proc mixed data = data;
  class usubjid trtpn avisitn;
  model chg = trtpn|avisitn base / ddfm=KENWARDROGER solution covb;
  repeated avisitn / subject=usubjid(trtpn) type=ar(1);
  lsmeans trtpn*avisitn / diff cl;
run;

proc mixed data = data;
  class avisitn trtpn usubjid;
  model chg = trtpn|avisitn base / ddfm=KENWARDROGER solution covb;
  repeated avisitn / subject=usubjid(trtpn) type=ar(1);
  lsmeans trtpn*avisitn / diff cl;
run;