An emission model tracking the life cycle pathways of human pharmaceuticals in Korea
Environ Health Prev Med
An emission model tracking the life cycle pathways of human pharmaceuticals in Korea
Eun Jeong Han 0
Hee Seok Kim 0
Dong Soo Lee 0
0 E. J. Han H. S. Kim D. S. Lee (&) Environmental Planning Institute, Graduate School of Environmental Studies, Seoul National University , Gwanak-ro 1, Gwanak-ku, Seoul 151-742 , Republic of Korea
Objectives Pharmaceuticals in the environment are of growing public health concern. The main objectives of this study were to develop a new emission estimation model, identify factors critical to reducing emission, and demonstrate the model's applicability for screening and priority setting. Methods A new emission estimation model was developed covering the life cycle pathways of pharmaceuticals from supply to discharge into surface water. The emission estimates of the model were assessed by coupling with SimpleBox to give predicted concentrations and by comparing the predicted concentrations with measured concentrations in Korean surface waters for five selected pharmaceuticals (acetaminophen, cephradine, ibuprofen, mefenamic acid, and naproxen). Results The sensitivity analysis revealed that the biodegradation rate in the sewage treatment plant and the excretion rate of pharmaceuticals were the most important factors influencing the emission rate. The uncertainty of the emission estimate was found to increase with increases in
Pharmaceuticals; Emission model; Pathway; Uncertainty; Sensitivity
the value of the emission estimate. Once the intrinsic
properties of a pharmaceutical (excretion rate,
biodegradation rate, and removal rate by sludge separation) were
given, the patient behavior parameters, such as
participation in a Take-back program and rate of administration,
were determined to have a strong influence on the emission
estimate. In our study, the predicted and measured
concentrations agreed with each other within one order of
magnitude. Several management implications were drawn
from the analysis of model outcomes.
Conclusions The model outcomes, alone or in
combination with toxicity data, may potentially be used for the
purposes of screening, priority setting, and the design of
The presence of pharmaceuticals in aquatic environments
was initially determined in the late 1990s [
], and since
this time concerns regarding their presence in the
environment among public and scientific communities have
been rapidly increasing [
]. Increasing evidence of the
potential ecological risks posed by the levels of certain
pharmaceuticals that have often been found in aquatic
] testifies to the need to develop
management options prior to or following the discharge of these
pharmaceuticals to minimize adverse health effects. To
develop such management programs, a knowledge of
emission rates of pharmaceuticals is essential.
The emission rate can be determined either by direct
measurement or by using estimation methods. Direct
measurement of numerous pharmaceuticals, however, may
be neither practical nor applicable considering the myriads
of pharmaceutical compounds being used and also the
difficulties in performing analytical measurements. In fact,
the quantity and quality of existing environmental
occurrence data on pharmaceuticals are insufficient for
]. In Korea, only a limited number of studies
have been published on the occurrence of pharmaceuticals
in surface water [
] or in sewage treatment plant (STP)
7, 10, 14
]. While providing valuable information,
these studies are fragmentary in nature and do not provide
sufficient data for estimating an emission rate at either the
national or local level.
This lack of data has led to the development of various
emission estimation methods which do not require
measurement data. One of the simplest estimation models is
that of U.S. Food and Drug Administration (FDA) which
employs a formula to calculate the expected introduction
concentration (EIC) of pharmaceuticals. Using a dilution
factor of 10, one can estimate the predicted environmental
concentration (PEC) from the EIC [
]. The FDA method
is based on the very conservative assumption that 100 % of
each individual pharmaceutical that is produced is
consumed and enters the publicly owned treatment works
system. The European Agency for the Evaluation of
Medical Products (EMEA) of the European Union has also
developed a formula to estimate PEC [
]. While the
EMEA method appears to be more realistic than the FDA
method in that important fate processes, such as the
excretion rate and STPs, are factored in for the emission
estimation, this approach also has room for improvement.
For example, this method does not consider the factors
affecting the emission rate in the life cycle stages of
pharmaceuticals, including distribution in the supply chain,
consumption, disposal, and waste treatment of
pharmaceuticals. Specifically, the quantity of disuse and treatment
efficiency of the disuse step in the supply chain
(pharmaceutical companies, importers, wholesalers, hospitals, and
pharmacies) can significantly alter the emission rate. The
excretion rate from the human body and the biodegradation
rate in STPs varies depending on the properties of
individual pharmaceuticals, thereby directly influencing their
emission rates. Consequently, taking these factors into
account in the emission estimation process will improve
the accuracy of the estimate. In addition, the consideration
of life cycle factors can provide information on the
contribution of individual stages or factors to the overall
emission rate, which is essential to the development of an
efficient emission reduction strategy. Although the need for
considering the life cycle of pharmaceuticals in emission
estimation has been suggested [
], it still remains to
be realized because the scientific data on their source, fate,
and transport are limited and uncertain [
of important factors in the life cycle of pharmaceuticals is
therefore an important challenge in developing emission
models of improved accuracy.
The main objectives of our study were to (1) develop a
new emission estimation model, (2) identify factors critical
to reducing emission, and (3) demonstrate the model’s
applicability for screening and priority setting.
Emission estimation model development
For model development, we first constructed a framework
of life cycle of pharmaceuticals in Korea by tracking all of
the pathways from distribution following production and/or
import to final discharge into surface water. Although the
life cycle in this present study is incomplete due to
exclusion of the production and import stages, the
uncertainty of the exclusion was assumed to be negligible
relative to the total emission rate. After the construction of the
life cycle framework, a set of equations was formulated for
calculating the amounts of pharmaceuticals involved in the
pathways and stages of the life cycle using the parameters
and variables identified to be necessary for the calculation.
To assess the accuracy of the emission estimates, the PEC
was calculated by using the emission estimates and
compared with the measured environmental concentration
(MEC) available for surface waters in Korea [
A modified version of SimpleBox (ver. 3.24a) was used
to calculate the PEC from the emission estimates. The
modification systematically included two aspects. First, the
transport of chemicals between the regional compartment
and the continental/global compartment was nullified
because it is not a relevant factor for surface water quality
in Korea, especially with chemicals of low vapor pressure.
Second, many parameter values given in the original
SimpleBox were replaced with those representing Korea’s
environmental and meteorological settings. A modified
version of SimpleTreat (ver. 3.1) was used to calculate the
biodegradation rate, removal rate by sludge separation, and
volatilization loss rate depending on the chemical
properties of the selected pharmaceuticals and average operation
conditions of STPs in Korea.
For assessing the accuracy of the model estimates, we
selected five target pharmaceuticals (acetaminophen,
cephradine, ibuprofen, mefenamic acid, and naproxen)
because (1) their MECs [
] were available to compare
with the PECs in our study, (2) they were considered to
have high management priority in Korea [
], and (3)
they were primarily used for human consumption. Details
of these pharmaceuticals are presented in Electronic
Supplementary Material (ESM) 1. The total production volume
in 2009 was calculated from the production data [
information on the active ingredient(s) in each medicinal
]. The excretion rate was obtained from the
American Society of Health–System Pharmacist’s DI ,
and the biodegradation rate in STPs and the removal rate
by sludge separation in STPs were calculated by the
Uncertainty and sensitivity analysis
For the uncertainty assessment, Monte-Carlo calculations
were conducted by using Crystal Ball (ver. 18.104.22.168.00;
Oracle Corp., Redwood, CA). As no prior information was
available on the distribution shape of the
parameters/variables used in the model, a uniform distribution was assigned
to each of parameters/variables. We performed 100,000
trials for each Monte-Carlo run and recorded five statistics
(minimum, maximum, range, median, and skewness) to
assess the uncertainty of the model estimate of the emission.
To identify sensitive parameters/variables that strongly
affect the model estimates, we used the rank correlation
coefficient normalized to 100 % [
] as an index of
sensitivity. The rank correlation coefficient was provided in
the sensitivity analysis function of Crystal Ball .
Emission estimation model
The pathways of human pharmaceuticals in the model are
depicted in Fig. 1. Pharmaceuticals produced or imported
are primarily supplied to domestic buyers, including
pharmaceutical companies, pharmacies, hospitals,
wholesalers, and others, and finally to patients through hospitals
and pharmacies which are the exclusive official pathways
to patients in Korea. All disuse from wholesalers, hospitals,
and pharmacies is principally incinerated. A
pharmaceutical supplied to a patient would be excreted following
administration or, if not administered, wasted or taken back
[Take-back program (TBP)]. A certain fraction of the
administered pharmaceutical is excreted (ER) into a toilet
connected with a septic tank (SEPT) and subsequently
transported to the nightsoil treatment plant (NISO). All of
the pharmaceuticals collected through the TBP are
]. Of the wasted pharmaceuticals, the portion
disposed of in the toilet is sent to NISO, the portion
disposed of in the sink enters the sewer to reach the STP, and
the portion disposed of in the waste bin eventually reaches
the landfill leachate treatment plant (LEACH). The residue
after incineration of the disused or taken-back
pharmaceuticals undergoes treatment by STP or LEACH. It is to
be noted, therefore, that all of the marketed human
pharmaceuticals in Korea are subject to one or more treatment
processes before entering surface waters (Fig. 1).
A total of 100 parameters/variables are used in the
model (ESM 2). The important parameters/variables
summarized in Table 1 are divided into three groups, i.e., (1)
variables for which the value is known or can be estimated
for individual pharmaceuticals, (2) parameters for which a
single fixed value [
] was used commonly for all
pharmaceuticals, and (3) parameters for which the range
was assumed due to the lack of sufficient information .
A single value was assigned to the supply rates (SR) in
Table 1 because the value is not expected to vary
considerably among pharmaceuticals.
In our study, we made four assumptions regarding the
parameters/variables in Table 1 and ESM 2. First, the
removal rate by sludge separation in LEACH and NISO, for
which values were unavailable, were assumed to be the
same as those in the STP (SLR.stp) because the sludge
removal processes are often similar. Likewise, the
biodegradation rate in LEACH was assumed to be the same as that
in STP (BR.stp). Second, the biodegradation in NISO was
assumed to be negligible. Most NISOs in Korea are
designed to perform preliminary treatments, such as solid
separation, and are connected to STPs for further treatment.
Third, the removal by incineration (INCN) was assumed to
be complete. Due to the public concern for dioxins in Korea,
the incineration temperature is required to be maintained
above 850 C, at which temperature pharmaceuticals would
be completely destroyed. Consequently, as the removal by
INCN is assumed to be complete, the landfill rate of
incineration residue (LFR.incn) becomes zero in our study.
Finally, although the return rate to the Take-back program
(TBR) appeared to vary annually, the ratio among the three
waste rates [waste bin (WR.wb), sink (WR.sink), and toilet
(WR.toilet)] were assumed to be constant at 86:7:7 as found
in the survey of 2009 [
]. By using the inputs and
assumptions described above, we identified a total of 57
model outputs, as summarized in ESM 2.
As shown in Fig. 2, the PECs calculated using the emission
estimates of the model were compared with the MECs [
The median and range of PECs were obtained from using
those of the emission rates estimated by the model and
adjusted by the modified SimpleTreat for removal
efficiency, respectively, as inputs to the modified SimpleBox.
Figure 2 shows that the PECs of the selected
pharmaceuticals agreed with the MECs for the median within one
order of magnitude.
Fig. 1 Schematic of the
estimation model in the present
study. See ESM 2 for definition
of parameters/variables in the
Mass flow along the pathways of pharmaceuticals
The emission estimation model can be used to estimate the
amounts of pharmaceuticals in various steps along the
pathways as well as the final emission into surface water.
For the model application, 14 pharmaceuticals were
selected in addition to those shown in Fig. 2. These
pharmaceuticals also meet the priority criteria applied in our
study to assess the model accuracy except that they are also
used extensively for veterinary purposes. The mass flows
of the 19 selected pharmaceuticals are summarized in
Table 2. The value in each step is the median of predicted
distribution by Monte-Carlo runs of 10,000 repetitions with
the sum of production and import (TS) of 100. The median
of TE.water was found to range from 0.6 to 40.3 % of the
TS, with the medians for roxithromycin, trimethoprim,
ciprofloxacin, cephradine, and cefadroxil having the five
highest values ([20 %).
Risk characterization and priority setting
Using the emission estimation model enabled the risk
characterization to be performed in combination with
toxicity data. For example, hazard quotients (HQ) were
calculated for the 19 pharmaceuticals used in the model
application, as shown in Fig. 3. All of the HQs of these 19
Value or equation
a Inpatients are hardly allowed to waste prescribed medications
b The minimum value of 39.1 % for AR.outpt was taken from the rate of people who take all the medications they purchase in a survey
performed in 2009 [
c According to a survey performed in 2009 [
], these rates were reported to be \10 %
d Assumed to be negligible
e 2009 data from Sewerage of Korea statistics [
f Assumed to be complete
g 2009 data from Korean pharmaceutical industry statistics [
h The ratio of 86:7:7 was taken from a survey in 2009 [
pharmaceuticals were found to be smaller than unity even
with the maximum PECs, indicating that each of the
pharmaceuticals may not pose significant ecological risk.
Nonetheless, precaution needs to be taken against the
potential combined effects [
], particularly for those
pharmaceuticals with a HQ value near one. The ranks by
HQ, PEC, toxicity, and emission of the 19 selected
pharmaceuticals are summarized in Table 3.
Factors critical to emission estimates
Due to a lack of information we were unable to assign a
single value to each of six input parameters (administration
rate of outpatients, return rate to Take-back program,
disuse inventory ratio in total
supplies/wholesalers/hospitals/pharmacies). Therefore, the influence of uncertainties
associated with these six parameters needs to be assessed
on the emission estimates of the model. Moreover, for the
assessment of uncertainty of the model estimate when no
specific pharmaceutical is specified (termed ‘‘general
uncertainty’’ hereafter), the influence of the variability of
two pharmaceutical-dependent variables (ER and BR.stp,
SLR.stp) should also be assessed. An arbitrary value of 100
for the sum of production and import (TS) was assigned to
assess the general uncertainty of the model estimate of the
As shown in Fig. 4a, the general uncertainty of the
model estimate for emission (TE.water) could vary from
0.0 to 83.0 % (median value 15.0 %) of TS. The
distribution is positively skewed, i.e., half of the TE.water
values are below 17.2 % of the range. The uncertainty of this
magnitude strongly suggests a need to acquire accurate
values for the uncertain parameters/variables, particularly
for those of high sensitivity. Based on the magnitude of the
rank correlation coefficients, the two most sensitive
parameters/variables were identified to be ER and BR.stp,
with a large gap between these and the following
parameter, TBR, as shown in Fig. 4b. The impacts of the
remaining parameters/variables were negligible.
To investigate further the influence of BR.stp and ER on
TE.water, we calculated a probability distribution of
TE.water using the Monte-Carlo technique for each of nine
(3 9 3) combinations of BR.stp and ER values of 10, 50,
and 90 %, respectively. As shown in Fig. 5a, the nine
distributions appear to differ substantially in their median
and range. For example, under conditions where ER is
90 % and BR.stp is 10 %, the median and variation are
about 98-fold greater and 12-fold wider, respectively, than
those in the case where ER is 10 % and BR.stp is 90 %.
This comparison clearly demonstrates the strong influence
of the two variables on the emission estimate. Furthermore,
as shown in Fig. 5b, both the magnitude (as represented by
the median of the distribution) and the uncertainty (as
represented by the width of the distribution) of TE.water
vary in the same direction with ER or BR.stp. For example,
the value of TE.water and its uncertainty increase with an
increasing ER or decreasing BR.stp. Therefore, greater
TE.water will tend to be predicted with a greater
uncertainty by the model. It follows that accurate values for
ER and BR.stp are particularly critical to the use of the
model because (1) they are sensitive variables which could
strongly influence the model estimate of emission for any
pharmaceutical and (2) without these accurate values, the
model estimate would be associated with larger
uncertainty, particularly for pharmaceuticals with a higher
emission potential (i.e., greater TE.water due to greater ER
and/or lower BR.stp).
Once the intrinsic properties of a pharmaceutical (ER,
BR.stp, and SLR.stp) are given, patient behavior
parameters, such as participation in a Take-back program and
administration rate of outpatient (AR.outpt), have strong
influence on the emission estimate. When the value of ER
and BR.stp is fixed at 90 and 10 %, respectively, (i.e., the
worst case of emission where TE.water ranges up to 75 %
of TS), the uncertainty of TE.water remains fairly constant,
as seen in Fig. 6, regardless of the TBR and AR.outpt
levels because the uncertainty of TE.water is primarily
governed by ER and BR.stp. As shown in Fig. 6, TE.water
decreases with TBR more sensitively at lower AR.outpt,
obviously suggesting that a consumer Take-back program
would have a lower potential for emission reduction for
pharmaceuticals with a greater administration rate.
Furthermore, the curve of TE.water at AR of 90 % in Fig. 6
indicates that take-back is likely to be of little practical
significance for emission reduction when both AR.outpt
and ER are high. For these pharmaceuticals, emission
Fig. 4 a Predicted distribution
of total emissions into surface
water, b sensitivity of the model
Sewage treatment plant
reduction can be theoretically achieved by increasing the
removal rate in STP and/or reducing their use. Increasing
the removal rate of pharmaceuticals, however, is of
secondary concern in STP operation. Therefore, reducing their
use appears to be the only viable option within the
pathways in Korea.
The uncertainties in the PECs found in our study (Fig. 2)
arise due to (1) the emission estimation model itself and the
various data used in the model and (2) the modified
SimpleBox and SimpleTreat and their input data. Furthermore,
as monitoring data on pharmaceuticals are very limited, it
is not certain if the MECs adopted in our study truly
represent the contamination levels in surface waters. Taking
these sources of uncertainty into account, the emission
model that we have developed appears to have a potential
to provide reasonable emission estimates for human
pharmaceuticals used in Korea.
Mass flow along the pathways of pharmaceuticals
As listed in Table 2, the median of TE.water for
roxithromycin, trimethoprim, ciprofloxacin, cephradine, and
cefadroxil are [20 %. These high emission rates suggest a
strong need to reduce the emission of these five
pharmaceuticals, which may be used as a rationale to prioritize
their management. The mass flow studies further showed
that the high emission rates resulted from high inflows into
NISO and subsequently through to STP. This provides
useful information for efficient management, i.e., the focus
should be placed on the means to reduce the NISO inflows.
However, it should also be noted that no difference in
INCN and LEACH resulted among the pharmaceuticals
because—due to the lack of information—the supply and
the disuse inventory ratios among suppliers and the waste
rates of outpatients were assumed to be independent of
pharmaceuticals. Once this information becomes available,
therefore, the significance of INCN or LEACH could be
discriminated in a pharmaceutical-dependent manner.
Fig. 5 a Probability distributions of TE.water at various ER and
BR.stp, b TE.water or uncertainty of TE.water with respect to ER and
BR.stp. Filled symbols TE.water, open symbols and uncertainty.
Model parameters are defined in Table 1
Risk characterization and priority setting
As can be noted in Table 3, the emission ranking and the
HQ ranking are not in accordance with each other. As the
HQ is a function of two factors, i.e., PEC and toxicity, this
discordance could arise from either or both of the two
factors. It was noted that the ranking by PEC tends to
follow that by emission, indicating that the emission rate
dictates the PEC of these 19 pharmaceuticals in water.
Therefore, the discordance between the rankings by
emission and by HQ should largely be accounted for by the
toxicity of the pharmaceuticals. These 19 pharmaceuticals
may be divided into three groups from a management
perspective. The first group includes pharmaceuticals of
high HQ ranking due to high emission (e.g., cimetidine,
roxithromycin, and amoxicillin). For this group, the
management focus should be placed on emission reduction
measures, such as usage control or Take-back programs
The second group is that of high HQ ranking primarily due
to high toxicity despite emission not being as high (e.g.,
acetaminophen, trimethoprim, and erythromycin). The use
or development of less or non-toxic alternatives would be a
solution if emission is already low. The third is the group
of pharmaceuticals of medium to low HQ ranking for
which the need of monitoring, as the first step of further
management action, should be determined depending on
the level of the respective HQ. More details on the
management approaches for each of the three groups are
presented in ESM 3.
To summarize, we have developed an emission
estimation model covering the pathways of pharmaceuticals,
including the supply chain, patient administration and
personal handling, and various treatment and disposal
processes. Based on the uncertainty and sensitivity
assessments, we have not only identified the most
influencing parameters/variables but have also drawn their
management implications. The model estimates, as
assessed using PECs, were in agreement with measured values
with a disparity less than one order of magnitude. We have
demonstrated that the model may potentially be used for
the purposes of estimating the emission rates to surface
waters and identifying factors critical to reducing these
emission rates, as well as be applied to the screening and
priority setting of pharmaceuticals.
Acknowledgments This study was funded by KEITI, NRF, and
KEI under research grants with contract numbers 412-111-003,
2011-0016767, and 2013-063, respectively.
Conflict of interest None.
1. Heberer T , Stan HJ . Determination of clofibric acid and N-(phenylsufonyl)-sarcosine in sewage, river and drinking water . Int J Environ Anal Chem . 1997 ; 67 ( 1-4 ): 113 - 24 .
2. Buser HR , Mu¨ller MD, Theobald N. Occurrence of the pharmaceutical drug clofibric acid and the herbicide mecoprop in various Swiss lakes and in the North Sea . Environ Sci Technol . 1998 ; 32 ( 1 ): 188 - 92 .
3. Buser HR , Poiger T , Mu¨ller MD. Occurrence and fate of the pharmaceutical drug diclofenac in surface waters: Rapid photodegradation in a lake . Environ Sci Technol . 1998 ; 32 ( 22 ): 3449 - 56 .
4. Daughton CG . Chemicals from the practice of healthcare: challenges and unknowns posed by residues in the environment . Environ Toxicol Chem . 2009 ; 28 ( 12 ): 2490 - 4 .
5. Rodriguez-Mozaz S , Weinberg HS . Meeting report: pharmaceuticals in water-an interdisciplinary approach to a public health challenge . Environ Health Perspect . 2010 ; 118 ( 7 ): 1016 - 20 .
6. Kim JW , Jang HS , Kim JG , Ishbashi H , Hirano M , Nasu K , et al. Occurrence of pharmaceutical and personal care products (PPCPs) in surface water from Mankyung River, South Korea . J Health Sci . 2009 ; 55 ( 2 ): 249 - 58 .
7. Choi K , Kim Y , Park J , Park CK , Kim M , Kim HS , et al. Seasonal variations of several pharmaceutical residues in surface water and sewage treatment plants of Han River, Korea . Sci Total Environ . 2008 ; 405 ( 1-3 ): 120 - 8 .
8. Choi K , Kim Y , Jung J , Kim MH , Kim CS , Kim NH , et al. Occurrences and ecological risks of roxithromycin, trimethoprim, and chloramphenicol in the Han river , Korea. Environ Toxicol Chem . 2008 ; 27 ( 3 ): 711 - 9 .
9. National Institute of Environmental Research, Korea (NIER). Development of analytical method and study of exposure of pharmaceuticals and personal care products in environment. 2006 (in Korean) . NIER, Incheon.
10. National Institute of Environmental Research, Korea (NIER). Development of analytical method and study of exposure of pharmaceuticals and personal care products in environment (II). 2007 (in Korean) . NIER, Incheon.
11. National Institute of Environmental Research, Korea (NIER). Environmental risk assessment of residual pharmaceuticals (I). 2008 (in Korean) . NIER, Incheon.
12. National Institute of Environmental Research, Korea (NIER). Environmental risk assessment of residual pharmaceuticals (II). 2009 (in Korean) . NIER, Incheon.
13. National Institute of Environmental Research, Korea (NIER). Environmental risk assessment of residual pharmaceuticals (III). 2010 (in Korean) . NIER, Incheon.
14. Behera SK , Kim HW , Oh JE , Park HS . Occurrence and removal of antibiotics, hormones and several other pharmaceuticals in wastewater treatment plants of the largest industrial city of Korea . Sci Total Environ . 2011 ; 409 ( 20 ): 4351 - 60 .
15. U.S. Food and Drug Administration (FDA). Guidance for industry-environmental assessment of human drug and biologics applications . 1998 . FDA, Washington D.C.
16. European Agency for the Evaluation of Medical Products. Guideline on the environmental risk assessment of medicinal products for human use . 2006 . European Agency for the Evaluation of Medical Products , London.
17. Halling-Sørensen B , Nielsen SN , Lanzky PF , Ingerslev F , Lu¨tzhøft HCH, Jørgensen SE . Occurrence, fate and effects of pharmaceutical substances in the environment-a review . Chemosphere . 1998 ; 36 ( 2 ): 357 - 93 .
18. Koehler A , Wildbolz C . Comparing the environmental footprints of home-care and personal-hygiene products: the relevance of different life-cycle phases . Environ Sci Technol . 2009 ; 43 ( 22 ): 8643 - 51 .
19. Daughton CG , Ruhoy IS . Environmental footprint of pharmaceuticals: the significance of factors beyond direct excretion to sewers . Environ Toxicol Chem . 2009 ; 28 ( 12 ): 2495 - 521 .
20. National Institute of Environmental Research, Korea Republic. Development of analytical method and study of exposure of residual pharmaceuticals (II). 2009 (in Korean) . NIER, Incheon.
21. Korea Pharmaceutical Manufacturers Association (KPMA). Pharmaceutical production in 2009. 2010 (in Korean) . KPMA , Seoul.
22. Korea Pharmaceutical Information Center. Monograph. Available at: http://www.health. kr. Accessed 4 June 2012 .
23. Druginfo . Monograph. Available at: http://www.druginfo.co. kr. Accessed 4 June 2012 .
24. American Society of Health-System Pharmacists. AHFS DI monographs . Available at http://www.drugs. com/monograph. Accessed 4 June 2012 .
25. Gentry B . Oracle crystal ball user manual , release 22.214.171.124 .00. Oracle ; 2008 .
26. Korea Zero Waste Movement Network (KWMN). Proper usage of domestic pharmaceuticals and pilot project of pharmaceutical Take-Back program. 2009 (in Korean) . KWMN , Seoul.
27. Korea Zero Waste Movement Network (KWMN). Proper usage of domestic pharmaceuticals and pilot project of pharmaceutical Take-Back program. 2010 (in Korean) . KWMN , Seoul.
28. Korea Pharmaceutical Manufacturers Association (KPMA). Statistics of Pharmaceutical Industry . 2010 (in Korean) . KPMA , Seoul.
29. Ministry of Environment, Korea. 2009 Statistics of sewerage. 2010 (in Korean) . Ministry of Environment , Seoul.
30. Korea Zero Waste Movement Network (KWMN). Treatment method of pharmaceuticals disposed by pharmaceutical-using institutes. 2009 (in Korean) . KWMN , Seoul. Available at: http:// www.waste21.or.kr.
31. Cleuvers M. Mixture toxicity of the anti-inflammatory drugs diclofenac, ibuprofen, naproxen, and acetylsalicylic acid . Ecotoxicol Environ Saf . 2004 ; 59 ( 3 ): 309 - 15 .