Reducing fracture incidence in children with osteogenesis imperfecta: contribution of orthotics to bisphosphonates treatment

Abstract Purpose To determine whether there was a reduction in fracture incidence amongst children with OI who were treated with both bisphosphonates and orthoses. Objective Was there an additional reduction in fracture incidence amongst children with Osteogenesis Imperfecta (OI) treated with both bisphosphonates and Hip-Knee-Ankle-Foot-Orthosis (HKAFO)? Materials and methods Of the 129 OI patients treated from 1990 to 2017, retrospective data from 48 patients who participated in the bisphosphonates-orthosis regime were analyzed including the incidence of fractures and modalities of fracture treatment. Results Bisphosphonates usage was more frequent than bracing and there were more positive changes (smaller or equal number of fractures each year) than negative changes (more fractures each year); negative changes were scarce, explained by non-compliance with the use of bracing. Poisson regression models were significant for positive changes, whereas the interaction between them was borderline significant. The main finding is that the association between bisphosphonates usage and the number of positive changes was stronger among the patients who used braces more frequently and weaker among patients who used bracing less frequently. Conclusions Bracing of OI patients has an additive effect on bisphosphonate treatment in fracture prevention which should lead to the reconsideration of a hybrid approach to OI management. IMPLICATIONS FOR REHABILITATION Two key goals of treatment of Osteogenesis Imperfecta (OI) patients include decreasing fracture incidence and improving function and independence as supported by a multi-disciplinary approach that combines medical, orthopaedic and rehabilitation treatments. Although the literature provides evidence that bisphosphonates reduce the frequency of fractures, there have not been reports of its effect when used with orthoses. Orthoses for OI patients have an additive effect on bisphosphonate treatment in fracture prevention. These results contribute to making an informed decision regarding this hybrid approach to OI management.


Introduction
Osteogenesis imperfecta (OI) is a group of hereditary diseases characterized by increased bone fragility due to a defect in collagen formation which is the major structural protein of bone tissue [1,2].Collagen forms a protein network on which the bone mineral settles which gives the bone its strength and resistance to the load [3].The primary manifestations of the disease include depletion of bone density, decrease in bone strength, multiple fractures, limb and spinal deformities, hyperlaxity, and low stature due to decreased bone growth [4].
Sillence et al. [5] defined four clinical types of OI: Type I is expressed as single fractures, especially during the accelerated growth period, normal or almost normal stature, and minor or no limb and spinal deformities.Type II is the lethal form of the disease.Type III is characterized by multiple fractures including intra-uterine fractures and limb and spinal deformities that usually require surgical intervention.Type IV patients sustain multiple fractures, their stature is low and there are limb and spinal deformities; all may have dentinogenesis.Even within these four types, there are additional differences, heterogeneity, and variability.Sillence's [5] classification has since been expanded due to genetic diagnosis although OI patients meet the clinical and radiological characteristics of one of the original subtypes [6,7].At the 2009 meeting of the International Nomenclature group for Constitutional Disorders ICHG of the Skeleton (INCDS) [8], a decision was made to group the known OI syndromes into five groups, that is, preserving the primary four groups and adding OI type V. OI type V patients have moderate to severe bone fragility with progressive calcification of the inter-osseous membranes in the forearms and increased propensity to develop hyperplastic callus [9].
The skeletal manifestations of OI are mainly fractures, most common in long bones but also vertebrae; most severe types manifest fractures of one or more ribs.Limb deformity and bowing can present prior to fracture or as a result of one.Spinal deformation and short stature are also common, especially with more severe types of OI [4,7].It is important to note that many children with OI experience complex psychosocial difficulties associated with the disease, pain and functional limitations [10].Dos Sontos et al. [10] suggested that many feel apprehension and even fear of their susceptibility to frequent fractures.
Until the introduction of treatment with cyclic administration of bisphosphonates, no effective medication for OI was available [11].Thus, treatment strategies focused primarily on non-operative (orthoses) and surgical interventions (rodding).Historically there have been reports of long-leg orthoses for children with OI [12].These types of orthotic systems were typically recommended to support weak muscles, control joint alignment, or reduce motion and pain.Although long-leg orthoses appear to promote upright balance, they likely do not reduce the load on the limb [13,14].Clinical experience using Hip-Knee-Ankle-Foot orthosis (HKAFO) in children with types III and IV OI has shown reduced tibial bowing, improvement in gross motor skills and some return to functional activity following fracture and surgical treatment [14,15].Such reports support the notion that long-leg orthoses do not significantly restrict function.Protected gait by orthoses helps to break the cycle of attempted walking leading to a fracture, immobilization, further osteoporosis and re-fracturing.Indeed, children who do not use orthoses spent more time in minimal activity and less time in moderate and strenuous activity compared to their peers who wore orthoses [14].Moreover, muscle atrophy and a decline in strength were not associated solely with orthoses; muscle strength declined during intervals when children did or did not use orthoses, even somewhat more so during the latter interval.The relative preservation of strength during the intervals with orthoses may be a result of the child's upright stance and weight bearing during functional activities [14].
Gerber et al. [14] have shown that children who wear long-leg orthoses are more independent and have more stamina while performing functional activities than those without orthoses.Gerber et al. [14] evaluated the effects of withdrawal of long-leg orthoses on activity and ambulation in young children with type III or IV OI who were able to walk with an orthosis.Ten children were divided into two groups where one group of five children currently using orthoses continued to use them for 16 months whereas their age-matched pairs had their orthoses withdrawn.Descriptive statistics pointed to differences in the functional level and activity intensity when the children used orthoses compared without orthoses; seventeen fractures of lower extremities occurred during all the periods without orthoses, and only eight occurred during the intervals with orthoses.Gerber et al. [14] indicated that treatment should be aimed at improving the child's functional abilities since this is the ultimate measure of good healthcare.Engelburt et al. [15] recommended ongoing physical rehabilitation and therapy for optimal OI management.Nuijuis et al. [16] and Aubry et al. [17] also strongly promoted the importance of personalized training programs designed to maintain each patient's physical activity throughout life.Indeed, in the absence of a cure for OI, physical activity is the primary way to achieve optimal mobility, functional independency and participation in society [18].
Improvement of clinical outcomes, reduced bone resorption and increased bone density following bisphosphonates have been reported [4,19].Although some information regarding the natural course of the various aspects of OI has been presented in cross-sectional and prospective studies [6,14], detailed follow-up covering a longer period is lacking.From Engelbert et al.'s [11,20] research it may be concluded that OI severity is manifested by a decline in muscle strength, joint range of motion (ROM), functional ability, and motor development.Prognostic factors for walking independently have also been suggested [21].
In summary, the two key goals of treatment of OI patients include decreasing fracture incidence and improving function and independence as supported by a multi-disciplinary approach that combines medical, orthopedic and rehabilitation treatments [6].
Although the literature provides evidence that bisphosphonates reduce the frequency of fractures [2,4], there have not been reports of its effect when used with orthoses.
At ALYN Hospital Pediatric and Adolescent Rehabilitation Center, a hybrid, preventive care approach has been the goal of treatment for children with OI; in addition to intravenous bisphosphonates, the children are fitted with lower extremity orthoses (HKAFO with anterior shells).The addition of a long-leg orthosis with a pelvic extension could support the integrity of the limb throughout the child's main growth period.However, some parents query the use of orthoses since such devices may cause discomfort to the child.
The objective of the retrospective data reported here is to determine whether there was a reduction in fracture incidence amongst children with OI who were treated with both bisphosphonates and orthoses.This will help identify the contribution of HKAFOs and help parents make an informed decision regarding this hybrid approach to OI management.

Methods
Design: Retrospective analysis of data from patient records.

Participants and data collection
Data collection was based on files of patients who attended ALYN hospital clinics from 1990 to 2017 inclusive.For each patient the following data were recorded, if available: 1) Demographic and background data that included age, gender, diagnosis, year of diagnosis, year started walking and first visit to ALYN hospital; 2) yearly longitudinal data that included number of fractures, location of fractures, recommendations for orthoses and bisphosphonates intravenous treatment.Note that since this study was based on convenience sampling of retrospective data, no calculation of sample size was made.Also note that there have been some changes in the type/model of the orthoses used over the long period of the retrospective study; these changes reflected only very small differences in the materials and fasteners.

ALYN management protocol
OI management at ALYN Hospital involves treatment with intravenous bisphosphonates (since 2000) and consideration of an HKAFO (since 1980).Bisphosphonates are primarily given based on the clinical findings and can be started as early as the age of 3 months.Multiple fractures, vertebral fractures and generalized bone pain, supported by low Dual-X-ray Absorptiometry (DEXA) indices, are indicators for treatment.Patient follow-up occurs every 3-4 months.Since 2000 the drug of choice at ALYN Hospital was pamidronate and since 2017 treatment first choice was replaced by zoledronate (zoledronic acid).
According to the ALYN Hospital protocol, HKAFO is indicated if there are more than 3-4 fractures, particularly of the lower limbs, as occurs in almost all the type III patients but in the other types as well.The patient is fitted with a lightweight, polypropelene HKAFO orthosis.The hips (femoral neck) are protected from shear forces by the addition of the pelvic band, and anterior shells are fitted over the femorae and tibiae to prevent direct trauma and encourage remodeling of anterior bowing.Initially, the ankles are fixed, and the knee joint has drop-locks to enable flexion while sitting.Patients are treated in physical therapy until they attain a stable gait while wearing the HKAFO.When the gait is stable, the ankle joints are opened.If gait continues to be stable, the knee joints are also opened.The decision to remove a protective device such as an HKAFO is primarily based on the absence of fractures for a number of years, depending on the patient's age; for adolescents, it is 2-3 years without a fracture and for preadolescents, it is 3-4 years without a fracture.If there is any doubt, the DEXA value helps in the decision making of removal of the HKAFO (i.e., Z-scores higher than −2 standard deviations [22,23].This process begins indoors and then increases gradually in distance outdoors.Once the child walks freely, has not had fractures and the DEXA score is at an acceptable level, then weaning from the HKAFO is started.If the patient undergoes surgery of the lower limbs, the orthosis is continued post-surgery.Intramedullary nailing of the lower limb long bones is indicated if the lower limbs cannot be braced due to marked deformities, or if there are recurring fractures at one particular site.

Outcome measures and data analysis
The outcome variable was based on the decrement or increment in the number of fractures between two consecutive years for each patient.Positive change was defined as a smaller or equal number of fractures in a certain year compared to the previous year (e.g., if the patient "John Doe" had three fractures recorded in 1990 and one fracture recorded in 1991, the event was tabulated as a positive change.If the patient "Jane Doe" had three fractures recorded in 1990 and three fractures recorded in 1991 the event was also tabulated as a positive change).A negative change was defined as a greater number of fractures in a certain year compared to the previous year.Thus, for each patient positive or negative changes with regard to the number of fractures between two consecutive years were noted.
It is important to note that an inherent feature of OI, and the natural course of the disease, is that patients have fractures; if not treated, there will always be fractures, and their occurrence may increase [24].Given this bleak state of affairs, a treatment that prevents the natural course of OI (namely, no change) should also be considered to be a successful outcome, just as a reduction in the occurrence of fractures is considered to be successful outcome.The clinical rationale for why "no change" should be viewed as a positive event is that the absence of fractures or, at least, no increase in fractures, is an indication that the goal of treatment has been accomplished.On the other hand, this goal would not be accomplished if there had been an increase in fractures in successive years (i.e., a negative event).
Since the data were recorded over 28 years, 27 change events were tabulated.These change events were counted and summed across the years resulting in two count variables that were the outcome variables in the current study: total number of positive changes (i.e., decreases and "no change") and total number of negative changes (i.e., increases).
The main predictors in the models were use of an orthosis and use of bisphosphonates.These data were counted and summed across the years resulting in two count data variables: total number of patients who had an orthosis and total number of patients who received bisphosphonates.The age of the participants was also included in the models as a controlling variable.
The data were described by means and standard deviations (SD), medians, ranges and inter-quartile ranges (IQR) and histograms.A Poisson regression model was applied in order to test the associations between the total use of orthoses and bisphosphonates and the total number of positive and negative changes.Prior to this analysis, Poisson distribution assumptions were tested by comparing the values of the mean and the variance of the outcome variables and by applying the Kolmogorov-Smirnov test for Poisson distribution for each outcome variable.All analyses were conducted using SPSS V25 and R software V4.0.2, variables were considered statistically significant under p-value ≤ 0.05.

Descriptive statistics
A total of 129 patients were treated for OI from 1990-2017.Fifty-nine did not participate in the orthosis-bisphosphonates regime and were excluded from this analysis.Of the 78 eligible patients, sufficient data were available in only 48 patients (a common occurrence in retrospective studies).At the time of the retrospective data collection, the patients' mean age was 12.33 years (SD: 5.32 years), the median age was 12.60 years (Range: 0.5 − 22.3 years; IQR: 8.93-16.08years).
Sixty percent (n = 29) of the analysed sample were boys and 40% (n = 19) were girls.All were diagnosed on clinical and radiologic criteria according to the Sillence classification.Forty-two percent (n = 20) have OI Type I, 17% (n = 8) have OI Type III and 2% (n = 1) have OI Type IV; in 40% (n = 19) of the analysed sample, the OI type was not recorded.
Table 1 presents the descriptive statistics of the predictors and outcome variables of the 48 patients who were included in the prediction models.Figure 1 presents the equivalent histograms.The data suggest that bisphosphonates usage was more frequent than the use of orthoses (Figure 1(a, b)), and that there were more positive changes than negative ones, the latter of which were found to be scarce (Figure 1(c, d)); note the difference with regard to zero events shown as grey bars in these two graphs.
The means and the variances of the outcome variables were 1.90 and 2.61 (respectively) for the number of positive changes and 0.75 and 0.66 (respectively) for the number of negative changes.The Kolmogorov-Smirnov test for Poisson distribution yielded insignificant p-values for both outcome variables (p = 0.42, for the number of positive changes, p = 1.00, for the number of negative changes).These results indicate that the distributions of the outcome variables may fit a Poisson distribution and thus Poisson regression models can be applied.

Prediction models
Table 2 presents the prediction model for the total number of positive changes.The whole model was significant (omnibus test: χ 2 (4) =18.68, p = 0.001), but neither of the two predictors (orthosis, bisphosphonates) was statistically significant.However, the interaction term yielded a p-value of p = 0.100, and is depicted in Figure 2(a, b).The association between the use of bisphosphonates and the number of positive changes was stronger among the patients who used an orthosis more frequently (figure 2(a)) and weaker    among the patients who used an orthosis less frequently (Figure 2(b)).The prediction model for the total number of negative changes was insignificant (omnibus test: χ 2 (4) =7.42, p = 0.120) and none of the predictors were statistically significant either (Table 3).This is not surprising due to the small variance of this outcome variable.

Discussion
This retrospective study examined relationships between the frequency of treatment with bisphosphonates and orthoses on the one hand and the reduction in fracture incidence in children with OI on the other.While bisphosphonates were used more frequently (mean ± SD = 11.2 ± 7.3), the number of children managed with an orthosis (mean ± SD = 3.5 ± 4.0) was considerable.Since the main variables (number of uses of orthoses and bisphosphonates) were well fit by a Poisson distribution, Poisson regression models were applied to the data.The key finding is the demonstration of a stronger association between bisphosphonate usage and positive events (i.e., a smaller or equal number of fractures each year compared to the previous year) among those who made greater use of orthoses, i.e., the goal of treatment was accomplished.In contrast, this goal was not accomplished if negative changes (i.e., more fractures each year compared to the previous year) occurred and was explained by non-compliance with the use of an orthosis (i.e., fractures occurred when the child was not in his/her orthosis).
The regression model was significant for positive changes; however, neither of the two predictors (number of uses of an orthosis and bisphosphonates) was statistically significant.However, the interaction effect was significant at p = 0.100 and thus suggests that the association between the use of bisphosphonates and the number of positive changes was stronger among the patients who used an orthosis more frequently and weaker among the patients who used an orthosis less frequently.
The prediction model for the total number of negative changes was insignificant meaning no associations were found with either treatment (orthoses and bisphosphonates) and an increase in the number of fractures.These results have considerable rehabilitative-therapeutic significance for the clinical community.
The patients whose data are shown in Figure 2(a) represent those whose total number of orthotic usage was above the median.That is, this group included twice the number of patients who are characterized as severe compared to the patients whose data are shown in Figure 2(b).The severe patient group also had a significant positive effect on the use of orthoses which does not exist in the milder patient group.In other words, the more severe the patient is, the more treatment helps since the patient has more opportunities to improve his condition.This is also of importance to clinicians.This provides further support for the use of an orthosis as an adjunct to bisphosphonates in severe OI.
There were several limitations of the study.As is the case with many retrospective studies, the sample size was relatively small.A second limitation, possibly related to the small sample size, is that the modelled interaction was significant at p = 0.10.However, this does not detract from the support for orthoses, a topic that has not been sufficiently studied in the literature.That is, the finding regarding the stronger association between bisphosphonate therapy and positive events only among those who made greater use of orthoses appears to be important.Though bisphosphonates alone are associated with positive events, the addition of an orthosis further enhances the positive events.This is important for therapists and physicians to understand so that adequate information is given to parents.
Complex psychosocial difficulties including apprehension and even fear of frequent fractures are associated with OI [10,19].These concerns by children and their families support the use of conservative means such as bracing as a way of reducing their distress.This may give them greater confidence and readiness to engage in activities that entail limb movement and even walking.
There is some concern that the use of bracing will lead to decreased mobilization and possibly increase osteoporosis [25].Although Gerber et al.'s [14] research predated the use of biphosphates, it does indicate that the use of an orthosis does not appear to cause weakness nor interfere with function.Since all treatments that serve to reduce stress merit consideration, the results of the present study suggest that the potential benefit of bracing is worthy of consideration.Subramanian [26] suggested that continuing education activities can be used to support the use of bracing for patients who have more severe types of OI.
Mueller et al. [27] are strong proponents of interventions that help children achieve independent seating by the age of 12 years since they will likely be able to walk and achieve independence in later life and, hence, greater quality of life; when a child starts to weight bare one should then consider whether bracing will help.Thus, the goal of therapy is to assist each child with OI to achieve developmental milestones and a level of independence while trying to prevent fractures and deformities.In the most severe patients, bracing may be considered a treatment modality even if only for short periods.
In the current study, there were seven patients who had more than one negative event (i.e., a fracture).This in-depth study cannot explain this as due to the severity of OI.In almost all cases there is detailed documentation regarding the events of the fractures.In all but one case, the fractures occurred when the patients were without an orthosis -falling out of bed, bathing in the bath, cessation of use of devices, and poor cooperation.This reinforces this study's claim about the additive effect of an orthosis to bisphosphonate medication.
In summary, the results demonstrate that there is a stronger association between bisphosphonate usage and positive events among those who made greater use of orthoses, i.e., the goal of treatment was accomplished.In contrast, this goal was not accomplished if negative changes occurred.These retrospective results appear to provide a basis to justify a future prospective study that will inform the healthcare community about the relative roles of surgery, orthoses and bisphosphonate treatment.It is recommended that concerns related to the children's self-image while wearing the described orthotics be addressed.In addition, physical evaluation should address the extent to which orthoses may contribute to the prevention of anterior bone angulation.Finally, future research should examine whether the increasing frequency of surgical intervention will reduce the need for bracing.

Figure 1 .
Figure 1.(a-d).histograms of the predictors and the outcome variables in the model.(a) number of uses of orthoses; (b) number of uses of bisphosphonates; (c) number of positive changes; (d) number of negative changes.

Figure 2 .
Figure 2. (a-b).(a) association between total use of bisphosphonates and total number of positive changes for patients whose total number of orthotic use was the median or lower.(b) association between total use of bisphosphonates and total number of positive changes for patients whose total number of orthotic use was above the median.note: the fit line shown in both graphs is the locally weighted scatterplot smoothing (loWess) line.

Table 1 .
Descriptive statistics of the predictors and the outcome variables in the model (n = 48).

Table 2 .
Regression model for prediction of the total number of positive changes.

Table 3 .
Regression model for prediction of the total number of negative changes.
b: regression coefficient; se: standard error; Wald: Wald test to assess constraints on statistical parameters; df: degrees of freedom; oR: odds ratio; Ci: confidence interval.