J Clin Pathol 1988;41:370-377 Use of morphometry in cytological preparations for diagnosing follicular non-Hodgkin's lymphomas M W STEVENS, K S CROWLEY, N L FAZZALARI, A E WOODS From the Division of Tissue Pathology, Institute of Medical and Veterinary Science, Adelaide, South Australia, and the South Australian Institute of Technology North Terrace, Adelaide, South Australia SUMMARY Morphometric techniques were applied to cytocentrifuge smears from 27 patients with histologically confirmed follicular non-Hodgkin's lymphoma to determine the usefulness of this method in diagnosis. Analysis of quantitative data confirmed that most subtypes were classified correctly on the basis of the proportion of large cells, small cleaved, and large non-cleaved cells, and nuclear, nucleolar, and cytoplasmic features. Morphometry is a more objective and reproducible technique than manual cell counting methods, and the use of cytocentrifuge smears has several advantages compared with conventional histological sectioning. Data derived with the aid of morphometry may be of value in follow up clinicopatho- logical studies. The Working Formulation for Clinical Usage separates follicular or nodular lymphomas into three distinct morphological subtypes.' Follicular, predominantly small cleaved cell, and mixed (small cleaved and large cell) are low grade tumours, sensitive to chemotherapy, but associated with short lived complete remission. Patients are often managed by adopting a "watch and see" approach.2 Follicular mixed lymphoma is often characterised by a more aggressive clinical course than its counterparts within the low grade category, however, and aggressive treatment with curative intent is controversial.2 Con- versely, more labile clinical features, including a shorter mean survival, are a feature of follicular, predominantly large cell lymphoma, and this subtype is classified by the Working Formulation to be of intermediate grade. According to the Working Formulation, patients with intermediate grade lymphoma are usually treated with multiagent chemotherapy regimens on initial presentation. Clearly, there are degrees of low, intermediate, and high grade lymphoma, and the proper management of patients, with appropriate choice of treatment regimens, relies on adequate clinical assessment and correct pathological diagnosis.2 The difficulties which pathologists experience in providing a reproducible subjective evaluation was highlighted in a recent study of follicular lymphomas,5 with major disagreement Accepted for publication 16 December 1987 evident in 37% of cases. The authors of this study emphasised the inherent difficulties in making a correct diagnosis including determination of the proportion of large cells, classification of cells into small, intermediate, and large size categories, and differentiation between cleaved and non-cleaved cells. To assist pathologists in the categorisation of these lymphomas more objective methods including the manual estimation method of Berard67 and semiautomated techniques8 have been proposed. Dar- dick et al analysed morphometrically derived data, obtained on the basis of nuclear variables in his- tological sections from 54 cases of follicular centre cell lymphoma, to determine whether subclassification could be achieved using this technique.8 To further assess the usefulness of morphometric techniques in the classification of these lymphomas we extended the study to incorporate not only nuclear data but quantitative information relating to nucleolar and cytoplasmic variables. Furthermore, to minimise problems inherent in quantitative studies of his- tological sections, semiautomated image analysis was performed on cytocentrifuge preparations, obtained from cell suspensions of representative lymphoid tissue. Material and methods Cytological specimens from patients with his- tologically confirmed non-Hodgkin's lymphoma were analysed. The histological diagnosis was made in 370 on July 18, 2021 by guest. Protected by copyright. http://jcp.bmj.com/ J Clin Pathol: first published as 10.1136/jcp.41.4.370 on 1 April 1988. Downloaded from
Transcript
J Clin Pathol 1988;41:370-377
Use of morphometry in cytological preparations fordiagnosing follicular non-Hodgkin's lymphomasM W STEVENS, K S CROWLEY, N L FAZZALARI, A E WOODS
From the Division of Tissue Pathology, Institute ofMedical and Veterinary Science, Adelaide, South Australia,and the South Australian Institute of Technology North Terrace, Adelaide, South Australia
SUMMARY Morphometric techniques were applied to cytocentrifuge smears from 27 patients withhistologically confirmed follicular non-Hodgkin's lymphoma to determine the usefulness of thismethod in diagnosis. Analysis of quantitative data confirmed that most subtypes were classifiedcorrectly on the basis of the proportion of large cells, small cleaved, and large non-cleaved cells, andnuclear, nucleolar, and cytoplasmic features.Morphometry is a more objective and reproducible technique than manual cell counting methods,
and the use ofcytocentrifuge smears has several advantages compared with conventional histologicalsectioning. Data derived with the aid of morphometry may be of value in follow up clinicopatho-logical studies.
The Working Formulation for Clinical Usageseparates follicular or nodular lymphomas into threedistinct morphological subtypes.' Follicular,predominantly small cleaved cell, and mixed (smallcleaved and large cell) are low grade tumours, sensitiveto chemotherapy, but associated with short livedcomplete remission. Patients are often managed byadopting a "watch and see" approach.2 Follicularmixed lymphoma is often characterised by a moreaggressive clinical course than its counterparts withinthe low grade category, however, and aggressivetreatment with curative intent is controversial.2 Con-versely, more labile clinical features, including ashorter mean survival, are a feature of follicular,predominantly large cell lymphoma, and this subtypeis classified by the Working Formulation to be ofintermediate grade. According to the WorkingFormulation, patients with intermediate gradelymphoma are usually treated with multiagentchemotherapy regimens on initial presentation.
Clearly, there are degrees of low, intermediate, andhigh grade lymphoma, and the proper management ofpatients, with appropriate choice of treatmentregimens, relies on adequate clinical assessment andcorrect pathological diagnosis.2 The difficulties whichpathologists experience in providing a reproduciblesubjective evaluation was highlighted in a recent studyof follicular lymphomas,5 with major disagreement
Accepted for publication 16 December 1987
evident in 37% of cases. The authors of this studyemphasised the inherent difficulties in making acorrect diagnosis including determination of theproportion of large cells, classification of cells intosmall, intermediate, and large size categories, anddifferentiation between cleaved and non-cleaved cells.To assist pathologists in the categorisation of these
lymphomas more objective methods including themanual estimation method of Berard67 andsemiautomated techniques8 have been proposed. Dar-dick et al analysed morphometrically derived data,obtained on the basis of nuclear variables in his-tological sections from 54 cases of follicular centre celllymphoma, to determine whether subclassificationcould be achieved using this technique.8To further assess the usefulness of morphometric
techniques in the classification of these lymphomas weextended the study to incorporate not only nucleardata but quantitative information relating to nucleolarand cytoplasmic variables. Furthermore, to minimiseproblems inherent in quantitative studies of his-tological sections, semiautomated image analysis wasperformed on cytocentrifuge preparations, obtainedfrom cell suspensions of representative lymphoidtissue.
Material and methods
Cytological specimens from patients with his-tologically confirmed non-Hodgkin's lymphoma wereanalysed. The histological diagnosis was made in
370
on July 18, 2021 by guest. Protected by copyright.
http://jcp.bmj.com
/J C
lin Pathol: first published as 10.1136/jcp.41.4.370 on 1 A
Use ofmorphometry to diagnosefollicular non-Hodgkin's lymphomasaccordance with the Working Formulation for Clin-ical Usage' and the Rappaport scheme.9 They includedseven cases of follicular, predominantly small cleavedcell (nodular, poorly differentiated lymphocytic)(NPDL); nine cases of follicular, mixed small cleavedand large cell (nodular, mixed lymphocytic-"his-tiocytic" (NM); six cases of diffuse, predominantlylarge cell (DL) (diffuse "histiocytic" (DH); and fivecases of follicular, large cell (FL) (nodular "his-tiocytic") (NH). For the purposes of this study, casesof diffuse and follicular, predominantly large celllymphomas (DL + FL), were combined with those ofdiffuse and nodular "histiocytic" (DH + NH).
Histological diagnosis was made by an experiencedpathologist of this specialty (KSC).
Fresh lymphoid tissue was forwarded to thelaboratory with minimal delay, and before fixationwas divided for histological, immunological, andultrastructural studies. A representative sample oftissue, measuring 5-10 mm3 was collected into RPMI1640 and cytological smears were prepared aspreviously described.'" Briefly, cell suspensions wereobtained by gently teasing lymphoid tissue through a1 mm gauge wire mesh. The cell suspension wasresuspended in 30% bovine serum albumin andcytological smears were prepared with thecytocentrifuge (Shandon Scientific Ltd, London). Thecentrifuge was spun at 250 rpm for five minutes andthe preparation was air dried before staining withJenner's giemsa.Random non-overlapping fields of the cell prepara-
tion were photographed on 35 mm film and thephotographic negatives were projected by Leitzprojector on to a digitiser pad (HP9874A). Thenuclear, nucleolar, and cytoplasmic contours ofall lymphoid cells within the field, excluding those withdisrupted cell membranes, were traced. Themorphometric features and statistical variables (table1) were processed by a Hewlet-Packard 9000 (series200) computer, interfaced with the digitiser pad.To determine the optimal number of cells to be
analysed, computer generated plots of the cumulativemean of nuclear area were derived for each case; aTable I Morphometricfeatures and statistics
Feature Statistical variables
Nuclear area MeanNuclear shape* SDNucleolar area Skewness (skew)Nucleolar: nuclear
ratio t KurtosisNo nucleoli/nucleus 25th percentileCytoplasmic area 75th percentileNuclear: cytoplasmic
ratio
*Mean feret diameter/perimeter equivalent diameter, tarea of allnucleoli per cell/area of nucleus'; all statistics listed calculated foreach feature.
minimum of 150 cells were digitised and themeasurement procedure was completed when thefluctuation in the cumulative mean was within 5%over at least 50 separate measurements. The totalnumber of cells traced for each case ranged from 150-300 cells.
Quantitation of nuclear shape was achieved using ashape descriptor previously applied in the field ofpowder technology." Convexity-concavity (C-C) wasused as it has a number of advantages over conven-tional shape indices, such as the nuclear contour index(NCI)-divergent irregular shaped nuclear profiles arenot assigned similar index values.'2For the purposes ofthis study, a cleaved nucleus was
defined as one possessing one or more invaginationswhere the margins ofthe invagination were parallel, orthe angle subtended by the walls of the invaginationwas less than 30°. Only those invaginations showingclear separation of contours were traced, as we havefound poor reproducibility associated with digitisinglinear grooves or infoldings of the nuclear membrane(unpublished data). A nucleus with an assigned indexvalue of less than 0 9 was classified as cleaved.To exclude normal mature lymphocytes from the
cell population, a total of 250 cells present in six casesof non-Hodgkin's lymphoma were traced; these cellswere selected on the basis of morphological criteria.'3A mean nuclear area of 36-4 pm2 was obtained for thesample and the upper 95% confidence interval of 37-6gm' was chosen as the cut off point for exclusion ofthese morphologically benign cells. Likewise, a similarapproach was followed to select the threshold valuefor nuclear shape. Hence all small regular shaped cellswith a nuclear area of < 37-6 pm2 and a C-C of > 0-9were excluded from study; exclusion of these cells wasa continuous piocess during the collection of data.
Separation oflymphocytes into small and large cellswas achieved by comparing the nuclear size of lym-phoma cells with that of the tissue macrophage, acomparison that often assists pathologists in makingthis distinction. The nuclear contours of 100 macro-phages, selected on the basis of cytological criteria,"were traced. A mean nuclear area of 133 pm2 wasobtained for the population and the lower 95%confidence interval of 87 pm2 was chosen as the cut offpoint for separation of small and large cells.
Using nuclear size as a criterion, subdivision ofnodular lymphomas into follicular small cleaved cell(FSC), follicular mixed cell (FM), and diffuse andfollicular large cell (DL + FL) was made when theproportion of large cells was less than 20%, between20-50%, and greater than 50%, respectively.'4
Further separation of the cell population intocategories representing the main types of follicularcentre cells was achieved using data on nuclear areaand shape. Computer generated scatterplots were
371
on July 18, 2021 by guest. Protected by copyright.
http://jcp.bmj.com
/J C
lin Pathol: first published as 10.1136/jcp.41.4.370 on 1 A
372derived for each case and the relative proportions ofsmall cleaved, large cleaved, and large non-cleavedcells were calculated using the threshold valuespreviously mentioned.
Inter- and intraobserver variability were assessedfor the cell variables under investigation.Reproducibility was within 5% for all variables withthe exception of nucleolar area, which showedvariability in individual measurements of less than10%.
Multiple comparisons for the morphometricfeatures and statistics studied between the three sub-types of nodular lymphoma was performed by oneway analysis ofvariance (ANOVA). Where significantdifferences were shown to be present by the variance F-test (p < 0 05), the results were compared by theNewman-Keuls test.'5 Significance was set at p < 0-05and all differences referred to in the text met thiscriteria.
Results
Table 2 lists those morphometric features andstatistics that show significant differences (p < 0 05)between subtypes. Table 3 shows the mean (SEM) forthe raw data.
Analysis of the results provided useful informationregarding classification of these groups. Separation ofFSC compared with DL + FL and FM comparedwith DL + FL was achieved using several variables;fewer differences were noted between FSC and FMlymphomas. This finding was related to the varyingproportions oflarge cells that comprised each subtype.Fig I depicts the relation between histological diag-nosis and the percentage of large lymphoid cellsdetermined by morphometry, using the thresholdvalues defined by Warnke et al."4 On the basis of thiscriterion two cases each of FSC, FM, and DL + FLwere classified as FM, DL + FL, and FM, respec-tively. Moreover, there was a significant difference
100
50-
20-
0-0
Stevens, Crowley, Fazzalari, Woods
Diffuse andfollicularlarge cell
0S
0Follicularmixed cell
0
* 00
00
S0
Follicular smallcleaved cell
00
70
0
0S
Table 2 Significant morphometric features and statistics (p < 0-05)
Follicular small cleaved cell Follicular small cleaved cell Follicular mixed cellvfollicular mixed cell v diffuse andfollicular large cell v diffuse andfollicular large cell(nodular poorly-differentiated (nodular poorly-differentiated (nodular mixed v
Feature v nodular mixed) v diffuse andfollicular large cell) diffuse andfollicular large cell)
Nuclear area (uM2) SD Mean, SD, 25th, 75th Mean, SD, Skewness, 25th, 75th% large cells Mean Mean MeanNuclear shape% Small cleaved Mean Mean% Large non-cleaved Mean Mean Mean% Large cleaved
Nucleolar area (um2) Skewness, Kurt Mean, SD, 25th, 75th Mean, 25th, 75thNucleolar:nuclear ratio Mean, SD, 75th Mean, 75thCytoplasmic area (um2) SD, Skewness Mean, SD, 25th, 75th Mean, Skewness, Kurt, 25th, 75thNuclear:cytoplasmic ratio SD
000
Fig 1 Distribution ofproportion oflarge cells (nucleararea > 87 pmu2) for each case grouped according tolymphoma subtype.
I I
on July 18, 2021 by guest. Protected by copyright.
http://jcp.bmj.com
/J C
lin Pathol: first published as 10.1136/jcp.41.4.370 on 1 A
between the mean percentage of large cells present ineach subtype.
Separation of FSC compared with DL + FL andFM compared with DL + FL was achieved with mostof the statistical variables analysed for nuclear area.The large cell lymphomas (DL + FL) were character-ised by a range of cell types with a predominance ofcells with nuclear area measurements of > 87 pm2;hence cell distributions were typified by higher valuesfor mean, 75th percentile (fig 2), standard deviation,25th percentile, and with a lower value for kurtosis,suggesting a relatively flat distribution, although therewas no significant difference between subtypes for thisstatistic.
Follicular mixed cell lymphomas occupy an inter-mediate position between FSC and DL + FL lym-phomas, except for skewness and kurtosis, whichreflect nuclear area distributions more positivelyskewed, and with a more pronounced peak with
respect to FSC and DL + FL lymphomas. Follicular,small cleaved lymphomas comprise a monotonouspopulation of small cells, and standard deviation ofnuclear area was of value in differentiating thissubtype from FM and DL + FL lymphomas (fig 3).
Nuclear shape was useful in differentiating groups,and fig 4 depicts the proportion of small cleaved, largenon-cleaved, and large cleaved cells within each sub-type. The percentage of small cleaved cells wassignificantly different between FSC compared withFM and between FSC compared with DL + FL, withFSC lymphomas comprising a greater proportion(108 (SEM) (2 7%), of these cells. No significantdifferences were noted between FM and DL + FLsubtypes nor was the proportion of large cleaved cellsof value in differentiating groups.
Greater differentiation was achieved by analysingthe proportion of large non-cleaved cells within eachsubtype and a trend towards greater numbers of large
373
on July 18, 2021 by guest. Protected by copyright.
http://jcp.bmj.com
/J C
lin Pathol: first published as 10.1136/jcp.41.4.370 on 1 A
Fig 2 Distribution ofmean and 75th percentile ofnuclear area, as determinedfor each case and subtypeofnon-Hodgkin's lymphoma. (Mean SEM).
non-cleaved cells in diffuse and follicular large celllymphomas was discerned.
Large non-cleaved cells contain one to multiplenucleoli, often varying in size, and nucleolar datawould be expected to reflect this large cell component.This was confirmed by analysing the results for thenucleolar features: nucleolar area, nucleolar: nuclearratio, and the number of nucleoli in each nucleus.Lymphomas with increasing proportions oflarge non-cleaved cells showed nucleolar distributions that weremore skewed in favour of larger nucleoli, with increas-ing numbers of nucleoli in each nucleus. This showedan increased standard deviation and a less pronouncedpeak, reflecting the wide range of nucleolar sizeidentified. Only selected variables, however, weresignificant (table 2).The mean and 75th percentile of nucleolar area is
shown in fig 5. A case of FM lymphoma was typifiedby increased nucleolar measurements and these werein the upper range of those obtained for DL + FLlymphomas. Furthermore, this case was characterisedby increased nuclear measurements, including themean and 75th percentile of nuclear area (fig 2) andcomprised a greater proportion of large cells (fig 1).
Stevens, Crowley, Fazzalari, WoodsAs with nuclear area, fewer differences were noted
between FSC and FM lymphomas with respect tonucleolar data. The shape of the distribution,however, as described by skewness and kurtosis, wassignificantly different. Follicular mixed cell lym-phomas were characterised by nucleolar distributionsthat were more positively skewed in favour of largernucleoli and with a more pronounced peak, suggestinga greater incidence of both small and large nucleoli.
Differentiation of subtypes was also achieved withmeasurements of cytoplasmic area and results were
similar to those obtained for nuclear area. Differencesbetween groups were again related to the varyingproportions of small and large cells. Nuclear: cyto-plasmic ratio was unhelpful in classifying subtypes,but the standard deviation of this feature separatedFSC and FM lymphomas, as large cells with more
abundant cytoplasm are a feature ofFM lymphomas.
Discussion
The results ofour study confirm that morphometry is auseful adjunct in categorising follicular lymphomas,
70-
60-
50-
40-
30-
20-
10-
0O
Diffuse andfollicularlarge cell
0
0~
Follicularmixed cell * _
0
Follicular smallcleaved cell
'.S
Fig 3 Standard deviation ofnuclear area calculatedfor eachcase, grouped according to subtype. (Mean SEM).
374
I II
on July 18, 2021 by guest. Protected by copyright.
http://jcp.bmj.com
/J C
lin Pathol: first published as 10.1136/jcp.41.4.370 on 1 A
Use ofmorphometry to diagnose follicular non-Hodgkin's lymphomas 375Diffuse and mixed cell lymphomas occupying an intermediatefollicular position. Considerable overlap between groups waslarge cell discerned. Additional quantitative data can be
derived, however, from nuclear area measurements90 and the lymphoid population can be separated into
small and large cells by using the nuclear dimensions80 of tissue macrophages, which provides more meaning-
ful information than pooled mean data. Using the70- percentage of large cells as a criterion, two cases each
of FSC, FM, and DL + FL lymphoma were incorrec-. tly classified.
60- : We agree with Dardick et ar that case to case
FollicuLar 4 ~ evaluation of follicular centre cell lymphoma is impor-50- mixed cell tant and that this approach allows the pathologist to
identify the borderline cases. Moreover, the threshold40- values chosen for separation of subtypes may be
Follicular small refined in further studies which relate quantitative cellcleaved cell data to clinical variables including survival.
30- Nuclear shape data derived in this study were not
directly comparable with the findings of previous20- .- : investigations as we used a new shape descriptor for
0 _ . 1 : the quantitation ofnuclear shape. The shape index (C-10- T_4 C) is specific for invaginations where the nuclear
.- ir-:- . I ; * : " contour exhibits an abrupt deviation, as seen with0- cleaved profiles, and is less influenced by the overall
Nodular- Nodular Diffuse andpoorly mixed nodular
differentiated histiocyticFig 4 Distribution ofproportion ofcell types, as determinedby nuclear area and shape datafor each case and subtype.(Mean SEM).
although more differences were noted between FSCcompared with DL + FL and between FM comparedwith DL + FL than between FSC and FM subtypes.Differences between groups were related to the varyingproportions of large cells within each subtype, andquantitative data derived for nuclear, nucleolar, andcytoplasmic features substantiate these findings.Quantitation of the large cell component is importantas recent cell kinetic studies suggest that the propor-tion of these cells may determine responsiveness tochemotherapy, with subsequent prolonged diseasefree survival.3"67The distribution of nuclear area measurements for
cases in our study was similar to that obtained byDardick et al,8 who concluded that on the basis ofmorphological criteria follicular lymphomas representa continuum of one disease process, with nodular
4*0-
3.5 -
3-0-
2 5-
2-0-
1 5-
1i0-
Follicularmixed cell
I I
FRllicular small
Diffuse andfollicularlarge cell
I I
*-*T 4-cleavea celt T .
Mean 75th Mean 75th Mean 75th
_ .
0° *
Fig 5 Distribution ofmean and 75th percentile ofnucleolararea, as determinedfor each case and subtype ofnon-Hodgkin's lymphoma. (Mean SEM).
on July 18, 2021 by guest. Protected by copyright.
http://jcp.bmj.com
/J C
lin Pathol: first published as 10.1136/jcp.41.4.370 on 1 A
376 Stevens, Crowley, Fazzalari, Woodsshape of a nucleus as is the case with conventionalshape indices such as the NCI.'2Computer algorithms were devised to partition cells
into distinct categories on the basis ofnuclear size, andshape and information regarding the proportion ofsmall cleaved, large cleaved, and large non-cleavedcells was derived. These morphological terms wereoriginally proposed in the classification of Lukes andCollins`8 and recently incorporated in the WorkingFormulation.' As with nuclear area data we con-sidered this approach to provide more useful informa-tion than pooled data.
Analysis of the proportion of cell types within eachsubtype (fig 4) provided some interesting observations.There was a trend towards fewer numbers of smallcleaved cells in diffuse and follicular large cell lym-phomas with follicular small cleaved cell comprisingthe highest percentage of this cell type. About 10% ofcells in this subtype possessed nuclear invaginations,which met the criteria for classification as cleavedprofiles. It must be emphasised that only nuclearinvaginations, where intervening cytoplasm was dis-cerned, were traced; unacceptable inter- and intra-observer reproducibility was associated with digitisinglinear grooves or infoldings of the nuclear membrane.The proportion of small cleaved cells has been
shown to be underestimated on histological sections ofpatients with follicular small cleaved cell lymphomas.On the basis ofcomputer modelling studies, the 4-5%of cells classified as cleaved on sections actuallyrepresent 25-30% ofthe cell population.8 Our value of10% small cleaved cells is slightly higher than the 4-5% calculated by Dardick, as the cytocentrifugetechnique tends to accentuate subtle variations innuclear shape,'0 and tangentially sectioned profileswould be expected to display less morphological detailthan cells flattened on to a glass slide.`' This value isstill not an accurate representation of the small cellcomponent.The proportion of large non-cleaved cells differed
significantly between subtypes with an expected trendtowards greater numbers of these cells in diffuse andfollicular large cell lymphomas. Likewise, a similartrend was noted for large cleaved cells, although thesecells comprised a smaller percentage of the large cellpopulation. No significant difference in the proportionof these cells was discerned.
Small non-cleaved cells were not analysed in thisstudy as tumours composed predominantly of this celltype almost invariably present as a diffuse pattern.Follicular lymphomas of this cell type have beendescribed.' The nuclear dimensions of these cellsapproximate to that of a tissue macrophage, andfurther studies are required to formulate appropriatecomputer algorithms to differentiate these cells fromthe large non-cleaved population. Quantitative data
relating to mean nucleolar area, number of nucleoliper nucleus, and location of nucleoli within a nucleusmay be of value in making this distinction. Further-more, the upper and lower confidence limits, derivedfrom the nuclear area of tissue macrophages, could beselected as threshold values to define the size range forthis cell.The importance of nucleolar data as a prognostic
indicator is unclear, although nucleoli are identified inthose cells which have been known to have an adverseinfluence on survival. Nucleolar area distributions(table 2, fig 5) show features similar to those obtainedfor nuclear area with follicular small cleaved celllymphomas, characterised by fewer and smallernucleoli, ranging to greater numbers of larger nucleoliin the diffuse and follicular large cell subtype. Whetherthe case ofnodular mixed cell lymphoma, typified by agreater percentage of large cells with more prominentnucleoli, follows a more unfavourable clinical coursewill need to be determined in follow up clinicopatho-logical studies.Manual counting ofcell types in follicular centre cell
lymphomas forms the basis of the standard andmodified Berard methods, and this more objectiveassessment of the percentages of each cell type hasbeen found to be of value in classifying these lym-phomas.67 None the less a certain degree of subjectiveinterpretation is associated with these methods,including the distinction between small and large cellsand the classification of cells on the basis of nuclearshape.Morphometry is an objective and reproducible
method that may be of value to the pathologist inclassifying this disease entity. Moreover, the techniqueprovides quantitative data which may predict clinicaloutcome, may select those patients who require moreaggressive treatment, and may allow for more meanin-gful comparisons of the results of clinicopathologicaltrials between centres.
We thank Ms Erica Fairbanks for preparation of theillustrations.
References
1 National Cancer Institute Sponsored study of classification onnon-Hodgkin's lymphoma: summary and description of aworking formulation for clinical usage. The non-Hodgkin'spathologic classification project. Cancer 1982;49:21 12-35.
2 Gaynor ER, Ultmann JE. Non-Hodgkin's lymphoma: man-agement strategies. N Engi J Med 1984;6:1506-8.
3 Longo DL, Young RC, Hubbard SM, et al. Prolonged initialremission in patients with nodular mixed lymphoma. Ann InternMed 1984;100:651-6.
4 Glick JH, Barnes JM, Ezdinli EZ, Berard CW, Orlow EL, BennettJM. Nodular mixed lymphoma: results of a randomized trialfailing to confirm prolonged disease-free survival with COPPchemotherapy. Blood 1981 ;58:920-5.
5 Metter GE, Nathwani BN, Burke JS, et al. Morphological
on July 18, 2021 by guest. Protected by copyright.
http://jcp.bmj.com
/J C
lin Pathol: first published as 10.1136/jcp.41.4.370 on 1 A
Use ofmorphometry to diagnosefollicular non-Hodgkin's lymphomas 377subclassification of follicular lymphoma: variability of diag-nosis among hematopathologists, a collaborative study betweenthe Repository Center and Pathology Panel for Lymphoma andClinical Studies. J Clin Oncol 1985;3:25-38.
6 Mann RB, Berard CW. Criteria for the cytologic subclassificationof follicular lymphomas: a proposed alternative method.Hematol Oncol 1983;1:187-92.
7 Nathwani BN, Metter GE, Miller TP. What should be themorphologic criteria for the subdivision of follicular lym-phomas? Blood 1986;68:837-45.
8 Dardick I, Cardwell DR. Follicular center cell lymphoma:morphologic data relating to observor reproducibility. Cancer1986;58:2477-84.
9 Rappaport H. Tumours of the hemopoietic system. Atlas oftumour pathology, section 3, fascicle 8. Washington, DC: USArmed Forces Institute of Pathology, 1966:91-206.
10 Stevens MW, Fazzalari NL, Crisp DJ. Lymph node cellularmorphology: comparative study ofimprints and cytocentrifugesmears. J Clin Pathol 1987;40:751-5.
11 Tsubaki J, Jimbo G. A proposed &:w characterization of particleshape and its application. Powder Technology 1979;22:161-9.
12 Stevens MW, Fazzalari NL, Crisp DJ. Quantitation of nuclearshape in non-Hodgkin's lymphoma: alternate shape descrip-tors. Analyt Quant Cytol Histopathol (in press).
13 Lennert K. Malignant lymphomas other than Hodgkin's disease:histology, cytology, ultrastructure, immunology. Berlin,Springer-Verlag, 1978:33-71.
14 Warnke RA, Kim H, Fuks Z, Dorfman RF. The co-existence ofnodular and diffuse patterns in nodular non-Hodgkin'slymphomas. Significance and clinicopathologic correlations.Cancer 1977;40:1229-33.
15 Snedecor GW, Cochran WG. Statistical methods. 6th ed. Ames,USA: The Iowa State University Press, 1967:258-96.
16 Hansen H, Koziner B, Clarkson B. Marker and kinetic studies inthe non-Hodgkin's lymphomas. Am J Med 1981;71:107-23.
17 Hellman S, Reinke U, Batnick L, Mauch P. Functional organiza-tion of the hematopoietic stem cell compartment: implicationsfor cancer and its therapy. J Clin Oncol 1983;1:277-84.
18 Lukes RJ, Collins RD. Immunologic characterization of humanmalignant lymphomas. Cancer 1974;34:1488-503.
19 Ball PJ, Van der Valk P, Kurver PHJ, Lindeman J, Meijer CJLM.Large cell lymphoma II. Differential diagnosis of centroblasticand B-immunoblastic subtypes by morphometry on cytologicpreparations. Cancer 1985;55:486-92.
20 Haroske G, Ehlers U, Ehlers HM, Kunze KD. Cytomorphometriccharacterization of NHL (non-Hodgkin's) of low and highgrade malignancy. Archiv fur Geschwulstforschung 1984;54:41-51.
Requests for reprints to: Mr MW Stevens, Division ofTissuePathology, Institute of Medical and Veterinary Science, POBox 14, Rundle St, Adelaide, South Australia 5000.
on July 18, 2021 by guest. Protected by copyright.
http://jcp.bmj.com
/J C
lin Pathol: first published as 10.1136/jcp.41.4.370 on 1 A
