Osteogenesis imperfecta (OI) is a genetically heterogeneous group of inherited disorders with brittle bones and typical extraskeletal characteristics.1 In the majority of patients, a dominant mutation in one of the two type I collagen genes leads either to reduced production or abnormal quality of type I collagen.2 The genotype-phenotype correlation in OI is limited.3, 4 Osteopenia and frequent peripheral and spinal fractures result in skeletal deformities during growth.5 Deformation of the craniocervical junction is a known complication of OI and can lead to life-threatening compression of the medulla and cervical spine.6, 7 To our knowledge, only two previous studies on the prevalence of craniocervical junction anomalies consist exclusively of pediatric patients.5, 8 The natural course of craniocervical junction anomalies in growing patients is for the most part unexplored despite the importance of early intervention before serious symptoms arise.
Pathology of the craniocervical junction can be divided into 1) basilar invagination (protrusion of uppermost vertebral structures into foramen magnum); 2) basilar impression (relative lowering of the cranial base with consequent positioning of the uppermost vertebral structures above the caudal border of the skull); and 3) platybasia (flattening of the cranial base).9–11 These complications can occur separately or simultaneously (Fig. 1).10 Craniocervical junction pathology can be diagnosed from lateral skull radiographs and with higher precision from lateral midsagittal magnetic resonance (MR) and computed tomography (CT) images.8, 10 In OI, lateral skull radiographs have been recommended as a screening method at individually adjusted intervals.8 A baseline cephalometric study has been suggested for all OI patients before school age.12
The original Sillence classification of OI into types I through IV is based on clinical and radiological findings. Type II is considered perinatally lethal. Type I is associated with increased risk of fractures and a normal or nearly normal height. Type III is delineated by short stature and progressive severe deformity, and type IV by moderate deformity.13 New genetic forms, with often autosomal recessive mode of inheritance, have recently been recognized.14–16 This increased understanding of the genetic background has led to extension of the classification into type XI.2 OI patients, particularly those with severe forms, are currently treated with bisphosphonates that improve bone mineral density (BMD) and may reduce the number of fractures.17, 18 It remains unknown whether bisphosphonates affect the pathological processes in the craniocervical junction.
Our purpose for this retrospective, longitudinal, and cross-sectional one-center study was to analyze the prevalence and natural course of craniocervical junction anomalies from lateral skull images in growing OI patients. We used a material, the collection of which had already been started in 1967, that represents a true sample of longitudinal data from times before bisphosphonates were widely used in the treatment of OI. Our specific aims were to establish guidelines for screening and follow-up of young OI patients, as well as to provide reference data for future work, assessing the effects of medical therapies. We also addressed the question of whether cranial base anomalies are associated with a particular OI type or other certain clinical characteristics that indicate disorder severity.
Materials and Methods
The study cohort was obtained from the records of the Institute of Dentistry, University of Helsinki, and Children's Hospital, Helsinki University Central Hospital. The diagnosis of OI was based on clinical and radiographic findings and a family history consistent with autosomal dominant inheritance. In several patients, the diagnosis had been confirmed also by COL1A1 and COL1A2 mutation analysis in the patient or in other affected family members.19 We included all OI patients of whom an initial lateral skull radiograph or MR image had been obtained before the age of 25 years, the age up to which vertebral bodies may normally grow.20 Altogether, 76 patients including 39 females and 37 males from 69 families met these inclusion criteria. Classified according to Sillence,13 47 of the patients had type I, 13 type III, and 16 type IV OI (Table 1). The study cohort was ethnically homogenous; all but 2 patients were of Finnish ancestry (Caucasian Europeans). The Ethics Board of the Institute of Dentistry, University of Helsinki, and the joint Ethical Committee of the Helsinki University Central Hospital had approved the study.
Table 1. Clinical Characteristics of the OI Patients
N = number of patients with information available on the parameter; n = number of patients with positive finding.
Average height (Z-score)
Average head circumference (Z-score)
Scoliosis, n/N (%)
Bisphosphonate treatment, n/N (%)
The study was conducted on single and serial lateral skull images, and the ages at imaging for each patient are displayed in Fig. 2. Longitudinal data were available for 31 patients (41%). In total, 143 radiographs and 7 MR images were analyzed.
Clinical data were collected from hospital records. Height measurements (as Z-score) were considered if they were obtained within 6 months before or after imaging, with the exception of 8 patients, for whom the height measurement was obtained at a later time point. Of those patients, 5 had type I and 3 had type IV OI. The height Z-score for these type I patients was obtained from the most recent available measure and was expected to remain similar through growth. The type IV patients were aged 14, 22, and 30 years at the time of imaging, and no significant change in height Z-score was expected to have occurred in the maximum 2-year interval between the height measurement and skull imaging. Height and head circumference Z-scores were obtained from Finnish standard growth curves.21
Scoliosis was clinically assessed as being present or absent. History of medical treatment at the time of imaging was recorded. Treatment history with bisphosphonates was negative in 62 and positive in 14 patients. The latter were excluded from the longitudinal analysis to reveal the natural course of craniocervical anomalies.
Based on our earlier observations indicating that cranial base measurements are age-dependent in young children,22 we used age-matched control groups. Reference values comprised those of 53 healthy individuals (aged 3 to 26 years) from our previous study22 and of 15 healthy children aged 0 to 3 years, examined at the Children's Hospital, Helsinki University Central Hospital, for suspected accidental skull fractures possibly affecting calvarial or facial but not skull base structures. We divided the 68 control subjects into 2-year age groups until 24 years of age, after which the subjects were grouped as 25 years and older. Total number of control images was 324.
Cephalometric analysis was performed identically on lateral skull radiographs and MR images, using points situated at the midsagittal plane. To improve the reliability of observations, two experienced examiners located the cephalometric landmark points on all images together. The examiners were blinded to the patient's medical history and OI type when analyzing the images. The examiners had a high degree of concordance, and inter-reader differences in image analyses did not influence the results, as previously reported.23
The measurements are shown in Fig. 3. To evaluate whether the odontoid process is pathologically protruded into foramen magnum as a sign of basilar invagination, we measured the perpendicular distance of the tip of the odontoid process to McRae's reference line (foramen magnum line).9 To detect whether the anterior cranial base angle is abnormally flat as a sign of platybasia, we measured the angulation between cranial base anterior to the pituitary fossa and the clivus.10 To evaluate whether the cervical spine was situated abnormally superiorly in relation to the caudal borders of the skull as a sign of basilar impression, we measured the perpendicular distance of the tip of dens to a line drawn through the most caudal point of occipital curve parallel to sella-nasion line (D-M distance).10 We also constructed a novel angular measure termed D-M angle, insensitive to radiographic magnification, to assess basilar impression. D-M angle was defined as the angle between a line running through the lowest point of occipital curve parallel to the sella-nasion line (as for D-M distance) and a line drawn from the tip of the dens to the lowest point of occipital curve (Fig. 3).
The radiographic magnification was corrected to measure the linear distances. When magnification was unknown, we used only angular measures (cranial base angle and D-M angle) and assessed the relative location of the odontoid process as being either below or above McRae's line. Measurement values were regarded as positive when the tip of dens was situated above the reference line. Results of males and females were combined to ensure adequate group size for comparison of different OI types. No difference has been detected between sexes in these measures in previous studies.10, 22
To obtain reference values for the novel D-M angle, we measured this angle on 290 lateral skull radiographs of longitudinal series in a subgroup of 65 subjects (aged 0 to 26 years) from the control cohort. Analysis using Pearson's correlation coefficient indicated a statistically significant positive relationship between D-M angle and D-M distance (r = 0.980, 95% confidence interval [CI] 0.967 to 0.990, p < 0.01, n = 290).
Threshold values of anomaly
Basilar invagination was diagnosed when McRae's measure was above zero, as originally presented by McRae and later confirmed by Kovero and colleagues and Cheung and colleagues.9–11 In previous studies, the D-M distance has been singled out as a reliable and thus recommended measure in evaluating basilar impression also in growing individuals.10, 11, 22 Radiographic criterion for basilar impression was defined as D-M distance or D-M angle +2.5 standard deviations (SD) from the average of healthy age-matched controls.22 Cranial base angle exceeding +2.5 SD from the average of healthy age-matched controls was considered to imply platybasia. Table 2 presents the age-appropriate threshold values for cranial base anomalies. Sampling variation in our control age groups was somewhat large because of the limited sample size. Same threshold values were applied for subjects older than 8 years, whose odontoid process has been shown to lie at a mean level comparable to that of adults.22
After age 9 years, children and adults have the same threshold values.
D-M distance (mm)
D-M angle (degrees)
Anterior cranial base angle (degrees)
The cranial base measures were converted to age-specific Z-scores based on the reference data. The relationship between patient characteristics and cranial anomaly was evaluated by means of logistic regression analysis. In this analysis, we used only one image of each individual (the first image in case of serial data) to avoid bias caused by more frequent imaging among those with severe OI types. Explanatory variables were age, sex, and height Z-score. Craniocervical junction pathology was a dichotomous (yes/no) variable. Data on scoliosis and head circumference were not available for all patients and were left out of the analysis. Statistical calculations were done using the SPSS software for Windows (version 13; SPSS Inc., Chicago, IL, USA) and R language (version 2.13.0; R Development Core Team, R: A Language and Environment for Statistical Computing, Vienna, Austria, 2008).
We analyzed longitudinal data of cranial anomalies using generalized linear mixed model (Douglas Bates, Martin Maechler, and Ben Bolker, lme4: Linear mixed-effects models using S4 classes, 2011, http://CRAN.R-project.org/package=lme4.). Age and sex were used as fixed effects and patient-specific intercept as random effect.
Craniocervical junction anomalies were observed in all OI types. Of the 76 patients, 28 (37%) had an abnormal finding in at least one of the four measures in the initial image. Table 3 displays the number of patients with abnormal findings. Of the 14 patients with only one cranial anomaly, 12 had platybasia, 1 isolated basilar invagination, and 1 isolated basilar impression. Six patients, of whom 1 had type I, 1 type IV, and 4 type III OI, displayed all three pathological conditions of the cranial base. Altogether, 11 of the 25 patients with anomaly (44%) presented with more than one type.
Table 3. Number and Percentage of Patients With an Abnormal Finding, Defined as a Positive McRae's Measure or a Value Above +2.5 SD From the Average of Healthy Controls for Each Measure and OI Type at Any Age
Type I n/N (%)
Type III n/N (%)
Type IV n/N (%)
N = total number of patients with measure value obtainable; n = number of abnormal findings.
Anterior cranial base angle
Most of the Z-score values for cranial base measures were positive (above zero), indicating abnormally high relative position of the uppermost vertebral structures or abnormally flat cranial base (Fig. 4). Patients with type III OI had a positive score in all measures. Comparison of the Z-scores in different OI types revealed that position of odontoid process tip in relation to foramen magnum was highest in type III before the age of 20 years and higher in type IV than in type I before the age of 18 years. D-M angle and D-M distance measure Z-scores were highest in type III between ages 6 and 21 years and second highest in type IV between ages 6 and 16 years, after which the scores were lower than in type I. Anterior cranial base angle Z-scores were highest in type III, with type IV following, and type I exhibiting the lowest scores in all age groups.
Logistic regression analysis showed that a higher risk of having any of the pathological conditions correlated with a lower height Z-score (odds ratio [OR] = 0.64; 95% CI 0.45 to 0.85; p < 0.004, for one SD unit) but the risk was not associated with age or sex (Table 4).
Table 4. Logistic Regression Analysis of Factors Associated With the Presence of Cranial Anomalya
OR (95% CI)
Sex effect of basilar invagination was inestimable because of the small number of abnormal findings.
Basilar invagination (n = 10/76)
Basilar impression (n = 11/75)
Platybasia (n = 21/76)
Any anomaly (n = 28/76)
Longitudinal series were available for 31 patients (Fig. 2) of whom 25 had not been treated with bisphosphonates and were included in the final analysis. The beginning of the series ranged from birth to 25 years. Onset of basilar invagination and basilar impression was found at all ages from 2 years onward. Platybasia was present from birth. Shedding some light on the natural course of basilar invagination, a normal McRae's measure in the beginning of follow-up changed into a pathological one in 2 of the 5 type III OI patients and in all 3 of the type IV OI patients later fulfilling the criterion for basilar invagination. Similarly, basilar impression developed during the follow-up in 2 of the 4 type III OI patients and 2 of the 3 type IV patients. Platybasia developed during follow-up in 1 of the 8 type III patients and in the only type IV patient diagnosed with it.
Once detected, basilar invagination persisted throughout the follow-up in all type IV patients, whereas in 2 type III patients McRae's measure interestingly returned to normal at older age. Basilar impression persisted in the follow-up, with the exception of 1 type I patient, in whom the D-M angle measure returned to normal. Platybasia persisted in all the patients diagnosed with it. Statistical analysis did not, however, confirm any association between age and cranial anomalies (p = 0.537).
In some patients, the cranial base anomalies associated with symptoms. One 18-year-old female with type III OI and both basilar invagination and impression suffered from numbness in the upper extremities. One 20-year-old male with type III OI and all three types of cranial base anomalies reported basilar headache induced by coughing; he died shortly after of pneumonia. Another male, with type III OI and all three types of anomalies, underwent decompression surgery and posterior stabilization at the age of 8 years because of Chiari I malformation with a sole clinical symptom of muscle weakness in lower limbs first detected a year before surgery. He was the only one receiving treatment for the cranial base pathology.
The sample of patients treated with bisphosphonates (n = 14) was too small for detailed statistical evaluation. Visual exploration of the developmental curves of the cranial base measurements revealed no difference between patients treated and not treated with bisphosphonates. Individual variation in measurement results was, however, notable. Interestingly, 3 patients free of cranial base pathology in the beginning of follow-up developed one of the cranial base anomalies after the onset of medication. Of these, two were newly acquired platybasias (1 patient with type I and 1 patient with type IV OI) and the third was basilar impression in a patient with type IV OI.
To our knowledge, this is the first report on the natural course of craniocervical junction anomalies in a widely growing OI population until adult age. Our findings agree with previous studies in that the most severely affected OI patients have a higher prevalence of craniocervical junction pathology.11, 24 Platybasia was the most prevalent anomaly in growing individuals (28%). It is often, however, asymptomatic and of disputable clinical significance.9
Cheung and colleagues reported a prevalence of 22% for craniocervical junction anomaly in pediatric and adult OI patients (187 patients of types I to VII aged 3 to 47 years), some of whom had undergone bisphosphonate treatment.11 In adults with OI, a prevalence of 25% has been documented (medication status unknown).10 Our finding of 37% is somewhat higher, possibly because our threshold limit was +2.5 SD from the average of healthy controls and in the above-mentioned studies the limit was set at +3.0 SD. A lower threshold limit ensures better sensitivity of the screening, whereas a higher limit yields better specificity of the diagnosis. In an adult population, the emphasis may be on limitation of false-positive findings, whereas in children it is beneficial to identify all subjects that need closer follow-up.
Janus and colleagues found basilar invagination in 6.2% of their pediatric patients (130 patients of types I to IV aged 0 to 18 years, medication status unknown).8 Charnas and Marini found basilar invagination in 11% of their pediatric and adult patients (76 patients of types I to IV aged 0 to 65 years, unknown medication status).25 Our results show 13% prevalence for basilar invagination. Sillence reported a 25% prevalence of basilar impression (age of patients unknown) using the McGregor's line as a reference.24 Our finding of 15% prevalence for basilar impression as measured with D-M angle is lower. Notably, however, comparison with most previous studies is only indicative because the reference lines and use of terms is different. Distribution of OI types I to IV is similar in the referred studies.
A novel measure, the D-M angle, was developed to overcome the problem created by unknown radiographic magnification and only seldom taken into consideration in the previous studies.10 The measure was tested and found to have high correlation with the previously successfully used D-M distance.
Cranial base anomalies are believed to result from the weight of the brain-inducing microfractures and/or bending in the underlying skeletal structures.10, 26, 27 Our findings agree with this apprehension in that the onset of basilar invagination and impression occur after an age when a child usually has adopted an upright position, suggesting a causal relationship. Based on a cross-sectional study, Sillence24 recommended delaying the upright posture of a child with a severe form of OI to postpone the onset of what he suspected a progressive pathology. Progression of craniocervical junction pathology in our longitudinal series was not, however, statistically evident. Our findings were similar to those of Janus and colleagues,8 who reported a progressive worsening of basilar impression in only 1 of 7 asymptomatic patients as evaluated with Chamberlain's line. McRae's measure, as an indicator of basilar invagination, interestingly changed from positive to negative in some of our patients. This probably occurs as a result of a growing number of microfractures lowering the posterior fossa and leading to upward infolding of the occipital condyles and posterior edge of foramen magnum,6 thus raising McRae's line.
Presence of platybasia already at birth suggests intrauterine development of the anomaly likely by bending of the anterior cranial base as the brain grows in volume, possibly by adaptive changes in the cartilaginous synchondroses. The cranial base angle decreases with age both in healthy subjects and in OI patients.11, 22, 28 The distance measures used, in contrast, have been shown to vary insignificantly with age in growing OI patients.11 McGregor considered an angle greater than 148 degrees to be highly suggestive of platybasia in young adults.26 Our threshold limit was lower in all age groups. Large SD in our measurement results is the result of the interindividual variation in a limited study cohort.
Head circumference may grow abnormally in OI patients, resulting in macrocephaly.25, 29 Similar to findings of Charnas and Marini,25 we found the average head circumference (Z-score) to be close to that of population values in most patients of all OI types. Scoliosis was present more often in the clinically severe OI types, as in previous studies.5
On detection of cranial base pathology, an MR scan and neurological examination were recommended. Our patients with radiologically defined pathology were mostly free of clinical neurologic symptoms. Two patients had mild headache or upper limb numbness. Only one patient, an 8-year-old boy, was operated on for cranial base pathology; he had mild muscle weakness in lower limbs before diagnosis.
The increasing use of bisphosphonates from an early age has considerably affected the BMD and incidence of fractures,30 whereas its effects on skull and craniocervical junction in growing patients are unknown. Some of our patients developed cranial base pathology during bisphosphonate treatment. However, because the number of treated patients was low in our cohort, the effect of bisphosphonates in the development of craniocervical junction anomalies still needs to be documented.
This study provides longitudinal and cross-sectional data on craniocervical junction in growing OI patients. We introduce a novel measure for diagnosing basilar impression insensitive to radiographic magnification: the D-M angle. We found that severe growth failure suggests presence of craniocervical junction anomaly in OI patients. On the other hand, shortened distance between skull base and shoulders as caused by craniocervical abnormality can result in shortened overall height. Careful follow-up of subjects with low height Z-score is warranted.
Sillence24 recommends radiographic screening every 2 to 3 years from the age of 5 years onward in severe forms of OI. Based on our findings we suggest that radiological analysis of cranial base dimensions with measurements described should be carried out for all patients before school age. In severe forms of OI, a radiograph or MR image is often taken already in infancy to facilitate diagnosis and classification and should also be subjected to analysis of craniocervical junction. In case of normal findings from the image(s) taken before school age, further imaging is unnecessary in symptomless patients to keep the patient radiation dose at a minimum. In case of abnormal findings in a radiograph or MR image, an individually adjusted plan for follow-up and treatment is warranted (Fig. 5).
All authors state that they have no conflicts of interest.
This study was supported by grants from Biomedicum Helsinki Foundation and Orthodontic Section of Finnish Dental Society (to HA); from the Foundation for Pediatric Research, the Sigrid Juselius Foundation, the Academy of Finland, and the Helsinki University Hospital Research Funds, Helsinki, Finland (to OM); and from the National Graduate School of Clinical Investigation, Finland (to MM).
Authors' roles: Study design: HA, OM, JH, and JWS. Study conduct: HA, OM, and JWS. Data collection: OM, HR, ME, MM, IK, and JWS. Data analysis: HA and JWS. Data interpretation: HA, OM, JH, and JWS. Manuscript preparation: HA, OM, and JWS. Approving final version of manuscript: HA, OM, JH, HR, MM, IK, and JWS. The authors take responsibility for the integrity and accuracy of the data analysis.