Intention-to-Treat Analysis and Accounting for Missing Data in Orthopaedic Randomized Clinical Trials

Transcription

1 2137 COPYRIGHT Ó 2009 BY THE JOURNAL OF BONE AND JOINT SURGERY, INCORPORATED Intention-to-Treat Analysis an Accounting for Missing Data in Orthopaeic Ranomize Clinical Trials By Amir Herman, MD, MSc, Itamar Busheri Botser, MD, Shay Tenenbaum, MD, an Ahron Chechick, MD Investigation performe at the Department of Orthopeic Surgery, Chaim Sheba Meical Center, Ramat-Gan, Israel Backgroun: The intention-to-treat principle implies that all patients who are ranomize in a clinical trial shoul be analyze accoring to their original allocation. This means that patients crossing over to another treatment group an patients lost to follow-up shoul be inclue in the analysis as a part of their original group. This principle is important for preserving the ranomization scheme, which is the basis for correct inference in any ranomize trial. In this stuy, we examine the use of the intention-to-treat principle in recently publishe orthopaeic clinical trials. Methos: We surveye eight leaing orthopaeic journals for ranomize clinical trials publishe between January 2005 an August We etermine whether the intention-to-treat principle was implemente an, if so, how it was use in each trial. Specifically, we ascertaine which methos were use to account for missing ata. Results: Our search yiele 274 ranomize clinical trials, an the intention-to-treat principle was use in ninety-six (35%) of them. There were significant ifferences among the journals with regar to the use of the intention-to-treat principle. The relative number of trials in which the principle was use increase each year. The authors ahere to the strict efinition of the intention-to-treat principle in forty-five of the ninety-six stuies in which it was claime that this principle ha been use. In forty-four ranomize trials, patients who ha been lost to follow-up were exclue from the final analysis; this practice was most notable in stuies of surgical interventions. The most popular metho of ajusting for missing ata was the last observation carrie forwar technique. Conclusions: In most of the ranomize clinical trials publishe in the orthopaeic literature, the investigators i not ahere to the stringent use of the intention-to-treat principle, with the most conspicuous problem being a lack of accounting for patients lost to follow-up. This omission might introuce bias to orthopaeic ranomize clinical trials an their analysis. Conucting a ranomize clinical trial is a consierable challenge. In an ieal trial, patients woul enter the stuy, comply with the assigne treatment, an complete the follow-up protocol, but this is rarely the case. Common problems in ranomize clinical trials inclue patients insistence that they receive a treatment to which they were not originally assigne or their failure to comply with the follow-up protocol i.e., either skipping a scheule appointment or ropping out from the stuy altogether. The intention-to-treat principle is intene to eal with some of these issues. The intention-to-treat principle ictates that all patients who ha been ranomly allocate to treatment(s) uner the auspices of the stuy are inclue in the final ata analysis accoring to the original treatment group to which they ha been ranomly assigne. Thus, the patients who crosse over to another treatment an those lost to follow-up are analyze accoring to their original treatment group 1-9. The aim of an intention-to-treat analysis is to preserve the ranomization scheme use to allocate the patients to the various treatment groups. This ranomization scheme forms the theoretical basis for the valiity of the statistical calculations. It is important that the conclusions of the stuy are capable of being generalize in orer to accommoate entire patient populations, not only the iniviuals inclue in a given stuy. Disclosure: The authors i not receive any outsie funing or grants in support of their research for or preparation of this work. Neither they nor a member of their immeiate families receive payments or other benefits or a commitment or agreement to provie such benefits from a commercial entity. No commercial entity pai or irecte, or agree to pay or irect, any benefits to any research fun, founation, ivision, center, clinical practice, or other charitable or nonprofit organization with which the authors, or a member of their immeiate families, are affiliate or associate. J Bone Joint Surg Am. 2009;91: oi: /jbjs.h.01481

2 2138 Fig. 1 Three types of intention-to-treat methoology were ientifie. With strict intention-to-treat analysis (sitt), patients are inclue in the ata analysis accoring to their original treatment allocation, regarless of whether they actually began the treatment, crosse over to another treatment group, or were lost to follow-up. Intention-to-treat analysis with exclusion of missing ata (exitt) is a metho in which patients lost to follow-up are not inclue in the final analysis. With the moifie intention-to-treat metho (mitt), patients are inclue in the follow-up analysis only if they actually receive the planne treatment. An alternative to the intention-to-treat principle is the per-protocol analysis or as-treate analysis, in which patients are analyze at the time of follow-up accoring to the treatment that they ha actually receive. The intention-to-treat analysis is sometimes referre to as an efficiency analysis, whereas the per-protocol analysis is referre to as an efficacy analysis 5,7. TABLE I Use of Intention-to-Treat Analysis Accoring to Year of Publication an Journal No. of Articles Reporting Intention-to-Treat Analysis/Total No. of Ranomize Clinical Trials 2005* 2006* 2007* 2008* Total J Bone Joint Surg Am 2/14 6/17 8/14 9/15 25/60 (42%) J Bone Joint Surg Br 3/12 2/12 4/12 4/10 13/46 (28%) J Arthroplasty 0/13 0/5 0/9 0/0 0/27 (0%) J Orthop Trauma 0/2 1/4 1/1 0/1 2/8 (25%) J Shouler Elbow Surg 0/0 1/1 0/2 1/3 2/5 (40%) J Han Surg Am 0/2 0/2 0/0 3/6 3/10 (30%) J Peiatr Orthop 0/2 0/1 0/1 0/0 0/4 (0%) Spine 13/28 14/32 20/34 4/19 51/113 (45%) Total 18/73 (25%) 24/74 (32%) 33/73 (45%) 21/54 (39%) 96/274 (35%) *A time-epenent tren was foun (p = 0.025) i.e., the proportion of ranomize clinical trials in which use of intention-to-treat analysis was reporte increase yearly. The highest proportions of stuies with use of intention-to-treat analysis were foun in J Bone Joint Surg Am an in Spine (p = 0.001).

3 2139 TABLE II Proportion of Patients Lost to Follow-up* Types of Interventions Compare Control Group Treatment Group 1 Treatment Group 2 Difference Total Surgical vs. nonsurgical (n = 11) 16.7 ± ± ± ± 6 Nonsurgical only (n = 65) 14 ± ± ± 20 1 ± ± 13 Surgical only (n = 17) 14.2 ± ± ± ± ± 12 *The values, which are percentages, are given as the mean an stanar eviation. The ifference in the proportion of patients lost to follow-up was calculate for each trial separately as the proportion of patients lost to follow-up in treatment Group 1 (or Group 2) minus the proportion of patients lost to follow-up in the control group. The mean an stanar eviation were then calculate. In this stuy, we examine the use of the intention-totreat principle in ranomize orthopaeic clinical trials an investigate whether the authors ha ahere to the strict efinition of this principle. Special emphasis was place on the hanling of missing ata i.e., the extent to which patients were lost to follow-up an the methos use to account for them. Materials an Methos We conucte a literature search of ranomize clinical trials publishe between January 2005 an August 2008, in eight leaing orthopaeic journals: the American an British volumes of The Journal of Bone an Joint Surgery, Spine, Journal of Peiatric Orthopaeics, Journal of Shouler an Elbow Surgery, The Journal of Arthroplasty, Journal of Orthopaeic Trauma, an The Journal of Han Surgery (American volume). The selection was base on a high citation inex, our wish to use journals from all orthopaeic subspecialties, our access to the journals, an the finings of a similar stuy on the use of levels of evience in orthopaeic journals 10. The reporte use of intention-to-treat analysis was etermine by reviewing the statistical methos section an searching for the string intent throughout the entire report of each ranomize clinical trial that we foun. We then evaluate in greater epth the trials in which the authors ha claime to have use intention-to-treat analysis. We ientifie three principal methos of application of the intention-to-treat principle: (1) strict aherence to the intention-to-treat principle i.e., stuies in which ata analysis inclue all ranomize patients accoring to their original treatment allocation (i.e., the authors ignore crossovers Fig. 2 Box-plot graphs showing the ifferences in the proportions of patients lost to follow-up accoring to the types of interventions in the clinical trials. The ifferences were calculate as the proportion of patients lost to follow-up in the treatment (Tx) group minus the proportion of patients lost to followup in the control group. The mile line of each plot is the meian, an the borers of the boxes represent the first an thir quartiles. The I-bars represent ata within 1.5 times the interquartile range, an the circles are outliers. It can be seen that more patients in the control group were lost to follow-up in clinical trials comparing surgical treatments with nonsurgical controls (p = 0.01).

4 2140 Fig. 3 The importance of the intention-to-treat principle is emonstrate by this example. In this trial, 400 patients were equally ranomize either to the surgical arm or to the nonsurgical arm an each of the interventions ha a similar effect. The surgery requires a two-week preparation perio in which the patients are treate with a splint, an uring this perio 10% of the fractures in each group isplace; 10% of the fractures in each group also isplace in the following year. Performance of an as-treate analysis will lea to the conclusion that the surgical treatment is superior to the nonsurgical therapy, with a relative risk reuction (RRR) of 0.6 (p = ). An intention-to-treat analysis will not lea to this erroneous conclusion. Aapte, with moifications, from: Montori VM, Guyatt GH. Intention-to-treat principle. CMAJ. 2001;165: an ajuste for missing ata); (2) intention-to-treat analysis with exclusion of missing ata i.e., stuies in which the ata analysis was conucte accoring to the patient s original treatment allocation (crossovers were ignore) but only patients who complete the follow-up protocol were inclue; an (3) moifie intention-to-treat analysis stuies in which the ata analysis inclue only the patients who starte the treatment being evaluate (e.g., those who attene at least the first session of physical therapy). Figure 1 presents a graphical efinition of the three types of intention-to-treat analysis. Each clinical trial was further classifie accoring to the nature of the interventions uner stuy: surgery compare with nonsurgical management, two nonsurgical interventions, an two surgical interventions. Articles that accounte for missing ata were evaluate for the manner in which the missing ata were ajuste. Four methos were use: (1) the last observation carrie forwar i.e., the last observe value is use as a replacement for the missing observations; (2) mean/meian imputation i.e., the mean or meian for the treatment group is use as a replacement for the missing observations; (3) worst outcome imputation i.e., all missing ata are replace by the worst outcome; an (4) longituinal regression imputation i.e., imputation is one accoring to a preictive regression moel base on several covariates. The proportion of patients lost to follow-up was calculate as the number of patients missing at the last follow-up point for each treatment arm, ivie by the number of patients who were originally allocate to that treatment arm. The total proportion of patients lost to follow-up was calculate as the total number of patients missing at the last follow-up point ivie by the total number of patients who unerwent ranomization. The ifference in the proportion of missing ata between groups was calculate as the proportion of patients lost to follow-up in the intervention group minus the proportion of patients lost to follow-up in the control group. In trials that compare nonsurgical an surgical management, the nonsurgical group was always consiere the control. For trials incluing two surgical (or nonsurgical) interventions, the previous gol stanar surgery was consiere as the control. The maximum ifference in the proportion between the treatment arms an the control group was use in stuies that inclue more than one intervention arm. The uration of follow-up reporte in each trial was recore as (1) until ischarge, (2) up to six months (incluing six months), (3) longer than six months to one year (incluing one year), or (4) longer than one year. We compare the total proportion of patients lost to follow-up in each stuy accoring to the uration of follow-up.

5 2141 TABLE III Effect of Type of Analysis on the Results of a Comparison of Surgical an Nonsurgical Treatment of Displace Mishaft Clavicular Fractures 18 * Surgery (N = 62) Original Analysis Nonsurgical (N = 49) P Value Surgery (N = 67) Intention-to-Treat Analysis with Last Observation Carrie Forwar Nonsurgical (N = 65) P Value Nonunion Malunion requiring further treatment Woun infection an/or ehiscence Harware irritation requiring removal Complex regional pain synrome Surgery for impening open fracture Transient brachial plexus symptoms Abnormality of acromioclavicular or sternoclavicular joint Early mechanical failure Other Total 23 (37%) 31 (63%) (34%) 31 (48%) 0.15 *In the original analysis, the patients who were lost to follow-up were exclue. The intention-to-treat analysis with use of last observation carrie forwar was conucte with the assumption that all of the missing patients i not have complications. The conclusions from the analysis accounting for missing ata iffer from those of the original analysis. Each ranomize clinical trial was evaluate by at least two of the four authors of the present stuy. Our agreement regaring the above-mentione classifications range from 80% to 95%. In cases in which there was isagreement, the ranomize clinical trial was reviewe by all of the authors an further iscusse until consensus was achieve. Statistical analysis was performe with use of R software (Vienna, Austria) 11. The chi-square test was performe to compare use of the intention-to-treat principle among the selecte journals. We checke for a time-epenent tren between use of the intention-to-treat principle an the year of publication i.e., a yearly increase in the proportion of ranomize clinical trials in which intention-to-treat-base analysis was use. This was one by performing a logistic regression analysis with the use of the intention-to-treat principle as the epenent covariate an the year of publication as the explanatory covariate. The proportion of patients lost to follow-up was reporte as a mean an stanar eviation. Analysis of both the total proportion an the ifference in the proportion of patients lost to follow-up accoring to follow-up time an intervention types (e.g., surgical compare with nonsurgical treatment, two nonsurgical treatments, or two surgical treatments) was one with the Kruskal-Wallis test. All of the p values reporte are two-sie. Source of Funing No external funing source finance this research. Results We foun 274 ranomize clinical trials, an the intentionto-treat principle ha been use in ninety-six (35%) of them (Table I). The highest proportions of stuies using the intention-to-treat principle were publishe in the American volume of The Journal of Bone an Joint Surgery an in Spine (42% an 45%, respectively), an the ifference among journals was significant (p = 0.001). There was a tren for a yearly increase in the proportion of ranomize clinical trials using intention-to-treat analyses between 2005 an 2008 (p = 0.025). A strict intention-to-treat analysis was use in forty-five (47%) of the ninety-six trials, an intention-to-treat analysis with exclusion of missing ata was use in forty-four (46%), a moifie intention-to-treat metho was use in six (6%), an the metho of intention-to-treat analysis was unclear from the escription in one. The investigators exclue the patients lost to follow-up in twenty-one (72%) of the twenty-nine trials in which surgical intervention was stuie. The authors use strict intention-totreat analysis in thirty-seven (56%) of the sixty-six trials in which only nonsurgical interventions were consiere, an the investigators exclue the patients lost to follow-up in twentythree (35%) of those sixty-six trials. All six of the trials in which a moifie intention-to-treat metho was use involve nonsurgical intervention groups. The surgical an nonsurgical ranomize clinical trials iffere significantly with regar to the metho of intention-to-treat analysis (p = 0.002). No patient was lost to follow-up in seventeen (38%) of the forty-five trials in which strict intention-to-treat analysis

6 2142 was use. In the other twenty-eight of these trials, the authors accounte for missing ata. Last observation carrie forwar was use in eighteen (40%) of the forty-five trials with use of strict intention-to-treat analysis, the mean or meian for the treatment group was use to replace missing ata in four (9%), a complex imputation metho (regression) was use in two, an worst-outcome imputation was employe in three. One article i not specify how the missing ata were accounte for. After exclusion of three stuies in which the exact number of patients lost to follow-up was not available, we etermine the proportions of patients lost to follow-up, accoring to treatment arm, in each ranomize control trial in which the intention-to-treat principle ha been use. We foun that 13.2% ha been lost from the clinical trials that compare surgical an nonsurgical interventions; 14.4%, from the trials comparing two nonsurgical interventions; an 12.6%, from the trials comparing two surgical interventions (Table II). These ifferences i not reach a level of significance (p = 0.9). The total proportion of patients lost to follow-up, however, was foun to increase significantly as the follow-up time increase (p = ). The means of the total proportions of patients lost to follow-up were 0.5% for the eight clinical trials that laste until hospital ischarge, 14% for the twenty-seven with a follow-up time of up to six months, 14% in the thirty-one with a follow-up time of more than six months to one year, an 17% in the thirty with a follow-up time of more than one year. Figure 2, which presents box-plot graphs, an Table II show that the proportions of patients lost to follow-up iffere accoring to the type of clinical trial (i.e., the interventions uner stuy). More patients in the control (nonsurgical) group were lost to follow-up in the clinical trials that compare surgical with nonsurgical treatments (p = 0.01). Discussion Our results showe that the use of the intention-to-treat principle in orthopaeic ranomize controlle stuies is still relatively sparse an not uniform. The authors of about half of the clinical trials in which it was claime that the intentionto-treat principle ha been use ha not ahere to its strict requirements. Most of the violations of the intention-to-treat protocol involve the hanling of missing ata i.e., the exclusion of patients lost to follow-up instea of ajustment for missing ata. This might introuce bias to the results an conclusions of the trials. The possibility of this bias is even more relevant to clinical trials comparing surgical with nonsurgical treatments, in which the proportion of patients lost to follow-up was shown to be larger in the control (nonsurgical) groups. We believe that patients in nonsurgical groups are less motivate to comply with a follow-up protocol, especially if they o not have any complications, an that patients who unergo surgery might feel more incline to comply with follow-up scheules. Several of the articles on methoology that have been publishe in the orthopaeic literature i not precisely elineate what is meant by the intention-to-treat principle Those publications efine intention-to-treat analysis as the analysis of patients accoring to the original treatment group to which they ha been allocate, regarless of whether they crosse over at any point. This efinition completely ignores the nee to account for missing ata. In a survey similar to ours, the authors examine the use of the intention-to-treat principle in articles publishe in 1997 in BMJ: British Meical Journal, The Lancet, JAMA: The Journal of the American Meical Association, an The New Englan Journal of Meicine 4. The authors foun that use of the intention-to-treat principle was mentione in 119 (48%) of the ranomize clinical trials. The authors of twelve of these 119 stuies i not inclue patients who ha not starte the assigne treatment. The survey i not inclue examination of the hanling of missing ata. Some primary outcome ata were reporte to be missing from eighty-nine (75%) of the trials in which the intention-to-treat principle ha been use, an >10% of the primary outcome ata were missing from twenty-nine (24%). Figure 3 provies an example illustrating the importance of implementing the intention-to-treat principle. In this example, a trial is esigne to compare a new surgical proceure with a cast immobilization technique for treatment of a nonisplace fracture. Displacement is the primary failure en point. Two hunre patients are ranomize to each treatment arm. The surgery requires a two-week preparation perio in orer for local eema to subsie an for skin conitions to be suitable for the operation. During this perio, the patients scheule for surgery are treate with a splint an unergo aily skin examination. The two approaches (surgical an cast treatment) have the same outcome: specifically, 10% of the fractures in each treatment arm isplace uring the two weeks after the injury an another 10% in each treatment arm isplace uring the one-year trial perio. With use of the as-treate approach, there are sixty treatment failures in 220 patients who were treate conservatively: 200 treate with a cast an twenty who unerwent fracture isplacement while being treate with a splint. These results are compare with twenty instances of fracture isplacement that occurre in 180 patients who unerwent surgery. Accoring to this analytic approach, the relative risk reuction is 0.6 (p = ), favoring the new surgical proceure. With use of intention-to-treat analysis, however, the 400 patients remain in the groups to which they were ranomize an forty in each treatment arm of 200 patients have fracture isplacement, so the relative risk reuction is now 0 (p = 1). Note that the researchers coul have ranomize the patients after the two-week skin-preparation perio. Note also that, with the intention-to-treat approach, the investigators are comparing the two treatment arms i.e., they are comparing surgery precee by splinting with cast treatment as oppose to comparing surgery alone with cast treatment. An example illustrating the importance of accounting for missing ata can be foun in a ranomize clinical trial, by the Canaian Orthopaeic Trauma Society 18, comparing surgical with nonoperative treatment for isplace mishaft clavicular fractures. The authors of that stuy state that they use the intention-to-treat principle. Careful scrutiny of the article, however, reveals that the analysis i not inclue patients lost

7 2143 to follow-up. Five (7%) of the sixty-seven patients in the surgical treatment group an sixteen (25%) of the sixty-five patients in the nonoperative group were lost to follow-up. This ifference was foun to be significant (p = 0.014). Two patients in the nonoperative group who ha complications were not followe an were omitte from the analysis. We performe our own analysis using last observation carrie forwar, which means that all of the patients who were free of complications at their last visit were consiere henceforth to be patients without complications. A comparison of that analysis with the analysis by the investigators from the Canaian Orthopaeic Trauma Society reners the original conclusions in favor of the surgical intervention less convincing (Table III). There are several methos with which to ajust for missing ata. Choosing one can be a challenging task, an the choice coul be closely relate to the reason for the missing ata. If, as hypothesize above, the reason for lack of compliance is lack of clinical nee (e.g., no complications), it woul be logical to use last observation carrie forwar. The influence of the ajustment for missing ata shoul, however, be examine by trying several ajustment methos, an approach terme sensitivity analysis 5, In our survey, the authors of only six articles mentione having compare more than one imputation metho for missing ata. It is important to note that the worstoutcome imputation metho has receive much criticism. Use of this approach might introuce bias when a large amount of ata is missing in one group, causing its treatment to seem unsuccessful. The last observation carrie forwar approach has also been criticize as being inappropriate because the assumption that the last observation is the long-term outcome is not always justifie. There are many other contemporary methos of hanling missing ata, incluing multiple imputation, expectation-maximization algorithms, an propensity ajustments. Missing-ata imputation can epen on other variables, such as the number of patients an the number of en points in a stuy. For the sake of brevity, we i not escribe all of the methos or use them all in our ata-set example. The intereste reaer can fin many references for these methos in the current literature. Over the four years stuie, the authors of only about a thir of the orthopaeic trials use some variation of the intention-to-treat principle. As has been mentione for ranomize trials in surgery 22, there is still room for improvement in the performance an analysis of ranomize clinical trials in orthopaeics. We conclue by quoting from Fisher et al. 3 : One of the great intellectual avances of the twentieth century [was] the concept of ranomization. We shoul go to greater lengths to preserve the benefits of ranomization through correct implementation of the intention-to-treat principle. n Amir Herman, MD, MSc Itamar Busheri Botser, MD Shay Tenenbaum, MD Ahron Chechick, MD Department of Orthopeic Surgery, Chaim Sheba Meical Center, Ramat-Gan 52621, Israel. aress for A. Herman: amirherm@gmail.com References 1. Begg CB. Ruminations on the intent-to-treat principle. Control Clin Trials. 2000;21: Bubbar VK, Kreer HJ. The intention-to-treat principle: a primer for the orthopaeic surgeon. J Bone Joint Surg Am. 2006;88: Fisher LD, Dixon DO, Herson J, Frankowski RH, Hearron MS, Peace KE. Intention to treat in clinical trials. In: Peace KE, eitor. Statistical issues in rug research an evelopment. New York: Marcel Dekker; p Hollis S, Campbell F. What is meant by intention to treat analysis? Survey of publishe ranomise controlle trials. BMJ. 1999;319: Lachin JM. Statistical consierations in the intent-to-treat principle. Control Clin Trials. 2000;21: Erratum in: Control Clin Trials. 2000;21: Montori VM, Guyatt GH. Intention-to-treat principle. CMAJ. 2001;165: Ellenberg JH. Intention to treat analysis. In: Armitage P, Colton T, eitors. Encyclopeia of biostatistics. Chichester, Englan: John Wiley an Sons; p Heritier SR, Gebski VJ, Keech AC. Inclusion of patients in clinical trial analysis: the intention-to-treat principle. Me J Aust. 2003;179: Chung KC, Burns PB. A guie to planning an executing a surgical ranomize controlle trial. J Han Surg Am. 2008;33: Obremskey WT, Pappas N, Attallah-Wasif E, Tornetta P 3r, Bhanari M. Level of evience in orthopaeic journals. J Bone Joint Surg Am. 2005;87: R Development Core Team. R: A language an environment for statistical computing. Vienna, Austria: R Founation for Statistical Computing; Boutron I, Ravau P, Nizar R. The esign an assessment of prospective ranomise, controlle trials in orthopaeic surgery. J Bone Joint Surg Br. 2007;89: Chan S, Bhanari M. The quality of reporting of orthopaeic ranomize trials with use of a checklist for nonpharmacological therapies. J Bone Joint Surg Am. 2007;89: Cowan J, Lozano-Calerón S, Ring D. Quality of prospective controlle ranomize trials. Analysis of trials of treatment for lateral epiconylitis as an example. J Bone Joint Surg Am. 2007;89: Dulai SK, Slobogean BLT, Beauchamp RD, Mulpuri K. A quality assessment of ranomize clinical trials in peiatric orthopaeics. J Peiatr Orthop. 2007;27: Petrie A. Statistics in orthopaeic papers. J Bone Joint Surg Br. 2006;88: Poolman RW, Struijs PA, Krips R, Sierevelt IN, Marti RK, Farrokhyar F, Bhanari M. Reporting of outcomes in orthopaeic ranomize trials: oes blining of outcome assessors matter? J Bone Joint Surg Am. 2007;89: Canaian Orthopaeic Trauma Society. Nonoperative treatment compare with plate fixation of isplace mishaft clavicular fractures. A multicenter, ranomize clinical trial. J Bone Joint Surg Am. 2007;89: Little R, Yau L. Intent-to-treat analysis for longituinal stuies with rop-outs. Biometrics. 1996;52: Salim A, Mackinnon A, Griffiths K. Sensitivity analysis of intention-to-treat estimates when withrawals are relate to unobserve compliance status. Stat Me. 2008;27: Wright CC, Sim J. Intention-to-treat approach to ata from ranomize controlle trials: a sensitivity analysis. J Clin Epiemiol. 2003;56: Jacquier I, Boutron I, Moher D, Roy C, Ravau P. The reporting of ranomize clinical trials using a surgical intervention is in nee of immeiate improvement: a systematic review. Ann Surg. 2006;244: