Professional Documents
Culture Documents
Leo T. Chylack, Jr, MD; John K. Wolfe, MD; David M. Singer, MD; M. Cristina Leske, MD; Mark A. Bullimore, MD; Ian L. Bailey, MD;
Judith Friend, MD; Daniel McCarthy, MD; Suh-Yuh Wu, MD; for the Longitudinal Study of Cataract Study Group
>
P. Isolated water clefts, vacuoles, retrodots,
lamellar separations, and sutural opacities
and retest sessions for NO/NC and C/P
groups of slides. Grading sessions were typ¬
-3
are ignored (not graded). If, however, such ically separated by 1 to 2 weeks.
irregularities are clustered and organized Each set of slides was graded by the same
2 into discrete arrays, they should be graded two experienced graders. Each grader wrote
1 as C. In many retroillumination images, the a score on a separate standardized form; the
O two graders then compared scores and ar¬
„
0
"O peripheral portions of the image manifest
5 I soft, radially oriented areas of variable rived at a consensus score that was record¬
contrast—almost like very soft opacities; ed on a third standardized form.
these zones lack the sharp, discrete edges of Method of Analysis.—We analyzed the
Fig 4.—Relationship between grades of pos¬ most cortical opacities. These soft hazy zones repeatability of LOCS III gradings and esti¬
terior subcapsular cataract (P) with the sub¬ should be ignored. We have not been able to mated the 95% tolerance limits for defining
jective Lens Opacities Classification System grade these areas consistently. change. For each photograph and for each
III and the objective measure of percentage To Grade P.—Only posteriorly focused characteristic, two grades may be compared
of area with opacity. The relationship be¬ retroillumination images are used in grading (either grades from each observer in the
tween percentage of area with opacity and P. The area of the opacity in the lens being same session or grades from different ses¬
the subjective grade is monotonie. The graded is compared with that in standards 1 sions determined by the same observer). The
square root of the percentage of area with through 5. The standard interval selected discrepancy or difference between the two
opacity is linearly related to the subjective should bracket the opacity of the ungraded scores can be simply calculated. The distri¬
grade. image. The assigned decimal grade should bution of these discrepancies is then plotted,
reflect the location of the cataract in the in¬ and the SD of the discrepancy distribution
terval; the assigned score may range from 0.1 provides a measure of the reproducibility of
to 5.9. Opacities visible only in the posterior¬ the grading. We defined change as a differ¬
ly focused image are graded as if they are ence greater than a specified distribution of
is slightly different from the rules used in the closer to the center than to the periphery of
LOCS II in that attention is being directed
grading errors. The approximate 95% toler¬
the area. ance limits16·24 are a useful specification of the
to the color of the entire nucleus, not just to
distribution of grading errors. They were
that of the posterior reflex. This is done to Evaluating the LOCS III System calculated as the next highest increment
avoid overestimating the severity of the above the 95th percentile of discrepancy
brunescent change, such as occurs occasion¬ Selection of Photographs to Test LOCS values. These limits become clinically useful
ally when the reflex alone is used. The over¬ III.—One hundred sixty sets of cataract im¬ as the established tolerance limits for defin¬
all view indicates whether there is signifi¬ ages were identified from the slide library of ing change. If the difference between se¬
cant brunescence of the nucleus, and the Longitudinal Study of Cataract at the Cen¬ quential observations exceeds these 95%
posterior reflex is the best location to judge ter for Clinical Cataract Research to repre¬
tolerance limits, it is assumed that a change
the quality of the brunescent color. The col¬ sent the full range of cataract types and
has occurred. If the discrepancy between
or in the observed lens should be compared severities based on their LOCS II classifica¬
with the color in NC standards 1 through 6 tion. Each set included three photographs:
sequential measures does not exceed the
tolerance limits, it is assumed there has been
and the grade assigned by using decimals to one 35-mm color slit-lamp transparency of a
no significant change in the condition. As¬
interpolate between the integer values of the cross-sectional view of the nuclear region of
reference standards. The assigned score for the lens and two black-and-white retroillu¬ suming no bias (the tendency to obtain con¬
NC may range from 0.1 to 6.9. In LOCS III, mination images, one focused on the anteri¬ sistently higher or consistently lower values,
either between observers or between ses¬
the color standards are boundaries, unlike in or lens at the pupillary plane and one on the
sions), the distribution of discrepancies
LOCS II where there are no color bound¬ posterior lens capsule. For the black-and- should have a mean of 0, about which there
aries. In LOCS III, the grader is asked to do white photographs, a flash intensity of 3, il¬ should be a relatively symmetrical distribu¬
the simpler task of deciding if the color is lumination aperture of 8 mm, and lens aper¬ tion of discrepancies.
more than or less than or equal to that in a ture of 3 are used. These images are used to
standard image. grade C and P. For the color transparency, RESULTS
To Grade C.—Cortical cataract is visual¬ the beam of the slit lamp is oriented 45° to the
ized in retroillumination images focused ei¬ line of vision, and the camera is focused in the The results are presented in the Ta¬
ther anteriorly (at the plane of the iris) or center of the nucleus. The illumination aper¬ ble. For each of the four cataract types,
posteriorly (at the plane of the posterior ture is set at 4 mm; the slit-lamp width at 0.2 the table shows the mean difference,
capsule). Small opacities are graded in LOCS mm; and the flash intensity at 3. The beam is the median difference, the SD of differ¬
III. To decide whether a small opacity is tall enough to just overlap the margins of the ence, and the 95% tolerance limits for
gradable, its size is compared with the size of pupil. One takes the film using an ASA speed the five different comparisons: for ob¬
the small dot opacity located at the 6 o'clock of 200. The color images are used to grade
servers 1 and 2, values obtained at ses¬
position in C standard 1. Any opacities that NO and NC. A slit-lamp image is also useful
size or larger are included. The grader com¬ in identifying the anteroposterior location of sion 1 were compared with those ob¬
pares the aggregate area of the opacity in the opacities seen in retroillumination photo¬ tained at session 2. For consensus,
image being graded with that in standards 1 graphs. values determined by both observers
through 6 and selects an interval that brack¬ Grading of Test Photographs.—The were averaged and then compared for
ets the amount of C in the unknown. In es¬ slides were coded by a technician, arranged the two sessions. For sessions 1 and 2,
timating the severity or aggregate extent of in random sequence, placed into slide carou¬ values obtained by observer 1 were
C, the observer should mentally compress sels, and projected onto a large screen for compared with those obtained by ob¬
the three-dimensional information into a grading. For grading NO and NC, the color server 2.
two-dimensional image and compare the ag¬ slit-lamp image was projected; for grading C For NO, the between-sessions 95%
gregate area of the opacity in the unknown and P, the two retroillumination images and
with that in the standard C images. Again, a slit-lamp image were projected. Between
tolerance limits were 0.7 for observer 1
decimalization should be used to interpolate experimental sessions a technician rear¬ and 1.0 for observer 2, and the between-
the integer values depicted in the sequence ranged the slide sequence into a new random observer tolerance limits were 0.7 and
of standard C images. The assigned score order. In each experimental session, either 0.8 at sessions 1 and 2, respectively. For
NC, the between-session limits were 0.7 sequential observations do not show In the past, grading brunescence has
and 0.8 for observers 1 and 2, respec¬ 95% concordance, the level of concor¬ not always been considered impor¬
tively, and the between-observers tol¬ dance required for the 95% tolerance tant. However, there is evidence that
erance limits were 0.6 at both sessions limit would be LO.15 increased NC may be related to con¬
1 and 2. For C, the between-sessions For each category (NO, NC, C, and P) trast sensitivity dysfunction." In ad¬
tolerance limits were 0.7 for observers LOCS III has some specific advantages dition, grading NC is necessary if one
1 and 2, and the between-observers tol¬ over LOCS II. For grading NO, the is to assess the protective effect of
erance limits were 0.6 for session 1 and rescaling has achieved approximately lens pigments against short-wave¬
0.5 at session 2. Finally, for P, the equal intervals between each reference length visible and long-wavelength
between-sessions tolerance limits were standard, and there are more standards UV light. There is also a well-known
0.9 for observers 1 and 2, and the for evaluation of the early stages of NO. but poorly quantified clinical relation¬
between-observers tolerance limits Removal of inconsistent variations in ship between NC and ease of nuclear
were 0.4 for both sessions 1 and 2. color over the range of NO standards phacoemulsification.
makes judgments about the severity of Nuclear opalescence and NC are cor¬
COMMENT NO much more straightforward. Im¬ related.9·11·21 In one analysis,2'"' 33% of the
The LOCS III, with its expanded sets proved opalescence grading may offer variability in NO grading (using LOCS
of reference photographs and decimal¬ opportunities to assess in vivo basic II) was due to variations in NC. If one
ized grading, is easier to use and pro¬ age-related biochemical processes such is to isolate the effect of opalescence on
vides more sensitive grading than as protein aggregation, and there is ev¬ visual function and define the relation¬
LOCS II when applied to photographic idence that NO, as judged using LOCS ship between NO and other biologic
images of cataracts. A fuller analysis of II, is correlated to some aspects of phenomena, one must be able to isolate
the benefits of decimalizing grading visual dysfunction (eg, contrast sensi¬ and eliminate the variability in opales¬
systems has been presented else¬ tivity loss).711 cence grading due to color in statistical
where,15 but the 95% tolerance limits for In LOCS II, there was only one analyses using regression models.
all of the LOCS III classes are much reference standard for NC, making Therefore, it is important to grade NC
smaller (0.4 to 1.0) than the 95% toler¬ assessment of NC the most subjective as well as NO.
ance limits for LOCS II, which uses in¬ of all the judgments in LOCS IL A For grading C, the reference stan¬
teger increments only. For each cate¬ much broader array of standard refer¬ dards in LOCS III are the same as those
gory of cataract or for NC, the LOCS II ence images in LOCS III greatly fa¬ used in LOCS II. The standards are
tolerance limits are at best 2.0. Because cilitates the assignment of NC grades. separated by intervals that increase