How to collect and report outcomes of heart valve surgery

Hanneke Takkenberg

Dept. of Cardio-Thoracic Surgery

Erasmus University Medical Center, Rotterdam

The Netherlands

East European Heart Valve Postgraduate Course Sep 2007

Survival after mechanical AVR relative to general population

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

0 5 10 15 20 25 30 35 40 45

Time (years since AVR

Su

rviv

al (

%/1

00)

survival 40-year old patient after AVR

40-year old male general population

Components of mortality after valve procedures

Operative mortality Valve-related events (Edmunds 1996 Guidelines/new Guidelines coming

soon!!!):

Structural valve deterioration

Non-structural valve deterioration

Endocarditis

Thrombo-embolism

Bleeding

Valve thrombosis

And their consequences: death, reop, invalidation

Excess mortality yet unexplained

Study designs

Randomized trial Highest level of evidence

Usually difficult to accomplish

Selection bias

Cohort study Prospective

Retrospective

Case-control study Useful when the outcome is rare

Study designs

Randomized trialRandomized trial Highest level of evidenceHighest level of evidence

Usually difficult to accomplishUsually difficult to accomplish

Selection biasSelection bias

Cohort study Prospective

Retrospective

Case-control studyCase-control study Useful when the outcome is rareUseful when the outcome is rare

How to collect heart valve surgery data?

Make a plan (and create a budget) first!

How to collect heart valve surgery data?

Make a plan (and create a budget) first!

Define your variables carefully and document this

Follow the Reporting guidelines (new ones coming up!!!)

Build a database (for example MS Access)

Obtain approval from your Institutional Review Board

Prospective rather than retrospective (recall bias, missing information)

Set up your annual prospective follow-up (using queries in MS Access)

Check your database periodically for quality and completeness of data (10% of entered data contains errors)

How to report outcome after heart valve operations

Descriptives: Describe the patient and procedure characteristics

Number of deaths/events (early and late)

Incidence rates of late events (number/year, Weibull,…)

Describe modes of failure, clinical status, last echo

Logistic regression: To assess factors that may influence early outcome, OR is calculated

Univariate versus multivariate

Kaplan- Meier analysis (time-to event model in the presence of censored cases): To describe freedom from death/events

Comparison of KM-estimates: log rank test

Cox regression: To assess factors that may influence outcome over time, HR is calculated

Univariate versus multivariate

Type of Data

Goal Measurement (from Gaussian Population)

Rank, Score, or Measurement (from

Non- Gaussian Population)

Binomial(Two Possible

Outcomes)

Survival Time

Describe one group Mean, SD Median, interquartile range Proportion Kaplan Meier survival curve

Compare one group to a hypothetical value

One-sample t test Wilcoxon test Chi-squareor

Binomial test **

Compare two unpaired groups

Unpaired t test Mann-Whitney test Fisher's test(chi-square for large

samples)

Log-rank test or Mantel-Haenszel*

Compare two paired groups

Paired t test Wilcoxon test McNemar's test Conditional proportional hazards regression*

Compare three or more unmatched groups

One-way ANOVA Kruskal-Wallis test Chi-square test Cox proportional hazard regression**

Compare three or more matched groups

Repeated-measures ANOVA

Friedman test Cochrane Q** Conditional proportional hazards regression**

Quantify association between two variables

Pearson correlation Spearman correlation Contingency coefficients**

Predict value from another measured variable

Simple linear regressionor

Nonlinear regression

Nonparametric regression** Simple logistic regression* Cox proportional hazard regression*

Predict value from several measured or binomial variables

Multiple linear regression*or

Multiple nonlinear regression**

Multiple logistic regression* Cox proportional hazard regression*

Standard methods of outcome assessment after heart valve operations

Descriptives: Describe the patient and procedure characteristics

Number of deaths/events (early and late)

Incidence rates of late events (number/year, Weibull,…)

Describe modes of failure, clinical status, last echo

Logistic regression:Logistic regression: To assess factors that may influence early outcome, OR is calculatedTo assess factors that may influence early outcome, OR is calculated

Univariate versus multivariateUnivariate versus multivariate

Kaplan- Meier analysis (time-to event model in the presence of censored cases):Kaplan- Meier analysis (time-to event model in the presence of censored cases): To describe freedom from death/eventsTo describe freedom from death/events

Comparison of KM-estimates: log rank testComparison of KM-estimates: log rank test

Cox regression:Cox regression: To assess factors that may influence outcome over time, HR is calculatedTo assess factors that may influence outcome over time, HR is calculated

Univariate versus multivariateUnivariate versus multivariate

Continuous variables:

Mean

(Median)

Standard deviation

Range

Discrete variables:

Proportion

Number

In your methods section

provide clear definitions

of your parameters

Do provide counts

but also explain!

Description of the 6 CABG complications:

Description of the causes of operative mortality:

Early versus late complications

Early complications (<30 days postop or during hospitalization): Describe using proportions (%) and numbers

Late complications (>30 days postop): Describe using numbers, incidence rates, or other functions (Weibull,

Gompertz, 2-period risk)

Incidence rate = number of complications / number of patient years

Example:

Incidence rates are also used by the FDA for measuring the OPCs of valves

OPC = Objective performance criteria

Example of a Weibull function for SVD

Age-dependent freedom from SVD after allograft aortic valve replacement

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

0 5 10 15 20 25Time since valve replacement (yrs)

Fre

edom

from

SV

D

65 yr old patient55 yr old patient45 year old patient35 year old patient25 year old patient

Standard methods of outcome assessment after heart valve operations

Descriptives:Descriptives: Describe the patient and procedure characteristicsDescribe the patient and procedure characteristics

Number of deaths/events (early and late)Number of deaths/events (early and late)

Incidence rates of late events (number/year, Weibull,…)Incidence rates of late events (number/year, Weibull,…)

Describe modes of failure, clinical status, last echoDescribe modes of failure, clinical status, last echo

Logistic regression: To assess factors that may influence early outcome, OR is calculated

Univariate versus multivariate

Kaplan- Meier analysis (time-to event model in the presence of censored cases):Kaplan- Meier analysis (time-to event model in the presence of censored cases): To describe freedom from death/eventsTo describe freedom from death/events

Comparison of KM-estimates: log rank testComparison of KM-estimates: log rank test

Cox regression:Cox regression: To assess factors that may influence outcome over time, HR is calculatedTo assess factors that may influence outcome over time, HR is calculated

Univariate versus multivariateUnivariate versus multivariate

Logistic regression

Is used to study factors that may influence a bivariate outcome that is not time-dependent

Outcome measure is Odds Ratio: OR

First perform univariate logistic regression analysis: One factor at once into the model

Repeat this for all factors that you think may affect outcome

Preferably use continuous measures of factors instead of categories

Then perform multivariate logistic regression analysis: All factors that were significant in the univariate model (p<0.05 or 0.10)

Note: do not put too many factors in (1 per 7-10 outcomes)

Avoid factors that represent approximately the same (example ECC time and aortic cross clamp time)

Example of logistic regression analysis  Univariate model

OR (95% CI) 

P-valueMultivariate model

OR (95% CI) 

P-value

Patient age (yrs) 1.07 (1.03-1.12) 0.001 1.07 (1.01-1.12) 0.016

Creatinin (mol/L) 1.005 (1.00-1.01) 0.004 1.005 (1.00-1.01) 0.002

Perfusion time (min) 1.007 (1.00-1.01) 0.004 1.004 (1.00-1.01) NS

Procedure-related CABG 9.85 (1.65-38.77) 0.01 13.14 (1.27-136.21) 0.03

NYHA class 1.67 (1.06-2.60) 0.03 1.46 (0.89-2.40) NS

Gender 2.52 (0.88-7.22) 0.09 -- --

Active endocarditis 2.46 (0.80-7.53) 0.12 -- --

Urgency (within 24 hrs) 2.16 (0.57-8.13) NS -- --

Concomitant procedures 2.25 (0.75-6.75) NS -- --

Circulatory arrest 0.53 (0.07-4.15) NS -- --

Prior operations 1.41 (0.47-4.28) NS -- --

SC vs ARR technique 1.06 (0.35-3.19) NS -- --

Ventilatory support 2.70 (0.56-13.17) NS -- --

Left ventricular function 0.79 (0.34-1.84) NS -- --

Independent risk factorsIndependent risk factors

Standard methods of outcome assessment after heart valve operations

Descriptives:Descriptives: Describe the patient and procedure characteristicsDescribe the patient and procedure characteristics

Number of deaths/events (early and late)Number of deaths/events (early and late)

Incidence rates of late events (number/year, Weibull,…)Incidence rates of late events (number/year, Weibull,…)

Describe modes of failure, clinical status, last echoDescribe modes of failure, clinical status, last echo

Logistic regression:Logistic regression: To assess factors that may influence early outcome, OR is calculatedTo assess factors that may influence early outcome, OR is calculated

Univariate versus multivariateUnivariate versus multivariate

Kaplan- Meier analysis (time-to event model in the presence of censored cases): To describe freedom from death/events

Comparison of KM-estimates: log rank test

Cox regression:Cox regression: To assess factors that may influence outcome over time, HR is calculatedTo assess factors that may influence outcome over time, HR is calculated

Univariate versus multivariateUnivariate versus multivariate

Example of a KM cumulative survival graph

Cumulative survival after subcoronary implantation versus root replacement

0

0,1

0,2

0,3

0,4

0,5

0,6

0,7

0,8

0,9

1

0 1 2 3 4 5 6 7 8 9

Time (years since operation)

Cum

ulat

ive

surv

ival

subcoronary implantation

root replacement

At Risk (N): SC ARR

1 year 89 1263 years 87 865 years 81 547 years 67 259 years 44 4

Always mention number of patients at

risk over time!

If <10% is still at risk estimates are no

longer valid

Log-rank test: p<0.01

Standard methods of outcome assessment after heart valve operations

Descriptives:Descriptives: Describe the patient and procedure characteristicsDescribe the patient and procedure characteristics

Number of deaths/events (early and late)Number of deaths/events (early and late)

Incidence rates of late events (number/year, Weibull,…)Incidence rates of late events (number/year, Weibull,…)

Describe modes of failure, clinical status, last echoDescribe modes of failure, clinical status, last echo

Logistic regression:Logistic regression: To assess factors that may influence early outcome, OR is calculatedTo assess factors that may influence early outcome, OR is calculated

Univariate versus multivariateUnivariate versus multivariate

Kaplan- Meier analysis (time-to event model in the presence of censored cases):Kaplan- Meier analysis (time-to event model in the presence of censored cases): To describe freedom from death/eventsTo describe freedom from death/events

Comparison of KM-estimates: log rank testComparison of KM-estimates: log rank test

Cox regression: To assess factors that may influence outcome over time, HR is calculated

Univariate versus multivariate

Cox regression analysis

Is simply logistic regression that has time as a covariable

Therefore it allows study of factors that may influence the

occurrence of complications (death/valve-related events) over time

Outcome measure = hazard ratio (HR)

Newer statistical methods

Actual versus actuarial (KM) method: The Kaplan Meier method is in general very useful for describing outcome over time

However, when a non-fatal event is described by means of the KM method there is

an overestimate of the risk that a patient may experience this event.

Why? Because the patient may die before the event occurs

The actual method corrects for the competing risk of death

Important:

The actual risk can be misused to make valve performance look betterThe actual risk can be misused to make valve performance look better

Actuarial method: describes valve outcomeActuarial method: describes valve outcome

Actual method : describes patient outcome (and not valve performance!)Actual method : describes patient outcome (and not valve performance!)

Simulation methodsSimulation methods

Longitudinal data analysis (for example echo data)Longitudinal data analysis (for example echo data)

Beware: Too much information!

More information or a copy of this presentation:

E-mail:

j.j.m.takkenberg@erasmusmc.nlOr download this presentation at:

www.cardiothoracicresearch.nl(first register)