safety and efficacy of computed tomography

21
Ying-Lie O Safety and efficacy of computed tomography a broad perspective EC-EURATOM 6 Framework Programme call 2003 Project no. FP6/002388 www.msct.info

Upload: others

Post on 12-Feb-2022

1 views

Category:

Documents


0 download

TRANSCRIPT

Ying-Lie O

Safety and efficacy of computed tomography

a broad perspective

EC-EURATOM 6 Framework Programme call 2003Project no. FP6/002388

www.msct.info

Ying-Lie O

Safety and efficacy of computed tomography WP1 Justification

UOXFStuart MeesonUHBSTilo Niemann

UOXFStephen GoldingUHBSGeorg Bongartz

UOXFChristopher AlveyAUHKaren Berenth Madsen

UoCNicholas TheocharopoulosAUHJolanta Hansen

UoCKonstantinos ChlapoutakisAUHAnne Grethe Jurik

UoCJohn DamilakisLUMCJob Kievit

UCMLaura Ruiz LopezLUMCJaap Sont

UCMIsabel Salmerón BérlizLUMCAlexander Meijer

UCMEduardo Fraile MorenoLUMCYing-Lie O

UCMAlfonso Calzado CanteraLUMCKoos Geleijns

UOXFUniversity of Oxford, Oxford, United Kingdom

UoCUniversity of Crete, Iraklion, Crete, Greece

UHBSUniversity Hospital Basel, Basel, Switzerland

UCMComplutense University, Madrid, Spain

AUHAarhus University Hospital, Aarhus, Denmark

LUMCLeiden University Medical Center, Leiden, the Netherlands

Ying-Lie O

Safety and efficacy of computed tomography WP1 Justification

Justification

• multi slice computed tomography MSCT

– new generation CT systems: accurate and fast

– at least 16 slices simultaneously

• safety of MSCT

– radiation effects: cancer and leukaemia

– contrast agents: severe allergic reaction

• efficacy of MSCT

– diagnostic accuracy: sensitivity and specificity

– correct diagnosis: appropriate treatment

– more diagnoses found in anatomic region

Ying-Lie O

Safety and efficacy of computed tomography WP1 Justification

Medical decision making: modelling to determine

the benefits of MSCT against the adverse effects

! recommendations and guidelines

• Current guidelines are based on previous

generation of CT

• Rapid technological advances in imaging

• The use of CT is growing

• Model-based guideline development

Ying-Lie O

Safety and efficacy of computed tomography WP1 Justification

• Need for diagnostic imaging

– Precise anatomy: spatial shape and location

– Pathology: intensity of tissue

• Benefits and risks: health-related utilities

• Groups at risk of radiation

– Children

– Pregnant women

– Genetic defects: BRCA

Ying-Lie O

Safety and efficacy of computed tomography WP1 Justification

• Patient groups who may benefit– good survival if treated properly

– age of interest: 13 - 80

– high risk if diagnosis is missed immediate

– temporary disfunction intermediate

– permanent disfunction long term

• Contrast agents– severe allergic reaction immediate

• Risks of diagnostic radiation– solid cancer long term

– leukaemia intermediate

• Radiosensitivity of Individuals and Susceptibilityto Cancer induced by ionizing RADiationRISC-RAD

Ying-Lie O

Safety and efficacy of computed tomography WP1 Justification

Medical imaging visualises anatomic regions,

organs, and pathology

• Anatomic regions and organs

– Abdomen: abdominal pain, haematuria, urolithiasis

– Thorax: chest pain

– Head: headache, minor head trauma

– Musculoskeletal: shoulder pain, low back pain

• Pathology

– inflammation and infection

– stenosis, calculus

– bleeding

– obstruction

– degeneration

– lesions, benign and malignant neoplasms

Ying-Lie O

Safety and efficacy of computed tomography WP1 Justification

Model-based guideline development

for symptom-based indications

Ying-Lie O

Department of Radiology

Department of Medical Decision Making

Leiden University Medical Center

Ying-Lie O

Model-based guideline development

Symptom-based: many possible diagnoses

Common modelling approach

• Medical diagnoses:

– probabilistic networks

– many diagnoses

– many tests

• Treatment choice and outcomes:

decision trees, influence diagram

– one disease

– many treatments

– outcome for each treatment

Ying-Lie O

Model-based guideline development

Model for guideline development

• Includes parts of the medical process

• Visible associations between clinical variables

• Valuation of outcomes to allow comparison

• Adaptation to local situation

And

• Graphical representation

• Understandable for medical and other experts

• Not for direct clinical use

• Suitable for training

Strategy: from comprehensive to restricted

Ying-Lie O

Model-based guideline development

The model should support the following

• Includes an imaging policy (and advanced tests)

• Includes a treatment policy containing (surgical)

interventions

• Outcomes for presence and absence of disease

• Ranking of probabilities of diseases according to

criteria

• Temporal order of groups (not within a group)

• Policy evaluation by ranking outcomes, not the

best policy

• Input variables may not be recalculated (control

value)

• Logical conditions that depend on variables

Ying-Lie O

Model-based guideline development

Influence diagram as model for guideline extraction

• Entry point: symptoms, patient groups, diseases

– symptoms in inclusion criteria

– (local) disease a priori

• Diagnostic part: general tests, imaging policy

– level of detail related to urgency

– causal relation between diseases and tests

• Treatment part: treatment policy

– focuses on outcomes

– includes negative treatment

• Outcomes part: disease and treatment outcomes,

health-related utilities

Ying-Lie O

Model-based guideline development

Probabilistic networks do not support ordering, but

it can be added on control and UI level

Order of probabilities of diseases and outcomes

• Ranking for each patient group and overall

• Criteria for ranking

• Ranking after each recalculation

Temporal order of groups of activities

• Order of activities can be imposed through the UI

• Marked activities may be skipped

• Logical conditions with respect to test results

Ying-Lie O

Model-based guideline development

Guideline extraction from influence diagram

• Diagnostic part: general tests, imaging policy

– Rank the probabilities after each test

– and for each imaging policy

• Treatment part: treatment policy

– Order the outcomes of most probable diseases

– including negative treatments

– for all treatment policies

Expected problems:

• Large numbers of combinations

• Decision paths not visible

Ying-Lie O

Model-based guideline development

Example: acute abdominal pain

• Common cause of presentation in emergency

• Many organs in abdomen, difficult to diagnose

• A third have appendicitis

of which a third of is equivocal without imaging

• Misdiagnosis lead to severe complications that

require hospitalisation and may lead to death

• Age group of adolescents and young adults

Categories of possible diseases

• Medical tract: digestive, urinary, …

• Pathology: inflammation and infections, …

• Urgency with respect to bad outcomes

Ying-Lie O

Ying-Lie O

Model-based guideline development

Example: acute abdominal pain

Order of probabilities for different test results

Order of outcomes of imaging policies and

treatment policies

Disease Appendicitis

Imaging policies only US gold standard decide MSCT gold standard always MSCT gold standard laparoscopy gold standard

present 0,36 0,98 0,90 1,00 0,91 1,00 0,93 1,00

US uncertain uncertain uncertain uncertain

MSCT maybepos maybepos maybepos maybepos

diagnostic laparoscopy positive positive

Treatment policies decide surgery decide surgery decide surgery laparoscopy

surgery no surgery surgery no surgery surgery no surgery laparoscopy missed

Outcomes T+| D+ T-| D+ T+| D+ T-| D+ T+| D+ T-| D+ T+| D+ T-| D+

death 0,00 0,62 0,01 0,10 0,01 0,09 0,01 0,07

disfunction 0,00 0 0,01 0 0,01 0 0,01 0

complication disfunction 0,02 0 0,04 0 0,05 0 0,05 0

complication 0,02 0 0,04 0 0,05 0 0,05 0

recovery 0,32 0 0,79 0 0,81 0 0,82 0

Ying-Lie O

Model-based guideline development

Example: acute abdominal pain

• Order of probabilities for different test results

" suspected appendicitis

• Order of outcomes of imaging policies and

treatment policies

1. Laparoscopy: combination of diagnosis and

intervention

2. Always MSCT followed by surgery

3. Decide for MSCT followed by surgery

4. Only ultrasound followed by surgery

• Not modelled:

– Consequences of time delay

– Specific adverse effects of laparoscopy

Ying-Lie O

Model-based guideline development

Guidelines for acute abdominal pain in case of

suspected appendicitis

• Conventional guidelines include ultrasound (low

sensitivity)

• “Always MSCT” gives a good outcome

• MSCT also detects other diseases

• Laparoscopy gives the best outcome, but may

have severe adverse effects (not modelled)

Ying-Lie O

Model-based guideline development

Modelling “prescription” for guideline development

• Needs support of the modelling tool

• Needs a good user interface for development

• Needs a tailored user interface for visualisation

and training

Challenges

• Specific modelling constructs

– Noisy-OR for the same underlying pathology

– Sub-models for different pathologies?

– Time delay in conditional probabilities

• Evidence-based data and expert knowledge

• Multi-attribute health-related utilities

Ying-Lie O

Model-based guideline development

? !