Deeplasia: deep learning for bone age assessment validated on skeletal dysplasias.

BACKGROUND: Skeletal dysplasias collectively affect a large number of patients worldwide. Most of these disorders cause growth anomalies. Hence, evaluating skeletal maturity via the determination of bone age (BA) is a useful tool. Moreover, consecutive BA measurements are crucial for monitoring the growth of patients with such disorders, especially for timing hormonal treatment or orthopedic interventions. However, manual BA assessment is time-consuming and suffers from high intra- and inter-rater variability. This is further exacerbated by genetic disorders causing severe skeletal malformations. While numerous approaches to automate BA assessment have been proposed, few are validated for BA assessment on children with skeletal dysplasias. OBJECTIVE: We present Deeplasia, an open-source prior-free deep-learning approach designed for BA assessment specifically validated on patients with skeletal dysplasias. MATERIALS AND METHODS: We trained multiple convolutional neural network models under various conditions and selected three to build a precise model ensemble. We utilized the public BA dataset from the Radiological Society of North America (RSNA) consisting of training, validation, and test subsets containing 12,611, 1,425, and 200 hand and wrist radiographs, respectively. For testing the performance of our model ensemble on dysplastic hands, we retrospectively collected 568 radiographs from 189 patients with molecularly confirmed diagnoses of seven different genetic bone disorders including achondroplasia and hypochondroplasia. A subset of the dysplastic cohort (149 images) was used to estimate the test-retest precision of our model ensemble on longitudinal data. RESULTS: The mean absolute difference of Deeplasia for the RSNA test set (based on the average of six different reference ratings) and dysplastic set (based on the average of two different reference ratings) were 3.87 and 5.84 months, respectively. The test-retest precision of Deeplasia on longitudinal data (2.74 months) is estimated to be similar to a human expert. CONCLUSION: We demonstrated that Deeplasia is competent in assessing the age and monitoring the development of both normal and dysplastic bones.

Cystic-fibrosis related-diabetes (CFRD) is preceded by and associated with growth failure and deteriorating lung function.

BACKGROUND: Impaired glucose metabolism and cystic fibrosis (CF)-related diabetes (CFRD) are associated with insufficient weight gain and impaired lung function in children and adolescents with CF. We have asked whether imminent CFRD may be a cause of poor growth in children and adolescents. METHODS: A retrospective case control study including 32 patients with CF with or without diabetes was conducted. Sixteen pairs, matched according to age, gender and exocrine pancreatic insufficiency, were analysed. Standard deviation scores (SDS) of height, growth, weight, body mass index (BMI), forced vital capacity (FVC), forced expiratory volume in the first second (FEV1) and forced expiratory flow at 75% of expired FVC (FEF75) were recorded during a mean observation period of 13 years per patient. RESULTS: SDS of height and weight were reduced in CF patients with diabetes compared to those without, not only at the point of diagnosis (both p<0.05) but years before the evidence of diabetes. Afterwards there was a significant decline in height (p<0.001) and weight (p<0.01) SDS in CFRD patients and an increasing difference between the height and weight of CF patients with or without diabetes. In contrast, no significant reduction of BMI-SDS was observed in CFRD patients. All analysed lung function parameters showed a marked decline in CFRD patients starting 1 year prior to the diagnosis of diabetes. CONCLUSIONS: Deteriorating growth, reduced weight and impaired lung function are related to the development of CFRD and are obvious several years before the actual diagnosis of diabetes.

Cancer Risks in Patients Treated With Growth Hormone in Childhood: The SAGhE European Cohort Study.

Context: Growth hormone (GH) is prescribed for an increasing range of indications, but there has been concern that it might raise cancer risk. Published data are limited. Objective: To examine cancer risks in relation to GH treatment. Design: Cohort study. Setting: Population-based. Patients: Cohort of 23,984 patients treated with recombinant human GH (r-hGH) in eight European countries since this treatment was first used in 1984. Cancer expectations from country-specific national population statistics. Main Outcome Measures: Cancer incidence and cancer mortality. Results: Incidence and mortality risks in the cohort were raised for several cancer sites, largely consequent on second primary malignancies in patients given r-hGH after cancer treatment. There was no clear raised risk in patients with growth failure without other major disease. Only for bone and bladder cancers was incidence significantly raised in GH-treated patients without previous cancer. Cancer risk was unrelated to duration or cumulative dose of r-hGH treatment, but for patients treated after previous cancer, cancer mortality risk increased significantly with increasing daily r-hGH dose (P trend < 0.001). Hodgkin lymphoma (HL) incidence increased significantly with longer follow-up (P trend = 0.001 for patients overall and 0.002 for patients without previous cancer). Conclusions: Our results do not generally support a carcinogenic effect of r-hGH, but the unexplained trend in cancer mortality risk in relation to GH dose in patients with previous cancer, and the indication of possible effects on bone cancer, bladder cancer, and HL risks, need further investigation.

Description of the SAGhE Cohort: A Large European Study of Mortality and Cancer Incidence Risks after Childhood Treatment with Recombinant Growth Hormone.

BACKGROUND: The long-term safety of growth hormone treatment is uncertain. Raised risks of death and certain cancers have been reported inconsistently, based on limited data or short-term follow-up by pharmaceutical companies. PATIENTS AND METHODS: The SAGhE (Safety and Appropriateness of Growth Hormone Treatments in Europe) study assembled cohorts of patients treated in childhood with recombinant human growth hormone (r-hGH) in 8 European countries since the first use of this treatment in 1984 and followed them for cause-specific mortality and cancer incidence. Expected rates were obtained from national and local general population data. The cohort consisted of 24,232 patients, most commonly treated for isolated growth failure (53%), Turner syndrome (13%) and growth hormone deficiency linked to neoplasia (12%). This paper describes in detail the study design, methods and data collection and discusses the strengths, biases and weaknesses consequent on this. CONCLUSION: The SAGhE cohort is the largest and longest follow-up cohort study of growth hormone-treated patients with follow-up and analysis independent of industry. It forms a major resource for investigating cancer and mortality risks in r-hGH patients. The interpretation of SAGhE results, however, will need to take account of the methods of cohort assembly and follow-up in each country.

Telephone-based adiposity prevention for families with overweight children (T.A.F.F.-Study): one year outcome of a randomized, controlled trial.

The one-year outcome of the randomized controlled T.A.F.F. (Telephone based Adiposity prevention For Families) study is presented. Screening of overweight (BMI-SDS > 90th centile) children 3.5-17.4 years was performed via the German CrescNet database, and candidates were randomized to an intervention group (IG) and control group (CG). The intervention consisted of computer-aided telephone counselling for one year, supported by mailed newsletters. The primary endpoint was change in BMI-SDS; secondary endpoints were eating behavior, physical activity, media consumption, quality of life. Data from 289 families (145 IG (51% females); 144 CG (50% females)) were analyzed (Full Analysis Set: FAS; Per Protocol Set: PPS). Successful intervention was defined as decrease in BMI-SDS ≥ 0.2. In the FAS, 21% of the IG was successful as compared to 16% from the CG (95% CI for this difference: (-4, 14), p = 0.3, mean change in BMI-SDS: -0.02 for IG vs. 0.02 for CG; p = 0.4). According to the PPS, however, the success rate was 35% in the IG compared to 19% in the CG (mean change in BMI-SDS: -0.09 for IG vs. 0.02 for CG; p = 0.03). Scores for eating patterns (p = 0.01), media consumption (p = 0.007), physical activity (p = 9 × 10-9), quality of life (p = 5 × 10-8) decreased with age, independent of group or change in BMI-SDS. We conclude that a telephone-based obesity prevention program suffers from well-known high attrition rates so that its effectiveness could only be shown in those who adhered to completion. The connection between lifestyle and weight status is not simple and requires further research to better understand.

Early detection of growth disorders with the CrescNet system at the Leipzig treatment center.

BACKGROUND: Growth hormone replacement therapy is more effective the earlier it is begun. This article adresses the question whether children with growth hormone deficiency (GHD) were able to begin treatment earlier through the use of the CrescNet system in the Department of Pediatrics at the Leipzig University Hospital. CrescNet is a network of pediatricians and endocrinological treatment centers, established in Leipzig in 1998, whose aim is to improve the early detection of growth disorders. METHODS: Pediatricians participating in CrescNet provided anonymized data on their patients' height and weight to the CrescNet database. Each participating pediatrician received a quarterly screening report with recommendations for the work-up of children with abnormal growth. Some patients with GHD who were treated in the Leipzig treatment center were referred in response to these recommendations, while others came spontaneously from the practices of pediatricians participating or not participating in CrescNet. We determined the age at the onset of treatment for the 139 patients treated for GHD in the University Children's Hospital Leipzig from 1999 to 2005 and compared the findings with data from other treatment centers in Germany. RESULTS: Patients from CrescNet practices began treatment at a younger age than those from non-CrescNet practices (8.77 ± 3.40 versus 9.50 ± 3.78 years, p = 0.26). Patients from CrescNet practices whose GHD was detected by screening began treatment earlier than those for whom no data had been sent (7.67 ± 3.21 versus 9.28 ± 3.39 years, p = 0.031). In the center in Leipzig, but not in other German GHD treatment centers, the age at onset of treatment dropped significantly over the period of the study and then remained steady till 2009 in the range of 7.61 ± 3.0 years. CONCLUSION: These descriptive results imply that the linking of pediatricians' practices with the CrescNet system resulted in earlier treatment of children with GHD.