Introduction
Traumatic skeletal fractures are commonly encountered in health care and present a large medical and socio-economic burden [
1,
2]. The majority of fractures occur in either the upper or lower extremity. For example, fractures of the wrist, hand, and ankle represent roughly 50% of all skeletal fractures [
3]. Due to the aging population, the incidence of extremity fractures is expected to increase in the coming decades [
4]. Current national and international protocols recommend frequent outpatient clinic visits at which radiographs of the fractured extremity are obtained. These radiographs can be used to check for (secondary) dislocation, assess bone healing, and provide early detection of complications [
5‐
8]. Other reasons for radiographic imaging include resident education, reassurance of patients, and medicolegal protection [
9]. The costs and cost-effectiveness of diagnostic imaging for traumatic skeletal fractures are becoming increasingly important factors in clinical decision making [
10]. Recent studies have assessed routine radiography use in patients with distal radius and ankle fractures. These studies suggested that radiographs obtained without a clinical indication do not lead to changes in treatment strategy whilst adding to treatment cost [
11‐
13]. The added value of radiographs for other fractures of the extremities and their consequences for treatment strategies are still unclear. Therefore, the aim of this review was to analyze studies that examine the influence of follow-up radiography for extremity fractures on treatment strategy. Specifically, we focused on whether omission of these more or less routine radiographs is associated with a delayed detection of complications and subsequently a possible deteriorated functional outcome.
Methods
This systematic review was conducted adhering to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines [
14]. Our methods include a comprehensive search of the literature, independent selection of studies, as well as assessment of the methodologic quality of these studies and extraction of the clinical outcomes by two of the authors.
Search strategy
A comprehensive literature search was conducted in multiple databases (i.e., Pubmed, Embase, and the Cochrane library) on October 9, 2017. The search strategies were developed with the guidance of a trained medical librarian and included combinations of different terms and synonyms for effectiveness, radiographs, and both upper and lower extremity fractures. In addition, the reference lists of the selected articles were screened for any other relevant studies not identified in the aforementioned electronic search. The search was limited to studies published in the English or Dutch language and was aimed at studies on adult, human subjects. The detailed search strategy is presented in Appendix 1.
The search was repeated on July 10, 2018. In total, 385 additional articles were identified and added to the screening process. No additional relevant studies were found, and thus, none were added to the analysis.
Criteria for considering studies included in this review
We included studies that described radiographic imaging in the follow-up of fractures of the upper and/or lower extremities. One of the outcome measures had to be either the influence of radiographic imaging on a change in treatment strategy, the association between radiographic imaging and complications (i.e., a lower number of complications detected, or a delayed detection of a complication due to the omission of radiographs) or a possible relation between the omission of radiographs and clinical outcomes (i.e., due to a possible missed complication) such as: range of motion, a functional outcome score (on a validated test/questionnaire), quality of life (using a validated questionnaire), or pain (using a validated instrument). Both randomized controlled trials and observational studies were eligible for inclusion. Case reports and small case series (< 20 subjects) were not included, as well as studies mainly describing patients with pathologic fractures, open fractures (Gustilo grade II/III), severely injured patients (ISS > 16), studies not reporting on the use of radiography in a follow-up setting (but rather in a diagnostic setting), and studies reporting the use of intra-operative control radiographs or their directly post-operative equivalents.
Selection of studies
After removal of duplicate records, the titles and abstracts of the remaining studies were independently screened by two authors (PvG, CN) using the online systematic review tool “Covidence” (
http://www.covidence.org, and Vertitas Health Innovation Ltd) Articles selected based on title and abstract were evaluated fully. If it was unclear whether a study met the inclusion criteria or if no abstract was available, but the title suggested relevance, the full text of the article was assessed for eligibility as well. In the case of a dispute, consensus between the two reviewers was reached by discussion or by consulting an arbiter (SMR), if necessary.
Assessment of methodological quality
Methodological quality of the included studies was assessed with the Newcastle–Ottawa scale (NOS) by two authors (PvG, CN) independently. In the case of inconsistent results, consensus between the two reviewers was reached by discussion. The Newcastle–Ottawa scale is a frequently used assessment tool for the methodological quality of nonrandomized studies [
15]. Separate scales are available for case–control and cohort studies. For this systematic review, we used the scale that evaluates cohort studies, since none of the included studies were randomized or had a case–control design.
The Newcastle–Ottawa scale assesses the methodological quality of studies on eight different criteria distributed over three domains: selection, comparability, and outcome. It is designed to measure the risk of selection bias, information bias, and confounding. Scoring is performed by allocating points when the criteria are met. A total of nine points equals a perfect score. The scale for cohort studies is presented in Appendix 2.
Data extraction and management
The following study characteristics were extracted: study design, country of origin, fracture location and/or type, number of participants, inclusion and exclusion criteria, participant demographics and study setting, number of (routine) radiographs, outcomes (including: changes in treatment strategy, the number of complications detected on a radiograph, radiographic changes compared to previous imaging or differences in clinical outcome), duration of follow-up, and results. Data extraction was performed by two authors independently (PvG, CN). In the case of a dispute, consensus between the two reviewers was reached by discussion.
Analysis of results
If the identified studies were clinically homogeneous, a meta-analysis was performed. If the studies were too heterogeneous to pool the data, we performed a descriptive review.
Assessment of level of evidence
The GRADE method was used to evaluate the overall quality of the evidence and weigh the recommendations [
16]. In GRADE, the levels of evidence are stratified high, moderate, low, and very low. Observational studies are primarily labelled ‘low’. A study can gain a ‘level’ if a large (e.g., RR < 0.5) or very large (RR < 0.2) effect was found if there is evidence of a dose–response effect (although this is not applicable to this systematic review) or if plausible residual bias or confounding would only result in study findings being more distinct. On the other hand, a study might drop a ‘level’ if there were limitations in the study design and execution and if there was inconsistency, indirectness, imprecision, or publication bias.
Discussion
In total, we identified 11 retrospective studies that examined the possible relation between radiographic imaging and treatment strategy. Several studies also described the influence of the omission of radiographs on functional outcome or detection of complications. Unfortunately, these studies were clinically so diverse that it was not possible to pool the data. Based upon the descriptive analysis, it appears that all studies come to essentially the same conclusion. They all suggest that omitting some, or even all, follow-up radiographs of extremity fractures does not have important clinical consequences, such as changes in treatment strategy, a deterioration of clinical outcomes, or missed complications. From the studies we included in this systematic review, no distinction could be made between different fracture locations or fracture types. However, all conclusions were based upon retrospective studies, introducing a high risk of bias and confounding. The level of evidence was low, indicating that these results should be interpreted with caution. We did not identify any prospective studies. As a result, studies included in this review should be regarded as the best available evidence at present.
For other indications, such as low back pain [
29], knee osteoarthritis [
30], or following paediatric spinal surgery [
31] the added value of routine radiographs are being questioned as well. Apparently, for other indications than extremity fractures, radiographs are also obtained routinely and without great impact on treatment strategies. In addition, for direct post-operative check radiographs of, for instance, hip fractures, multiple retrospective studies exist that debate their usefulness or discourage their use [
32‐
35]. A randomized study investigating the usefulness of direct post-operative control radiographs for operatively treated wrist and ankle fractures is currently being conducted by Oehme et al. [
36]. Routine radiographs might resemble low-value care, and omitting them might lead to increased efficiency for the health care system. The American College of Foot and Ankle Surgeons released a consensus statement discouraging the use of routine radiographs to monitor fracture, osteotomy, and arthrodesis healing without a clinical indication in the foot and ankle [
37]. However, to date, prospective evidence to support this claim is lacking.
In all studies included in this review, the number of changes in treatment strategy based on radiography was low. As depicted in Table
3, it ranged from 0 to 2.6%. The number of complications detected on a routine radiograph, in the absence of clinical symptoms, was similarly low. Both patients and physicians tend to ascertain value to radiographic confirmation of a favourable recovery. However, this review suggests that findings on a routine radiograph that require a change in treatment strategy, in the absence of clinical symptoms, are rare. The presence of clinical symptoms could be a good predictor of an unfavorable outcome, and might justify the use of radiography to rule out a complication. In the randomized controlled trial we are currently conducting [
38], reasons to obtain radiographs include: a pain score higher than 6 on a 1–10 Visual Analog Scale, a loss in range of motion, neurovascular symptoms, or a successive trauma to the injured limb. It is clear from our overview that interest in this topic is growing. All but two studies were published in the last 6 years, and quality and precision of the studies improved over time. For example, the older two studies contributed just 2% to the total number of participants and scored poorly on the Newcastle–Ottawa scale (three and four points out of nine, respectively). The more recent studies included more participants and, on average, scored higher on the Newcastle–Ottawa scale.
Limitations and strengths
All studies included in this review had a retrospective design and several other limitations in their study design on the Newcastle–Ottawa scale. All studies but two had a non-comparative design, and no statistical testing of outcomes was performed. The risk of bias was high, confounding was likely, and the external validity was limited. This resulted in a ‘very low’ level of evidence according to GRADE.
Conclusions in systematic reviews are dependent on the quality and design of studies included. The fact that only retrospective studies were identified and the level of evidence was very low hinders us in making strong recommendations. A second potential limitation was the tool used for assessment of the methodological quality of the included studies. The Newcastle–Ottawa scale is best suited for comparative and prospective non-randomized studies; therefore, this tool might not deliver the best assessment of risk of bias in the current setup. Finally, we limited our search to English and Dutch; therefore, language bias may affect our conclusions. However, no studies in Dutch were identified by the search strategy, and manual screening of the reference lists of included studies did not yield any references in a language other than English. Consequentially, the chance that language bias played a substantial role in the selection process of the systematic review was deemed low.
A strength of this study is presented by the fact that the percentage of included studies was very low (0.4%). This indicates that our initial search was broad, and as a result, the risk that important publications were missed was low.