learning programs to accelerate the biopharma transition network meta-analysis what is a network...
TRANSCRIPT
Learning Programs to Accelerate the BioPharma Transition
Network Meta-analysis
• What is a network meta-analysis?
• GRADE approach to confidence in estimates
• Determining credibility of NMA
Learning Programs to Accelerate the BioPharma Transition
Comparing Multiple Treatments:Introduction to Network Meta-Analyses
• Many disease areas where many alternatives exist
• Clinicians/patients need to know about relative merits
• Impractical to test each comparator directly
• Simultaneous comparison of multiple treatments
• “Network meta-analysis”, “mixed treatment comparisons”, “adjusted indirect comparisons”
Learning Programs to Accelerate the BioPharma Transition
Conventional meta-analysisPooled Estimate assumption
Patients Interventions Outcomes Methodology
Assumes effect similar across
“Homogeneity assumption”
Single best estimate of treatment effect
Learning Programs to Accelerate the BioPharma Transition
Relative Risk with 95% CI for Vitamin D Non-vertebral Fractures
Chapuy et al, (2002) 0.85 (0.64, 1.13)
Pooled Random Effect Model 0.82 (0.69 to 0.98)p= 0.05 for heterogeneity, I2=53%
Chapuy et al, (1994) 0.79 (0.69, 0.92)
Lips et al, (1996) 1.10 (0.87, 1.39)
Dawson-Hughes et al, (1997) 0.46 (0.24, 0.88)
Pfeifer et al, (2000) 0.48 (0.13, 1.78)
Meyer et al, (2002) 0.92 (0.68, 1.24)
Trivedi et al, (2003) 0.67 (0.46, 0.99)
Favors Vitamin D Favors Control
Relative Risk 95% CI
'
'
'
'
'
'
'
'
0.1 1
Learning Programs to Accelerate the BioPharma Transition
Alendronate (A) Risedronate (B)
Placebo (C)
Interested in A versus B available data A vs C, B vs C
Less confidence than direct?Why?
Indirect Comparisons
Learning Programs to Accelerate the BioPharma Transition
Vulnerability of Indirect comparison
• Effect modifiers– Patients– Optimal interventions– Comparator– Cointerventions– Outcome measures– Risk of bias
Learning Programs to Accelerate the BioPharma Transition
Combine direct and indirect comparisons- additional assumption mediators same in direct and indirect- “consistency” or “coherence” assumption
Learning Programs to Accelerate the BioPharma Transition
Learning Programs to Accelerate the BioPharma Transition
Comparison with
TreatmentsNRT
.5 1 2.75 1.33
Odds ratio
4
Antidepressants
NRT + NRT
1.54(1.02-2.31)
.5 1 2.75 1.33
Odds ratio
4 .5 1 2.75 1.33
Odds ratio
4
Varenicline
Antidepressants
Varenicline1.70
(1.41-2.04)
NRT + antidepressant1.62
(1.00-2.63)
1.28(0.82-1.99)
.5 1 2.75 1.33
Odds ratio
NRT + antidepressant
1.34 (0.71-2.56)
Comparison with
TreatmentsNRT
.5 1 2.75 1.33
Odds ratio
4
Antidepressants
NRT + NRT
1.54(1.02-2.31)
.5 1 2.75 1.33
Odds ratio
4 .5 1 2.75 1.33
Odds ratio
4
Varenicline
Antidepressants
Varenicline1.70
(1.41-2.04)
NRT + antidepressant1.62
(1.00-2.63)
1.28(0.82-1.99)
.5 1 2.75 1.33
Odds ratio
NRT + antidepressant
1.34 (0.71-2.56)
Network Meta-analysis Case Study: Which Approach to Nicotine Addiction Works Best
Learning Programs to Accelerate the BioPharma Transition
Network Meta-analysis Case Study
Combines effect estimates from direct and indirect comparisons
Placebo
Varenicline
Antidepressants + NRT
Nicotine replacement
treatment (NRT)
Antidepressants
Learning Programs to Accelerate the BioPharma Transition
Comparison with
TreatmentsNRT
.5 1 2.75 1.33
Odds ratio
4
Antidepressants
NRT + NRT
1.54(1.02-2.31)
.5 1 2.75 1.33
Odds ratio
4 .5 1 2.75 1.33
Odds ratio
4
Varenicline
Antidepressants
Varenicline1.70
(1.41-2.04)
NRT + antidepressant1.62
(1.00-2.63)
1.28(0.82-1.99)
.5 1 2.75 1.33
Odds ratio
NRT + antidepressant
1.34 (0.71-2.56)
Comparison with
TreatmentsNRT
.5 1 2.75 1.33
Odds ratio
4
Antidepressants
NRT + NRT
1.54(1.02-2.31)
.5 1 2.75 1.33
Odds ratio
4 .5 1 2.75 1.33
Odds ratio
4
Varenicline
Antidepressants
Varenicline1.70
(1.41-2.04)
NRT + antidepressant1.62
(1.00-2.63)
1.28(0.82-1.99)
.5 1 2.75 1.33
Odds ratio
NRT + antidepressant
1.34 (0.71-2.56)
Comparison with
TreatmentsNRT
1.01 (0.88- 1.15)
.5 1 2.75 1.33
Odds ratio
4
Antidepressants
NRT + NRT 1.35 (1.04- 1.75)
1.34(1.00- 1.78)
.5 1 2.75 1.33
Odds ratio
4 .5 1 2.75 1.33
Odds ratio
4
Varenicline
Antidepressants
Varenicline 1.17 (0.98- 1.39)
1.17 (0.98- 1.39)
1.16(0.97- 1.38)
1.16(0.97- 1.38)
NRT + antidepressant 1.30(0.98- 1.73)
1.30(0.98- 1.73)
1.29 (0.96- 1.74)
1.29 (0.96- 1.74)
1.12 (0.81- 1.55)
.5 1 2.75 1.33
Odds ratio
1.16 (0.85- 1.58)
NRT + antidepressant
1.04 (0.71- 1.52)
Learning Programs to Accelerate the BioPharma Transition
Antide-pressants
NRT
Direct Comparison
control
buspirone
rimonabant
varenicline
antidepressants +NRT
clonidine
NRT+NRT
1.63, I2=0% 4 comparisons
1.36, I2=0% 2 comparisons
0.73, I2=0% 2 comparisons
2.68, I2=82% 5 comparisons
1.28, I2=0%3 comparisons
4.851 comparison
1.54, I2=46%5 comparisons
1.14, I2=63%6 comparisons
1.70, I2=0% 3 comparisons
1.281 comparison
1.121 comparison
1.88, I2=19%29 comparisons
1.85, I2=13%67 comparisons
1.5 1 1.5 2 2.5NRT superior Antidepressants
superior
Direct evidence (3 trials)
1.34 (0.71, 2.56)I-squared=43.7%
Learning Programs to Accelerate the BioPharma Transition
Antide-pressants
NRT
Indirect Comparison 1
control
buspirone
rimonabant
varenicline
antidepressants +NRT
clonidine
NRT+NRT
1.63, I2=0% 4 comparisons
1.36, I2=0% 2 comparisons
0.73, I2=0% 2 comparisons
2.68, I2=82% 5 comparisons
1.28, I2=0%3 comparisons
4.851 comparison
1.54, I2=46%5 comparisons
1.14, I2=63%6 comparisons
1.70, I2=0% 3 comparisons
1.281 comparison
1.121 comparison
1.88, I2=19%29 comparisons
1.85, I2=13%67 comparisons
1.5 1 1.5 2 2.5NRTsuperior Antidepressants
superior
Direct evidence (3 trials)
Indirect evidence
1.34 (0.71, 2.56)I-squared=43.7%
1.01 (0.81,1.27)
Learning Programs to Accelerate the BioPharma Transition
Antide-pressants
NRT
Indirect Comparison 2
control
buspirone
rimonabant
varenicline
antidepressants +NRT
clonidine
NRT+NRT
1.63, I2=0% 4 comparisons
1.36, I2=0% 2 comparisons
0.73, I2=0% 2 comparisons
2.68, I2=82%5 comparisons
1.28, I2=0%3 comparisons
4.851 comparison
1.54, I2=46%5 comparisons
1.14, I2=63%6 comparisons
1.70, I2=0%3 comparisons
1.281 comparison
1.121 comparison
1.88, I2=19%29 comparisons
1.85, I2=13%67 comparisons
antidepressants +NRT
1.14, I2=63%6 comparisons
1.5 1 1.5 2 2.5NRTsuperior Antidepressants
superior
Direct evidence (3 trials)
Indirect evidence
1.34 (0.71, 2.56)I-squared=43.7%
0.85 (0.38, 1.92)
Learning Programs to Accelerate the BioPharma Transition
5 Paths to Indirectly CompareAntidepressants vs NRT
placebo andnonplacebo control
buspirone
rimonabant
vareniclineantidepressants
NRT
antidepressants +NRT
clonidine
NRT+NRT
4 comparisons
2 comparisons
2 comparisons
5 comparisons29 comparisons
67 comparisons
3 comparisons
1 comparison
5 comparisons
6 comparisons
3 comparisons
3 comparisons
1 comparison
1 comparison
1
2
3
4
5
Learning Programs to Accelerate the BioPharma Transition
5 Paths to Indirectly Compare Antidepressants vs NRT
placebo andnonplacebo control
buspirone
rimonabant
vareniclineantidepressants
NRT
antidepressants +NRT
clonidine
NRT+NRT
4 comparisons
2 comparisons
2 comparisons
5 comparisons29 comparisons
67 comparisons
3 comparisons
1 comparison
5 comparisons
6 comparisons
3 comparisons
3 comparisons
1 comparison
1 comparison
1
2
3
4
5
1.01 (0.81, 1.27)
0.85 (0.38, 1.92)
0.89 (0.29, 2.77)
1.56 (0.54, 4.49)
1.31 (0.25, 6.76)
Learning Programs to Accelerate the BioPharma Transition
Antide-pressants
NRT
Comparative effectiveness of NRT vs. Antidepressants on prolonged abstinence (≥6 months)
control
buspirone
rimonabant
varenicline
clonidine
NRT+NRT
1.63, I2=0% 4 comparisons
1.36, I2=0% 2 comparisons
0.73, I2=0% 2 comparisons
2.68, I2=82%5 comparisons
1.28, I2=0%3 comparisons
4.851 comparison
1.54, I2=46%5 comparisons
1.14, I2=63%6 comparisons
1.70, I2=0%3 comparisons
1.281 comparison
1.121 comparison
1.88, I2=19%29 comparisons
1.34, I2=44%3 comparisons
1.85, I2=13%67 comparisons
Indirect evidence
Direct evidence
I-squared = 43.7%
1.34 (0.71, 2.56)
1.5 1 1.5 2 2.5
3 trials pooled
NRT superior Antidepressants superior
0.89 (0.29, 2.77)
1.56 (0.54, 4.49)
0.85 (0.38, 1.92)
1.31 (0.25, 6.76)
1.01 (0.81, 1.27)
1.5 1 1.5 2 2.5
Antidepressants superior
NRT superior
1
2
3
4
5
Path
Learning Programs to Accelerate the BioPharma Transition
Antide-pressants
NRT
Comparative effectiveness of NRT vs. Antidepressants on prolonged abstinence (≥6 months)
control
buspirone
rimonabant
varenicline
clonidine
NRT+NRT
1.63, I2=0% 4 comparisons
1.36, I2=0% 2 comparisons
0.73, I2=0% 2 comparisons
2.68, I2=82%5 comparisons
1.28, I2=0%3 comparisons
4.851 comparison
1.54, I2=46%5 comparisons
1.14, I2=63%6 comparisons
1.70, I2=0%3 comparisons
1.281 comparison
1.121 comparison
1.88, I2=19%29 comparisons
1.34, I2=44%3 comparisons
1.85, I2=13%67 comparisons
Indirect evidence
Direct evidence
I-squared = 43.7%
1.34 (0.71, 2.56)
1.5 1 1.5 2 2.5
3 trials pooled
NRT superior Antidepressants superior
0.89 (0.29, 2.77)
1.56 (0.54, 4.49)
0.85 (0.38, 1.92)
1.31 (0.25, 6.76)
1.01 (0.81, 1.27)
1.5 1 1.5 2 2.5
Antidepressants superior
NRT superior
1
2
3
4
5
Path
0.98 (95% 0.85-1.13)
Learning Programs to Accelerate the BioPharma Transition
Antide-pressants
NRT
Comparative effectiveness of NRT vs. Antidepressants on prolonged abstinence (≥6 months)
control
buspirone
rimonabant
varenicline
clonidine
NRT+NRT
1.63, I2=0% 4 comparisons
1.36, I2=0% 2 comparisons
0.73, I2=0% 2 comparisons
2.68, I2=82%5 comparisons
1.28, I2=0%3 comparisons
4.851 comparison
1.54, I2=46%5 comparisons
1.14, I2=63%6 comparisons
1.70, I2=0%3 comparisons
1.281 comparison
1.121 comparison
1.88, I2=19%29 comparisons
1.34, I2=44%3 comparisons
1.85, I2=13%67 comparisons
Indirect evidence
Direct evidence
I-squared = 43.7%
1.34 (0.71, 2.56)
1.5 1 1.5 2 2.5
3 trials pooled
Antidepressants superior
0.89 (0.29, 2.77)
1.56 (0.54, 4.49)
0.85 (0.38, 1.92)
1.31 (0.25, 6.76)
1.01 (0.81, 1.27)
1.5 1 1.5 2 2.5
Antidepressants superior
NRT superior
1
2
3
4
5
Path
0.98 (95% 0.85-1.13)
pooled estimate1.01
(95% 0.88-1.15)
NRT superior
Alendronate n=5,084 Raloxifene
n=10,975
Zoledronate n=4,954
Denosumab n=3,933
Vitamin D n=12,469
Calcium n=3,896
Ibandronate n=1,912
Placebo n=41,548
Teriparatide (PTH) n=1,093
Risedronate
n=6,850
Hip Fractures# of trials =40# of participants =139,647# of hip fracture =2,567
Vitamin D and Calcium n=46,933
Confidence in Estimates
Step 1: Present direct and indirect estimate for each comparison of the evidence network
Step 2: Rate the confidence in the direct and indirect estimate
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
0.53(0.30 to 0.90)
ExampleApproach
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
Step 3: Present the NMA estimate
Step 4: Rate the confidence in the NMA estimate
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
0.51 (0.29-0.87)
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
0.51 (0.29-0.87)
MODERATE
Step 1: Present direct and indirect estimate for each comparison of the evidence network
Step 2: Rate the confidence in the direct and indirect estimate
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
0.53(0.30 to 0.90)
ExampleApproach
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
Step 3: Present the NMA estimate
Step 4: Rate the confidence in the NMA estimate
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
0.51 (0.29-0.87)
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
0.51 (0.29-0.87)
MODERATE
Starting level of indirect
• Intutitively lower confidence– Previous GRADE guidance start at moderate
• NMA enthusiasts argue no different– New guidance: Start at high– Only rate down for lack of similarity
(intransivity)
Alendronate n=5,084 Raloxifene
n=10,975
Zoledronate n=4,954
Denosumab n=3,933
Vitamin D n=12,469
Calcium n=3,896
Ibandronate n=1,912
Placebo n=41,548
Teriparatide (PTH) n=1,093
Risedronate
n=6,850
Hip Fractures# of trials =40# of participants =139,647# of hip fracture =2,567
Vitamin D and Calcium n=46,933
Alendronate n=5,084 Raloxifene
n=10,975
Zoledronate n=4,954
Denosumab n=3,933
Vitamin D n=12,469
Calcium n=3,896
Ibandronate n=1,912
Placebo n=41,548
Teriparatide (PTH) n=1,093
Risedronate
n=6,850
Hip Fractures# of trials =40# of participants =139,647# of hip fracture =2,567
Vitamin D and Calcium n=46,933
Dominant link
Confidence in indirect estimate
• Lowest of direct estimates
• A versus B comparison of interest
• Through C dominant link
• A versus C high; B versus C moderate or low
• Confidence based on B versus C
Step 1: Present direct and indirect estimate for each comparison of the evidence network
Step 2: Rate the confidence in the direct and indirect estimate
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
0.53(0.30 to 0.90)
ExampleApproach
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
Step 3: Present the NMA estimate
Step 4: Rate the confidence in the NMA estimate
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
0.51 (0.20-1.32)
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
0.51 (0.20-1.32)
MODERATE
Step 1: Present direct and indirect estimate for each comparison of the evidence network
Step 2: Rate the confidence in the direct and indirect estimate
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
0.53(0.30 to 0.90)
ExampleApproach
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
Step 3: Present the NMA estimate
Step 4: Rate the confidence in the NMA estimate
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
0.51 (0.20-1.32)
Comparison Direct evidence OR
(95% confidence
interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence
confidence in estimates
NetworkOR (95% credible interval)
Network confidence in
estimates
Alendronate vs. Raloxifene
0.49 (0.04 to 5.45)
LOW
0.53(0.30 to 0.90)
MODERATE
0.51 (0.20-1.32)
MODERATE
Confidence in NMA estimate
• If only direct or indirect use that
• If both use higher
Confidence in NMA estimate
• Direct and indirect very different
• Terminology differs
• Ours (for now)– Heterogeneity for direct– Transitivity for indirect– Incoherence for direct versus indirect
Judging incoherence
• Difference in point estimates
• Extent of overlap in confidence intervals
• P-value for test of incoherence
P-value for test of incoherence 0.02
Alendronate n=5,084 Raloxifene
n=10,975
Zoledronate n=4,954
Denosumab n=3,933
Vitamin D n=12,469
Calcium n=3,896
Ibandronate n=1,912
Placebo n=41,548
Teriparatide (PTH) n=1,093
Risedronate
n=6,850
Hip Fractures# of trials =40# of participants =139,647# of hip fracture =2,567
Vitamin D and Calcium n=46,933
Credibility of the Process of NMA
• Usual criteria for systematic review– Explicit sensible questions– Search exhaustive– Selection and assessment reproducible– Present results ready for clinical application
• Address confidence in effect estimates
Presentation of Results
• Often presented with rankings
• Potentially very misleading– Small difference between ranks– Everything low or very low confidence– First ranked lower than others
Comparison Direct evidence OR (95%
confidence interval)
Direct evidence
confidence in estimates
Indirect evidence OR (95% credible
interval)
Indirect evidence confidence in
estimates
NetworkOR (95% credible
interval)
Network confidence in
estimates
Teriparatide vs. Placebo --- --- 0.42 (0.10-1.82) very low 3, 6 0.42 (0.10-1.82) very low
Zoledronate vs. Placebo --- --- 0.50 (0.33-0.74) high 0.50 (0.34-0.73) high
Risedronate vs. Placebo 0.17 (0.05 to 0.59) moderate 1 0.54 (0.36-0.75) low 6 0.48 (0.31-0.66) moderate
Conclusion
• NMA will play important role in EBCP
• Needs usual criteria for SR/MA credibility
• If includes confidence in estimates interpretable
• If no confidence very difficult to interpret– Probably shouldn’t bother