Propensity Score Matching to Assess the Effect of Single versus - - PowerPoint PPT Presentation

▶
propensity score matching to assess the effect of single
SMART_READER_LITE
LIVE PREVIEW

Propensity Score Matching to Assess the Effect of Single versus - - PowerPoint PPT Presentation

Apr 05, 2024 295 likes •869 views

Propensity Score Matching to Assess the Effect of Single versus Double Embryo Transfer for In Vitro Fertilization Cassandra Wolos Pattanayak and Maurizio Macaluso Friday, March 16, 2012 Joint work with Donald B. Rubin and the Division of

slide-1
SLIDE 1

Propensity Score Matching to Assess the Effect of Single versus Double Embryo Transfer for In Vitro Fertilization

Joint work with Donald B. Rubin and the Division of Reproductive Health at the Centers for Disease Control and Prevention Cassandra Wolos Pattanayak and Maurizio Macaluso Friday, March 16, 2012

slide-2
SLIDE 2

Multiple Birth Rates

United States, 1973-2008

Twin Birth

(per 103 Live Infants Born)

High Order Birth

(per 105 Live Infants Born)

7 14 21 28 35

Calendar Year

50 100 150 200

Calendar Year

slide-3
SLIDE 3

The Mastera twins, in a family photo at 2 months old, each weighed only about 3 pounds at birth. They are doing well now, but the path was difficult. By STEPHANIE SAUL Published: October 10, 2009

The Gift of Life, and Its Price

slide-4
SLIDE 4

What is Assisted Reproductive Technology?

Treatments or procedures that include the handling

  • f human eggs and sperm or embryos for the

purpose of establishing a pregnancy

slide-5
SLIDE 5

Specific ART Treatments

  • ART includes
  • In vitro fertilization with trans-cervical embryo transfer

(IVF)

  • Intracytoplasmic sperm injection (ICSI)
  • Gamete and zygote intrafallopian transfer (GIFT, ZIFT)
  • ART does NOT include
  • Assisted (artificial) insemination only
  • Use of ovarian stimulation medications without egg

retrieval

slide-6
SLIDE 6

What are the steps in a typical ART procedure?

slide-7
SLIDE 7

ART data collection in the U.S.A.

  • 1986: Society for Assisted Reproductive Technology (SART) established

a data collection and reporting program available to ART clinics

  • 1990's: Federal Trade Commission issued cease-and-desist orders

based on inaccurate advertising by an ART program

  • 1992: Congress enacts the Fertility Clinic Success Rate and Certification

Act (FCSRCA)

  • 1997: CDC publishes the 1995 ART Success Rates Report, the first

published under FCSRCA

slide-8
SLIDE 8

Use of ART in the USA over time

20 40 60 80 100 120 140 160 Number (thousands) Year

ART procedures Live infants

slide-9
SLIDE 9

Multiple Births Attributed to ART, US 1997–2007

11.2 12.9 13.5 13.6 15.5 17.1 17.9 18 17.3 17.9 17.8 2 4 6 8 10 12 14 16 18 20

1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

% Year

slide-10
SLIDE 10

Percentage of ART cycles using fresh own embryos, by numbers of embryo transferred, USA, 1996-2009

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

1 2 3 4+

slide-11
SLIDE 11

Percentage of live-birth deliveries from ART cycles using fresh own embryos, by plurality, USA, 1996-2009

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

1 2 3+

slide-12
SLIDE 12

e-SET eligible ART cycles USA, 2005-2006 (N=23,010)

2: 78% 3: 14% 1: 6.4% 4+:1.5% Cycles by no. of embryos transferred Cycles by no. of embryos transferred and no. of live-born infants

slide-13
SLIDE 13

e-SET eligible ART cycles Actual outcomes, USA, 2005-2006 (N=23,010)

1: 33% 2: 19% None: 47% 3+:1% Cycles by no. of live-born infants

37wk+, 2.5Kg+ 37wk+,1.5-2.4Kg 37wk+,<1.5Kg <37wk,2.5Kg+ <37wk, 1.5-2.4Kg <37wk, <1.5Kg

Infants (N=17,220) by delivery outcome 50% 6% 11% 26% 7%

slide-14
SLIDE 14

Cost estimates

Charges (2007 $, Millions) Actual If repeated e-SET policy adopted Charges to patients, ART 358 (260-515) 470 (351-659) Charges to patients, infant outcomes 383 (248-445) 136 (86-167) Charges to third-party payers 1,531 (993-1,781) 546 (344-667) Total charges 2,271 (1,501-2,741) 1,153 (781-1,493)

slide-15
SLIDE 15

Does the decision to transfer one versus two embryos affect pregnancy rates? Multiple birth rates? Rates of prematurity and other complications?

slide-16
SLIDE 16

Our Study Criteria

Fresh cycles, with patients’ own eggs First-time IVF cycles At least three embryos available for transfer Either one or two embryos transferred At least one embryo cryopreserved for possible later use

slide-17
SLIDE 17

Goal: Select a control group from the set of two- embryo cycles that is very similar, on background covariates, to the active treatment group of one-embryo cycles

No outcomes!

slide-18
SLIDE 18

US Assisted Reproductive Technology Database 2004-2005: 22,735 patients in 420 clinics

slide-19
SLIDE 19

US Assisted Reproductive Technology Database 1,025: single embryo transfer to uterus (SET) – low risk of twins 21,710: double embryo transfer to uterus (DET) – higher chance of success

slide-20
SLIDE 20

US Assisted Reproductive Technology Database 1,025: single embryo transfer to uterus (SET) – low risk of twins 21,710: double embryo transfer to uterus (DET) – higher chance of success

slide-21
SLIDE 21

Identifying Covariates

  • Baseline patient data
  • Clinic-level data
  • Medical treatment data
  • Complications prior to embryo transfer

Which are important for the decision to transfer one embryo v. two? Which are related to outcomes?

slide-22
SLIDE 22

61 important covariates identified:

  • 3 for exact matching
  • 21 “primary” binary
  • 16 “secondary” binary
  • 6 continuous
  • 5 transformations of continuous covariates
  • 10 pairwise interactions
slide-23
SLIDE 23

Summarize multiple covariates with estimated

propensity score: – Fit logistic regression model to predict treatment status given covariates – Fitted values from logistic regression model are estimated propensity scores – Match or subclassify based on estimated propensity scores and examine resulting balance – Iterate, choosing the model and matching process that give the best balance, as shown by diagnostics

slide-24
SLIDE 24

Randomly order SET patients For each SET patient, create a set of potential matches:

  • Same clinic
  • Same year
slide-25
SLIDE 25

Randomly order SET patients For each SET patient, create a set of potential matches:

  • Same clinic
  • Same year
  • Same category of number of days between egg retrieval and

embryo transfer (2-3 days or 4-7 days) Number of days: 2 3 4 5 6 7

slide-26
SLIDE 26

Randomly order SET patients For each SET patient, create a set of potential matches:

  • Same clinic
  • Same year
  • Same category of number of days between egg retrieval and

embryo transfer (2-3 days or 4-7 days)

  • Within fixed distance on logit of estimated propensity score

Select the potential DET match with the estimated propensity score logit closest to that of the SET patient, if it has not already been matched Logit Prop Score:

  • 3.50
  • 3.25
  • 3.00
  • 2.75
  • 2.50
  • 2.25

Propensity Score: 0.03 0.04 0.05 0.06 0.08 0.10

slide-27
SLIDE 27

When a satisfactory match did not exist within the year, matched across years Each DET patient was matched to no more than one SET patient Using the criteria above, we identified matches for 923 of the 1025 SET patients; the omitted SET patients were those who were not similar to any of the DET patients

slide-28
SLIDE 28

Measures

  • Entire distribution, graphically
  • Difference in means (standardized for

continuous covariates)

  • Odds ratio (for binary covariates)
  • Variances (for continuous covariates)
  • Extremes: minimum and maximum
slide-29
SLIDE 29

Difference in Means for Primary Binary Cov

difference in means

  • 0.05
  • 0.04
  • 0.03
  • 0.02
  • 0.01

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

slide-30
SLIDE 30

Difference in Means for Primary Binary Cov

difference in means

  • 0.05
  • 0.04
  • 0.03
  • 0.02
  • 0.01

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

RFAENDO RFAPCO RFAUTERI RFAUNEXP RFAMultFemaleonly RFAMultMaleFemale RFADIMOVonly RFAOTHERonly RFAMALEonly RFATUB GESTATIO.1 HighFSH HighFSHUnknown

slide-31
SLIDE 31

Difference in Means for Primary Binary Cov

difference in means

  • 0.05
  • 0.04
  • 0.03
  • 0.02
  • 0.01

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

RFAENDO RFAPCO RFAUTERI RFAUNEXP RFAMultFemaleonly RFAMultMaleFemale RFADIMOVonly RFAOTHERonly RFAMALEonly RFATUB GESTATIO.1 HighFSH HighFSHUnknown

Initial Final Matches Initial 1 5 Initial Final Matches Initial Final Matches Initial Initial Final Matches

slide-32
SLIDE 32

Difference in Means for Primary Binary Cov

difference in means

  • 0.05
  • 0.04
  • 0.03
  • 0.02
  • 0.01

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

RFAENDO RFAPCO RFAUTERI RFAUNEXP RFAMultFemaleonly RFAMultMaleFemale RFADIMOVonly RFAOTHERonly RFAMALEonly RFATUB GESTATIO.1 HighFSH HighFSHUnknown

Initial Final Matches Initial 1 5 Initial Final Matches Initial Final Matches Initial Initial Final Matches Initial Final Matches

slide-33
SLIDE 33

Difference in Means for Primary Binary Cov

difference in means

  • 0.05
  • 0.04
  • 0.03
  • 0.02
  • 0.01

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

RFAENDO RFAPCO RFAUTERI RFAUNEXP RFAMultFemaleonly RFAMultMaleFemale RFADIMOVonly RFAOTHERonly RFAMALEonly RFATUB GESTATIO.1 HighFSH HighFSHUnknown

Initial Final Matches

slide-34
SLIDE 34

Difference in Means for Primary Binary Cov

difference in means

  • 0.05
  • 0.04
  • 0.03
  • 0.02
  • 0.01

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

RFAENDO RFAPCO RFAUTERI RFAUNEXP RFAMultFemaleonly RFAMultMaleFemale RFADIMOVonly RFAOTHERonly RFAMALEonly RFATUB GESTATIO.1 HighFSH HighFSHUnknown

Initial Final Matches

slide-35
SLIDE 35

Difference in Means for Primary Binary Cov

difference in means

  • 0.05
  • 0.04
  • 0.03
  • 0.02
  • 0.01

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

RFAENDO RFAPCO RFAUTERI RFAUNEXP RFAMultFemaleonly RFAMultMaleFemale RFADIMOVonly RFAOTHERonly RFAMALEonly RFATUB GESTATIO.1 HighFSH HighFSHUnknown

Initial Final Matches

slide-36
SLIDE 36

Difference in Means for Primary Binary Cov

difference in means

  • 0.05
  • 0.04
  • 0.03
  • 0.02
  • 0.01

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

RFAENDO RFAPCO RFAUTERI RFAUNEXP RFAMultFemaleonly RFAMultMaleFemale RFADIMOVonly RFAOTHERonly RFAMALEonly RFATUB GESTATIO.1 HighFSH HighFSHUnknown

Initial Final Matches Initial 1 5 Initial Final Matches Initial Final Matches Initial Initial Final Matches Initial Final Matches

  • 0.038

0.005

slide-37
SLIDE 37

Absolute Difference in Means Primary and Secondary Binary Co

Absolute Difference in Means Initial Final Match 0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25

slide-38
SLIDE 38

Absolute Difference in Means Primary and Secondary Binary Co

Absolute Difference in Means Initial Final Match 0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25

+0.038 0.005

slide-39
SLIDE 39

Absolute Difference in Means Primary and Secondary Binary Co

Absolute Difference in Means Initial Final Match 0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25

slide-40
SLIDE 40

Absolute Difference in Means Primary and Secondary Binary Co

Absolute Difference in Means Initial Final Match 0.00 0.05 0.10 0.15 0.20 0.25 0.00 0.05 0.10 0.15 0.20 0.25

  • 37 decreased

10 increased

slide-41
SLIDE 41

Initial Final 0.00 0.05 0.10 0.15 0.20 0.25

Absolute Difference in Means Primary and Secondary Binary C

Absolute Difference in Means 0.00 0.05 0.10 0.15 0.20 0.25

slide-42
SLIDE 42

Initial Final

  • 0.10
  • 0.05

0.00 0.05 0.10 0.15 0.20 0.25

Difference in Means Primary and Secondary Binary C

Difference in Means

  • 0.10
  • 0.05

0.00 0.05 0.10 0.15 0.20 0.25

slide-43
SLIDE 43

Initial Final

  • 0.04
  • 0.02

0.00 0.02 0.04

Difference in Means Primary and Secondary Binary C

Difference in Means

  • 0.04
  • 0.02

0.00 0.02 0.04

slide-44
SLIDE 44

Initial Final

  • 0.04
  • 0.02

0.00 0.02 0.04

Difference in Means Primary and Secondary Binary C

Difference in Means

  • 0.04
  • 0.02

0.00 0.02 0.04

slide-45
SLIDE 45

Initial Final 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Odds Ratios Primary and Secondary Binary C

Odds Ratio 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

slide-46
SLIDE 46

All SET Units

Frequency 20 40 60 80 100 50 100 150

Matched SET Units

Frequency 20 40 60 80 100 50 100 150

All DET Units

Frequency 20 40 60 80 100 500 1500

Matched DET Units

Frequency 20 40 60 80 100 50 100 150

TRAN_OTHER.1

slide-47
SLIDE 47

All SET Units

Frequency 20 25 30 35 40 45 50 20 60 100

Matched SET Units

Frequency 20 25 30 35 40 45 50 20 40 60 80

All DET Units

Frequency 20 25 30 35 40 45 50 500 1500

Matched DET Units

Frequency 20 25 30 35 40 45 50 20 60 100

Patient Age

slide-48
SLIDE 48

Patient Age

Matched SET Units

Frequency 20 25 30 35 40 40 80

Matched DET Units

Frequency 20 25 30 35 40 40 80

slide-49
SLIDE 49

All SET Units

Frequency 0.0 0.2 0.4 0.6 0.8 1.0 100 300

Matched SET Units

Frequency 0.0 0.2 0.4 0.6 0.8 1.0 100 200 300 400

All DET Units

Frequency 0.0 0.2 0.4 0.6 0.8 1.0 4000 8000

Matched DET Units

Frequency 0.0 0.2 0.4 0.6 0.8 1.0 100 200 300 400

  • Est. Propensity Score
slide-50
SLIDE 50

Why these matches?

Why this propensity score model? Why this matching protocol?

slide-51
SLIDE 51

Choose the model leading to best balance, as shown by diagnostics, emphasizing the most important covariates: Not necessarily the best-fitting model Not necessarily the model that best reflects reality Not necessarily the model that creates best balance overall

slide-52
SLIDE 52

Difference in Means for Primary Binary Cov

difference in means

  • 0.05
  • 0.04
  • 0.03
  • 0.02
  • 0.01

0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08

RFAENDO RFAPCO RFAUTERI RFAUNEXP RFAMultFemaleonly RFAMultMaleFemale RFADIMOVonly RFAOTHERonly RFAMALEonly RFATUB GESTATIO.1 HighFSH HighFSHUnknown

Initial Final Matches

slide-53
SLIDE 53

Difference in Means for (Important) Binary Co

difference in means

  • 0.08
  • 0.06
  • 0.04
  • 0.02

0.00 0.02 0.04 0.06 0.08

RFAENDO RFAPCO RFAUTERI RFAUNEXP RFAMultFemaleonly RFAMultMaleFemale RFADIMOVonly RFAOTHERonly RFAMALEonly RFATUB GESTATIO.1 HighFSH HighFSHUnknown

Initial MatchYear MatchYear.Day

Difference in Means for Primary Binary Covariates

slide-54
SLIDE 54

Difference in Means for (Important) Binary Covariates

difference in means

  • 0.08
  • 0.06
  • 0.04
  • 0.02

0.00 0.02 0.04 0.06 0.08

RFAENDO RFAPCO RFAUTERI RFAUNEXP RFAMultFemaleonly RFAMultMaleFemale RFADIMOVonly RFAOTHERonly RFAMALEonly RFATUB GESTATIO.1 HighFSH HighFSHUnknown

Initial Matches WithinYear (OrigModel) WithinYear (SepModels)

Difference in Means for Primary Binary Covariates

slide-55
SLIDE 55

For data through 2008: Matched based on continuous time instead of within year Included transfers on days 3 and 5 only Exact matched within age groups: <35, 35+

slide-56
SLIDE 56
  • SART clinicians reviewed and

approved proposed matches

  • Protocol for outcome analysis

approved Outcome analysis currently underway