FYI . 

Cwt is current weight
bwt is base weight. 



data test;
input subject visit cwt bwt 8.;
datalines;
1 1 78 80
1 2 79 .
1 3 70 .
1 4 72 .
1 5 74 .
1 6 88 .
1 7 70.
2 1 51 50
2 2 58 .
;
run;

proc sort data=test;
by subject visit;
run;

data test1;
set test;
by subject visit;
f= first.subject;
if f ne 1 then call missing(bwt);
if n(cwt, bwt) = 2 then pdiff = round((cwt - bwt)*100/bwt,0.1);
if . < pdiff <= 10 then dwt = bwt; else if pdiff > 10 then dwt = cwt;
run;

data test2;
do i = 1 by 1 until(last.subject);
set test1;
array bbwt bwt1 - bwt4;

  bbwt(i) = lag(dwt);
  by subject visit;
  if f ne 1 then bwt = bbwt(i);
if n(cwt, bwt) = 2 then pdiff = round((cwt - bwt)*100/bwt,0.1);
if . < pdiff <= 10 then  dwt = bwt; else if pdiff > 10 then  dwt = cwt; 
output;

end;
run;