We will conduct a pragmatic multicenter, randomized, controlled, outcome assessor-blinded,
clinical superiority trial. The objective is to compare surgical fixation of humeral shaft
fracture to non-surgical treatment with early identification and treatment of delayed union
by a patient-reported outcome after 52 weeks. .
Null-hypothesis:
The DASH score at 52 weeks after surgical treatment is not superior to non-surgical treatment
with the option of early crossover surgery in patients with humeral shaft fractures
The trial population is divided in two age-groups due to the changes in DASH score by age.
The definition of delayed union differs in the young and elderly population to consider
dissimilarity in bone healing rates:
SHAFT-Y for the young with an age cut-off of 18 to 64 years. The early identification
and treatment of delayed union is set to 6 to 12 weeks
SHAFT-E for the elder with an age cut-off +65 years. The early identification and
treatment of delayed union is set to 12 to 26 weeks
Sites from Denmark, Sweden and Norway have been recruited and spans from academic level I to
level III trauma centers
287 patients (n=163 for SHAFT-Y, n=124 for SHAFT-E) with a humeral shaft fracture will be
equally randomized to surgical treatment or non-surgical treatment in each group.
Patients admitted to the emergency department in one of the trial sites and fulfil the
eligibility criteria, will be invited to enroll into the trial. They will be given time to
consider and be scheduled for a consultation with a trial worker within 10 days. If written
consent is obtained at the consultation, randomization will occur immediately after.
Treatment will be performed within 14 days after injury
Surgical treatment. The specific treatment is decided by the treating surgeon/department
Non-surgical treatment with the option of early secondary surgery from 6-12 weeks for
SHAFT-Y and 12-26 weeks for SHAFT-E
Patients can be offered to undergo early crossover fixation with a surgical procedure of the
surgeon's choice, if one of these criteria are met:
Unacceptable pain experienced by the patient
Severe pain with gross instability of the fracture site assessed by:
Unable to en bloc elevate the arm due to clear fracture instability
Gentle manipulation of the fracture site. Gentle manipulation should respect the risk of
callus breakage
Severe problems tolerating the brace, e.g. discomfort, skin irritation, wounds, hygiene
problems.
The patients that undergo early crossover surgery will have the reason for crossover
thoroughly noted. We anticipate the surgical procedures will be similar to the ones
previously mentioned with the possible addition of bone graft.
A computerized database software, Research Electronic Data Capture (REDCap) will be used to
generate an irreversible random allocation sequence and perform block randomization with
selected block sized of 2 and 4, which will be stratified on site and age (18-64 and +65).
Patients will be assigned to the trial with an allocation of 1:1 to either surgical treatment
or non-surgical treatment. The trial worker acquires the allocated treatment from the central
coordinator with randomization rights to REDCap. The trial worker then initiates the
treatment, either by scheduling the surgery date or applying the chosen non-surgical method.
The two groups (SHAFT-Y and SHAFT-E) require individual sample size calculations. Two
standard deviations (SDs) were obtained from the data of the FISH trial(13) and were
separated in age groups of 18-64 years and 65 years and above. By the distribution-based
approach, one half a SD corresponds to the minimal important change (MIC). The calculations
are powered to detect a MIC of 7 points in the young and 10 points in the elderly group in
DASH, respectively. Two independent means sample size calculation were performed. For SHAFT-Y
the following data were included: Mean difference= 7.0, SD= 14.91, α= 0.05 and power= 0.8.
For SHAFT-E the following data were included: Mean difference= 10.0, SD= 18.59, α= 0.05 and
power= 0.8. Based on the preceding assumptions and including an attrition of 15%, the total
sample size is estimated to 163 patients for SHAFT-Y and 124 patients for SHAFT-E.
Primary analysis Descriptive statistics will be used to report demographic data. Demographic
data and outcome measures will be tested visually and statistically (i.e. Shapiro Wilks
test). Numeric variables will be summarized by means, standard deviations and 95% confidence
intervals (95% CI). Median and interquartile ranges will be used when normal distribution is
not met. Categorical variables will be summarized by frequency and proportion. For group
comparison with numerical data, a student's t-test will be used if data is normally
distributed, otherwise a non-parametric test will be used. For categorical data a Chi-square
test will be used for group comparison. An intention-to-treat (ITT) analysis of the primary
outcome will be conducted by univariable linear regression, including all patients that do
not meet the withdrawal criteria and will be conducted to minimize bias within results. A
sensitivity analysis will test the effects of non-adherence to protocol by conducting a
per-protocol analysis and includes only patients who comply with the protocol. For missing
data points in an outcome measure, a multiple imputation analysis using predictive covariates
(age, sex, smoking, alcohol, UCLA activity, ASA grade)(50-52) will be conducted to deal with
nonresponse bias. For comparison we will carry out a sensitivity analysis excluding all the
missing values.
Data will be considered statistically significant if p-values < 0.0471.
Secondary analysis In order to validate data a linear regression analysis will be computed
with DASH score as the dependent variable and treatment modality as the independent variable.
Additional regression analysis will be carried out between the early crossover group and the
primary treatments. A multivariate regression analysis will be conducted to adjust for
potential confounders. Variables adjusted for are: age, sex, smoking, alcohol, UCLA activity,
ASA grade. Furthermore, we will analyze the longitudinal observations by applying a linear
mixed effects regression model, including modality and time as well as a modality-time
interaction as fixed effects and a random intercept for each patient. Data will be summarized
as coefficients with 95% CIs and variance will be summarized as r-squares, adjusted
r-squares, predicted r-squares, standard errors. Coefficients will be considered
statistically significant if p-values < 0.05.
Outcome timepoints Subjective and objective outcome measures will be obtained at following
time points: pre-injury, baseline, 6 weeks, 12 weeks, 26 weeks, 52 weeks, 2 years and 5
years.