Ann. occup. Hyg., Vol. 45, No. 7, pp. 569-576, 2001
© 2001 British Occupational Hygiene Society
Published by Elsevier Science Ltd. All rights reserved
PII: S0003-4878(01)00011-4
A Model for Predicting Endotoxin Concentrations in
Metalworking Fluid Sumps in Small Machine Shops
DONGUK PARKf, KAY TESCHKEij:* and KAREN BARTLETT§
^Department of Environmental Health, Korea National Open University, Dongsung-dong, Jongro-ku,
Seoul, South Korea, 110-791; ^Department of Health Care and Epidemiology, Mather Building, 5804
Fairview Avenue, University of British Columbia, Vancouver, BC, V6T 1Z3, Canada; §School of
Occupational and Environmental Hygiene, University of British Columbia, Vancouver, BC, V6T 1Z3,
Canada
Methods: In British Columbia, Canada, nineteen small machine shops which used waterbased metalworking fluids (MWF) were examined. One bulk MWF sample was taken from
each independent sump (/V=140) and tested for endotoxin using the Limulus Amoebocyte
Lysate assay. Factors that might influence the MWF sump endotoxin concentration were
investigated using mixed effect multiple regression modelling to control for repeated measures within shops.
Results: The geometric mean (GM) endotoxin concentration was 6791 EU/ml. Contamination of MWF with tramp oil, MWF pH, MWF temperature, and MWF type were significant predictors of sump fluid endotoxin concentration (model P=0.0001, ordinary least
squares R2 =0.36). Concentrations of endotoxin in sump fluids were increased by MWF contamination with tramp oils such as hydraulic oils, preservative oils, spindle oils, slidway lubricants, gear lubricants, and greases (model predicted GM=17 400 EU/ml vs. 1600 EU/ml without tramp oil). Concentrations were also elevated where pH was lower than 8.5 (predicted
GM=10 600, vs 3600 EU/ml for pH 8.5 to 9.5), where soluble fluids were used (predicted
GM=11 800 vs. 2800 EU/ml for synthetic fluids), and where sump fluid temperatures were
higher (predicted GM=2600 EU/ml at 11°C vs. 21 500 EU/ml at 32°C). The within-shop correlation of sump bulk fluid endotoxin concentrations was 38%.
Conclusions: Minimizing tramp oil contamination, using synthetic fluids, and monitoring
pH and temperature would be valuable tools for controlling endotoxin contamination in
MWF sumps. In addition, since there was correlation within-shop, contamination of one
sump in a shop may suggest changing the fluids in all. © 2001 British Occupational Hygiene
Society. Published by Elsevier Science Ltd. All rights reserved
Keywords: metalworking fluid (MWF); industrial oils; endotoxins; occupational exposure; statistical models
INTRODUCTION
. . . , , ,
,
,
,•
Water-based and water-contaminated metalworking
J /w^rr^
-i u
i
J- c
•
a
fluids (MWF) easily become growth media for micro, , .
,
,
,
,
organisms
and their products, such as endotoxins and
c
.
_,,
.
.
,
exotoxins. These organisms may become normal
,
a > r L. .nirr/crTAnn m n J . , . , . ,
flora of the MWF (SHARP, 1997). Microbial growth
..„,.„
,
,
,
• j
i_
m MWF sumps has long been recognized as a prob,
,
• , ,
r • JZ
f
lem, though mainly because of its effect on metal prot
• • ,TT.,i ,r>o-, o i
i
cessing& characteristics (Hill, 1983; Salmeen et al,
Received 29 June 2000; in final form 1 December 2000.
^Author to whom correspondence should be addressed. Tel,
+1-604-822-2772; fax:+1-604-822-4994; e-mail: teschke®interchange.ubc.ca
569
1987; Mattsby-Baltzer et al., 1989), rather than on the
possible health risks for exposed workers.
^ ,
•
,• .
, ,
• .
,
Endotoxins, high molecular weight complexes
, ., ,
,
°
.
associated with the outer membrane of Gram-negative
,
•
,
,
•
r
bacteria, are known to be causative agents of occu„
,
,
.
ai
pationally-related respiratory effects, e.g. chronic
f
, •• u
,
l-f , ,.
•
,
bronchitis, abnormal cross-shift declines in pulmon.
.
, , ,
j , 1 ^ ,„..,
r
ary
J function, asthma, and other long-term effects (Hill
, .. _ , ., i r V 7 n TT . n 1f . o ,r
,
and Al-Zubaidy, 1979; Hill, 1983; Kennedy
et al,
,„„„ , , ^ , „ .
. i n o r i _ ,J
,
1987; Mattsby-Baltzer et al, 1989a; Gordon et al,
1992; Robins et al, 1997; Smid et al, 1992). Endotoxins are heat stable compounds containing lipid,
, , , .
,
. . J,u
. .7 K j
carbohydrate and protein. They are constantly shed
int0 the
environment by living Gram-negative bacteria, and are completely released from bacterial cells
570
D. Park et al.
undergoing lysis (Pearson, 1985). MWF that have
high concentrations of Gram-negative bacteria have
been shown to have high concentrations of endotoxin
(Bennett, 1972; Mattsby-Baltzer et al, 1989; Milton
and Johnson, 1995; Rossmoore and Rossmoore,
1980). They may be an etiologic agent in the deterioration in respiratory health of workers exposed to
MWF (Holdom, 1976; Hill and Al-Zubaidy, 1979;
Kennedy et al., 1989; Sprince et al, 1997).
In industries that use MWF, most airborne endotoxin to which workers are exposed is in aerosols generated by high speed machining operations (e.g. drilling, grinding, lathing, sawing, milling) cooled by
MWF circulated through a sump. Several studies have
found that airborne endotoxin concentrations were
strongly correlated with the MWF sump endotoxin
concentration and were particularly high when
microbial growth in the MWF was excessive (Milton
and Johnson, 1995; Thome et al, 1996).
To prevent adverse health effects related to workers' exposure to endotoxin, it seems prudent to properly manage MWF sumps to control microbial
growth. However, to date, there is little data available
to indicate what specific conditions lead to the
accumulation of endotoxin in MWF sumps. MWF
usage patterns, operating characteristics, chemistry,
and toxicology are very complex (NIOSH, 1998). The
microbiology of contaminated MWF is similarly
complicated, with different organisms dominant at
different times, each with a different spectrum of biochemical activity (Hill, 1977; Bennett, 1972, 1974;
NIOSH, 1998). For these reasons, it has been difficult
to identify factors affecting microbial growth and the
accumulation of their biological products in MWF.
The object of this study was to measure MWF
sump endotoxin concentrations in a sample of small
machine shops located in British Columbia, Canada
and to identify factors significantly influencing the
measured concentrations. Small machine shops like
those in this study often differ from those common
in large industries (e.g. the automobile industry) in
that each machine is supplied with fluid from an independent sump.
METHODS
MWF sump bulk sample collection
This study was conducted in 19 small machine
shops that use water-based MWF. A total of 140
machines, all with independent sumps, were studied.
One bulk sample of sump MWF was taken from a
flowing stream at the cutting points of each machine
when the circulation system was in operation. If the
system was not in operation, the MWF circulation
system was run for at least 10 min prior to sampling,
according to Bennett's (1972) recommendation. The
sample was collected in a 50-ml sterile, tissue-culture
grade, centrifuge tube (Fisher Scientific cat. #05538-55).
The temperature of the MWF was measured in the
field using a thermometer (model UEI PDT 300,
range —40 to 150°C). Factors that may have been
related to MWF sump endotoxin concentrations were
recorded at the time the MWF bulk samples were collected. To determine whether the MWF circulation
system could be contaminated by foreign substances,
the MWF circulation systems (trenches, lines, pits,
sumps, etc.) were checked to determine whether they
were in closed systems or covered. The presence of
circulation system fittings (e.g. oil skimmers, chip
conveyors, filters, chillers, dissolved air float systems)
was checked. The type of machine tool (e.g. drill,
lathe, grinder), and the type of machining fluid
(synthetic or soluble) were recorded. Finally, information on the replacement of old MWF with fresh,
topping up MWF losses, biocide use, the operating
times of machines, and machine ages was sought
from interviews with shop personnel or shop records.
The MWF samples were brought to the laboratory
in a refrigerated container. Upon receipt at the laboratory, the pH (Fisher Scientific, Accument pH meter)
and MWF concentration (T/C Refractometer, model
104311, Buffalo, New York) were measured.
Assays for tramp oil and endotoxin contamination
The MWF samples (-20 ml) were centrifuged at
3000 rpm for 15 min, separating the bulk samples
into up to three parts: a bottom section consisting of
the settled suspended solids, the metal working fluid
itself above that, and finally, in contaminated fluids
only, an upper section of tramp oil. Contamination
with tramp oil was determined by visual inspection
for separated oil at the top of the sample.
After removing any tramp oil, the supernatant
above the solids was transferred to depyrogenated,
screw-capped glass vials (Fisher Scientific, cat #03339-25B) with a sterile, pyrogen-free transfer pipette
(Fisher Scientific, cat. #13-711-20) and refrigerated
until the endotoxin assay was performed. Endotoxins
in all samples were analyzed within 4 days. Endotoxin concentrations were measured using the kinetic
turbidimetric Limulus Amoebocyte Lysate (LAL)
method (Pyrogent-5000, BioWhittaker, Walkersville.
MD), using the time of onset protocol. The Pyrogent
5000 reagent was chosen because the lysate is modified to be less reactive to the glucan-sensitive pathway of the Limulus assay. All glassware was made
pyrogen-free by baking it at 200°C for 1 h. None of
the glassware, pipette tips (Gilson, type Tipac), or
microtiter plates (96-well, flat-bottomed, sterile polystylene, Costar, Cambridge, MA, 3596) had been previously used.
A standard curve was constructed using a stock solution of endotoxin (E. coli O55:B5, lot number 9L
4650, BioWhittaker, Walkersville, MD) made by
reconstituting the supplied 50 ng of endotoxin with
3.8 ml of LAL reagent water to produce a concentration of 100 EU/ml (giving a conversion factor of
Endotoxin concentrations in metalworking fluid sumps
7.6 EU/ng). This stock solution was shaken vigorously for 15 min at high speed on a vortex mixer. A
standard curve was constructed using serial dilutions
of the stock solution to produce a range of concentrations from 0.1 to 50 EU/ml. The curve showed a
strong linear correlation (Pearson r>0.99) between the
log endotoxin concentration and the log mean reaction time.
All samples were vortexed for 30 s and diluted by
serial dilutions of 1:10 and 1:100 using 900 |J,1 LAL
reagent water blanks in depyrogenated 10x75 mm test
tubes (Fisher Scientific cat. #14-925C). In each assay,
MWF bulk samples were tested for enhancement or
inhibition by spiking a sample with a known endotoxin amount. Thus, 10 (al of endotoxin standard (50
EU/ml) was spiked into 100 (0,1 of MWF samples with
each dilution factor. This should yield more than 5
EU/ml if there is no inhibition or enhancement from
the MWF bulk sample. All spikes within the dilution
range assayed were fully recovered.
100 |il volumes of all samples diluted and spiked,
and standards were dispensed into the appropriate
wells of a 96-well tissue culture microtitre plate
(Costar, Fisher Scientific cat. #3594). The assay plate
was incubated at 37°C in a plate reader (Thermomax,
Molecular Devices Corp., Menlo Park, CA) for 15
min, after which 100 jul of LAL reagent was added
to each well of the microtitre plate. The LAL reagent
was reconstituted with 5.2 ml of Pyrogent-5000 LAL
Reconstitution buffer reagent (BioWhittaker cat.
#N383).
The endotoxin content of the samples was determined by comparison to the standard curve. The kinetic
turbidimetric assay measured the time of onset of the
clotting reaction of the lysate incubated at 37°C,
initiated by the presence of endotoxin in the sample
or standard. The increase in turbidity is measured as
an increase in optical density (OD) determined at 340
nm. Computer software (SOFTMAX Pro version 2.0,
Molecular Devices Corp., Menlo Park) was used to
calculate the standard curve and to assign endotoxin
concentrations to the unknown samples.
Data analysis
Descriptive statistics were used to summarize both
the endotoxin concentrations as well as the shop,
machine, and machining fluid characteristics. Some
characteristics, including information on topping up
or replacing MWF, biocide use, and sump age were
not included in statistical analyses because the
responses were not sufficiently complete or reliable.
The distribution of MWF sump endotoxin concentrations was positively skewed and approximately
log-normal, so endotoxin concentrations were logtransformed prior to analysis to improve the
efficiency of the model.
Simple linear regression was used to test the
association between the log-transformed sump endotoxin concentration and each continuous variable.
571
Associations with dichotomous variables were tested
using ?-tests. The following variables with P-values
less than 0.25 were offered in a mixed effect multiple
regression analysis:
• MWF temperature;
• MWF concentration;
• MWF pH [1 if less than 8.5, 0 if 8.5-9.5, based
on the optimum pH range for MWF recommended
by Safety and Health Assessment Research for
Prevention (SHARP, 1997)];
• contamination with tramp oil (0 if yes; 1 if no);
• the possibility of MWF contamination by foreign
substances (0 if no; 1 if yes);
• MWF type (1 if soluble; 0 if synthetic); and
• machine tool type (1 if grinder, 0 if other machine
type) This grouping was based on the fact that
grinding operations have characteristics different
from other machine types — material machined,
operating characteristics, temperature, chip size
produced, and chip removal — which may influence microbial growth (Miyoma, 1993). Note that
endotoxin concentrations were not significantly
associated with machine tool type, so no alternate
basis for grouping was suggested by the data
(Tukey's multiple comparison test).
Correlations between the selected independent
variables were evaluated to identify potential problems prior to offering to the multiple regression
model. Correlations involving at least one continuous
variable were tested using Spearman's rho, and
between two dichotomous variables using Kendall's
tau. The absolute values of all correlation coefficients
were less than 0.5, so all of the above variables were
considered appropriate for input to a single model.
Multiple linear regression mixed effect analysis
was used to model MWF sump endotoxin concentrations while controlling for correlation between
repeated measurements within shops. A backwards
elimination stepwise regression procedure was used
to create the final model. The variables with the highest P-values greater than 0.05 were eliminated one at
a time, then the model refitted until all included variables had P-values of 0.05 or less. The strength of
the linear relationship among the independent variables and Cook's distance were evaluated to confirm
that the assumptions of the final model had not been
violated. Descriptive statistics, correlations, Mests,
and simple linear regressions were carried out using
SPSS 9.0 Standard Version (SPSS Inc., Chicago, IL)
and mixed multiple regression modelling was done
using SAS (ProcMixed, SAS Institute Inc., Cary,
NC).
RESULTS
General characteristics of machine shops investigated
Table 1 summarizes the characteristics of the metal
working machines in the machine shops investigated.
572
D. Park et al.
Table 1. Characteristics of the machines (JV=140) where
bulk samples were taken in the 19 machine shops
Variable
MWF type
Soluble MWF
Synthetic MWF
Straight oil
Semi-synthetic MWF
MWF machine tool type
Grinder
Saw
Lathe
Vertical/horizontal mills
Drill/tap/rim
Broach/bore/mill/turn
Contamination of MWF with
tramp oil
Yes
No
No. of
machines
(samples)
%
86
54
0
0
61.4
38.6
9
11
51
61
6
2
6.4
7.9
36.4
43.6
85
55
60.7
39.3
46
94
40
20
32.9
67.1
28.6
14.3
32
108
22.9
77.1
0
0
4.3
1.4
Contamination of MWF by
foreign substances considered
possible
Yes
No
1
Sumps with chip conveyor
Sumps with tramp oil
skimmer1
Frequency of machine
operation2
Intermittent
During all working hours
These variables were not offered in the multiple regression
models because they were not associated with log-transformed endotoxin concentration (P>0.25) in simple linear
regression.
Of the machines investigated, 61.4% used soluble,
and 38.6% used synthetic MWF. Straight oils and
semi-synthetic MWF were not used in any of the
machines examined for this study. The majority of
the machine tools were lathes (36.4%) and
vertical/horizontal mills (43.6%). The machines were
operated independently, each with its own circulation
system and sump. MWF maintenance, including
refilling or replacement of fresh MWF was usually
carried out by individual machine operators. The
machine shops in the study were run for a regular
workday duration of 8 hours. MWF circulation systems in all sumps were shut down when the machines
were not in use. 22.9% of machines were run intermittently. Although chip conveyors were installed in
28.6% of sumps, no sumps had MWF circulation systems with devices such as filters, chillers, air-floatation devices, etc. Skimmers that remove floating
tramp oil were installed on 20 (14.3%) of the MWF
sumps.
Endotoxin concentrations
The geometric mean (GM) endotoxin concentration
of the 140 sumps was 6791 EU/ml, with a geometric
standard deviation (GSD) of 7.53 illustrating the wide
variability of endotoxin concentrations. Table 2 summarizes the pH, temperature, tramp oil contamination,
and endotoxin concentrations of the sump fluid
samples.
The variables that were associated with endotoxin
concentration in univariate analyses are shown in Fig.
1 (dichotomous variables only) and Table 3
(dichotomous and continuous variables). 60.7% of the
MWF sumps were contaminated with tramp oils. The
GM endotoxin concentration (17 964 EU/ml) in these
sumps was significantly higher than in the sumps that
were not contaminated with tramp oils (1567 EU/ml).
The GM endotoxin concentration in sumps where
contamination of MWF by foreign substances could
have occurred was 14 852 EU/ml, significantly higher
than that in other sumps (4582 EU/ml). The average
pH of the MWF was 8.16 (range 5.90-9.56), lower
than the optimum range (8.5-9.5) recommended by
SHARP (SHARP, 1997). The GM of endotoxin in
sumps where the pH was lower than 8.5 was 13 064
EU/ml, significantly higher than that in sumps with
pH higher than 8.5 (2307 EU/ml). The GM endotoxin
concentration was 8554 EU/ml in sumps using soluble fluids, significantly higher than that of synthetic
MWF sump samples (4698 EU/ml). MWF sumps of
grinding machines had GM endotoxin concentrations
of 3164 EU/ml, lower than that of sumps with other
machine tool types (7415 EU/ml), but not statistically
significant. Endotoxin concentrations were inversely
related to MWF concentration, but the relationship
was also not statistically significant. To check for
non-linearities in the relationship, we also dichotomized MWF concentration into the optimum range
(2.5-5%, Bennett, 1974) versus not, but these categories were more poorly associated with endotoxin
concentrations (P = 0.02, R2 = 0.0004) than the continuous variable. Increasing temperature of the fluid
was significantly associated with increasing endotoxin concentrations.
The following variables were eliminated from
further consideration because, in simple regression,
they were not associated with sump endotoxin concentration (P>0.25): presence of a tramp oil skimmer,
presence of a chip conveyor in the sump, and the frequency of machine operation.
Development of model for predicting MWF sump endotoxin concentrations
Seven variables were selected as potential predictors of MWF sump endotoxin concentration from the
simple linear regression analyses (Table 3); these
were initially offered in the multiple regression analysis.
Table 4 shows the final multiple regression model.
Contamination of MWF with tramp oil, MWF temperature, MWF pH, and MWF type remained as significant predictors of endotoxin concentration, after
controlling for within-shop correlation. Within-shop
Endotoxin concentrations in metalworking fluid sumps
573
Table 2. pH, temperature, tramp oil concentration, and endotoxin concentration in bulk MWF samples from sumps
(At 140)
Mean (SD)
MWF pH
MWF Temperature (°C)
MWF concentration (%)
Endotoxina (EU/ml)
8.16
20.4
5.60
26 329
(0.76)
(2.93)
(4.18)
(44 056)
GM (GSD)
Range
6791 (7.53)
5.90-9.56
11.1-31.8
<0.2-26
1.10-346 736
a
Endotoxin in MWF (EU/ml) as referenced to an equivalent amount of US Pharmacopeia Standard Endotoxin. To calculate
ng/ml, divide EU/ml by 7.6 (see Methods for standard solution preparation).
100000
*\a
10000
1000
Contamination
Potential
pH < 8.5
Soluble MWF*
with Tramp Oil Contamination
with Foreign
Substances
Machine, Sump, and Fluid Characteristics
Grinder*
Fig. 1. The endotoxin concentration in MWF sump associated with various machine, sump, and fluid characteristics (*yes=soluble; **yes=grinder, no=other tool types). Table 3 indicates the explanatory power and statistical significance of each of these
dichotomous variables.
Table 3. Results of simple linear regression analyses for selection of variables as potential predictors of MWF sump
endotoxin concentrations (log-transformed, base 10), variables with P<0.25
Variables
Variable type
Coefficient
Intercept
CV a
R2
No
contamination
with tramp oil
Possibility of
MWF
contamination
by foreign
substances
Grinder
machine tool
Soluble MWF
MWF pH lower
than 8.5
MWF
concentration
Dichotomous
-1.134
4.254
3.05
0.29
0.0001
Dichotomous
0.484
4.146
6.60
0.05
0.0093
Dichotomous
0.314
3.550
7.10
0.01
0.2013
Dichotomous
Dichotomous
0.310
0.857
3.932
3.253
6.98
5.06
0.02
0.16
0.0878
0.0001
Continuous
-0.033
3.976
7.10
0.01
0.1939
Continuous
0.088
2.008
6.43
0.06
0.0034
MWF
temperature
a
P-value
CV=coefficient of variation of prediction.
correlation was moderate, despite the fact that every
machine had an independent sump (r = 0.38). Contamination with tramp oil, a pH below 8.5, soluble
fluid (vs. synthetic), and increasing MWF temperature
were all associated with increased endotoxin concentrations. The following variables were not retained in
the final model: MWF concentration, machine tool
type, and the possibility of contamination of MWF
by foreign substances (P>0.05). There was no multicollinearity among the predictor variables (all VIF
values < 5 , all tolerances <0.2). All Cook's distances
were < 1, indicating that there were no outliers in the
model (Graeme and Nick, 1999).
574
D. Park et al.
Table 4. Summary of independent variables, coefficients, and standard errors of the multiple linear regression model2 for
predicting MWF sump endotoxin concentrations (log-transformed, base 10)
Variables included in model
No contamination with tramp oil
Soluble MWF (vs. synthetic)
MWF pH less than 8.5 (vs. >8.5)
MWF temperature (°C)
Intercept
Coefficient
-1.03
0.62
0.47
0.044
2.68
Standard error
0.16
0.15
0.14
0.021
0.49
R
2b
0.26
0.007
0.043
0.057
-
P-value
0.0001
0.0001
0.0011
0.042
0.0001
"Mixed effect model with shop designated as the random variable.
Proportion of variance explained, based on same variables in ordinary least squares regression model without taking
into account within-shop correlation (R2 cannot be calculated for the mixed model).
DISCUSSION
MWF sump endotoxin concentrations
The MWF sump endotoxin
concentration
(GM=6791 EU/ml; range 1 to 347 000 EU/ml) in this
study can be compared to endotoxin concentrations
measured elsewhere. Milton and Johnson (1995)
reported a range of sump endotoxin concentrations of
13 100-124 000 EU/ml in two automotive plants.
Woskie et al. (1996) conducted a study in automotive
parts manufacturing facilities where machines had
independent sump systems and were run intermittently. The GM MWF sump endotoxin concentration
was 39 000 EU/ml.
These comparisons suggest that the endotoxin concentrations in the automotive machining plants were
higher than those measured in our small machine
shops. The difference may be explained in part by
the different nature of the manufacturing processes
involved. Automotive plants which had large connected sump systems and high-production manufacturing
lines had higher endotoxin concentrations than plants
where most machines had their own sump systems,
like those in the present study. In addition, differences
in production rates, operating times of machines,
management of MWF sumps, as well as endotoxin
assay methods may have contributed to the differences in endotoxin concentrations among these studies.
A model to predict MWF sump endotoxin concentrations
Few studies have developed statistical models to
identify factors that influence endotoxin concentrations in sump fluid. Many factors may influence
microbial growth and accumulation of biological contaminants such as endotoxins and exotoxins in MWF
sumps: oil-water ratio, presence of inorganic salts,
urine, pH, temperature (Bennett, 1972), presence of
sludge, protein, feces, soil, shop sweepings and
hydraulic fluids (Fabian and Pivinick, 1957), fluid
type, system size and particular metalworking operations (Rossmoore, 1981). However, no studies have
been published on the quantitative effects of these
factors on microbial growth or the formation of
microbial products.
This study used a mixed effect model to estimate
the log-MWF sump endotoxin concentrations. A
basic assumption of regression analysis is that all
observations are statistically independent. Ignoring
correlations among measurements could lead to false
conclusions because of incorrect variance estimates
(Liang and Zeger, 1993; Khattree and Dayanand,
1999). Our mixed effect model identified correlation
between measurements within a single shop, even
though each machine in every shop had its own dedicated sump. A source of the correlation could simply
be airborne transport of the microbial flora from one
sump to another in aerosolized MWF, though similarities in other shop characteristics, such as the fluid
types used or their maintenance, could also be
important. The final mixed model which controlled
for within-shop correlation was substantially different
from an ordinary least squares model produced by the
same backwards stepwise regression procedure. One
variable was dropped from the mixed effects model
(the possibility of the MWF contamination by foreign
substances) and another included (temperature).
The final model indicated that contamination of
MWF with tramp oil was the most important factor
contributing to increases in MWF sump endotoxin
concentrations. If tramp oils such as hydraulic oils,
preservative oils, spindle oils, slidway lubricants, gear
lubricants, greases and wire rope lubricants (NIOSH,
1998) contaminate the surface layer of the MWF, they
can prevent aeration and encourage the growth of
anaerobic bacteria (SHARP, 1997). Hill (1977)
observed that it is often not sufficient to confine antimicrobial measures such as biocides or pH control to
MWF, since tramp oils may provide an additional
route of infection for MWF circulation systems.
pH also played a significant role in predicting the
MWF sump endotoxin concentration. MWF sumps
with pHs lower than 8.5 had higher endotoxin concentrations. Similarly Bennett (1972) found that the
tendency for fluid spoilage increased due to the
growth of microorganisms when pH was between 7
and 9, and much less when fluid pH was between 9
and 9.5. Wort et al. (1976) found that in contaminated
MWF, bacteria were usually the dominant microorganisms, and MWF pH could be expected to fall
Endotoxin concentrations in metalworking fluid sumps
because of acidic metabolic products resulting from
bacterial overgrowth (Hill, 1977; NIOSH, 1998). This
suggests that low pH may be a result, rather than a
cause, of bacterial growth and increased endotoxin
levels.
MWF temperature was also a significant predictor
of MWF sump endotoxin concentration. Others have
identified that MWF temperature can influence the
viability of Gram-negative bacteria (Bennett, 1972;
Wathes et al., 1986; Mattsby-Baltzer et al, 1989).
Bennett (1974) found that temperatures above 36°C
can decrease bacterial growth and favor fungal
growth, the development of musty odors, and the generation of sulfides. In the fluid temperature range in
our study (11.1-31.8°C), increased temperature contributed to increased MWF sump endotoxin levels.
These results suggest that measurements of pH and
temperature might be used as convenient indicators
of potential bacterial overgrowth and the accumulation of endotoxin (SHARP, 1997).
The use of soluble MWF resulted in higher endotoxin concentrations than synthetic fluids. Others
have found results which support ours. MattsbyBaltzer et al. (1989) reported that the total bacterial
count in a tank containing soluble MWF was higher
than that in a tank containing synthetic MWF. Petroleum-based MWF, including soluble fluids, are susceptible to bacterial degradation; synthetic MWF has
been reported to be susceptible to degradation by coliform bacteria and fungi Bennett, 1972, 1974;
Rossmoore and Holtzman, 1974). These findings may
be explained by the difference between the ratio of
oil to water (Bennett, 1972, 1974), and the chemical
components (petroleum base oil, petroleum sulfonates, and fatty acids) used as a carbon source (FoxallVanaken et al., 1986). Soluble MWF typically have
a combination of 35-85% severely refined petroleum
base oils, emulsifiers, and water (SHARP, 1997). On
the other hand, synthetic MWFs contain no petroleum
base oils (NIOSH, 1998) and tend to reject tramp oils
because they lack emulsifiers (SHARP, 1997).
The possibility of contamination of MWF by
foreign substances was found to be a significant variable in a model that ignored within-shop correlation,
but was removed from the final model that controlled
for within-shop correlation. Since this variable was
based on observation of the potential for contamination, based on the presence of sump covers, etc., it
was not likely to have the same power as a direct
measurement of contamination. It has been reported
that contamination of MWF by foreign substances
such as soil, food, water (NIOSH, 1998), the intestinal tracts, respiratory tracts, and skins of animals
(Bennett, 1972) due to improper maintenance or
enclosure of MWF circulation systems could cause
changes to MWF which may contribute to a favorable
environment for microbial overgrowth (NIOSH,
1998).
The type of machine tool (i.e. grinder or not) was
575
not associated with MWF sump endotoxin concentration in the final model. This result is not consistent
with the suggestion that microbial concentrations in
MWF and biosusceptibility may vary depending on
the system in question and the type of operation
(Rossmoore, 1981; Rossmoore and Rossmoore,
1996).
Because microbial growth can cause chemical
breakdown of MWF and release byproducts such as
endotoxins (NIOSH, 1998) and bio-resistance is
related to the concentration of fluids (Rossmoore and
Rossmoore, 1996), MWF concentration may be suspected to be related to MWF sump endotoxin concentration. However, MWF concentration was not found
to be a predictor in the final model in this study.
Factors related to the replacement of old MWF
with fresh, or topping up lost fluid are often considered to be among the most important factors influencing microbial growth (Hill, 1977; Mattsby-Baltzer
et al., 1989; NIOSH, 1998). Although we attempted
to investigate this issue in our study, we could not
because of the difficulty of obtaining reliable data.
61% of the machine shops kept no record of the time
of MWF change or the amount of MWF replaced or
refilled. There were no shops that followed MWF
replacement guides which suggest that the entire circulating system and sump should be completely cleaned and disinfected before fresh MWF is introduced
(SHARP, 1997; NIOSH, 1998). Worse, 6% of the
sumps examined in this study had never been cleaned
or had MWF completely replaced. Interestingly, Woskie et al. (1996) were able to investigate this issue,
but found that pH, tramp oil concentration and endotoxin concentrations were not affected by how
recently soluble sump fluids were changed (<4 days,
4-21 days, >21 days).
It has been reported that biocides used to kill bacteria can result in the release of large quantities of
endotoxin due to cell lysis, despite reductions in the
numbers of viable cells (NIOSH, 1998). In our study,
we attempted to obtain data about biocide use, but
most of the participating shops were unable to provide information on the extent, timing, or sometimes
even the use of biocides in their machining fluids.
CONCLUSIONS
Our results indicate that contamination of MWF
with tramp oil, decreased MWF pH, increased MWF
temperature, and soluble fluids were related to
increased sump endotoxin concentrations. For costeffective sump management, maintenance or monitoring of these factors should be prioritized. Thus, MWF
sump endotoxin could be reduced by preventing contamination of MWF with tramp oil, and by using synthetic fluids. MWF pH and temperature could be used
as convenient monitoring tools because these factors
are related to bacterial overgrowth and endotoxin
concentration. In addition, we found that endotoxin
576
D. Park et al.
concentrations were correlated within shops even
though the machines had separate fluid sumps, suggesting that once the fluid in one sump in a shop is
contaminated, the fluid in all sumps may need to be
changed.
Further study is needed to describe how factors that
could not be examined in this study, such as production rate, sump change or refill frequency, biocide
use, and industry type, affect MWF sump endotoxin
concentrations.
Acknowledgements—The authors thank Andrew Ross and
David Bell for arranging visits to the machine shops, as well
as personnel from the machine shops that took part in the study.
We also appreciate the advice and support given by Drs. Susan
Kennedy and Steve Marion throughout this study. The work
reported in this article was supported in part by funds from
the British Columbia Lung Association and the post-doctoral
support program of the Korea Scientific and Engineering Foundation.
REFERENCES
Bennett EO. The biology of metalworking fluids. Lubr Eng
1972;28:237^7.
Bennett EO. The deterioration of metal cutting fluids. Progr
IndMicrobiol 1974;13:121^9.
Fabian FW, Pivinick H. Growth of bacteria in soluble oil emulsions. Appl Microbiol 1957; 1:199-203.
Foxall-Vanaken S, Brown JR, Young W, Salmeen I, Mccure
TS, Napier JR, Olsen RH. Common components of industrial
metal-working fluids as sources of carbon for bacterial
growth. Appl Microbiol 1986;51:1165-9.
Gordon T, Galdanes K, Brosseau L. Comparisons of sampling
media for endotoxin-contaminated aerosols. Appl Occup
Envir Hyg 1992;7:472-7.
Graeme H, Nick S. The multivariate social scientist. London:
Sage, 1999.
Hill EC. Microbial infection of cutting. Trib Int 1977;16:49-54.
Hill EC. Microbial aspects of health hazards from water-based
metal working fluids. Trib Int 1983; 16:136^0.
Hill EC, Al-Zubaidy T. Some health aspects of infections in
oil and emulsions. Trib Int 1979;8:161^.
Holdom RS. Microbial spoilages of engineering materials: Part
3. Are infected oil emulsions a health hazard to workers and
to the public? Trib Int 1976;9:271-80.
Kennedy SM, Christiani DC, Eisen EA, Wegman DH et al..
Cotton dust and endotoxin exposure-response relationship in
cotton textile workers. Am Rev Respir Dis 1987;135:194200.
Kennedy SM, Greaves IA, Kriebel D, Eisen EA, Smith TJ,
Woskie SR. Acute pulmonary responses among automobile
workers exposed to aerosols of machining fluids. Am J Ind
Med 1989; 15:627^1.
Khattree R, Dayanand NN. Applied multivariate statistics with
SAS software, 2nd ed. SAS Institute, Cary, NC, 1999.
Liang KY, Zeger SL. Regression analysis for correlated data.
Pub Hlth Rev 1993;14:43-68.
Mattsby-Baltzer I, Edebo L, Jarvholm B, Lavenius B. Serum
antibodies to Pseudomonas pseudoalcaligenes in metal work-
ers exposed to infected metal-working fluids. Int Arch Aller
Appl Immunol 1989;88:304-l 1.
Mattsby-Baltzer I, Sandin M, Ahlstrom B, Allenmark S. Edebo
M, Falsen E et al.. Microbial growth and accumulation in
industrial metal-working fluids. Appl Envir Microbiol
1989;55:2681-9.
Milton D, Johnson DK. Endotoxin exposure assessment in
machining operations, In: Symposium Proceedings: The
Industrial Metalworking Fluid Aerosols in an Automotive
Transmission Plant, Dearborn, MI, November 13-16.
Detroit, MI: American Automobile Manufacturers Association; 1995. p. 241-3.
Miyoma Y. Grinding operation and metalworking fluids (in
Korean). Machine and Lubrication 1993; 18:7—13.
National Institute for Occupational Safety and Health
(NIOSH). Criteria for a recommended standard. Occupational exposure to metalworking fluids. US Department of
Health and Human Services, Public Health Service, Centers
for Disease Control and Prevention, National Institute for
Occupational Safety and Health. Cincinnati. Ohio. 1998.
Pearson III FC. Endotoxin, In: Pyrogens, endotoxins. LAL
testing and depyrogenation. New York: Marcel Dekker
Inc; 1985..
Robins TG, Seixas N, Franzblau A, Abrams L, Minick S.
Burge H et al.. Acute respiratory effects on workers exposed
to metalworking fluid aerosols in an automotive transmission
plant. Am J Ind Med 1997;31:510-24.
Rossmoore HW. Antimicrobial agents for water-based metalworking fluids (1981). J Occup Med 1981 ;23:247-54.
Rossmoore HW, Holtzman GH. Growth of fungi in cutting
fluids. Dev Ind Microbiol 1974; 15:273-80.
Rossmoore HW, Rossmoore LA. Factors affecting selection of
metalworking fluid biocides. Lubr Eng 1996;23:23-8.
Rossmoore HW, Rossmoore LA. The identification of a defined
microbial inoculum for the evaluation of biocides in waterbased metalworking fluids. Lubr Eng 1980:36:16-20.
Safety and Health Assessment and Research for Prevention
(SHARP). Metalworking fluids: a resource for employers
and health and safety personnel in Washington State. Technical report number 46-2-1997. 1997:9.
Salmeen I, Brown JA, Foxall-Van Aken S, Olsen RH. Presence
of Actinetobacter species among the predominant bacteria
found in a contaminated metal-working fluid. Trib Int
1987;20:218-21.
Smid T, Heederik D, Houba R, Quanjer PH. Dust and endotoxin-related acute lung function changes and work-related
symptoms in workers in the animal feed industry. Am J Ind
Med 1992;25:877-88.
Sprince NL, Thorne PS, Popendorf W, Zwerling C, Miller ER.
DeKoster JA. Respiratory symptoms and lung function
abnormalities among machine operators in automobile production. Am J Ind Med 1997;31:403-13.
Thorne PS, DeKoster JA, Subramanian P. Environmental
assessment of aerosols, bioaerosols, and airborne endotoxins
in a machining plant. Am Ind Hyg Assoc J 1996;57:1163-7.
Wathes CM, Howard K, Webster AJF. The survival of Escherichia coli in an aerosol at air temperature of 15 and 30°C
and a range of humidities. J Hyg 1986:97:489-96.
Wort MD, Lloyd GI, Schofield I. Microbiological examination
of six industrial soluble oil emulsion samples. Trib Int
1976;9:35-7.
Woskie SR, Virji MA, Kribel D, Sama SR et al.. Exposure
assessment for a field investigation of the acute respiratory
effects of metalworking fluids.I. Summary of findings. Am
Ind Hyg Assoc J 1996;57:1154-62.