bG (blood glucose) meters are not as accurate as the readings you get
from them imply. For example, you might think that 108 means 108 mg/dl,
not 107 or 109. But in fact all meters made for home use have at least a
10-15% error under ideal conditions. Thus you should interpret "108" as
"probably between 100 and 120". (Similar considerations apply if you
measure in units of mmol/L.) This is a random error and will not be
consistent from one determination to the next. You cannot expect to get
exactly the same reading from two checks done one after the other, nor
from two meters using the same blood sample.

This is generally considered acceptable because variations in this range
will not make a major difference in treatment decisions. For example,
the difference between 100 and 120 may make no difference in how you
treat yourself, or at most might make a difference of one unit of
insulin. With present technology, more accurate meters would be much
more expensive. This expense is only justified in research work, where
such accuracy might detect small trends which could go undetected with
less accurate measurements.

This discussion applies to ideal conditions. The error may be increased
by poor or missing calibration, temperatures outside the intended range,
outdated strips, improper technique, poor timing, insufficient sample
size, contamination, and probably other factors. Contamination is
especially serious since it can happen so easily and is likely to result
in an overdose of insulin. Glucose is found in fruits, juices, sodas,
and many other foods. Even a smidgen can seriously alter a reading.

When comparing meter readings with lab results, also note that plasma
readings are 15% higher than whole blood, and that capillary blood gives
different readings from venous blood.

Visually read strips are slightly less accurate than meters, with an
error rate around 20-25%.

For some meters, strips are available from manufacturers other than the
meter manufacturer. Some m.h.d. readers have compared the strips side-
by-side and found those from one manufacturer to read consistently lower
than the strips from another. The differences are not likely to make a
significant difference in your treatment, but are large enough to be
noticeable and possibly confusing. For this reason it is not a good idea
to change strip manufacturers without comparing the readings from one
with the readings from the other.

I've seen no such direct comparison of meters, but the possibility
exists that some meters might read consistently lower than others. Be
careful when changing meters.

By "error rate" I mean twice the standard deviation from the mean. An
error rate of 15% says that about 95% of the readings will be within 15%
of the actual value.