Kinetics of lead in blood after the end of occupational exposure.

S, RANSTAM J, CHRISTOFFERSSON J-O. Kinetics of lead in blood after the end of occupational exposure. Scand J Work Environ Health 13 (1987) 221-231. The sum of two exponential functions was fitted to the decay of blood lead (PbB) level after the end of lead exposure. For two subjects who had not formerly been occupationally exposed to lead but who had been exposed to a single short heavy dose, the fast compartment (probably soft tissues) had a biological half-time of 27 and 44 d, respectively. For 20 lead workers after the end of occupational exposure, the corresponding median was 29 (range7-63) d. For 21ex-lead workers, the median biological half-time of the slowcom partment was 5.6 (range 2.3-27) years. There was significant interindividual variation in both the fast and the slowhalf-time This finding probably means a considerable variation in risk at a certain exposure level. In the lead workers, the PbB fraction corresponding to the slow compartment had a median as high as 1.8 (range 0.7-2.7) pmolll, which constituted more than half of the total PbB. This fraction was associated with exposure history, and with the lead levelin the skeleton, the latter determined in vivo byan X-ray fluorescence method. The data thus indicate a rather rapid turnover of the skeIetal lead pool, a phenomenon which may affect the PbB level considerably.

Reprint requests to: Dr A Schiitz, Departm ent of Occupational Medicine, University Hospital, S-22I 85 Lund, Sweden. ber 102) the re was a la ck of data between 3.5 an d 9.6 years a ft er the end of exposure.
In addi tion 17 male lead wor kers temporarily removed from expo sur e were investigated (table 2). Th eir mean age was 49 years, and their mean exposure time was II yea rs . The reason fo r removal from exposure was high PbB levels (generall y a bo ut 3.0 ILmol/l o r more) . Tw elve of the workers were tr ansferred from a smelte ry to a nea rb y plant, th e wor k in which d id not in vol ve lead exposure. The PbB level was generally determined when th e workers left the smelter y and then once a week for three week s; later PbS measurements were made once ever y two to four weeks.
Fu rthermore, two male vo luntee rs, who had been unexposed occupationally, but who had been exposed to a sing le sho rt heavy lead do se, were included (table  2) . Details on these tw o su bjects ha ve been published elsewhere (56).
Spo t determinations of "background" PbB levels were made for 47 healthy wo rkers not occupationally exposed to lead. They lived in th e same county as the exposed su bjects and were all blue-collar wo r kers . Their a verage PbB level was 0.3 ILmol/l. In th is co nnection, it may be mentioned that, for 15 wo rkers in a glue production pl ant located close to th e nonlead resort of the temporarily removed smeltery workers , the average PbB level was 0.5 (range 0.3-0.7) /Lmol/l.

M edical exam inations
For most o f th e su bjects , an occupational and medical history, including alcohol habits, was obtained. Venous blood sam ples were analyzed for lead (see th e section Blood Lead Det ermination s), hemoglobin , sedimentation rate; red and white cell counts ; calcium , phosphate, and creatinine concentrations; and alkaline phosphatase and gamma glutamyl transferase acti vities in serum. A urine sample was analyzed for albumin and glucose .
Among the ex-lead workers, detailed medical information was lacking for four. Among the 19 remaining, 12 had earlier been removed (at least once) temporarily from lead exposure because of a high PbB level and /or a high delta-aminolevulinic acid level in the urine. One subject was clinically diagnosed as lead poisoned (upper abdominal pain, constipation, neuropathy, and slight anemia) at the time when his exposure ended. No one else had been treated in a hospital because of lead poisoning. One worker had a clinically silent chronic lymphatic leukemia, and one had a type 2 diabetes treated with diet only. Three persons had slight increases in their serum creatinine levels, and two others showed slight albuminuria. Three subjects had somewhat increased gamma glut amyl transferase activities, in their serum, two of whom were known to abuse alcohol.
Among the 17 temporarily removed lead workers, detailed medical information was available for 14. Among these 14, seven had earlier been removed because of a high PbB level and/or a high deltaaminolevulinic acid level in their urine . None had been treated with drugs because of lead poisoning. Three subjects had slightly increased gamma glut amyl transferase activities in their serum, one of whom also had an increased alkaline phosphatase activity in his serum. One person had an isolated marginal increase in alkaline phosphatase activity in his serum.
The two subjects without previous occupational lead exposure were both in excellent health.

Blood lead determinations
Blood was obtained from the cubital vein. Durin g the first years of the study, acid-wa shed heparinized sampling tubes were prepared at our laboratory. Later on , evacuated, metal-free Vacutainer@ tubes were used.
Almost all of the PbB determinations were made in the same laboratory and by the same method. The samples were wet-ashed, and lead was complexed with dithizone, extracted, and determined by flame atomic absorption spectrometry (AAS) (55,56). The detection limit was 0.05 pmol /l (10 pg/l).
Each analytical series contained six samples, two blanks containing reagents only, and four "normal" blood samples (two of them with standard lead addition). All the samples were analyzed twice . The coefficient of variation calculated from duplicate analyses of 25 samples containing 0.5 pmolll or less was 6.6 070 of the mean, for 57 samples containing 0.5-1 pmolll it was 3.9 %, for 58 samples containing 1-2 pmolll it was 2.5 %, and for 60 samples containing 2-3.5 pmolll it was 2.0 %.
The accuracy was tested twice each year in a Nordic interlaboratory calibration program with 6-19 (mean 222 12)  During the first year of observation of subjects 104-115, the first three years of subject 102, and the six first years of subject 116, PbB was determined by a colorimetric method after extraction with dithizone in chloroform. The detection limit was about 0.05 pmolll. The results obtained by the colorimetric method averaged 105 (SO 6) % in the concentration range 0.3-1.9 pmolll and 100 (SD 5) % in the range 2.0-5.4 pmolll of results obtained with the flame AAS method.
From subject 116, for the following four years, determinations were made by flame AAS after precipitation of proteins with trichloroacetic acid (22). The detection limit was 0.2 pmolll, and the method error about 10 % .

Mathematical analysis
Three models corresponding to the sum of one, two, and three exponentials were considered for the PbB decay curves of each individual worker. A fixed "background" value of 0.3 p,mol/I was used for all the subjects and each model. The nonlinear regression procedure in the statistical package BMDP (21) was used. This program produces estimates of the parameters which minimize the unweighted residual sum of squares using a modified Gauss-Newton algorithm. Minimum and maximum values can be specified for each parameter. Thus two parameters and their asymptotic standard deviation were estimated: an elimation rate [transformed and quoted as half-time T Y2(I), TY2 (2), and TY2(3)] and the concentration corresponding to each compartment [YO), Y(2), and Y(3)]. Confidence intervals were estimated on the assumption of asymptotic normality of the estimates. The fit of the three models was judged from comparison of the fraction of total variance in the PbB values explained (R 2 %). In addition plots of residuals versus time were used for checks of the validity of the model and the accuracy of the individual curve fittings .
To describe accumulation, a function of the type yet) = A[I -exp( -B x t)l, where A is a scale constant, B an elimination constant, and t time , was fitted to the data by use of the nonlinear regression procedure in BMDP (21).  The fit of the two-compartment model (median 97 %, range 35-99 %) was considerably and significantly (P < 0.0001, Wilcoxon) better than that of the one-compartment model (median 86 %) (tables 1 and 2). The fit of the three-compartment model was similar (medians 97 versus 97 %) (table 1), though significantly (P<O.OI, Wilcoxon) better.
In the following presentation, only the simplest model with a good fit, ie, the two-compartment one, will be discussed .
With the use of the two-compartment model, the decay rate in the remaining 21 ex-lead workers (table  I)

Statistics
In general, non parametric tests were employed. For associations the Spearman's rank correlation (rs) was used, and for comparisons o f duplicate measurements in the same individual the Wilcoxon's matched-pairs ranked-sign test was used. Comparisons between groups were made by the Mann-Whitney U-test. In a few instances, a single or multiple linear regression analysis was made. For establishing interindividual variations, binomial tests were employed. When more than one observation series was available for a particular individual, the value corresponding to the calculations with the best fit (R 2 ) of the compartment analysis was used. All P-values are two-tailed . "Statistically significant" denotes P<0.05.

Bone lead levels
Bone lead levels were determined in vivo from the middle phalanx of the left forefinger of 37 subjects by an X-ray fluorescence method, as described earlier (16). The detection limit was 20 p.g/g, and the method error about 15 070 . Readings below the detection limit were assigned a value of 10 p.g/g in the calculations. In most cases, levels derived either at duplicate measurements (16) or calculated from a series of measurements (15) were employed.
The decline rate of PbB was, in most cases, rapid soon after end of exposure, but later on it was slower (figure 1). There was generally a good fit of the observed PbB to any of the three compartment models tested (tables 1 and 2). However, one subject (number 122) (table 1) displayed pronounced irregularities in the elimination pattern, which rendered serio us suspic ion of occasional, ongoing exposure. In addition the lack of PbB data during the first 94 d after the stated end of exposure probably contributed to the bad fit to any of the models tested. He has thus been disregarded in the following results. Another ex-lead worker (number 114) was excluded because of suspected lead exposure during the first month of the supposedly exposure-free period (rising PbB). Furthermore, before the second observation period (from year 7 on), his PbB had decreased to very close to the background level, and the T V2 (2) was determined merely from the observations made during the first year. No conclusions as regards the kinetics of a slow compartment seem to be justified in this case. For most of the ex-lead workers, the number of observations during the first period after the end of exposure was too few (less than four per two months) to allow reasonably accurate estimates of the decay rate of the fast compartment. Thus, for all but three (numbers 102, 103, and 123), an approximate half-time of the fast compartment [T Y2(I)] of 30 d was employed. (See the following text.) Table 1. Ki net ics of th e decrease of lead in blood (PbS) aft er th e end of th e occ upational ex posure o f 23 ex-lead workers obs erved for more t han one year. (R' '" degree of ex planat ion, TY,(1) '" half-time o f fast compartment , T'I2 (2) '" half-time of slo w com partment, 95 % CI '" 95 % confidence int erva l, Y(l ) '" Y inter cept for the fast compartment, and Y(2) '" Y intercep t for the s low com part men t)

One
Two-com part ment mode l Th ree Obser-Numb er com part-co m partsu b- a Nu mber 101 was a cast bron ze fou nder, num ber 102 a spray painte r, numbers 103-11 5 storage battery workers, number 116 a wir e lead coater, and num bers 117-123 smeltery workers. • At t he end of expos ure. c Num be rs 101 and 117 are ident ical wit h nu m bers 201 and 217, respectively, in tab le 2. d When fewe r than four samples were o btained du ring the fi rst two months after th e end of exp osu re, the T 'I2(l ) was ass um ed 10 be 30 d. e Sample ob tai ned 94 d after the end of exposure. Table 2. Kin etic s o f th e dec rease of the blo od lead (PbS) levels du ring a temporary cessation of occupational exp os ure among 17 lead wo rkers ob served to r less than one year and among two vo lunteers who had a short, heavy ex posure. (R' '" degree of explanation, T Y2(1)'" half-time for the fast compa rt men t, T 'I2(2) '" half -ti me of the slow co mpartment, 95 % CI '" 95 % co nf idence interval, Y(l) '" Y inte rcept for the fast co mpartment, Y(2) '" Y int erc ept fo r the s low com part ment)

One
Two -compart me nt model  workers (numbers 102, 103, and 123) (table I) and in the 17 temporarily removed lead workers (table 2). The ob servat ion period in the latt er group was too short for an accurate estimation of the half-time of the slow compartment [T Vz (2 )]. Thus , in the calculat ion of T Vz(l) for these subjects, an approximate T Vz(2) of 5 years (see the preceding text) was used as being the best estimate. The median T I/2(l ) for the 20 subjects was 29 d. The ir median observati on time was 155 d . The range of TVz(l) was considera ble (7-63 d) (table  2) for subjects 203 and 208, the T Vz(l ) with the best fit, 49 d and 24 d , respectively, being used . The decay curves with the shor test and next longest T I/2(l ), respectively, are shown in figure 2.
When, in two subjects (numbers 203 and 208) (table 2), the decay patt ern was studi ed during two periods of temp orary removal fro m exposure , 0.5 and 1.0 years apar t, the decline rat es of PbS were compatibl e.
The two formerly occupationally unexposed subjects had a T 1/ 2(l ) of 27 and 44 d (tabl e 2).
The    (1). This result showed that there was a statistically significa nt (P = 0.0004, bino mial test) inte rindividual variati on for T V2(l).
When Y (2), for all 38 lead workers and the two unexposed subjects , was plotted against exposure time (figure 5), there was no significant nonp arametric associa tion . Neither was there any linear co rrelatio n . However , when an exponential acc umulat ion curve was fitted to th e data, there was a reasonable fit (R 2 = 49 % , P< O.OOl). The elimination constant was 1.2, corresponding to a half-time of 0.6 years, and Y(2) leveled off at 1.8 /lmol/l. Moreover, there was a tendency for the workers temporarily removed fro m exposure to have a higher Y(2) at a particular expos ure time than the ex-lead workers. However, the difference was not statistically significant in th e mult iple regression analysis. Neither did the observatio n time display any significant association with Y(2).
For the 35 lead wo rkers and the two unexposed subjects, there was a significant correlation between Y(2) and bone lead conte nt (r s =0. 36, P=O.OI) (figur e 6). Multiple linear regression analysis displa yed th at th ere was an increase in Y(2) of 0.008 /lmol/I per p.g/g of bone-Pb (P = 0.005). Furth ermore, the temporarily removed workers had a Y(2) that, on the average, was 1.0 /lmol/I higher than that of the ex-lead wor kers. This finding was also obvious fro m the decrease in Y(2) of 0.09 p.mol/I per year of the postexposure ob servation time (P< O.OOO I) . Among the active workers, there seemed to be a leveling of f of Y(2) when the bon e lead content increased. There was no such clear corresponding tendency amo ng the retired workers. relation between the air lead level and PbB (eg, in references 12,13,19), between PbB and plasma lead (20,42,45), and between PbB and the lead level in urine (12,13,56,67) could favor the choice of a nonlinear model. However, elimination rates similar to ours have been reported earlier, both for subjects far less (I I , 13,53) and far more (14,46) exposed. In addition there was no indication in the present data that the elimination rates were faster in subjects with a high initial PbS; instead the T IIz(I) increased with increasing Y(I). Thus there are at present no serious objections against the use of a multiple exponential model, at least not in the PbB concentration range that we have studied. For a few subjects, the fit was not as good. For five subjects the R 2 was <800/0, and for one of them it was 0 %. This result may be due to the fact that, in these subjects, the PbB was low during a major part of the observation period and that thus the analytical error had a great impact.
It is relevant to consider whether the number of compartments is two . Different authors have proposed one (23,68), two (I , 7, 63, 64), three (4,33,43,53,66), four (13,23,44,45,60), and even five (6) compartments in human metabolic models. The simplest model that gave a good fit in the present study was the twocompartment one. Thus there was no reason to choose a more-complicated model. Of course, from a theoretical point of view, a larger number of compartments is probable. Thus data from the two subjects exposed only to a single heavy lead dose may indicate an initial very fast decay of PbB (56). For most of the subjects in the present study, such a phenomenon would have remained undetected. There was probably a continuous absorption of lead from the lungs and gastrointestinal tract for some time after the end of exposure. (See the following discussion .) In addition the early observation s were few. However, for the two subjects (numbers 208 and 2 I6) from whom frequent obser vations were made during the first few days after the end of exposure, the data did not indicate an y rapid initial decrease. Moreover, in addition to the relatively small number of observations, the limited observation period and the analytical method error may obscure other compartments, especially small or very slow ones . Indeed, some ob servations in the present study may indicate that the slow compartment really has more than one component. (See the following discussion.) The parts of the body which constitute the two compartments must also be considered . The only organ containing lead amounts sufficiently large to cover the considerable excretion associated with the decay of the slow compartment [on the order of 0.05-0. I mg124 h in urine only, for several years (60)] is the skeleton, which contains hundreds of milligrams (I, IS, 16,39,59). The identity between the slow compartment and the skeletal pool is also strongly supported by the association between Y(2) and the bone lead content. The soft tissues, including the lung, contain only a few milligrams (10,59), an amount which fits with the excre-tion during the emptying of the fast compartment.
The median half-time of the fast compartment was about one month. There may be errors that affect this estimate. It is difficult to be absolutely sure that the exposure did stop totally at a fixed date; some exposure may have continued after the formal end of exposure. For example , for the temporarily removed smeltery ' workers, the work site after removal was located only a few hundred meters from the plant. Also, the homes of the workers may have been contaminated. In addition, a worker may have a pool of lead in the lungs and the gastrointestinal tract and thus continue to absorb lead for some time after lead inhalation and ingestion have ceased. The limited data on hand -frequent measurements of two subjects -may indicate such an absorption, but mainly up to one week after the end of exposure, which is in accordance with earlier observations (9,12,13,31,32). This delayed absorption is probably the reason of the present weak positive correlation between TV2(1) and Y(I). These possible sources of error all tend to give a somewhat too long an estimate as compared to the true half-time. Furthermore, the workers temporarily removed from exposure were not randomly selected. They were removed from exposure because of high PbB levels, and this occurrence might partly be the result of a slow elimination rate in those particular individuals. Another possible explanation of the slight positive association between T !h(l) and Y(I) is a bias introduced by the disregarding of possible intermediate compartment s.
PbB is mainly present in the red cells. It could thus be suspected that the lifetime of these cells would determine the TV2(1). However, the calculated T!h(I) is considerably shorter than would be expected if lead were eliminated from blood only at the normal death of these cells. But lead is known to cause hemolysis, and the question can be raised of whether it could have affected the T V2 (1). Hardly, at least not considerably, as there was no correlation between T V2(1) and the initial PbB. A negative correlation would be expected if hemolysis were important. For the lead workers temporarily removed from exposure, we had to employ an estimated T!h(2) of five years. However, this procedure did not affect the T !h(l); even a T V2(2) as short as one year, or as long as 10 years , would cause only slight changes in the T !h(I).
There was a considerable interindividual variation in T V2 (1). To some degree this occurrence may be explained by various errors in the estimates of individual decay curves . Thus the two subjects studied twice had a different TV2(1) on the first and second occasion, a finding which could not reasonably be accounted for by real variation. It should be noted though that the half-times were not statistically significantly different, as the confidence intervals overlapped. It is difficult to explain all the variation within the population by errors; there is very probably a true interindividual variation in turnover rate , as has been observed for lead in dogs (23), as well as for other heavy metals in man [methylmercury (58), cadmium (34,71), chromium (70)}, though the true range of the T V2(1) may perhaps not be as large as 10 times. The interindividual variation maybe due to differences in excretion, in urine and/or in feces. However , it was not associated with the serum creatinine levels, and there was no association between elimination rate and urinary lead excretion (60). One possibility is a variation in fecal excretion due to different elimination through bile and /or intestinal reabsorption. In fact, there are indications of a considerable interindividual difference in gastrointestinal absorption (8,32).
The median half-time of the slow compartment was about five years. Some of the estimates are probably not too accurate because, at lower PbB levels, the analytical method error is more important. In addition, the true "background" PbB may vary somewhat between individuals, even in a population such as the Swedish one , which has low and rather homogeneous PbB levels. Thus a somewhat lower "background" was indicated for subject 104 in the decay curve in figure I. Furthermore, the number of observations was relatively low, and the observation period should preferably have been a couple of decades. For most of the ex-lead workers, we had to employ an estimated T V2(1) of 30 d. This procedure may have affected the TV2(2), but only slightly. However, we consider the median of about five years as reasonably correct. Furthermore, there was definitely an interindividual variation in the T Vz (2). But the true range between individuals cannot be considered firmly established.
There was an increase in the TV2(2) with increasing age. This phenomenon does not necessarily mean a decrease in the turnover rate of skeletal lead, but may have other explanations. First, senile osteoporosis may add significantly to the mobilization of lead from the skeleton and thus cause a gradual increase in PbB with time, which then leads to a spuriously high TV2(2). However, for the subjects followed by repeated X-ray fluorescence measurements of bone lead (15,61), there was no association between age and T V2 (2), and this finding is consistent with the lack of acceleration in the decrease of the bone lead content with time. Second, as lead is excreted by glomerular filtration (69), the decrease of this parameter with increasing age may cause a gradual relative increase in the PbB content, which would result in a spuriously high T Vz(2). We did not see any association between the serum creatinine level and the TVz(2), but the glomerular filtration rate may decrease considerably without any rise 228 of serum creatinine. Another possible explanation for the slight association between TVz (2) and age is that older subjects were exposed to higher lead levels in the past than at present and consequently have a larger pool of "old lead" in a slow pool , from which lead is now returning to the blood. However, the fact that there was no significant association between time of exposure and T V2(2) speaks against the importance of this phenomenon.
The presence of a lead pool in the body for a long time after the end of exposure has been obvious from a series of investigations (eg, those reported in references 15,16,26,29,47,52). The main lead pool is in the skeleton (1,2,15,16,39,57,59). Much has been obscure considering the kinetics of the skeletal lead pool. The T V2(2) (ie, elimination from the skeleton) of five years reported in the present investigation is faster than most earlier estimates of the half-time of skeletal lead (8-71 years) by different techniques (6, 18,23,30,35,51 ,62,63,66). However , it is in accordance with data from some other studies (3,17,25,68,72,73). In addition it fits remarkably well the results of repeated X-ray fluorescence measurements of finger bone lead in ex-lead workers [a few of whom are identical with the present subjects (15, 61)} . There are several possible explanations for the differences, ie, the method employed , the part of the skeleton assayed (in our case a reflex of the lead release from the whole skeleton), and the observation time in relation to exposure.
The Y intercepts Y(l) and Y(2) reflect the influence on PbB by the fast and the slow compartment, respectively. The relation between Y(l) and Y(2) varied considerably between different individuals, probably because of the exposure history. Thus, in an individual with a short, heavy exposure, the Y(I) should dominate, while in subjects exposed for many years , thou gh only slightly lately, Y(2) should be more important. It is interesting to note that, overall in the present study, Y(2) made up for as much as 1.8 f(mol/I, corresponding to 65 % of the PbB concentration, and that it ranged upward as far as 2.7 f(mol/I and 100 070 . This is a larger fraction than has earlier been assumed (l, 12,30,41,53). The difference may be due to the intensity and time of exposure and the time of observation. The large impact of skeletal lead on PbB is important from a practical point of view, as the fraction of PbB which mirrors the slow compartment decreases only slowly, even after complete removal from exposure.
As has already been mentioned, the slow compartment probably mainly reflects the lead level in the skeleton, which is strongly supported by the association between Y(2) and the lead level in finger bone. The retired workers were lower in Y(2) in relation to finger bone lead than the active ones. In the former, Y(2) may mainly reflect the lead pool in the slow bone lead compartment (53,63), represented by the mainly cortical (or compact) finger bone; in the latter the fast trabecular (or cancellous) bone pool may be more important (57). But, in addition, in the active workers, Y(2) may also be affected by a small, intermediately fast pool, perhaps contained mainly in the liver, the kidneys (10, 59), and the skeleton (an even faster pool than the trabecular one). When the data of subject 103 in figure 1 is closely examined, one may, in fact, anticipate more than one component in the slow compartment.
The impact of organs other than the skeleton on Y(2) may be indicated by the accumulation pattern of Y(2) with increasing exposure time. A steady state was reached already within a couple of years . This period is faster than in trabecular bone (57); the turnover of cortical bone is much slower (15,61). The accumulation pattern may also be affected by the nonlinear behavior of lead in erythrocytes (19,40,43). This phenomenon may also be the cause of the apparent leveling off of Y(2) upon lead in finger bone in active workers (but not in the retired ones, who had lower PbB levels).
There was a considerable interindividual variation in Y(2) at a particular exposure time . One obvious explanation is variations in the intensity of exposure. An additional explanation may be the interindividual variations in lead metabolism seen in this study, as well as in earlier ones (8,32,60).
The variation in the kinetics of lead metabolism should mean a considerably varying risk for different individuals exposed at the same level, which, of course, is important from a practical point of view. A short TIIz(I), as the result of a rapid excretion, is probably an advantage for the worker. On the contrary, a long T liz (2) may be good, as it means that the net endogenous exposure from the skeleton is low.
In many countries, lead workers are removed from lead exposure when they reach a PbB "trigger level" ("removal level"; in Sweden at present 3.0 ",molll), and they are not allowed to return until the PbB concentration has decreased to a "safe level" ("return "level" ; 2.0 /lmolll) . In our "typical" lead worker, who had an Y(2) of 1.8 ",molll and a T Y2(1) of one month, this process will take as much as about six months. However, newly employed workers, who have a small Y(2) , would display the same decay in less than a month. On the other hand, as many as 57 % (24 of 42) of our workers had a Y(2) of more than 1.7 ",mollI. On the assumption of a "background" level of 0.3 ",mol, they would reach 2.0 ",mol/l only after a sufficiently long time had elapsed to affect the slow compartment. In the workers who were followed for a long time, 30 % (7 of 23) would require more than a year to reach the "safe level." These assumptions are in accordance with observations of workers removed from exposure (49) .