T-24-1, P-8-87
What is the Meaning of Thresholds in Laser Injury Experiments?
1
D. Sliney1, J. Mellerio2, and K. Schulmeister3
U.S. Army Center for Health Promotion and Preventive Medicine, Aberdeen Proving Ground, MD 21010-5422,
USA
2
University of Westminster, School of Biosciences, W1M 8JS London, UK
3
Austrian Research Centers Seibersdorf, A-2444 Seibersdorf, Austria
ABSTRACT
Human exposure limits for laser radiation are based upon experimental ocular injury studies. The limits
are derived by committees of ophthalmic experts through a review of all available threshold data and an
understanding of mechanisms of laser-tissue interaction. A major point of discussion in this derivation process
relates to the level of uncertainty of the threshold of injury. An indication of the level of uncertainty relates to
the slope of the transformed dose-response curve, or the "probit plot" of the data. The most important point on
the probit plot is the exposure that represents a 50 % probability of injury: the ED-50. It is this value that is
frequently referred to as the "threshold," even though some experimental damage points exist below this
"threshold." The slope is related to the reciprocal of one standard deviation of the normal distribution of
experimental data and thus it reflects not only natural biological variation, but also the impact of experimental
errors. The class of damage mechanism will also alter the steepness of the probit plot. When the steepness is less,
this indicates an increased standard deviation, which in turn may be due to increased experimental error. The
techniques of probit analysis come from toxicology (1), and certain inherent assumptions are carried over to
laser safety studies.
An analysis of any number of example data sets reveals that the slope in most experiments could not be
explained by biological variation alone. This type of critical analysis is essential in deriving exposure limits.
For example, if the slope is not very steep, as with some retinal injury studies, the probit curve may suggest that
at one-tenth the ED-50 energy value, there might be a 0.1 % risk of injury--a risk generally not acceptable in the
laser safety community. Yet, from fundamental biophysical principles, this result could be shown clearly to be
flawed. If the ED-50 energy level for thermal injury corresponds to a retinal temperature elevation of 15°, an
energy of 10 % of the ED-50 must correspond to 1.5° (10 % of the ED-50 temperature elevation), which could
not produce photocoagulation. This aptly illustrates that any derivation of human exposure limits for laserinduced injury requires one to estimate the true biological variation and separate this from the added
experimental errors, which reduces the probit slope.
Analysis of reported experimental data indicates that the thermal and thermoacoustic damage
mechanisms apparently have an intrinsic slope of approximately 1.15 to 1.2. However, experimental threshold
data from retinal studies give slopes that are often much greater (e.g., 1.5 - 1.7), which is really not surprising.
The enormous difficulty of seeing a minimally visible lesion and focusing the laser beam to produce the nearly
diffraction-limited image leads to this greater spread of data and shallower slopes. If a probit curve is applied to
probabilistic risk analyses, it should have a slope of 1.2 or less with the ED-50 point shifted to a lower value.
INTRODUCTION
Setting Maximum Permissible Exposure Limits
Human exposure limits (ELs) for laser radiation have been derived by committees of experts in
ophthalmic biophysics and occupational health through a review of all available threshold data and an
understanding of mechanisms of laser-tissue interaction (2-4). A major part of this derivation procedure relates
to the level of uncertainty of the threshold of injury. The uncertainty is greatest for studies of retinal injury in the
400 - 1400 nm retinal hazard region. The review of uncertainties normally centers on an examination of the
slope of the transformed dose-response curve, or the "probit plot" of experimental data. It is generally agreed
that the most important reference point on the probit plot is the exposure that represents a 50 % probability of
injury: the ED-50. Indeed, this value is frequently referred to as the "threshold," even though some
experimental damage points exist below this "threshold." The steepness of the curve is not only related to the
type of damage mechanism and variation among individual animals, but also indicates problems in conducting
the experiment. Traditionally, all of these factors are assessed collectively in order to derive an appropriate
"safety factor" for deriving the EL from the ED-50. The safety factor is not a true ratio between a true "threshold
of injury" and the EL, but represents a committee judgment of uncertainties. Greater safety factors have been
applied when uncertainties are greatest, and considerations of extrapolation from animal to human are also part
of the process. Because the slopes of some experimental probit curves have sometimes been made part of some
risk analyses, this paper examines the EL derivation process and the meaning of the damage probability curve.
The goal of the review is to clarify why the "safety factor" defined as the ED-50 divided by the maximal
permissible exposure (MPE) necessarily varies based upon careful reviews of each set of experimental data. For
1
T-24-1, P-8-87
retinal limits, this factor is most often taken as one order of magnitude, and based upon consideration of:
•
the overall level of uncertainty in the data,
•
the experimental details,
•
the sources of potential error,
•
the differences between animals and humans,
•
the state of knowledge of the injury mechanism, and
•
the biological sequelae.
The slope of the probit plot is also considered in recognizing the overall uncertainty and quality of the
experimental data, but there is some disagreement as to whether this slope could be considered an index of risk
to a human population. In addition to experimental uncertainties, the committees adjust the MPE values to
minimize complexity and do not attempt to mimic every spectral and temporal variation in threshold. Figure 1
shows an example of an early, hand-drawn damage probability curve where the actual tick marks indicating the
absence or presence of damage are recorded above and below the curve. It also shows the lowering of the
threshold values when histological criteria for damage are applied.
Figure 1. Example of an early, hand-drawn damage probability curve where the actual tick marks indicating the
absence or presence of damage are recorded above and below the curve (5). It also shows the lowering of the
threshold values when histological criteria for damage are applied (Adapted from Ref. 4)
Method of Determining a Risk of Injury in Awake Humans
To explore the issues related to laser damage probability further, we undertook a review of the extensive
database from experimental animal threshold ocular injury studies. This review focused on the variation in the
probit slopes as well as the ED-50 values, where available. The review showed that there are some difficulties in
relating the animal research probit slope and ED-50 data to the human situation as result of experimental
uncertainties. As noted above, all of the sources of experimental error - other than false-positive data - will
increase the ED-50 value and the slope of the probit plot. The examination of the published ED-50 data with
slopes revealed that slopes S ranged from about 1.04 to 2.5, where the slope S is defined as ED-84 divided by
ED-50, i.e. a large number for S corresponds to a shallow slope and a sharp true threshold would have a slope of
1. As the slope values increase, corresponding to an increased spread of data, one implication is that
experimental difficulties have also increased. For example, one early series of experiments to determine ED-50
values for a 30-ns q-switched ruby laser (694.3 nm) had particularly large slopes of 2.2 for the smallest spot sizes
which dropped to 1.3 for large (0.89 mm) retinal image sizes. From a biophysical standpoint, one would not
expect the dose response curve (and therefore probit slopes) to vary so much for different retinal image sizes. It
is well known that ruby laser beams did not have clean Gaussian profiles and therefore, a consistent, minimal
image diameter was difficult to achieve, and the spread of data were particularly great for these small spot sizes.
Recent studies typically report probit slows of 1.1 to 1.4. Indeed, the choice of the "safety factor" used by
committees for derivation of MPEs recognized that experimental difficulties led to a skewed probability
distribution and ED-50 values that are too high. The choice of large "safety factor" values, as great as 10 to 20,
2
T-24-1, P-8-87
resulted from this recognition of experimental uncertainties. As experimental techniques improved (often
indicated by steeper probit slopes), safety factor values less than 10 have sometimes been applied. It is
interesting to note that in deriving exposure limits for skin and cornea, smaller safety factors have been possible
because of the reduced uncertainties in experimental determinations of exposure parameters and damage.
An analysis of a number of example data sets reveals that the slope in some experiments could not be
explained by biological variation alone. This type of critical analysis is essential in deriving exposure limits.
For example, if the slope is not very steep, as with some retinal injury studies, the probit curve may suggest that
at one-tenth the ED-50 energy value, there might be a 0.1% risk of injury--a risk generally not acceptable in the
laser safety community. Yet, from fundamental biophysical principles, this result could be shown clearly to be
flawed. If the ED-50 energy corresponds to a retinal temperature elevation of 15° an energy of 10 % of the
ED-50 must correspond to 1.5° (10 % of the ED-50 temperature elevation), which could not produce
photocoagulation (6). This is true despite the fact that individual proteins, although denatured at normal body
temperatures, are repaired or replaced by cellular maintenance processes that insure proper cellular function (79). This fits with one’s own experience, that an elevation of body temperature (as with a mild fever of 1.5° C)
although unpleasant, is not fatal. The above example aptly illustrates that any derivation of human exposure
limits requires the committee to estimate the true biological variation and separate this from the added
experimental errors that reduce the probit slope. This leads to the central question: What probit slope would
realistically predict the human risk of ocular injury in any real exposure condition?
An awake, task-oriented human with good vision should more consistently achieve a smaller retinal
image than that achieved in a laboratory experiment. For such a human, the proper probabilistic injury function
would have a smaller slope (steeper real slope). A far better indication of the true slope in an ideal laboratory
study is reflected in the experimental data for threshold determination of corneal injury, since the errors related
to the passage of the beam through the ocular media (as well as uncertainties in the retinal image size) are
obviated. The total uncertainty in threshold studies of CO2 laser corneal thermal injury is frequently reported as
10 % or less. For example, Bargeron and colleagues (1989) explained that they did not need to apply probit
analysis because the final bracket between a lesion and no-lesion was approximately 10% of the working power
level. As a consequence, they required fewer animals to determine their thresholds (10). Such corneal studies
imply a probit slope less than 1.1 because experimental error is minimal and the underlying biological variation
for photocoagulation is revealed.
Controlling for Refractive Errors
With regard to retinal injury studies, a review of the data shows that probit slopes tended to be smaller
for the more recent retinal injury studies where laser quality and experimental techniques have improved. It
follows that the distributions of experimental data points leading to each probit plot are more tightly clustered in
these more recent studies and should more closely approach the ideal experimental exposure conditions with
minimal experimental error. The steepest slopes reported ranged from 1.01 for a large spot 3 µs threshold (11) to
about 1.2 for small images (e.g., Ref. 11) for visible wavelengths. This variation in slope with retinal spot size
results for two reasons. There are greater experimental difficulties in consistently achieving the smallest
possible retinal image due to variations in corneal clarity and intra-ocular light scattering and achieving optimal
refraction. This is aptly demonstrated by the studies of Birngruber and colleagues who employed a contact lens
delivery system for exposing both rabbits and monkeys with a contact-lens delivery system which eliminated
most of the experimental error related to refraction (6); they consistently obtained steeper slopes of 1.1 to 1.4 in
the visible. Also, the small retinal structural and pigmentary inhomogeneities have dimensions approximately
the same size at the minimum retinal image and therefore would influence localized energy absorption from
exposure to exposure.
To estimate the magnitude of effect upon the probit slope of refractive errors in an experimental study
of retinal thresholds for small images, we mathematically simulated a range of likely refractive errors. It is
generally accepted that the best refractive correction in a laboratory setting is approximately 0.25 diopter (0.25
D). In the model, a series of hypothetical exposures were delivered in one human eye having an effective focal
length in air of 17.00 mm and the minimal, nearly diffraction-limited image of 10 µm was achieved for every
exposure and the ED-50 was set at 1 µJ with a probit slope of 1.1. Next, a similar series of simulated exposures
were made under the assumption that all were delivered with varying degrees of refractive error, but not
exceeding 0.25 D. A maximal refractive error of 0.25 D resulted in a retinal image blur circle of the order of 30
µm. When it was assumed that the thermal injury threshold varied linearly with spot size (12), then the slope
increased to 1.5 and the ED-50 increased to 2.3 µJ. When it was assumed that the threshold was constant with
retinal irradiance (as postulated for some sub-nanosecond durations), then the slope increased to values greater
than 2.0 and the ED-50 increased to approximately 6 µJ. Because the monkey eye has a shorter focal length
(about 15 mm) and the rabbit eye has a focal length of about 10 mm, these results would be even more dramatic.
The results are illustrated in Figure 2.
3
T-24-1, P-8-87
1
0.9
0.8
0.7
Probit ( E )
Theory ( E )
0.6
0.5
0.4
0.3
0.2
0.1
0
1
2
3
4
5
E
ED50: 1.0 -> 2.5
slope: 1.1 -> 1.4
Figure 2. The expected random variation of the refractive power of an experimental eye between ± 0.25 D
around emmetropia results in slightly larger retinal images leading to an increased ED-50 and reduced probit
slope. To model this shift, the refractive state was varied in small increments up to 0.25 D, leading to an ED-50
increase from 1.0 to 2.5 and a decrease in the probit slope from 1.1 to 1.4 relative to an experiment where the
refraction is perfect for every exposure.
Pigmentary Inhomogeneities
Pigmentary inhomogeneities are more pronounced in the near-infrared (IR-A) spectrum beyond 780 nm,
where the choroid, which underlies the retinal pigment epithelium (RPE), plays an increasingly important role.
As wavelengths increase from 400 nm to 1300 nm, the absorption of melanin steadily decreases with the result
that the RPE absorbs less with increasing wavelengths. For example, RPE absorption is approximately 90 % at
450 nm, but drops to less than 10 % at 780 nm (13). The underlying choroid, with a more mottled distribution of
melanin therefore absorbs the greatest fraction of laser energy in the infrared and this leads to a wider
distribution of local injury thresholds and a shallower slope for the near-infrared (IR-A) wavelengths. Gabel and
associates also studied the pigmentary patterns in retina and choroid in rabbit, monkey and human and showed
greater mottling of pigment in the rabbit. All human races appeared to have the same RPE pigmentation, but
there was a variation among individuals, which could be considerable (13). The impact of pigmentary spatial
non-uniformities is aptly shown by comparing the published probit slopes in threshold studies at visible and
IR-A wavelengths. Whilst slopes of less than 1.2 are common at 532 nm, slopes of 1.4-1.7 are more common at
wavelengths such as 850 nm and 1064 nm (Nd:YAG). To demonstrate these factors, we collected published
threshold data from a variety of studies. This shows that the probit slope generally increases with increasing
wavelength for the same range of pulse duration.
For visible wavelengths, it would seem appropriate to apply a slope of no greater than 1.2 to the
probabilistic injury function. It was initially decided after examining all of the experimental data for
wavelengths less than 650 nm, that the best slope to apply for the probabilistic injury function would be about
1.1. As a consequence of this, the ED-50 value for the "ideal" experiment would be reduced by as much as
two-fold from the value reported in past laboratory experiments. The impact would be that the probabilistic
injury function would predict a higher probability of achieving a retinal injury at a level above the ED-50, but a
lower probability at exposures below the downward shifted ED-50. One study of Gabel and Birngruber (Ref.
14) made use of a contact lens delivery system to substantially reduce the refractive errors. They exposed
human volunteers to argon laser radiation achieving a slope of 1.16, which compared to the slope of 1.28 they
obtained for the same type of exposures in a rabbit eye. This result fits with the greater mottling of pigment in
the rabbit retina. As noted above, the spatial variations in retinal threshold were greater as one approaches the
near infrared. Indeed, it is a common experience of clinical ophthalmologists using photocoagulators, that the
spatial variations in thresholds in one retina are small with argon lasers (514.5 nm) and considerable at the diode
laser wavelength of 850 nm.
After considering all of the above variabilities, it was decided to apply Monte Carlo techniques to vary
the slope in the visible from 1.05 to 1.2 for the probabilistic injury function and from 1.2 to 1.4 for wavelengths
greater than 650 nm. Figure 3 shows plots resulting from varying the probit slope along with the associated shift
in ED-50 values with the reported values for 599 nm 500 ns pulses of ED-50 = 5.2 µJ and slope S = 1.6 (14).
4
T-24-1, P-8-87
0.9
0.8
Probability for minimal lesion
Probability for minimal lesion
1.0
1.4
1.1
0.7
S=1.6
0.6
0.5
0.4
0.3
0.2
0.1
0.0
1
10
Ocular Energy (µJ)
100
1
10-1
10-2
-3
10
-4
10-5
10
-6
10
-7
10
-8
10
10-9
10-10
10-11
-12
10
10-13
-14
10
10-15
0.1
S=1.6
1.4
1.1
1
Ocular Energy (µJ)
Figure 3. Model of the potential shift of experimental parameters due only to the expected variations of the
retinal image blur arising from a random change in refractive power ± 0.25 D around perfect focus, resulting in
varied probit slope along with the associated shift in ED-50 values. Note that the low risk levels for shallower
probit slope can be relatively high, but this is biophysically unrealistic, since the level of change due to
pigmentary inhomogeneities and slight changes in refraction are limited.
Extrapolation from Animal to Human
If refractive errors are eliminated, the best evidence indicates that the rhesus monkey is more
susceptible to retinal injury than humans. This is based upon very limited studies of human volunteers (15) and
studies of retinal pigmentation. The studies of retinal pigmentation indicate that the pigment density in humans
of all races is approximately the same in the RPE. However, the pigmentation in the rabbit and monkey is more
mottled.
CONCLUSIONS
In an ideal experiment where experimental errors (such as those introduced by the beam propagation
through the ocular media) are not present, the true spread of retinal threshold values would be much smaller than
that obtained by current experimental methods aimed at determining the minimum visible lesion. Thus, the
probit slopes reported in retinal injury studies reflect experimental difficulties as well as a true range of
biological variability. Furthermore, fundamental studies of biophysical damage mechanisms support the
argument that a lower value must exist below which any macroscopic injury could take place. However, in
practice it is impossible to determine an absolute specification for this lowest value. Nevertheless, in setting
laser safety limits, committees have always agreed that human exposures at levels less than 10 % of the apparent
ED-50 assured safe exposure. In other words a “safety factor” of 10 was quite adequate even though the “real
safety factor ” between a more accurate ED-50 would be less. In the future, committees deriving human
exposure limits must take full account of the impact of improved experimental techniques. Thus, as techniques
improve and uncertainties reduce, probit slopes will become steeper, and it must be recognized that the former
“safety factor” of 10 is no longer justified, and should be reduced accordingly without sacrificing safety. Clearly,
an improvement in experimental techniques leading to a lower ED-50 and resulting in a “safety factor” of 5
rather than 10 does not change the level of safety as existed previously.
We have shown that by applying reasonable assumptions regarding the experimental errors inherent in
the design of the retinal injury study, it is possible to derive a more reasonable value for the "safety factor." With
improved experimental techniques now achieving steeper probit slopes, it is not necessary to lower exposure
limits if the safety factor falls below 5 - 10. In the derivation of corneal exposure limits such large “safety
factors” have never been required and are not employed in the UV spectrum nor in most of the far-infrared
spectrum.
REFERENCES
1.
2.
Finney DJ, Probit Analysis, 3rd Edn., Cambridge, Cambridge University Press, 1971
American Conference of Governmental Industrial Hygienists ACGIH, Documentation for the Threshold
5
10
T-24-1, P-8-87
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
Limit Values, 4th Edn., American Conference of Governmental Industrial Hygienists, Cincinnati, OH. 1992
International Commission on Non-Ionizing Radiation Protection (ICNIRP), Guidelines on Limits for Laser
Radiation of Wavelengths between 180 nm and 1,000 µm, Health Phys, 71:804-819, 1996
Sliney and Wolbarsht, Safety with Lasers and Other Optical Sources, New York, Plenum Publishing Corp.
1980
Bresnick, G. H., Frisch, G. D., Powell, J. 0., Landers, M. B., Holst, G. C., and Dallas, A. G., Ocular effects
of argon laser radiation I. Retinal damage threshold studies, Invest Ophth, 9: 901-910 (November 1970)
Birngruber R, Hillenkamp F, Gabel VP. Experimentelle und theoretische Untersuchungen zur thermischen
Schädigung des Augenhintergrundes durch Laserstrahlung. GSF-Bericht, AO 251, Munich, Gesellschaft
fur Strahlen- und Umweltforschung mbH, 1978
Henriques, Jr. F.C., Studies of thermal injury V. The predictability and the significance of thermally
induced rate processes leading to irreversible epidermal injury, Amer J Pathol, 23:489-502, 1947
Barnes, F. S. Biological damage resulting from thermal pulses, in (M.L. Wolbarsht, Ed.) Laser
Applications in Medicine and Biology, Vol. 2., Plenum Press, New York, pp 205-221, 1974
Hillenkamp, F. Interaction between laser radiation and biological systems, in F. Hillenkamp, R. Pratesi,
C.A. Sacchi, Eds., Lasers in Biology and Medicine, NATO Advanced Study Institutes Series A #4. Plenum
Press, New York, 1980.
Bargeron CB, Deters DJ, Farrell RA, McCally RL. Epithelial damage in rabbit corneas exposed to CO2
laser radiation, Health Phys, 56:85-95, 1989
Zuclich J.A., Lund D. J., Edsall P., Hollins R. C., Smith P. A., Stuck B. E., Mclin L. N. Laser induced
retinal-damage threshold as a function of retinal image size, in: Stuck BE and Belkin M, 1997, Laser and
Noncoherent Ocular Effects: Epidemiology, Prevention, and Treatment. SPIE Proceedings, 3591:335-343,
1999
Beatrice. E. S. and Frisch, G. D, Retinal laser damage thresholds as a function of image diameter, Arch
Environ Health, 27:322-326, 1973
Gabel VP, Birngruber R and Hillenkamp F. Visible and near infrared light absorption in pigment
epithelium and choroid. (K. Shimizu and JA Oosterhuis, Eds.) XXIII Concilium Ophthalmologicum,
International Congress Series No. 450, pp. 678-662, 1978.
Lund DJ, Stuck BE, Beatrice ES. Biological research in support of project MILES. Presidio of San
Francisco, CA: Letterman Army Institute of Research, 1981. Institute Report No. 96
Gabel VP and Birngruber R. Comparative experiments on laser threshold in rabbits and human volunteers.
(I. Kaplan and P. W. Asher, Eds.) Laser Surgery, pp. 246-249, 1980
6