44 |
Klinisk Biokemi i Norden · 1 2015
the degree of change from a given INR (CDs). In
patient B with an unexpected supra-therapeutic INR
(INR 5.9 in paper I and 4.8 in paper II), the varia-
tion in dose reduction during the two first days and
number of days until a new INR measurement after
the supra-therapeutic INR was large. No correlation
with the physicians’ 48-hour bleeding risk estimate
was found. Large variations within countries were
also found regarding the suggested new weekly dose
and time until a new INR measurement after an INR
reduction to 2.9. Answers on the practical handling of
the patients varied substantially within and between
countries regardless of level of care and type of VKA
used. The variation was especially large regarding
risk estimate of bleeding and ischemic stroke in atrial
fibrillation with and without VKA treatment and risk
estimates of bleeding after a supra-therapeutic INR,
and was substantially overestimated by doctors both
in primary and secondary care.
Use of standardised care was not prevalent among
the responding physicians with the exception of the
Netherlands and the United Kingdom where anti-
coagulation clinics and computer dosing are used
to some extent and in Denmark and Norway where
about 50% used manual dosing algorithms. Very
few within-country differences were found between
those using clinical experience versus users of com-
puter dosing programs or manual dosing algorithms,
but only 27% of those who stated they use manual
algorithms or computer dosing programs in clinical
practice, actually used them when answering the
questionnaire.
Within-subject variation of D-dimer in preg-
nancy
During pregnancy there is a change in haemostasis
towards a more hypercoagulable state, probably to
decrease the risk of bleeding during delivery. The aim
of paper III was to develop a method to calculate bio-
logical variation in constituents which change during
the period studied, as exemplified with D-dimer
during pregnancy. D-dimer was chosen since a bet-
ter estimate of changes in D-dimer during pregnancy
and post-partummight be important when assessing
the probability of suspected VTE in pregnant women.
D-dimer was measured every 4
th
week in 20 pregnant
and 19 non-pregnant women to calculate biological
variation. The D-dimer concentrations were transfor-
med by dividing each D-dimer value by the median
for the corresponding pregnancy period, creating
“multiples of the median”, MoM, and thereafter
by calculating the natural logarithms of the MoM
results (lnMoM). Analysis of variance (ANOVA)
with the statistical model for repeated sub-sampling
(nested design) was used to calculate the average
within-subject variation (SDws), between-subject
variation (SD
BS
) and within-series analytical varia-
tion (SD
AW
) for the transformed D-dimer values. The
median D-dimer concentrations increased five-fold
during pregnancy, to values high above the diagnostic
cut-off value for VTE, peaked in conjunction with
delivery and then gradually decreased post-partum
(Fig. 2A). The considerable change in D-dimer was
compensated for by the transformation to MoM,
which established a kind of steady-state (Fig 2B).
However, only values for lnMoMwere normally dist-
ributed and showed variance homogeneity, for which
performance characteristics for biological variation
was estimated (Fig 2C). Within-subject variation in
pregnancy (SD
WS
), for lnMoMD-dimer, was estima-
ted to 0.27 (95%CI 0.24 - 0.32), while between-subject
variation (SD
BS
) was slightly higher (0.35, 95%CI 0.26
- 0.54)). These values were not significantly changed
if calculated in pregnancy and postpartum combined
or if calculated in non-pregnant women. Index of
individuality (II) and reference change values (RCV)
were also comparable in pregnant and non-pregnant.
Summary
Large variation both within and between countries
were found in practical handling of patients trea-
ted with VKAs. This may be caused by the limited
use of standardised algorithms for this treatment.
Implementation of manual dosing algorithms and
computer dosing programs may improve the quality
of care, especially in practices with little experience
in this field.
Biological variation can be calculated in some non-
steady state situations (e.g. D-dimer in pregnancy)
after transformation of the original data, which
might be important clinically in assessing changes
in D-dimer. Further studies should be done to eva-
luate the method of calculating biological variation in
non-steady state situations, both to explore its general
validity and especially its use in pregnant women with
suspected VTE.
