The D-score (D for discriminant) is derived from morphometric analysis of an endometrial biopsy showing hyperplasia. It can be used to predict the likelihood of progression to adenocarcinoma. Dr Baak is from the Free University Medical Centre in Amsterdam.


(1) 4 micron thick paraffin sections are prepared routinely and the most complex or atypical area is selected for analysis.

(2) Nuclear morphometry is performed on photomicrographs taken at 1,000x magnification (oil-immersion) and printed to exactly 3,500x. These are measured on a graphic tablet for circumference, area, longest axis and shortest axis. At least 50 and usually 100 nuclei from the selected area are analyzed.

(3) Glandular architecture is analyzed using a projection microscope and a morphometric point grid. This is analyzed using a stereologic software module to calculate 1, 2 and 3 dimensional histologic features.


Parameters measured (see Table I, page 336, Baak 1988):

(1) volume percent stroma, with units in percent

(2) standard deviation for the shortest nuclear axis, with units in microns

(3) glands outer surface density, with units in square mm per cubic mm


D –score =

= 0.6229 + (0.0439 * (volume percent stroma)) – (3.9934 * (LN(standard deviation of the shortest nuclear axis))) – (0.1592 * (outer surface density of the glands))



• minimum score: -4 (little stroma, highly variable nuclei, very glandular)

• maximum score: +4



Risk of Endometrial Adenocarcinoma

<= 0.0

high risk (unfavorable)

> 0.0 to 1.0

uncertain risk

> 1.0

low risk (favorable)


index of likelihood for NO progression to cancer (from Baak 1988, page 340) =

= 4.5 * EXP((0.57 * (D-score)) – 0.11)



• The index is an exponential curve that ranges from 0 (at D-score of –4) to 32 (at a D-score of +4).



• The D-score is highly reproducible between observers, but this assumes that the pathologists are using exactly the same analytical set-up (morphometric grid, workstation, software, etc.). This would be expected from the level of automation required. The only human variables would be the selection of the site for analysis.

• The probability of misclassification is given in Table II, page 340 Baak 1988 for various D scores.



• The requirement for a morphometric computer workstation and quite a bit of time would preclude the use of this score in routine practice.

To read more or access our algorithms and calculators, please log in or register.