Classification of bone flap resorption after cranioplasty: a proposal for a computed tomography-based scoring system

The proposed score and grading schemes were designed collaboratively by the authors, with the ultimate goal of obtaining a robust and objective system that could be used to reliably define BFR, which has been described to be a diverse complication with various subtypes occurring even in the same bone flap [23]. Therefore, universal markers of BFR are required for the scoring system. Variables proposed to constitute the scoring system are “Extent” (estimated remaining bone volume), “Severity” (possible perforations and their measured diameter), and “Focus” (the number of BFR foci within the bone flap). The theoretical basis for the selection of these variables and their cutoff values is addressed in detail in the “Discussion” section of the present manuscript.

Each variable was further divided into four subclasses and scored accordingly, and the resorption score was calculated as the sum of the subclass scores, with the minimum score being 0, and the maximum 9. Increasing values indicate more serious BFR (Fig. 1). A summary of the proposed three-variable scoring system is presented in Table 1.

The new system was tested on our patient cohort. We identified all the patients on whom a decompressive craniectomy and subsequently a primary autologous cranioplasty had been performed at Oulu University Hospital, a tertiary-level teaching hospital, between 2004 and 2014 (n = 45). In order to ensure no data were missed, secondary centers were queried for additional CT scans. Patients were invited for a follow-up CT scan at Oulu University Hospital or a secondary center if their latest available CT scan was more than 1 year old. If the patient’s bone flap had already been removed, the pre-removal CT scan was used for the evaluation. Four patients (8.9%) had to be excluded, three (6.7%) due to death before the required CT scans could be taken, and one (2.2%) due to insufficient CT data, leaving a series of 41 patients (91.1% of total). A study on the current patient cohort has been published earlier [11].

Using the latest CT data for each patient, two of the authors (TKK and JN) independently evaluated and graded each patient’s bone flap status using the new scoring system. For comparison purposes, the raters were allowed access to the earliest post-cranioplasty head CT scan available. If no sufficient post-cranioplasty comparison scan was found, a pre-craniectomy CT scan was used for comparison. Prior to the classification process, the evaluators held a meeting to identify and resolve any discrepancies and ambiguities in the classification system. TKK re-scored the CT data in random order 1 month after the initial evaluation. The mean resorption scores were calculated based on these three evaluation rounds, and the mean scores were used for further statistical analyses.

Any classification system should confer value for clinical decision-making. In order to assess the viability of the scoring system, two neurosurgeons, each with at least 10 years of experience (ST and NS), independently assessed the same head CT data set to ascertain whether any signs of BFR were present, and whether the BFR was relevant, e.g., whether neurosurgical consultation or a re-cranioplasty evaluation should be recommended. A consultation was considered necessary if at least either of the evaluators recommended it.

The neurosurgeons’ evaluations on the relevance of the BFR findings were compared with the mean resorption scores of the patients in order to obtain preliminary threshold values for the scoring system. To produce a more user-friendly scoring system, the patients’ score results were divided into grades 0 (no BFR), I (non-relevant BFR), II (relevant BFR), and III (bone flap failure due to BFR).

The CT data used for the analyses were evaluated using the hospital’s clinical radiology software (neaView Radiology, Neagen Ltd., Helsinki, Finland) with bone window settings (width 2800 and level 600 in Hounsfield Units). An approximate estimate of the bone flap size was obtained from the 2D scout image using the same software, which allowed for manual outlining of the bone flap and calculation of its area. All the CT data were stored in a picture archiving and communication system.

The patients’ baseline characteristics were collected from the hospital records and consisted of sex, age at cranioplasty, dates of the craniectomy and cranioplasty procedures, complications (bone flap resorption, surgical site infections (SSI), hematoma/seroma, cerebrospinal fluid leak, poor cosmesis, hydrocephalus, implant migration, or exposure), and outcome. Additionally, data on smoking and primary diagnosis was collected.

Summary baseline measurements are reported as mean with standard deviation (SD) and range or median with interquartile range (IQR). The neurosurgeons’ agreement on the presence and relevance of BFR was evaluated using Cohen’s kappa (κ), which is a statistical measure of observer agreement strength that also accounts for agreement by chance. The κ values are reported with 95% confidence intervals (CI). The values of κ were interpreted according to Landis and Koch [13], who classify values of < 0.0 as poor, 0.0–0.20 slight, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 substantial, and 0.81–1.00 almost perfect. Categorical variables were evaluated using the χ² and Fisher’s exact tests with grade II BFR as the dependent variable. A two-tailed p value of < 0.05 was considered statistically significant. The analyses were performed using SPSS for Windows (IBM Corp. Released 2016. IBM SPSS Statistics for Windows, Version 24.0. Armonk, NY: IBM Corp).

The study was performed in accordance with the declaration of Helsinki on ethical principles for medical research. The protocol was approved by the ethics review committee of the Northern Ostrobothnia hospital district, and patient consent was acquired for the follow-up CT scans. Patients whose follow-up CT scans that were used for the analyses showed alarming findings were contacted, examined clinically and counseled on the findings by the senior author.

Forty-one patients included in this study had undergone primary autologous cranioplasty after decompressive craniectomy. The baseline characteristics are reported in Table 2, and the postoperative complications in Table 3. The mean age at cranioplasty was 41 years (range 15–65 years, SD 14.8). The bone flaps underwent a mean freezer time of 207 days (range 1 to 538 days, SD 124). The CT images from which the BFR was evaluated were taken at a mean 4.3 years (SD 3.14, range 0.13–11.55 years) after the cranioplasty operation. The median 2D lateral bone flap area was 91.7 cm² (IQR 34.8 cm²).

When evaluated using the proposed scoring system, nine (22.0%) of our patients had no signs of BFR, and thus their mean score was 0. A mean resorption score higher than zero, implying the presence of some degree of BFR, was found in 32 (78.0%) of the cases (Table 4).

The neurosurgeons’ assessments pointed to some degree of BFR in the bone flaps in 31 (75.6%) and 34 (82.9%) cases. BFR was considered relevant by both neurosurgeons in 6 cases (14.6%) and by only one surgeon in 5 (12.2%). Thus it may be said that in total, relevant BFR was found by the neurosurgeons in 11 out of the 41 cases (26.8%).

The surgeons’ agreement on the presence of any level of BFR and the relevance of BFR is reported in Tables 5 and 6. The agreement demonstrated substantial strength concerning both the presence and relevance of BFR (respective κ values of 0.63 (95% CI 0.34 to 0.92) and 0.64 (95% CI 0.36 to 0.91)).

All of the 11 patients whose BFR status was considered relevant by the neurosurgeons had a mean resorption score ≥ 5 (Table 4). Additionally, 4 patients with a mean resorption score ≥ 5 were evaluated by the neurosurgeons as having non-relevant BFR. In order for the system to identify all patients with relevant BFR, we propose a resorption score of ≥ 5 as the threshold value for relevant BFR (grades II and III). Thus, 15 patients (36.6%) had relevant BFR, which indicated radiological BFR severe enough to likely demand clinical interventions as evaluated by the neurosurgeons. Both of the patients for whom the neurosurgeons recommended to consider re-cranioplasty evaluation due to bone flap failure had reached a mean resorption score of 9, which is proposed as the threshold value of grade III BFR (Table 4).

The evaluation of risk factors for relevant BFR (resorption score ≥ 5) is depicted in Table 2. No statistically significant associations between relevant BFR and sex, age, primary diagnosis, smoking habits, 2D defect size, or the duration of freezer time between craniectomy and cranioplasty were observed.

The variables proposed to constitute the Oulu resorption score system (Table 1) are based on the three previously published attempts to classify radiological BFR findings [4, 21, 23] and a recent volumetric study [11].

All of the previous CT-based attempts to classify BFR consider thinning of the bone as a marker of BFR [4, 21, 24]. In a previous study [11], quantitative bone flap volumes of less than 32.2% of the original volume manifested as aseptic BFR severe enough to indicate removal of the bone flap. Based on this, the lower cutoff value for the estimated remaining bone volume, the “Extent” variable, was chosen to be 25.0%. In the same study, 73.2% of the patients had a bone flap volume of ≥ 75.0%, which is suggested to be the upper cutoff value for the estimated bone volume variable.

In addition to the loss of bone volume, perforations in the bone flap deteriorate both the functional [8] and cosmetic outcomes of cranioplasty and are accounted for in the score using the “Severity” variable. A perforation of the size of a burr hole may already produce cosmetic issues [3]. Thus, a diameter of 1.0 cm, which has been used as a marker of BFR in a previous study [9], is proposed as the cutoff value for the “Severity” variable. As bone flaps may contain burr holes or other iatrogenic defects, we suggest that only new perforations or enlargement of the existing ones are taken as markers of BFR.

The total portion of the bone flap affected by the resorption process is also reflected in two of the previous classification systems [4, 23], but a more objective description of the stage of BFR would probably better correlate with the total integrity of the bone flap. Therefore, we propose that the number of BFR foci within the bone flap is accounted for in the scoring system by way of the “Focus” variable, as one focal resorption change is clinically less significant than a diffusely resorbed bone flap. As a moderate degree of BFR is associated with the revitalization process of the bone flap after cranioplasty [16], a remaining bone volume of over 75.0% and the presence of at most one focal BFR change were both interpreted physiological and therefore scored 0 points.

The cutoff value of ≥ 5 for relevant BFR was obtained by comparing the patients’ mean resorption scores with the independent evaluations on the relevance of BFR conducted by two neurosurgeons. To ensure the viability of this threshold, the agreement of the neurosurgeons was tested with the κ statistic. The agreement demonstrated substantial strength with a κ value of 0.64. Additionally, the neurosurgeons recommended re-cranioplasty evaluation based on the CT images for two patients with a resorption score of 9. This was defined as the cutoff value for grade III BFR, which represents bone flap failure.

When the CT images of our autologous cranioplasty patients were evaluated independently using this scoring system, some degree of BFR (grades I–III) was found in 32 out of 41 cases (78.0%). Correspondingly, the two neurosurgeons independently noted any level BFR in 79.3% of cases on average. Radiological BFR with an Oulu resorption score of ≥ 5 was defined relevant, and it was found in 36.6% of our patients, which suggests that the Oulu resorption score successfully ruled out mild cases of BFR from the relevant BFR group. Though more robustly defined, the prevalence of BFR using the Oulu resorption score is in line with previous works assessing the radiological manifestations of BFR [4, 5, 8, 9, 20, 21, 23].

Based on our findings, we recommend that patients with grade II or III BFR should be referred to a neurosurgeon for consultation. Further, grade III BFR indicates failure of the bone flap, and replacement of the autologous cranioplasty with a synthetic implant should be considered by the neurosurgeon in at least these cases.

The proportion of patients with grade II or III BFR (Oulu resorption score ≥ 5) seemed to decrease with increasing patient age (Table 2), but this result did not reach statistical significance due to the small size of the age groups. Smoking was not associated with an increased prevalence of grade II or III BFR, but an earlier report demonstrated smoking to have a detrimental effect for autologous cranioplasty outcome mainly through increased SSI rates [12]. Additionally, no associations between relevant BFR and sex, primary diagnosis, freezer time (< 90, 90–180, > 180 days), or 2D craniectomy area (over or under the median 91.7 cm²) were found. Of these, young age is a commonly accepted predictor of BFR, and the results reported earlier on the other factors are thus far inconclusive [5, 12, 14, 17, 19].

The reliability and clinical applicability of the present scoring system would be an interesting topic for future research, since adoption of the proposed resorption score should offer possibilities for limiting the number of CT follow-ups and avoiding unjustified neurosurgical consultations and referrals arising from unclarity in the interpretation of BFR findings, which is an important consideration with the increasing patient volumes. Additionally, a standardized BFR classification system facilitates accurate identification of risk factors of BFR and enhances inter-study comparison thus ultimately improving future research quality.

As the essence of the process of developing radiological grading systems and treatment protocols lies in the fact that it is an iterative process, future comments and suggestions for modifications to the presently proposed scoring system will be important and are to be welcomed. Further, the scoring system requires validation in terms of reliability using independent patient cohorts. The validation of the present scoring system is the subject of a subsequent study, and we are looking to welcome additional centers for a multicenter validation study of the scoring system.

The strength of this work is that it accurately represents our autologous cranioplasty patient cohort from 2004 to 2014, as it includes 91.1% of the patients operated on during that period in the Oulu University Hospital. The proposed system for BFR scoring requires a minimum of subjective decisions, and it is based on the relevant previously published classification and bone flap volumetry studies.

The clinical data was collected retrospectively in the present study, and thus the inherent weaknesses of retrospective study design apply. The present study population, although accurately represented, is limited in size and thus the effect of chance may be prominent in both the prevalence of BFR and observer agreement. Additionally, the length of follow-up varied between patients and consequently the time of recording the Oulu resorption score was not constant, which may have influenced the results. Nevertheless, differences were recognized by the scoring system. The youngest patient who underwent autologous cranioplasty after decompressive craniectomy was 15 years old. Thus, we could not extend our results to pediatric patients younger than 15 years of age. The cutoff age for emergency decompressive craniectomy in our institution is 65 years, and the study cohort did not contain patients over that age. Further, patient-dependent variables, especially age may have influenced the size of the bone defect and other variables measured in the present study. Despite being a good tool for evaluating CT images, the present system only accounts for BFR, and this score alone does not suffice for determining whether a re-operation is necessary, but a clinical evaluation is also required.