Kaplan Meier for left-censored data

[Clemence]

Hello,

I study data non detects and different methods for handling these data. I would like obtain the mean using Kaplan Meier method, but my data are left censored and not right censored. So, I found a code to estimate the mean, but when I compare the result of mean with a function in R for non detect data (package NADA, function cenfit), I don't find again the result. Do you think that this code is correct to calculate the mean ? Or, do you know a code to obtain the mean with Kaplan Meier method for left-censored data ?

Here is my code :

``` 

/**Put data to right censored**/

data KM;
if _N_=1 then set Max (keep = MAXC);
SET data.savona;
value_censure_droite = MAXC - Concentration;
run;

/**Estimator KM  : censored=1 if data censored and 0 if data observed**/
proc lifetest data=KM outsurv=out_km method=KM ;
time value_censure_droite*censored(1);
run;

proc sort data=out_km;
by SURVIVAL;
run;

data out_km1;
if _N_=1 then set MAX;
set out_km;
value_initiale = MAXC-value_censure_droite;

if _CENSOR_ IN (0,.);
DENS=DIF(SURVIVAL); 

if _N_ = 1 then DENS = SURVIVAL;

XX=LAG(value_initiale);
if _N_ = 1 then XX=0;

MPROD = XX*DENS;
run;

PROC MEANS DATA=out_km1 noprint;
VAR MPROD;
OUTPUT OUT=MMM SUM=MEAN; 
RUN;

```

Thanks for your help !

PS : When I run the proc lifetest, I have this note "The mean survival time and its standard error were underestimated because the largest observation was censored and the estimation was restricted to the largest event time."

[Clemence]

Nobody have a response ? It's for my internship .  Thanks !