0.005).
Comparative Trials
Eligible Literature
The highest quality score (11 points) was given to a prospective, randomized study that compared open reduction and internal fixation with immobilization in an above-the-knee plaster cast for the treatment of ninety tibial fractures sustained by rugby and football players in Britain1. All of the patients had been followed, the study was randomized, and the functional outcome (the ability to resume play) was reported. The authors found that internal fixation was safe, led to a faster return to premorbid activities (a mean of one rugby season was missed in the group that was treated with open reduction compared with 3.5 in the group that was treated non-operatively), and was associated with fewer complications (malalignment; pain in the ankle; and stiffness of the knee, ankle, or subtalar joint) than immobilization in a cast. The highly selected patient population (athletes) makes it difficult to apply these results to patients seen in a general orthopaedic practice, but it does not detract from the study's internal validity.
Edwards provided detailed information on the clinical characteristics and outcomes of treatment of 492 patients. He recommended that transverse fractures be treated with closed reduction, with or without intramedullary nailing, and that spiral or long oblique fractures be treated with open reduction and internal fixation. Many of the 492 patients, such as those who had an open fracture, were not eligible for our study; however, we were able to identify the eligible patients with use of the extensive tables in the report. Some patients had been assessed retrospectively with passive follow-up, and others had been assessed prospectively. We divided these two broad groups according to treatment and summarized the data for purposes of comparison with groups in other eligible reports. We treated two of the retrospective groups (fifty-two patients managed with immobilization in a cast and 121 managed with open reduction and internal fixation) as a comparative trial with a quality score of 6 points. We treated three of the prospective groups (forty-three patients managed with immobilization in a cast, twenty-seven managed with open reduction and internal fixation, and twenty managed with an intramedullary rod) as another comparative trial with a quality score of 9 points. Too few patients had been managed with an intramedullary rod to meet our eligibility requirements for an independent group.
Lucas and Todd performed a retrospective analysis of 102 patients who had been managed with a cast and eighty-seven patients who had been managed with open reduction and internal fixation. The patients were followed for a minimum of two years, and the data were published only as an abstract. With no randomization, no blinding, no patient-reported subjective outcomes, and no evidence that follow-up was either active or reasonably complete, this report earned a quality score of only 5 points. Although the report did not provide the time to union, the authors stated that "6 per cent [of the operatively treated group had] better union at twenty weeks and 10 per cent [had better union] at thirty-two weeks." The poor quality score reflects major problems with the design of the study, such as a high potential for bias; therefore, these results should be interpreted with caution.
In 1979, van der Linden and Larsson performed a prospective, randomized trial to compare the results of open reduction and internal fixation with an AO plate in fifty patients with those of immobilization in a cast in another fifty patients. The clinical characteristics and the outcome were carefully described, and all of the subjects were followed actively. Randomization was described as strict, but was not specified. Only one patient was lost to follow-up. Despite the fact that the study was not blinded and there was no subjective reporting of the outcome by the patients, this study was partially protected against bias because of the randomized design and active follow-up. It earned a quality score of 9 points. There were more complications in the patients who had been managed operatively (fifteen) than in those who had worn a cast (six); however, the time to osseous union was less in the patients who had been managed operatively (twelve weeks) than in those who had worn a cast (seventeen weeks).
In 1975, surgeons at a Swedish hospital changed their procedure for treatment of closed displaced fractures of the tibia from open reduction and internal fixation with AO compression plates to non-operative measures25. Their non-randomized, retrospective study included seventy-three patients who had been managed with a cast and sixty-two who had been managed with internal fixation. The quality score for this study was 6 points since it was not blinded, lacked subjective data from the patients, and used passive follow-up. Treatment with plates necessitated a median duration of hospitalization of eighteen days (range, five to 106 days) compared with eight days (range, two to 140 days) for non-operative treatment; however, the median time to healing was twelve weeks after operative treatment compared with seventeen weeks after non-operative treatment. The ranges for time to healing were not reported.
Results of Pooling
Only one comparative study included a group treated with an intramedullary rod. Therefore, we analyzed only the results of immobilization in a cast and open reduction and internal fixation. The minimum of three trials that was necessary for the comparative analysis was available for only six outcomes: mortality (three trials), superficial infection (six trials), deep infection (six trials), union by twenty weeks (five trials), non-union (four trials), and reoperation or an operative procedure after treatment with a cast (five trials) (Table I).
As expected, treatment with a cast was associated with a lower rate of superficial infection: the odds ratio was 0.20 (95 per cent confidence interval, 0.08 to 0.50), and the risk difference was -5.81 per cent (p = 0.02). There was also a trend toward a lower rate of deep infection following treatment with a cast. This trend was large, with an odds ratio of 0.50 (95 per cent confidence interval, 0.13 to 2.03) and a risk difference of -0.99 per cent. However, no significant difference could be detected (p = 0.36), probably because of the very low rate of deep infections in either group (a total of only ten infections in 756 patients).
Open reduction and internal fixation was superior to treatment with a cast with regard to union by twenty weeks, with an odds ratio of 0.21 (95 per cent confidence interval, 0.06 to 0.68) and a risk difference of -18.07 per cent. This difference was significant (p = 0.008).
The trends in favor of open reduction and internal fixation with regard to the rates of mortality (p = 0.82), reoperation or an operative procedure after treatment with a cast (p = 0.45), and non-union (p = 0.47) were not found to be significant, with the numbers available. There were insufficient data for us to determine other outcomes of interest such as time to union and functional status.
Case Series
Treatment Groups
Combined with the five comparative trials described earlier1,8,19,24,25, fourteen other eligible reports2-4,6,9,11-14,16,20-23 generated twenty-seven patient groups. The descriptive characteristics of the three treatment groups were analyzed (Table II). The same number of series was not analyzed for each characteristic. The three treatment groups were similar with regard to most of the characteristics. However, they differed significantly with regard to the site of the fracture. The median rate of distal fracture was 10 per cent for the ninety-nine patients managed with an intramedullary rod, 66 per cent for the 193 patients managed with open reduction, and 42 per cent for the 608 patients managed with a cast (p = 0.04). Another area of difference (but not of significant difference) was comminution: the median rate was 46 per cent for the 407 patients managed with an intramedullary rod, 7.4 per cent for the 276 patients managed with open reduction, and 8 per cent for the 740 patients managed with a cast. A median of 50 per cent of the fractures in the twenty patients who were managed with an intramedullary rod had been caused by a high-velocity injury compared with 24.6 per cent of the fractures in the 255 patients managed with open reduction and 28.9 per cent of the fractures in the 747 patients managed with a cast. The median age was 26.5 years for the twenty patients who had an intramedullary rod, 38.5 years for the 220 who had open reduction, and 34.7 years for the 1423 who had a cast. With the numbers available, we could detect no significant difference in quality scores across the three treatment groups (p = 0.52).
Outcomes
The outcome data were analyzed with regard to twelve variables, including time to union in weeks, mortality, superficial infection, deep infection, late reoperation or an operative procedure after treatment with a cast, union by twenty weeks, delayed union, non-union, slow union, amputation, refracture, and duration of hospitalization in days (Table III). The same number of series was not analyzed for each variable in each treatment group.
Time to Union
The median time to union was 20.0 weeks after fixation with an intramedullary rod, 14.7 weeks after immobilization in a cast, and 13.0 weeks after open reduction and internal fixation. The small difference of eleven days between the time to union after open reduction and that after immobilization in a cast is not likely to be due to chance alone, but it was not found to be significant (p = 0.06) (Fig. 1). The other measures of union (union by twenty weeks, p = 0.15; the rate of delayed union, p = 0.15; the rate of non-union, p = 0.99; and the rate of slow union, p = 0.08) also were not found to differ significantly among the three treatment groups, although they did tend to favor open reduction and internal fixation. The data must be viewed with caution since the methods for determining or reporting rates of union were not uniform in the eligible reports.
Failure of Treatment
The median rate of an operative procedure after primary treatment with a cast was 4 per cent; it ranged from zero of twenty-seven patients8 to 17 per cent of eighty-one patients11. Most of the operative procedures were performed more than three weeks after the injury was sustained. The median rate of reoperation after primary open reduction and internal fixation was 0.8 per cent (range, 0 to 6 per cent). Only one study8 provided the rate of reoperation after treatment with an intramedullary rod; that rate was one of twenty patients.
The median rate of refracture after treatment with a cast was 2.3 per cent and the range was 2 per cent of fifty24 to 4 per cent of seventy-three25 in the four series that were analyzed. The median rate of refracture after open reduction and internal fixation was 0.8 per cent and the range was zero of forty-nine24 to 11 per cent of twenty-seven9 in the four series that were analyzed. No refractures were reported after fixation with an intramedullary rod.
Complications
Most authors did not report the prophylactic use of antibiotics. The rates of superficial and deep infection were highest in the patients who had been managed operatively. The rate of superficial infection ranged from 0 to 4 per cent (two of forty-five patients24) after closed treatment and from 0 to 22 per cent (six of twenty-seven patients9) after open reduction and internal fixation (p = 0.05). The prevalence of superficial infection after fixation with an intramedullary rod in the only study in which it was recorded was zero of twenty patients8. A similar but non-significant trend was noted for deep infection. The rate of deep infection ranged from 0 to 2 per cent (one of forty-five patients1) after treatment with a cast, from 0 to 15 per cent (four of twenty-seven patients8) after open reduction and internal fixation, and from 0 to 1 per cent (three of 308 patients13) after fixation with an intramedullary rod (p = 0.35).
Rates of pulmonary embolism, deep-vein thrombosis, and compartment syndrome were reported infrequently. The prevalence of pulmonary embolism was reported in two series of patients managed with a cast; the prevalence was one (1 per cent) of eighty-one patients in one study11 and one (1 per cent) of seventy-three patients in the other25. In one study25, pulmonary embolism developed in one (2 per cent) of the sixty-two patients who had had open reduction and internal fixation and deep-vein thrombosis developed in one (1 per cent) of the seventy-three patients managed with a cast. The prevalence of deep-vein thrombosis in three groups that had been treated with open reduction and internal fixation ranged from 2 per cent (one of sixty-two patients)25 to 7 per cent (two of twenty-seven patients)9. The rates of compartment syndrome as reported in one comparative study1 were one of forty-five patients who had been managed with a cast and none of forty-five patients who had been managed with open reduction and internal fixation. None of the authors who studied fixation with an intramedullary rod reported the rates of any of these complications. Because of the lack of uniformity in the definitions and methods used to determine the rates of complications, these data must be interpreted with caution.
No deaths had occurred in most of the cohorts in which the mortality rate was reported, and the median death rate was 0 per cent (Fig. 2). The maximum reported death rate was 6 per cent of fifty-two patients managed with immobilization in a cast8. The higher death rates tended to be in the earlier studies, although the trend over time was not found to be significant, with the numbers available.
The rate of amputation was reported in two studies from the 1960s, but it was not reported in any of the recent studies. The maximum rate of amputation was 4 per cent (three of eighty-one patients) after treatment with a cast11 and 1 per cent (one of 113 patients) after open reduction and internal fixation8. The rate of amputation after fixation with an intramedullary rod was zero of twenty patients8.
Functional Outcomes
The duration of hospitalization was similar for the three treatment groups. The median duration was 12.5 days after treatment with a cast, 15.0 days after open reduction and internal fixation, and 14.0 days after fixation with an intramedullary rod (Table III).
Only two of the nineteen studies provided data on time lost from work as a result of the closed tibial fracture. In one randomized, comparative study, forty-five patients who had been managed with a cast lost a mean of eight months from work and forty-five patients who had been managed with open reduction and internal fixation lost a mean of four months1. In a case series of ninety-one patients who had been managed with a cast, a mean of four months was lost from work22.
Clinical decision-making is enhanced when direct comparisons can be made between methods of treatment. The choice is not whether to treat tibial fractures but rather which treatment provides the best risk-to-benefit ratio. Our aim was to document the outcomes of non-operative and operative treatment of closed tibial fractures to help practitioners to compare these therapeutic options.
Quality of the Literature
There were few randomized clinical trials, and a total of only 189 subjects were included in such studies. Fewer than 800 patients were studied in comparative trials, and our entire data set of case series included fewer than 3000 patients. The quality of the literature was poor and, even though our scoring system was not rigid or stringent, none of the articles received a perfect quality score. Points were lost because of a lack of a comparison group, a lack of blinding, and a lack of patient-reported outcomes. The first two criteria protect against bias, and the third makes it possible to address the issue that is most important to the patient: how he or she will feel and perform after treatment. The literature was largely inadequate in these areas, and we acknowledge that several elements of the quality-score methodology, such as patient-reported subjective outcome data, have not generally been a part of clinical reports in orthopaedics. This flaw is likely to continue until the design of orthopaedic clinical studies is altered. We believe that our quality-scoring system is relevant even though it resulted in low scores.
A recurrent problem in the orthopaedic literature is the lack of standardization for reporting results and complications. Common areas of concern related to fracture-healing, such as time needed for union and rates of infection, are not consistently reported. Many different individual and composite scales are used to define these outcomes8,12. A recent review of the results of treatment of closed fractures of the tibia revealed that the percentage of patients who had an unsatisfactory result could be 4 or 42 per cent, depending on the scale used for the evaluation5.
Limitations of the Analysis
All meta-analyses are prone to publication and detection bias. If we failed to include relevant reports, our estimates may be biased. We believe that we avoided these biases by adhering strictly to our protocol. First, we performed a thorough search of the published literature, and it is unlikely that a sizable reported cohort of tibial fractures escaped our attention. However, we acknowledge that our failure to use reports written in a foreign language or unpublished reports as well as errors in our search could have resulted in missing data. Second, during the analysis, we found that many of our estimates were stable since they persisted across several data sets (controlled trials and case-series cohorts) with small standard deviations. In particular, our estimate of the difference in union by twenty weeks between open reduction and internal fixation and treatment with a cast has a very narrow confidence limit (Table I). Third, we can estimate how many trials would be needed to change our estimates. For example, in the case series, the median time to union was 13.0 weeks after open reduction and internal fixation and 14.7 weeks after non-operative treatment. If we were to find some additional case series in which patients managed with a cast had shorter times to union, those values would bring the median time nearer to that after open reduction and internal fixation. However, five such series would be necessary before the median time to union after treatment with a cast would be shorter than that after open reduction. We believe that it is highly unlikely that five such studies will appear.
Much of our analysis is based on non-randomized data; therefore, it is possible that differences in the characteristics of the patients before treatment may account for some of the differences between the groups. Most of the reports that were studied did not provide data with which to evaluate this factor. This bias can be controlled only with randomized, controlled trials.
Many reviewers discount the use of data from non-randomized clinical trials in a meta-analysis because such data are prone to confounding and bias. Certainly, we would prefer to review well designed randomized, controlled trials, but they are not available. Nevertheless, tibial fractures continue to occur, and surgeons must use the best available data to make decisions regarding treatment. If data from case series are ignored, decisions must be made on the basis of intuition and personal experience. We prefer to use the best available data, no matter how limited and inadequate the designs of the studies. On the basis of our review, we are sufficiently reassured that, where they overlap, the comparative data and the case-series data tend to give similar results.
Results of the Analysis
We have fairly strong comparative data on the rates of superficial infection and the rates of union by twenty weeks to support the conclusions that non-operative treatment is more likely to prevent infections and open reduction and internal fixation is more likely to result in union. These statements are supported by the analysis of comparative trials in which 756 patients were evaluated for the presence of superficial infection and 567 patients were evaluated for union by twenty weeks. For evaluation of all other outcomes, and for evaluation of any outcome of fixation with an intramedullary rod, we had to resort to much weaker, uncontrolled data.
Our finding that treatment with a cast leads to a longer time to union could be based on the inconsistent definitions of union in the various studies. However, comparative trials avoid this error, and the evaluation of union by twenty weeks showed the same superiority of open reduction and internal fixation compared with treatment with a cast.
Future Directions
To improve clinical decision-making, surgeons and patients need data based on randomized, controlled trials with unselected patients drawn from community practice17. Details regarding the selection of patients, preoperative preparation such as antibiotic prophylaxis, the operative procedure, follow-up, and outcomes should be fully reported. Outcomes should include mortality, infection, failure of a device, manipulations, reapplication of casts, reoperations or operative procedures performed after treatment with a cast, union of the fracture, angulation, rotation, shortening of the limb, pain, satisfaction of the patient, time to return to activities, function, duration of hospitalization, and costs. To avoid random errors in these estimates, enrollment of patients must be sizable. To avoid bias in the assessment of outcomes, the investigators should use predetermined criteria and be blinded to the type of treatment whenever possible. To avoid bias introduced by comorbidity or previous functional status, treatment should be assigned in a randomized manner. To ensure high-quality data, participating investigators must be trained in accurate research methodology and must adhere to a written protocol. To provide reliable measures of complications, active follow-up at fixed intervals should be established. To provide data that can be used universally, subjects should be enrolled from community practices.
A good trial should have at least an 80 per cent chance of detecting a clinically important difference with a p value of less than 0.05. To detect the same difference in the median rate of union by twenty weeks that we found in our case series (84.7 per cent after treatment with a cast and 96.3 per cent after open reduction), about 140 subjects would be required in each arm of the randomized, controlled trial. Other outcomes would require larger samples. About 1000 subjects would be required to detect the same difference in median time to union that we found in the case series (14.7 weeks after treatment with a cast and 13.0 weeks after open reduction), assuming a standard deviation of thirteen weeks. A larger clinical effect would require fewer subjects. A long time may be needed for a large practice to enroll enough patients (100, for example) for even a small trial. We suggest a multicenter trial to address these issues. However, if standardized methods of study and evaluation instruments are used for smaller groups, a meta-analysis of such studies would mimic a multicenter trial.
The present meta-analysis of the currently available literature suggests that the time to union after open reduction and internal fixation may be shorter than that after treatment with a cast for most closed fractures of the tibia. However, non-operative treatment is associated with a lower rate of infection. The data were insufficient for us to evaluate fixation with an intramedullary rod or to compare functional status, level of pain, or other patient-reported outcomes among the three treatments. Although the quality of reports seems to be improving with time, the current published literature is highly imperfect for sophisticated clinical decision-making.
NOTE: The authors acknowledge the contributions of Susan Utterback and Chad Munger of The American Academy of Orthopaedic Surgeons.
Abdel-Salam, A.; Eyres, K. S.; and Cleary, J.: Internal fixation of closed tibial fractures for the management of sports injuries. British J. Sports Med.,25: 213-217, 1991.25213
1991
Albert, M.: Delayed union in fractures of the tibia and fibula. J. Bone and Joint Surg.,26: 566-578, July 1944.26566
1944
Bostman, O. M.: Rotational refracture of the shaft of the adult tibia. Injury,15: 93-98, 1983.1593
1983
[PubMed]
Brady, T. A.: Flexible intramedullary bar treatment of unstable tibial shaft fractures. Am. J. Orthop. Surg.,12: 17-19, 1970.1217
1970
[PubMed]
Bridgman, S. A., and Baird, K.: Audit of closed tibial fractures: what is a satisfactory outcome?. Injury,24: 85-89, 1993.2485
1993
[PubMed]
Christensen, J.; Greiff, J.; and Rosendahl, S.: Fractures of the shaft of the tibia treated with AO-compression osteosynthesis. Injury,13: 307-314, 1982.13307
1982
[PubMed]
DerSimonian, R., and Laird, N.: Meta-analysis in clinical trials. Control. Clin. Trials,7: 177-188, 1986.7177
1986
[PubMed]
Edwards, P.: Fracture of the shaft of the tibia: 492 consecutive cases in adults. Importance of soft tissue injury. Acta Orthop. Scandinavica, Supplementum 76, 1965.
Gebuhr, P.; Larsen, T. K.; and Petersen, O. C.: Displaced tibial shaft fractures treated with ASIF compression internal fixation. Acta Orthop. Belgica,56: 531-534, 1990.56531
1990
Hasselblad, V.; Mosteller, F.; Littenberg, B.; Chalmers, T. C.; Hunink, M. G.; Turner, J. A.; Morton, S. C.; Diehr, P.; Wong, J. E.; and Powe, N. R.: A survey of current problems in meta-analysis. Discussion from the Agency for Health Care Policy and Research Inter-PORT Work Group on Literature Review/Meta-Analysis. Med. Care,33: 202-220, 1995.33202
1995
[PubMed]
Hoaglund, F. T., and States, J. D.: Factors influencing the rate of healing in tibial shaft fractures. Surg., Gynec. and Obstet.,124: 71-76, 1967.12471
1967
Johner, R., and Wruhs, O.: Classification of tibial shaft fractures and correlation with results after rigid internal fixation. Clin. Orthop.,178: 7-25, 1983.1787
1983
[PubMed]
Klemm, K. W., and Börner, M.: Interlocking nailing of complex fractures of the femur and tibia. Clin. Orthop.,212: 89-100, 1986.21289
1986
[PubMed]
Kristensen, K. D.; Kiaer, T.; and Blicher, J.: No arthrosis of the ankle 20 years after malaligned tibial-shaft fracture. Acta Orthop. Scandinavica,60: 208-209, 1989.60208
1989
Laird, N. M., and Mosteller, F.: Some statistical methods for combining experimental results. Internat. J. Technol. Assess. Health Care,6: 5-30, 1990.65
1990
Lambdin, C. S.: Simple technique for managing tibial shaft fractures. Am. Fam. Phys.,8: 178-180, 1973.8178
1973
Laupacis, A.; Bourne, R.; Rorabeck, C.; Feeny, D.; Wong, C.; Tugwell, P.; Leslie, K.; and Bullas, R.: The effect of elective total hip replacement on health-related quality of life. J. Bone and Joint Surg.,75-A: 1619-1626, Nov. 1993.75-A1619
1993
Light, R. J., and Pillemer, D. B.: Summing Up: The Science of Reviewing Research. Cambridge, Harvard University Press, 1984.
Lucas, K., and Todd, C.: Closed adult tibial shaft fractures. In Proceedings of the British Orthopaedic Association. J. Bone and Joint Surg.,55-B(4): 878, 1973.55-B(4)878
1973
Netz, P.; Olsson, E.; Ringertz, H.; and Stark, A.: Functional restitution after lower leg fractures. A long-term follow-up. Arch. Orthop. and Trauma Surg.,110: 238-241, 1991.110238
1991
Oni, O. O. A.; Hui, A.; and Gregg, P. J.: The healing of closed tibial shaft fractures. The natural history of union with closed treatment. J. Bone and Joint Surg.,70-B(5): 787-790, 1988.70-B(5)787
1988
Peter, R. E.; Bachelin, P.; and Fritschy, D.: Skiers' lower leg shaft fracture. Outcome in 91 cases treated conservatively with Sarmiento's brace. Am. J. Sports Med.,16: 486-491, 1988.16486
1988
[PubMed]
Sarmiento, A.: A functional below-the-knee cast for tibial fractures. J. Bone and Joint Surg.,49-A: 855-875, July 1967.49-A855
1967
van der Linden, W., and Larsson, K.: Plate fixation versus conservative treatment of tibial shaft fractures. A randomized trial. J. Bone and Joint Surg.,61-A: 873-878, Sept. 1979.61-A873
1979
Wihlborg, O.: The effect of a change in management of displaced tibial shaft fractures. Rigid fixation versus conservative treatment. Helvetica Chir. Acta,53: 191-199, 1986.53191
1986