1
Considerations for Comparative
Observational Studies of Implantable
Medical Devices
Jessica J Jalbert, PhD
Director of Pharmacoepidemiology, LASER Analytica
Assistant Professor (courtesy) of Healthcare Policy and Research, Weill Cornell Medical College
CADTH Symposium
Ottawa, Canada
24 April 2017
2
Device Characteristics
3
IMD Example: Stents
4
Medical Device Lifecycles
> Lifecycle of medical devices << drugs
• Different regulatory landscape
• Iterative improvements of existing technologies
> Incremental changes to device and procedure are common
• Improve performance/overcoming shortcomings
> Devices may be on market for a short period before being
discontinued or replaced
5
Medical Devices “Systems”
> Devices can be marketed as individual, separate
components or as “systems”
6
Identifying IMDs
> Administrative claims data
• Procedure codes (e.g. ICD-9, CPT-4 codes)
• Varying granularity
• Device-specific information generally lacking
> Electronic healthcare records
• Additional device/procedure information potentially
available (in addition to procedure codes)
• Lack of standardization/limited follow-up
> Device registries
• Detailed device/procedure/clinical data
• Smaller population/limited follow-up
7
What are the implications?
> Unobserved heterogeneity
• Multiple versions of device in population
• Multiple combinations of components of a device in a
population
• Differences in safety/effectiveness
> Stratify by device (manufacturer, design, lot, version) or
combinations
• Too few patients/too little data to study all permutations
of a device or combinations of device systems
8
Characteristics of Providers
9
Role of Provider
> Operator skill and learning curve
• Related to procedural complexity
• Affect patient outcomes (short- and long-term)
• Affect choice of treatment
> Intervention setting
• Affect choice of treatment
• Affect patient outcomes
10
0
0.5
1
1.5
2
2.5
3
3.5
Mo
rtal
ity
Ris
k (%
) 30-Day Mortality Risk and Past-Year
Physician CAS Volume
CAS Volume 0 1-4 5-9 10-19 20-24 ≥25
p-value for trend: <0.0001
Ref: Jalbert JJ et al. Circ Cardiovasc Qual Outcomes 2015;8(6 Suppl 3):S81-9.
Adjusted RR Ref 0.80 0.74 0.60 0.52 0.46 (0.62-1.04) (0.56-0.97) (0.46-0.80) (0.34-0.80) (0.33-0.65)
11
0
0.5
1
1.5
2
2.5
3
3.5
Mo
rtal
ity
Ris
k (%
)
CAS Volume <10 10-19 20-39 ≥40
p-value for trend: < 0.0001
Adjusted RR 1.0 0.81 0.74 0.54 (0.63-1.04) (0.58-0.93) (0.42-0.71)
30-Day Mortality Risk and Past-Year
Hospital CAS Volume
Ref: Jalbert JJ et al. Circ Cardiovasc Qual Outcomes 2015;8(6 Suppl 3):S81-9.
12
CAS vs. CEA with Different Levels of
Adjustment
13
Proportion of CEA Procedures in HRRs
Performed in Hospitals with
<20 Past-Year CEA (2007 – 2008)
HRR = Hospital Referral Regions
From the Dartmouth atlas of health care
http://www.dartmouthatlas.org/data/region/
14
Proportion of CAS Procedures in HRRs
Performed in Hospitals with <20 Past-Year CAS
(2007 – 2008)
HRR = Hospital Referral Regions
From the Dartmouth atlas of health care
http://www.dartmouthatlas.org/data/region/
15
What are the implications?
> Real-world IMD performance
• Affected by operator and hospital proficiency
• Proficiency can be diverse in real-world settings
> Need to understand role of provider
• Describe volume-outcome relationship
• Adjust (potentially) for provider proficiency
• Consider other provider characteristics
16
Thank you to collaborators/mentors
> Art Sedrakyan, MD, PhD – Weill Cornell Medical College
> Soko Setoguchi, MD, PhD – Duke Clinical Research
Institute and Brigham and Women’s Hospital/Harvard
Medical School
> Natasha Chen, PhD – Brigham and Women’s
Hospital/Harvard Medical School
> Hiraku Kumamaru, MD, ScD – Harvard Medical School
17
Thank you for your attention
