As Patrick noted, it looks as if its an issue with the scale on which the marker is measured.  For example, if your risk measure is in inches -- convert to feet;  if it is in minutes -- convert to hours;  if it is grams -- try converting to "per 100 grams".  By the way, this conversion will not have any effect on your model fit characteristics; significance testing; or confounding evaluation

Paul

Range of this marker is  from 1259.52 to 24541.05. Mean 9853.90.  I divided values with 10, 100, and 10000. and then I got HR 1.002, 1.020, HR 1.220 respectively. Is this appropriate?

There are two possibilities

1. May be large number of cases as compared to controls or vice versa. Or may be sufficiently large sample size.

2. It is better to categorize  this marker in meaningful way.

If still these are not possibilities then please ignore the significance.