Fracture Prediction by Opportunistic Screening for Osteoporosis

OPPORTOS : Fracture Prediction by Opportunistic Screening for Osteoporosis

Fractures related to skeleton fragility (i.e. osteoporotic fractures) represent a growing health problem, as the life expectancy and thus the number of frail elderly subjects is increasing. These fractures are associated with individual and societal consequences. The fractures are responsible for increased disability, chronic pain, and loss of independency. The annual cost of either prevalent or incident osteoporotic-related fractures exceeds the same ratio calculation for many other serious chronic diseases. Mortality risk is increased following osteoporotic fractures. Several classes of osteoporosis therapies are proven to reduce fracture risk, based on placebo controlled trials of 3-5 years duration, including in elderly patients. These data are the rationale for screening of patients at risk of fracture, recognizing that the optimal approach is to identify subjects at risk for major fractures . Bone fragility is related to the decrease of both the quality and the quantity of bone. Bone mineral density (BMD) is a surrogate of bone fragility, with the advantage of being non-invasively measurable, at relevant sites, such as vertebrae and upper extremity of the femur. A low BMD, age, and prevalent fractures are the 3 main determinants of the risk of sustaining a fracture. A low BMD has also been reported as a determinant of all cause mortality risk in the general population. So far, screening of low BMD by QCT has not been recommended because of low availability of the devices, irradiation, and cost. However, a huge number of QCT are performed daily for various medical indications. These thoracic and abdominal QCT carry potential information about vertebral BMD. These data are already available, with no additional cost, patient time, nor radiation exposure. They can be retrospectively (in our study) or prospectively (in the future context of care) analyzed, and are the basis of an opportunistic screening for osteoporosis: this denotes the use of diagnostic QCT scans made for other medical indication to screen for patients at high fracture risk. There is no study of this QCT based measurement as an opportunistic screening for patients at short-term risk for fracture. Opportunistic screening of osteoporosis, by diagnosis of low BMD on abdominal QCT performed for various medical indications, is able to detect subjects at short-term (i.e. over 3 years) risk of fracture (necessitating an hospitalization).

Primary objective : Assessment of the performance of an automatic measurement of vertebral bone mineral density (BMD) on routine abdominal scans for the prediction of fractures (necessitating hospitalization) over 3 years.

Secondary objective :

• Assessment of the performance of an automatic measurement of lumbar vertebral bone density on routine abdominal scans (performed for other medical indications) for the prediction of fractures (necessitating hospitalization) over 3 years according to the location of fracture (spine, hip, non- spine-non-hip).
• Assessment of the performance of an automatic measurement of lumbar vertebral bone density on routine abdominal scans (performed for other medical indications) for the prediction of fractures (necessitating hospitalization) over 3 years according to age (>= 75 vs < 75) and sex.
• Assessment of the impact of a preventive treatment of patients at high risk of fracture in terms of number of avoided fractures (simulation study)
• Assessment of classical risk factors of fracture (using SNDS database) and added value of the automatic bone density measurement on fracture risk estimation.
• Prevalence estimation of vertebral fractures (as assessed automatically by the Zebra algorithm) in the population at index date
• Assessment of the impact of IV contrast on the values of the measurements provided by the Zebra algorithm.
• To establish mean values for zebra parameters according to age and gender in subjects without fracture.

The study is based on the patients of the Paris University Hospitals / Assistance Publique - Hôpitaux de Paris (AP-HP), the largest hospital entity in Europe, who had, whatever the medical indication, an abdominal CT; the CT related data (metadata and data issued from the analysis of abdominal CT with the Zebra software) and images are stored in the PACS workstation (available since 2007), and are available through the AP-HP Clinical Data Repository (CDR). After legal authorizations, data from the CDR will be merged with data from the national Programme de Médicalisation des Systèmes d'Information (PMSI) database within the national system of health data SNDS (Système National des Données de Santé), which records all discharge summaries of public and private hospitals in France (i.e. the national PMSI (Programme de Médicalisation des Systèmes d'Information) database) and reimbursed outpatient care (i.e. the DCIR database).

Each summary of the PMSI contains the age and sex of the patient, the motive of admission, described through the diagnosis codes from the International Statistical Classification of Diseases, 10th Revision (ICD-10) and secondary diagnosis, allowing to identify fractures and, among them, neoplastic (on bone metastasis or primary bone cancer) will not be considered as event of interest.

• All legible abdominal scans performed between 2007 and 2014 at AP-HP in subjects of 60 years and older (i.e. those at a priori higher risk of fragility fractures) during a hospital stay and stored on PACS workstation will be used. If a subject performed several abdominal scans, only the first one will be used.
• Vertebral BMD will be calculated automatically with a validated algorithm (in Hounsfield Unit - HU), with the lower attenuation (lower HU) representing less dense bone (i.e. more fragile), at each vertebra from L1 to L4. The algorithm automatically segments the L1-L4 vertebrae, and extracts their overall density based on a number of features extracted. The analysis of abdominal CT with the Zebra software will also identify vertebral fractures.
• Clinical data of APHP patients are available in the CDR.
• Incident fractures will be identified in the national PMSI database within the SNDS (2007-2017). An event will be defined by the occurrence of an incident fracture (excluding polytraumatic fractures and neoplastic fractures) necessitating hospitalization, within 3 years after the abdominal scan.
• Prediction of the occurrence of incident fracture (as defined above) by vertebral BMD will be assessed.

The source population will be patients 60 years and older (i.e. those at a priori higher risk of fragility fractures because of age) who had abdominal QCT with images stored centrally in the Picture Archiving and Communication System (PACS) of our institution (Assistance Publique Hôpitaux de Paris, AP-HP). AP-HP is the largest university hospital entity in Europe (Greater Paris University Hospitals, 39 hospitals, 23 of which are for acute care, 20,700 beds including 11700 beds for acute care, more than 7 million patients treated and 1.2 million hospitalizations each year in acute care). The PACS has been implemented in 2007, and stores all the images acquired in the AP-HP institution and related data. These data are available in the CDR.

After legal authorizations, data from the CDR will be merged with data from the SNDS (Système National des Données de Santé) which includes data from the national Programme de Médicalisation des Systèmes d'Information (PMSI).

Inclusion Criteria:

• Men and women 60 years and older (i.e. those at a priori higher risk of fragility fractures because of age). We do not suggest an upper limit for age, as anti-osteoporotic treatments are effective in elderly individuals on the basis of a high risk of fracture.
• having an abdominal scan* in the PACS workstation, whatever the medical indication (Axial or Sagittal slices / Slice thickness of 4mm or better / Presence of the L1 to L4 vertebrae) made during a hospitalization in an APHP hospital between 2007 and 2014.

We will use the term abdominal scan for any scan involving L1 to L4 vertebrae, i.e. abdominal scans, thoraco-abdominal scans, lombar scans…

Exclusion Criteria:

• Abdominal scan not legible by the software used to measure BMD (this includes patients with spine implants, screws and other spinal devices, scans with a part of L1 to L4 vertabrae missing for any reasons, or any other technical reason of non legibility)
• Impossibility to match the CT scan data with the SNDS (including the French Hospital National Database (Programme de Médicalisation des Systèmes d'Information: PMSI)*, which includes all hospitalizations occurring in public and private acute care settings in France), between 2007 and 2017. Matching will be performed using probabilistic matching, as described below. It will give incident fracture data during follow-up.
• Foreign resident