A COMPARATIVE STUDY O F T H E ACCURACY O F A R E A ASSESSMENT BY P O I N T C O U N T I N G F O R B R O N C H I A L MUCOUS GLANDS
P. W. LORD,B. J. BILES,W. MALINOWSKA* AND W. F. WHIMSTER* Clinical Section, M R C Toxicology Unit, St. Bartholomew's Hospital Medical College, Charterhouse Square, London, E C l M 6BQ * Department of Pathology, King's College Hospital Medical School, Denmark Hill, London SE5
PLATES XI11 AND XIV T H Edetermination of the distribution of mucous glands in the walls of human airways is of importance in relation to the causes and incidence of diseases characterised by abnormal mucus production. In this paper we describe a preliminary study of the method of gland area assessment to be used with histological preparations of human bronchial tissue. Our exploratory work showed that there were serious practical errors in applying the counting technique (Weibel, 1963) to area estimations of gland. At different magnifications and with different stains there were variations in the appearance of the tissue which gave rise to large differences in the estimates of gland area, even from adjacent sections of the same tissue block. Differences in area of the order of 20 per cent. between high and low magnifications were obtained, since using the former, inter-acinar tissue was excluded and using the latter it was included in the count (even when examining the same section). Similar differences (of the order of 15 per cent.) due to the same cause were noted by Hale, Olsen and Mickey (1968). These differences were so large that we could not neglect them. Though we considered the tissue definition problem to be the major cause of these discrepancies, the counting method per se was also carefully examined. When using histological preparations errors in the assessment of the amount of bronchial gland in vivo are due to three major factors: (a) The differential dimensional changes in the tissues from life to histological section which are due to processing. These will not be discussed here, but will be the subject of another paper. (b) Error due to the difficulty of defining the area to be measured when it is examined under the microscope. At low magnifications detail in the glands cannot be seen, and the glands often appear as homogenous units (fig. la); at high magnifications acini of secretory cells can be distinguished, together with inter-acinar tissues (fig. lb). At both high and low magnifications the appreciation of the position of the boundaries of both whole gland and of the acini Received 30 Dec. 1911; accepted 17 March 1978 J . PATH.-VOL.
126 (1978)
45
46
P . W .LORD, B. J. BILES, W . MALINO WSKA AND W . F. WHIMSTER
is subject to uncertainties due to differences between histological stains, to the degree of autolysis in the tissue, to the morphology of the gland, and to different interpretations of these factors by different observers. (c) Having defined the tissue there remain the variations in area measurement inherent in the method employed. For the point counting procedure described by Weibel well-known error formulae are cited. In practice however there is deviation from the theoretical error analysis due to the non-random distribution of gland tissue over the counting field. To eliminate the major problem of defining the edge of the tissue, the point counting method was assessed on a series of clearly outlined drawn shapes, and the results are presented in part 1. In Part 2 those results are compared with results from actual tissue sections in which errors due to inadequate gland edge definition and to different observers were examined. MATERIALS AND APPARATUS The point counting procedure was investigated using 24 drawn shapes representing the outlines of whole mucous glands, and one of these shapes was subdivided into 4, 16 and 64 sub-areas to represent gland acini. These shapes were reduced photographically by a factor of 47-75 (linear reduction) and the negative was mounted on a microscope slide. Post-mortem specimens were used for the investigation of the influence of histological stain and observer variation in gland area determination. From the cartilaginous region of trachea and the two main bronchi longitudinal slices of tissue were used and from the smaller airways transverse slices were taken. The tissues were dehydrated, cleared, embedded in wax and 5 pm sections were cut, mounted and stained. Four stains were employed : (a) RAB: Reticulin (Gordon and Sweet), followed by Alcian Blue (pH 2-6) (to stain mucin) with Tartrazine as a counterstain (fig. 2a). Used to show glandular and acinar basement membranes and intra-cellular acid mucus. (b) HE: Haematoxylin (Coles) and Eosin (fig. 2b). (c) PAS: Periodic Acid Schiff (to stain mucin) with Light Green as a counterstain (fig. 2c). (d) PBS: A lead sulphide method, where, after the dewaxed tissue had been impregnated with 1 per cent. aqueous lead acetate for two minutes the lead was precipitated as the sulphide by treatment with 5 per cent. sodium sulphide for 15 seconds. The sections were then washed and mounted in gelatine, which reduced the fading of the stain which occurred with standard mounting media. Initial differential absorption of the lead salt resulted in different staining densities in the various tissue components. Cartilage and mucous gland were shown dark brown against the paler connective tissue (fig. 2d). This last stain was used as it gives a transmission picture of a wax-block surface-stain method (Hegre and Brashear, 1947) used in parallel studies. The microscope used was a Leitz Ortholux equipped with a 0.5 mm squared eyepiece graticule, one point representing 0.149 per cent. of the microscope field. Sections were examined using x25, x 10, x 6.3, x4, ~ 2 . 5 ,x 1 nominal objectives. Drawings of the selected glands were made using a Leitz drawing arm which was also used in the assessment of the areas of the drawn shapes used in Part 1 . The areas of these drawn shapes and of drawings of gland tissue were measured independently using a carefully calibrated planimeter, and their perimeters were determined using an opisometer. Part I . Assessment of the Point Counting Procedure Point Counting of drawn shapes. The number (N) of points of intersection in the 0.5 mm grid which were seen superimposed on the image of the tissue of interest was counted. The
MUCOUS GLAND MORPHOMETRY
47
area of the image was proportional to N/M where M was the total number of points superimposed on the field by the eyepiece graticule. We assessed this method for bias by comparing area estimates with planimetry, and for variability. Variation between counts over the same microscope field were due to the position and orientation of the grid relative to the image being measured (Weibel, 1963). The standard description of statistical variation assumes the random distribution of the tissue of interest over the microscope field. Whole mucous glands are not randomly distributed; fig. la shows that a gland is a unit (or a group of units). The acini (fig. Ib), of which a gland is composed, present a somewhat more dispersed pattern, but are still non-random. All the counts described in this paper refer to single microscope fields. However because mucous glands are not evenly distributed in bronchial walls the total gland surrounding the airways will be assessed in our future work. This will require the use of multiple fields and the use of these is discussed later. Comparison of point counts with planimetry. The 24 drawn areas were each assessed five times by planimetry and five times by point counting using the drawing arm. The magnification in the graticule plane was x 0.0962. Before each of the five repeated point counts the eyepiece graticule was rotated a few degrees and the object was also moved a little-such repeated counts are called replicates. The drawn areas filled between 0.18 per cent. and 39.5 per cent. of the microscope field, and the means of five counts ranged from 1.2 to 265.6. Very close correlation between point counting and planimetry was obtained (r = 0.9999, p
P. W. LORD,B. J. BILES,W. MALINOWSKA* AND W. F. WHIMSTER* Clinical Section, M R C Toxicology Unit, St. Bartholomew's Hospital Medical College, Charterhouse Square, London, E C l M 6BQ * Department of Pathology, King's College Hospital Medical School, Denmark Hill, London SE5
PLATES XI11 AND XIV T H Edetermination of the distribution of mucous glands in the walls of human airways is of importance in relation to the causes and incidence of diseases characterised by abnormal mucus production. In this paper we describe a preliminary study of the method of gland area assessment to be used with histological preparations of human bronchial tissue. Our exploratory work showed that there were serious practical errors in applying the counting technique (Weibel, 1963) to area estimations of gland. At different magnifications and with different stains there were variations in the appearance of the tissue which gave rise to large differences in the estimates of gland area, even from adjacent sections of the same tissue block. Differences in area of the order of 20 per cent. between high and low magnifications were obtained, since using the former, inter-acinar tissue was excluded and using the latter it was included in the count (even when examining the same section). Similar differences (of the order of 15 per cent.) due to the same cause were noted by Hale, Olsen and Mickey (1968). These differences were so large that we could not neglect them. Though we considered the tissue definition problem to be the major cause of these discrepancies, the counting method per se was also carefully examined. When using histological preparations errors in the assessment of the amount of bronchial gland in vivo are due to three major factors: (a) The differential dimensional changes in the tissues from life to histological section which are due to processing. These will not be discussed here, but will be the subject of another paper. (b) Error due to the difficulty of defining the area to be measured when it is examined under the microscope. At low magnifications detail in the glands cannot be seen, and the glands often appear as homogenous units (fig. la); at high magnifications acini of secretory cells can be distinguished, together with inter-acinar tissues (fig. lb). At both high and low magnifications the appreciation of the position of the boundaries of both whole gland and of the acini Received 30 Dec. 1911; accepted 17 March 1978 J . PATH.-VOL.
126 (1978)
45
46
P . W .LORD, B. J. BILES, W . MALINO WSKA AND W . F. WHIMSTER
is subject to uncertainties due to differences between histological stains, to the degree of autolysis in the tissue, to the morphology of the gland, and to different interpretations of these factors by different observers. (c) Having defined the tissue there remain the variations in area measurement inherent in the method employed. For the point counting procedure described by Weibel well-known error formulae are cited. In practice however there is deviation from the theoretical error analysis due to the non-random distribution of gland tissue over the counting field. To eliminate the major problem of defining the edge of the tissue, the point counting method was assessed on a series of clearly outlined drawn shapes, and the results are presented in part 1. In Part 2 those results are compared with results from actual tissue sections in which errors due to inadequate gland edge definition and to different observers were examined. MATERIALS AND APPARATUS The point counting procedure was investigated using 24 drawn shapes representing the outlines of whole mucous glands, and one of these shapes was subdivided into 4, 16 and 64 sub-areas to represent gland acini. These shapes were reduced photographically by a factor of 47-75 (linear reduction) and the negative was mounted on a microscope slide. Post-mortem specimens were used for the investigation of the influence of histological stain and observer variation in gland area determination. From the cartilaginous region of trachea and the two main bronchi longitudinal slices of tissue were used and from the smaller airways transverse slices were taken. The tissues were dehydrated, cleared, embedded in wax and 5 pm sections were cut, mounted and stained. Four stains were employed : (a) RAB: Reticulin (Gordon and Sweet), followed by Alcian Blue (pH 2-6) (to stain mucin) with Tartrazine as a counterstain (fig. 2a). Used to show glandular and acinar basement membranes and intra-cellular acid mucus. (b) HE: Haematoxylin (Coles) and Eosin (fig. 2b). (c) PAS: Periodic Acid Schiff (to stain mucin) with Light Green as a counterstain (fig. 2c). (d) PBS: A lead sulphide method, where, after the dewaxed tissue had been impregnated with 1 per cent. aqueous lead acetate for two minutes the lead was precipitated as the sulphide by treatment with 5 per cent. sodium sulphide for 15 seconds. The sections were then washed and mounted in gelatine, which reduced the fading of the stain which occurred with standard mounting media. Initial differential absorption of the lead salt resulted in different staining densities in the various tissue components. Cartilage and mucous gland were shown dark brown against the paler connective tissue (fig. 2d). This last stain was used as it gives a transmission picture of a wax-block surface-stain method (Hegre and Brashear, 1947) used in parallel studies. The microscope used was a Leitz Ortholux equipped with a 0.5 mm squared eyepiece graticule, one point representing 0.149 per cent. of the microscope field. Sections were examined using x25, x 10, x 6.3, x4, ~ 2 . 5 ,x 1 nominal objectives. Drawings of the selected glands were made using a Leitz drawing arm which was also used in the assessment of the areas of the drawn shapes used in Part 1 . The areas of these drawn shapes and of drawings of gland tissue were measured independently using a carefully calibrated planimeter, and their perimeters were determined using an opisometer. Part I . Assessment of the Point Counting Procedure Point Counting of drawn shapes. The number (N) of points of intersection in the 0.5 mm grid which were seen superimposed on the image of the tissue of interest was counted. The
MUCOUS GLAND MORPHOMETRY
47
area of the image was proportional to N/M where M was the total number of points superimposed on the field by the eyepiece graticule. We assessed this method for bias by comparing area estimates with planimetry, and for variability. Variation between counts over the same microscope field were due to the position and orientation of the grid relative to the image being measured (Weibel, 1963). The standard description of statistical variation assumes the random distribution of the tissue of interest over the microscope field. Whole mucous glands are not randomly distributed; fig. la shows that a gland is a unit (or a group of units). The acini (fig. Ib), of which a gland is composed, present a somewhat more dispersed pattern, but are still non-random. All the counts described in this paper refer to single microscope fields. However because mucous glands are not evenly distributed in bronchial walls the total gland surrounding the airways will be assessed in our future work. This will require the use of multiple fields and the use of these is discussed later. Comparison of point counts with planimetry. The 24 drawn areas were each assessed five times by planimetry and five times by point counting using the drawing arm. The magnification in the graticule plane was x 0.0962. Before each of the five repeated point counts the eyepiece graticule was rotated a few degrees and the object was also moved a little-such repeated counts are called replicates. The drawn areas filled between 0.18 per cent. and 39.5 per cent. of the microscope field, and the means of five counts ranged from 1.2 to 265.6. Very close correlation between point counting and planimetry was obtained (r = 0.9999, p
