To the Editor:

We recently noticed several examples of differential attrition in orthopaedic publications and presentations (1-4). We applaud the authors of “Nonoperative Treatment Compared with Plate Fixation of Displaced Mid- shaft Clavicular Fractures. A Multicenter, Randomized Clinical Trial,” (1) for identifying a significant difference in the loss to follow-up seen between their operative (6%) and nonoperative (25%) groups (p = 0.008).

Differential attrition is a form of outcome selection bias due to a difference in loss to follow-up between study groups (5). Although discussed in other specialties (6-8), it is not given much attention in the orthopaedic literature despite its presence in some studies (1-4). As such, conclusions made based on analysis of available data may not be valid.

This is especially important in studies comparing operative to nonoperative treatment. The differential loss to follow-up in the nonoperative group may be selecting for those with worse outcomes. In a recent study by Judd et al. comparing nonoperative to operative management of midshaft clavicle fractures, a similar differential attrition was reported, with no comment in the discussion (2). Likewise, in a recent retrospective analysis favoring operative treatment of humeral shaft fractures, there was a 2.4% loss to follow-up in the operative group and 11% loss to follow-up in the nonoperative group (3,4). It is difficult to draw practice-changing conclusions from these data.

Strategies currently exist to minimize loss to follow-up, but we are unaware of any guidelines in the orthopaedic literature describing appropriate interpretation of results when differential attrition is present (9,10).

We only bring this up as a consideration when investigators and readers are designing, conducting, or reviewing a study. While we agree with Mckee et al. that their differential loss to follow-up likely did not compromise their results, we feel that this is an important point and should be addressed by all authors when present. We would like to thank Mckee et al. for bringing this topic to the reader’s attention in your journal.

The authors did not receive any outside funding or grants in support of their research for or preparation of this work. Neither they nor a member of their immediate families received payments or other benefits or a commitment or agreement to provide such benefits from a commercial entity.

References

1. Canadian Orthopaedic Trauma Society. Nonoperative treatment compared with plate fixation of displaced midshaft clavicular fractures. A multicenter, randomized clinical trial. J Bone Joint Surg Am. 2007;89:1-10.

2. Judd DB, Pallis MP, Smith E, Bottoni CR. Acute operative stabilization versus nonoperative management of clavicle fractures. Am J Orthop (Belle Mead NJ). 2009;38:341-5.

3. Tucker MC, Obremskey WT, Floyd M, Denard A. Operative versus nonoperative treatment of humeral shaft fractures: a retrospective review of 213 patients from two level 1 trauma centers. Read at the Annual Meeting of the Orthopaedic Trauma Association. 2009 Oct 9; San Diego, CA. Paper no 49.

4. Hayashi A. Surgical treatment may be more effective for humeral fractures. AAOS Now. Dec 2009. http://www.aaos.org/news/aaosnow/dec09/clinical3.asp. Accessed 2010 Apr 12.

5. Morshed S, Tornetta P 3rd, Bhandari M. Analysis of observational studies: a guide to understanding statistical methods. J Bone Joint Surg Am. 2009;91 Suppl 3:50-60.

6. Valentine JC, McHugh CM. The effects of attrition on baseline comparability in randomized experiments in education: A meta-analysis. Psychol Methods. 2007;12:268-82.

7. Heneghan C, Perera R, Ward AA, Fitzmaurice D, Meats E, Glasziou P. Assessing differential attrition in clinical trials: self-monitoring of oral anticoagulation and type II diabetes. BMC Med Res Methodol. 2007;7:18.

8. Amico KR, Harman JJ, O’Grady MA. Attrition and related trends in scientific rigor: a score card for ART adherence intervention research and recommendations for future directions. Curr HIV/AIDS Rep. 2008;5:172-85.

9. Sprague S, Leece P, Bhandari M, Tornetta P 3rd, Schemitsch E, Swiontkowski MF; S.P.R.I.N.T. Investigators. limiting loss to follow-up in a multicenter randomized trial in orthopedic surgery. Control Clin Trials. 2003;24:719-25.

10. Smith JS, Watts HG. Methods for locating missing patients for the purpose of long-term clinical studies. J Bone Joint Surg Am. 1998;80:431-8.

To The Editor:

We have read the letter from Dr. Flugsrud and are pleased to respond to his queries.

1. We did reliably record those who were excluded. Of all potentially eligible patients, 62% were included in the study and 38% were excluded or chose not to participate. 2. There were no demographic differences between participants and non- participants in the study. Patients who did not participate were individualized to treatment according to their and their surgeon’s wishes. Most were treated non-operatively. We do not know how many subsequent operations they received as a group although anecdotally we do know that there were both nonunion repairs and malunion corrections in the non- participants treated non-operatively. 3. We believe the different ratio of males versus females is due to chance. Sex was not a prognostic factor for outcome.

With the randomized format of our study, we have done our best to minimize bias and describe results that are applicable to the eligible population. On the basis of our study, we believe that primary operative repair has benefits for a specific group of patients with completely displaced fractures. We look forward to the publication of other trials in this area currently underway to refine our knowledge in this area.

Yours truly,

Michael McKee, MD, FRCS(C) On behalf of the Canadian Orthopaedic Trauma Society MDM

June 22, 2007

Dear Sir/Madame

We have read the letter from Dr. Flugsrud with interest and are pleased to respond to his queries.

1. We did reliably record those who were excluded. Of all potentially eligible patients, 62% were included in the study and 38% were excluded or chose not to participate. 2. There were no demographic differences between participants and non- participants in the study. Patients who did not participate were individualized to treatment according to their and their surgeon’s wishes. Most were treated non-operatively. We do not know how many subsequent operations they received as a group although anecdotally we do know that there were both nonunion repairs and malunion corrections in the non- participants treated non-operatively. 3. We believe the different ratio of males versus females is due to chance. Sex was not a prognostic factor for outcome.

With the randomized format of our study, we have done our best to minimize bias and describe results that are applicable to the eligible population. On the basis of our study, we believe that primary operative repair has benefits for a specific group of patients with completely displaced fractures. We look forward to the publication of other trials in this area currently underway to refine our knowledge in this area.

Yours truly,

Michael McKee, MD, FRCS(C) On behalf of the Canadian Orthopaedic Trauma Society MDM

To The Editor:

Having read the article comparing plating with non-operative treatment of displaced midshaft fractures of the clavicle(1)I would like to know the authors' opinion on the generalizability of their findings.

1. Number of eligible. (i)Did the authors reliably record patients eligible for the study (except for providing informed consent) who were not included? (ii) How many were not included?

2. When a patient declined to participate. (i)What was a patient told, when being invited to participate, about the sort of treatment he or she would receive after declining to participate in the study? (ii)Did the not-included patients differ in demographics or otherwise from the included patients? (iii)What treatment did the not-included patients, in fact, receive? (iv)How many reoperations did the not-included undergo?

4. Though the p value in Table I is 0.062 the randomization of males and females is striking: 53 of 87 males (61%) were randomized to surgical treatment, while 15 of 24 females (63%) were randomized to conservative treatment. (i)Are the authors confident that this is due to chance, or are they considering other explanations? (ii) Do the authors think that this randomization affects the interpretation or generalizability of their findings?

The authors showed that patients randomized to plating had interestingly better outcome than those randomized to conservative treatment. Are the included patients representative of the eligible population, or do the authors suspect any possible bias?

The author did not receive any outside funding or grants in support of his research for or preparation of this work. Neither he nor a member of his immediate family received payments or other benefits or a commitment or agreement to provide such benefits from a commercial entity. No commercial entity paid or directed, or agreed to pay or direct, any benefits to any research fund, foundation, division, center, clinical practice, or other charitable or nonprofit organization with which the author, or a member of his immediate family, is affiliated or associated .

References:

1. McKee MD, Canadian Orthopaedic Trauma Society. Nonoperative treatment compared with plate fixation of displaced midshaft clavicular fractures. J Bone Joint Surg Am. 2007;89:1-10.

We thank Dr Jenkins and his colleagues for their interest in our paper(1) and for the opportunity to respond.

1. (i) All nonunions in the non-operative group were confirmed at the time of operative repair. Therefore, the rate of nonunion in the non- operative group is at least a high as we reported: if there were fractures in this group misread as healed, then the true nonunion rate would be higher. None of the patients in the ORIF group have had any plate breakage or other untoward mechanical event to suggest the (unrecognized) presence of a nonunion.

1. (ii) We are not aware of any way to blind a reviewer as to whether there is a plate on the clavicle or not on a plain radiograph.

1. (iii) We agree that, with only four time points available for analysis, it is a relatively crude estimate of time to union. However, if we are able to show a difference with so few time points, it is probable that a true difference exists.

2. In the non-operative group (49 patients), there were 16 patients who eventually required operation for nonunion or symptomatic malunion. By our calculation, this means roughly 3 operations to prevent one failure of non-operative care. Our re-operation rate (for plate removal, the commonest reason for a repeat procedure) is less now that we routinely use a pre-contoured plate.

3. We have always been impressed by how economic analyses can find such disparate results from similar clinical scenarios, and decided that an economic analysis was not something we wished to use our limited resources on. Practically speaking, the earlier return to function, work, and recreation in the operative group (data not yet published) seemed a fair trade for an hour of operative time. We stress again that these patients represent a minority of clavicle fractures and that on a logistic level, we performed 62 operative procedures in 7 centers over a three year period. This is approximately one short operative procedure per center every four months, which hopefully won’t overload most reasonably-sized centers. This emphasizes the point that while a great many clavicle fractures are seen in the fracture clinic, only a small percentage are suitable for primary fixation.

4. Amongst other outcome measures, we used the DASH, which is a validated, responsive, patient-based outcome measure as free from bias as possible. While we agree that it would be helpful to separate the patients with nonunion and malunion, that is the point: we are at present unable to do so at the time of presentation, which is really the time when it matters. We agree that there are patients with displaced clavicular fractures who have good or excellent results with non-operative care; our current efforts are directed towards identifying prognostic variables at the time of presentation.

Secondly, we disagree that secondary reconstruction is as effective in restoring normal shoulder function as primary operative repair. While this has traditionally been taught, a recent study that we performed (to be published in the Journal of Shoulder and Elbow Surgery) suggests that late reconstruction for malunion or nonunion results in shoulder function inferior to that seen after primary fracture fixation. Thus, there is some drawback to waiting for patients to “declare themselves” as those who will have a poor outcome.

In essence, we agree that great care needs to be used when recommending primary operative care for these fractures. We believe that the traditional approach of sling treatment for all clavicle fractures, regardless of type or displacement, is incorrect. We hope that our study, and other well-designed prospective studies will help to identify the small but significant percentage that would benefit from primary fixation.

Reference:

1. McKee MD, Canadian Orthopaedic Trauma Society. Nonoperative treatment compared with plate fixation of displaced midshaft clavicular fractures. J Bone Joint Surg Am. 2007;89:1-10.

To The Editor:

We commend McKee et al. for performing a multi- centre trial comparing operative and non-operative treatments of clavicle fractures(1). Despite the lower mal/nonunion rates and improved functional outcomes, we believe their argument for primary plate fixation to be flawed. Some of our reservations are outlined below:

1. Regarding radiographic outcomes:

(i) Conventional radiography alone is notoriously inaccurate in the assessment of union for non-operatively treated clavicle fractures, and even less reliable after operative treatment (with anatomic reduction and direct bone healing).

(ii) In this unblinded trial there was potential for both observer bias and variability in the assessment of fracture union, yet no attempt was made to quantify these sources of error.

(iii) It is unclear how a continuous variable for the time to union was generated, when radiographic assessments appear to have only been performed at four fixed time points after injury.

2. Their argument that plate fixation reduces mal/nonunion rates is only superficially persuasive. A number-needed-to-treat analysis shows that 9 fractures require operative fixation to prevent one non-union, and 3.3 require fixation to prevent one symptomatic mal/nonunion, at the expense of a 34% complication rate and an 18% re-operation rate.

3. The absence of an economic analysis is a major omission from the study. Fixing mal/nonunions after non-operative treatment may be costly, but primary fixation of all displaced fractures costs much more, and is likely to overload burdened trauma services.

4. The reported functional data are largely subjective. In an unblinded study, with substantial losses to follow-up(2), these data should not be seen as clear evidence of benefit. The key question is: Does a patient with a fracture which heals after operative treatment have better shoulder function than one whose fracture heals after non-operative treatment? The two validated functional scores in this study (which are reported in graphic form only, with truncated y-axes and undefined error bars) suggest a small but statistically-significant benefit from plate fixation. However, the poorer overall scores in the non-operative group may have been due to a minority of outlying patients with poor scores from nonunions. It would be interesting to see a subset analysis excluding these patients. We invite the authors to generate a table comparing Constant and DASH scores for patients with healed fractures that have been primarily (i) operatively-treated, and (ii) non-operatively treated.

There is no doubt that there is a role for primary surgery in some younger, physically-active patients with displaced clavicle fractures. The results of this study will be useful in counseling these patients about its potential advantages, as well as its shortcomings. However, with non- operative treatment, most patients will heal a displaced clavicle fracture, and have good shoulder function; when they do not, operative fixation will reliably salvage a good outcome (3-5). This study does not provide sufficient evidence to support a radical departure towards the routine use of primary plate fixation. It is our concern that in their enthusiasm to recommend this technique, the authors have over-emphasized its benefits, whilst failing to highlight its drawbacks.

The authors did not receive any outside funding or grants in support of their research for or preparation of this work. Neither they nor a member of their immediate families received payments or other benefits or a commitment or agreement to provide such benefits from a commercial entity. No commercial entity paid or directed, or agreed to pay or direct, any benefits to any research fund, foundation, division, center, clinical practice, or other charitable or nonprofit organization with which the authors, or a member of their immediate families, are affiliated or associated.

References:

1. Canadian Orthopaedic Trauma Society. Nonoperative treatment compared with plate fixation of displaced midshaft clavicular fractures. A multicenter, randomized clinical trial. J Bone Joint Surg Am. 2007;89:1- 10.

2. Devereaux PJ, McKee MD, Yusuf S. Methodologic issues in randomized controlled trials of surgical interventions. Clin Orth Rel Res. 2003;413:25-32.

3. McKee MD, Wild LM, Schemitsch EH. Midshaft malunions of the clavicle. J Bone Joint Surg Am. 2003;85:790-7.

4. Kloen P, Sorkin AT, Rubel IF, Helfet DL. Anteroinferior plating of midshaft clavicular nonunions. J Orthop Trauma. 2002; 16: 425-30.

5. Collinge C, Devinney S, Hersovici D, DiPasquale T, Sanders R. Anterior- inferior plating of middle-third fractures and nonunions of the clavicle. J Orthop Trauma. 2006; 20: 680-86.

We would like to thank Dr. Chitgopkar and his colleagues for their interest in our recent article concerning displaced midshaft fractures of the clavicle. We agree completely that some or many of the patients randomized to the non-operative treatment group were not compliant with their sling treatment regimen. While ideally every patient would have been completely compliant with the sling, unfortunately it is the reality of clinical research that they were not, and practically speaking, their outcomes reflect the outcome that can be expected by the practicing orthopaedic surgeon.

Using the “intention to treat principle”, patients randomized to a particular treatment arm are assessed as if they had received that treatment. Therefore, failure of sling treatment in the non-operative group is reflected by the number of patients randomized to this treatment, not necessarily the number who completed the treatment. Again, practically speaking, this result is what the practicing orthopaedic clinician can expect with this treatment method.

We have extensive experience with midshaft malunions of the clavicle. In a previous publication we have described the presenting symptoms of such individuals. We find that their symptoms are distinct from rotator cuff impingement if this is what the authors mean by “classic impingement symptoms and signs”. We find that these patients do exhibit easy and rapid fatigability of the rotator cuff, especially with overhead work. They can present with thoracic outlet type symptoms and neurological sequelae in the arm, scapular winging and shoulder asymmetry.

We would agree with most of the operative indications that Dr. Chitgopkar and his colleagues list, especially shortening of more than 2 cm. We agree completely that mal- rotation (which has been poorly recognized in the past) typically leads to scapular winging and is an entity which is becoming more clearly characterized as a relative indication for operative intervention.

At the present time we are not aware of any objective evidence to support the use of a locking plate for the clavicle, although we do use a pre-contoured plated in this setting.

We thank Dr. Chitgopkar and his colleagues for their interest in our work.

To The Editor:

We would be interested in learning more about concomitant shoulder girdle injuries in patients who sustained displaced mid-clavicular fractures, and the sling used for conservative management(1).

We share the authors’ experience in that compliance with using a sling is variable and most patients discard the sling when pain subsides. Therefore, some of the patients randomised to the non-operative management group probably did not complete their treatment regimen.

The number of patients complaining of a sensitive and/or painful fracture site was equal in both the groups, but if hardware irritation and/or prominence are combined with the above symptom, then the number of patients in the operative group expressing these symptoms is double that of the non-operative group.

In our experience young patients with a malunited midshaft clavicle fracture present with classic impingement symptoms and signs. Our indications for operative fixation are: patients older than 16 years;patients who are compliant; overlap of fracture ends of 2 cm or more, even if the injury is due to low energy trauma;comminuted fractures; impending skin penetration from the fractured bone ends; open fractures;associated neurovascular injuries; and fractures with mal-rotation.

Our choice of implant is a contoured locking plate.

The authors did not receive any outside funding or grants in support of their research for or preparation of this work. Neither they nor a member of their immediate families received payments or other benefits or a commitment or agreement to provide such benefits from a commercial entity. No commercial entity paid or directed, or agreed to pay or direct, any benefits to any research fund, foundation, division, center, clinical practice, or other charitable or nonprofit organization with which the authors, or a member of their immediate families, are affiliated or associated.

Reference:

1. Canadian Orthopaedic Trauma Society. Nonoperative Treatment Compared with Plate Fixation of Displaced Midshaft Clavicular Fractures. A Multicenter, Randomized Clinical Trial. J Bone Joint Surg Am 2007; 89: 1-10.

We thank Dr. Bernstein for his interest in our recent article entitled “Nonoperative treatment compared with plate fixation of displaced midshaft clavicular fractures. A multicenter randomized clinical trial”(1). We would have the following comments to make in response:

1. Our study did not contain an economic analysis. While we agree that the short term costs of operative treatment are greater than that of nonoperative care, this is offset by the much more rapid return to gainful employment, everyday activities and lack of dependence on others that we found in the operative group.

2. It is unlikely that there will ever be a randomized study in orthopaedics where the two treatments (one operative, one nonoperative) are “completely identical in terms of cost and risk”. Therefore, practically speaking, we must do the best we can with the methods at our disposal. We felt that a randomized comparing operative to nonoperative treatment in this area would be appropriate.

3. These procedures were performed at seven university hospitals and one community hospital. We feel that plating of an acute clavicular shaft fracture is something which is within the technical grasp of the typical orthopaedic fracture surgeon.

4. While it would be ideal to blind the patients as to their chosen treatment method, practically speaking, this is impossible to do in a surgery/no surgery trial. While we are aware of the phenomenon that patients who deliberately choose or elect to have a surgical intervention, may represent an intrinsically different subgroup than those who choose not to undergo surgery, patients in our study did not “subject themselves” to a surgery: they were randomized to surgery and thus are no more “psychologically inclined” to surgery than the group that was randomized to a nonoperative care.

5. We did have a substantial prevalence of surgical complications in our group, and listed them--they include hardware failure, nonunion, and infection. We point out specifically that this procedure is not without risk. Fortunately we did not experience any of the catastrophic complications (pneumothorax, neurovascular injury) that have traditionally been associated with this procedure. However, they are, of course, (remotely) possible as with any surgical procedure.

6. We hope this initial foray into a randomized trial comparing nonoperative and operative care is not the “last word” on this topic. We are aware of several other similar randomized trials planned or in progress. We encourage other authors to investigate the same topic and feel strongly that there is still much more information to be obtained regarding this injury and its treatment. This would include the timing of surgery, prognostic indicators, the use of intramedullary fixation devices versus plates, etc. We eagerly look forward to other similar studies being presented and published.

We feel that traditionally the treatment of the displaced midshaft clavicular fracture in active individuals has been dominated by nonoperative treatment, against mounting evidence that such treatment often results in a suboptimal outcome. Our study provides important information regarding the improved outcomes that can be obtained in select patients (active, healthy individuals between 16 and 60 years of age) with completely displaced (mean displacement 2 cm.) midshaft clavicular fractures. We urge orthopaedic surgeons to read this article carefully and hope that it helps them in their care of such individuals. We stress that our study deals with a select group and that the results therein are not necessarily generalizable to all individuals with all types of clavicle fractures.

Thank you for the opportunity to respond to this letter.

Reference:

1. Canadian Orthopaedic Trauma Society, c/o Michael D. McKee, M.D., FRCS(C). Nonoperative treatment compared with plate fixation of displaced midshaft clavicular fractures. J Bone Joint Surg Am. 2007;89:1-10

To The Editor:

The article,“Nonoperative Treatment Compared with Plate Fixation of Displaced Midshaft Clavicular Fractures” (1) is an excellent prospective randomized trial, well conceived and well executed, on a question of current clinical interest. The authors deserve our congratulations and our gratitude.

I only have one small complaint with the conclusion, namely the authors’ claim that the study “supports primary plate fixation of completely displaced midshaft clavicular fractures in active adult patients”.

In fact, it does no such thing.

This study provides outcomes data to guide the surgeon and the patient regarding the management of a particular case at hand. Even if “operative fixation of a displaced fracture of the clavicular shaft results in improved functional outcome and a lower rate of malunion and nonunion compared with nonoperative treatment at one year of follow-up”, not all patients should select this option. Indeed, had the study simply mentioned the financial cost of treatment –with the same noted differences in outcome-- it very easily could have claimed to support “non-operative management of completely displaced midshaft clavicular fractures in active adult patients”.

As we knew even before this study was undertaken, operative fixation is to be chosen if and only if the balance of personal costs and benefits given the anticipated results tilts in that direction. This study helps us make that decision with greater confidence, but it does not “support” one treatment or another: a study “supports” a particular treatment approach over another only when the two treatments are completely identical in terms of cost and risk, and one offers a superior outcome. In all other cases, the study merely offers grist for the decision analytic mill.

(I should add parenthetically that when I cite this study in my discussions with patients-- and I will-- I intend to tell them three things: 1) that the reported surgical outcomes here in all likelihood represent an upper bound on the expected results, as this study was conducted at trauma centers; 2) that because there was no patient blinding, the datum regarding patients' perceived satisfaction is potentially biased to the point of meaninglessness by cognitive dissonance—ie, patients who subject themselves to surgery are psychologically inclined and indeed will convince themselves to believe that the surgery has helped; and 3) that the absence of major surgical complications does not mean that their risk is zero.)

One might say that I am like the man in the cliché, given an inch and now demanding a yard. I hope my comments are taken not in that spirit, as I am truly impressed with this work. Yet this excellent report serves to illustrate a glaring deficit in our literature: we have not grappled with (let alone solved) the problem of integrating outcomes data into a decision analytic model of patient preferences. Within the rubric of Evidence Based Medicine, this paper is probably the first, and not the last, word on how to manage clavicular fractures.

The authors of this did not receive any outside funding or grants in support of their research for or preparation of this work. Neither they nor a member of their immediate families received payments or other benefits or a commitment or agreement to provide such benefits from a commercial entity. No commercial entity paid or directed, or agreed to pay or direct, any benefits to any research fund, foundation, division, center, clinical practice, or other charitable or nonprofit organization with which the authors, or a member of their immediate families, are affiliated or associated.

Reference:

1. Canadian Orthopaedic Trauma Society, c/o Michael McKee, M.D. Nonoperative treatment compared with plate fixation of displaced midshaft clavicular fractures. A multicenter, randomized clinical trial. J Bone Joint Surg Am 2007;89:1-10