ndma formation from amine-based pharmaceuticals – impact from prechlorination and water matrix
TRANSCRIPT
ww.sciencedirect.com
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 7
Available online at w
journal homepage: www.elsevier .com/locate /watres
NDMA formation from amine-based pharmaceuticals e
Impact from prechlorination and water matrix
Ruqiao Shen*, Susan A. Andrews
Department of Civil Engineering, University of Toronto, 35 St. George St., Toronto, Ontario, Canada M5S 1A4
a r t i c l e i n f o
Article history:
Received 1 October 2012
Received in revised form
6 February 2013
Accepted 8 February 2013
Available online 17 February 2013
Keywords:
NDMA
Prechlorination
Ranitidine
Sumatriptan
NOM
* Corresponding author. Tel.: þ1 4169783141;E-mail addresses: [email protected]
0043-1354/$ e see front matter ª 2013 Elsevhttp://dx.doi.org/10.1016/j.watres.2013.02.017
a b s t r a c t
The presence of N-nitrosodimethylamine (NDMA) in drinking water is most commonly
associated with the chloramination of amine-based precursors. One option to control the
NDMA formation is to remove the precursors via pre-oxidation, and prechlorination is
among the most effective options in reducing NDMA formation. However, most of the
findings to-date are based on single-precursor scenarios using the model precursor
dimethylamine (DMA) and natural organic matter (NOM), while few studies have consid-
ered the potential interactions between water matrix components and the target pre-
cursors when investigating the prechlorination impact. Specifically, little is known for the
behaviour of amine-based pharmaceuticals which have recently been reported to
contribute to NDMA formation upon chloramination. This work demonstrates that pre-
chlorination can affect both the ultimate NDMA conversion and the reaction kinetics from
selected pharmaceuticals, and the nature and extent of the impact was compound-specific
and matrix-specific. In the absence of NOM, the NDMA formation from most pharma-
ceuticals was reduced upon prechlorination, except for sumatriptan which showed a
consistent increase in NDMA formation with increasing free chlorine contact time. In the
presence of NOM, prechlorination was shown to enhance initial reactions by reducing the
binding between NOM and pharmaceuticals, but prolonged prechlorination broke down
NOM into smaller products which could then form new bonds with pharmaceuticals and
thus inhibit their further conversion into NDMA.
ª 2013 Elsevier Ltd. All rights reserved.
1. Introduction drinking water (Health Canada, 2010). USEPA also placed it on
The presence of N-nitrosodimethylamine (NDMA) in drinking
water has been commonly associated with the disinfection
process, especially chloramination. NDMA is a highly muta-
genic compound and a potential human carcinogen, with a
10�6 lifetime cancer risk associated with a drinking water
concentration of 0.7 ng/L (EPA IRIS, 1993). The exposure to
NDMA through drinking water has become a concern espe-
cially for utilities that apply chloramine as the secondary
disinfectant. Health Canada has recently proposed a
maximum acceptable concentration for NDMA of 40 ng/L in
fax: þ1 4169783674..cn (R. Shen), sandrews@
ier Ltd. All rights reserve
the drinking water contaminant candidate list 3 (CCL3)
together with four other nitrosamines (USEPA, 2009). NDMA is
currently regulated in drinking water in several provinces and
states across North America, including Ontario (9 ng/L; MOE,
2003), Massachusetts (10 ng/L; MassDEP, 2004), and Califor-
nia (10 ng/L; OEHHA, 2006).
During drinking water treatment processes, NDMA is most
commonly formed via the slow reaction between chloramines
(especially dichloramine) and amine-based precursors
(Schreiber and Mitch, 2006a). NDMA is also formed through a
nitrosation mechanism at acidic pH (Choi and Valentine,
civ.utoronto.ca (S.A. Andrews).d.
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 7 2447
2003), but this mechanism is of less importance in drinking
water due to the generally low nitrite concentrations and the
neutral/basic pH. Moreover, ozonation can also lead to the
formation of NDMA (Andrzejewski et al., 2008; Oya et al.,
2008); an especially high yield was observed for dime-
thylsulfamide, a degradation product of the fungicide toly-
fluanid (Schmidt and Brauch, 2008).
Early studies on NDMA typically used the model precursor
dimethylamine (DMA; Gerecke and Sedlak, 2003) and natural
organic matter (NOM; Chen and Valentine, 2007; Dotson et al.,
2007; Gerecke and Sedlak, 2003; Mitch and Sedlak, 2004),
however chloramination of these precursors typically gave
low yields of NDMA and thus cannot always account for all of
the precursors present in natural waters. More recently,
higher levels of NDMA formation have been associated with
wastewater-impacted surface water (Krasner, 2009; Schreiber
and Mitch, 2006b; Shah et al., 2012), indicating the contribu-
tion from anthropogenic compounds. Several studies have
linked NDMA formation to quaternary amines used in per-
sonal care products (Kemper et al., 2010), pharmaceuticals and
pesticides (Le Roux et al., 2011, 2012a b; Shen and Andrews,
2011a,b), as well as some amine-based polymers and resins
(Kohut and Andrews, 2003; Mitch and Sedlak, 2004; Najm and
Trussell, 2001; Wilczak et al., 2003). Mitch and Schreiber (2008)
have proposed that the NDMA formation from tertiary amines
proceeds via a chlorine transfer reaction to release the DMA
which is then subsequently oxidized to NDMA. However,
NDMA formed fromDMA alone cannot explain the high yields
from some tertiary amines such as ranitidine. A recent study
by Le Roux et al. (2012b) has proposed an alternative mecha-
nism for NDMA formation from chloramination of ranitidine,
which involves a direct substitution on the DMA group of ra-
nitidine and can well explain its high yield of NDMA.
In practice, chloramine is usually used as the secondary
disinfectant following primary disinfection (e.g., Cl2, UV, and
O3). Moreover, instead of using preformed monochloramine,
most utilities that perform chloramination typically apply free
chlorine first, followed by the addition of ammonia to form
chloramine on site. Compared with chloramine alone, the
application of pre-oxidation may modify or destroy the pre-
cursors and release transformation products that may or may
not react with the subsequent secondary disinfectant.
Generally, the application of preoxidation prior to chlor-
amination has been reported to reduce the NDMA formation
from DMA and NOM. Preoxidation processes have included
prechlorination (Chen and Valentine, 2008; Charrois and
Hrudey, 2007; Mitch et al., 2010), O3 and O3/H2O2 (Chen and
Valentine, 2008; Pisarenko et al., 2012), ClO2 (Lee et al., 2007),
ferrate (Lee et al., 2008), and KMnO4 (Chen and Valentine,
2008). Among all the options, ozone and chlorine were found
to be most effective in reducing NDMA formation (Shah et al.,
2012). However, oxidation of precursors does not necessarily
lead to the reduction in NDMA formation. In some cases,
prechlorination may increase NDMA formation at lower ex-
posures due to insufficient oxidation (Chen and Valentine,
2008; Shah et al., 2012). It has also been observed that UV
and UV/H2O2 pretreatment have increased the NDMA forma-
tion from amine-based polymers (e.g., polyDADMAC, epi-
DMA), possibly due to the increased degradation of the poly-
mers releasing more NDMA precursors (Harvey, 2009). More
recently, Radjenovic et al. (2012) have reported that the
oxidation products of the pharmaceutical tramadol by UV and
UV/H2O2 have higher NDMA formation potentials (FPs) than
the parent compound. Thus, the impact of preoxidation on the
formation of NDMA requires further investigation.
Currently, most of the findings regarding the preoxidation
impact have been based on DMA and NOM. Several studies
used treated wastewater as a “precursor pool”, but few specific
compounds have been studied separately, especially
pharmaceutical-based precursors. Moreover, very little infor-
mation is available in terms of how the preoxidation process
might affect the NDMA formation kinetics. This study in-
vestigates the impact of prechlorination on the NDMA for-
mation from eight selected pharmaceuticals, especially the
impact on their reaction kinetics. The eight pharmaceuticals
were selected because of their relatively high NDMA FPs
among the twenty pharmaceuticals and personal care prod-
ucts (PPCPs) tested in a previous study by the authors (Shen
and Andrews, 2011a). This work also compares the prechlori-
nation impact with and without the presence of NOM, and
looks into how the interactions in between NOM, pharma-
ceuticals, and free chlorine could affect the NDMA formation
from selected pharmaceuticals. Relatively high pharmaceu-
tical concentrations (6.8e11.1 mg/L) compared with their ex-
pected environmental levels were applied in this study in
order to be able tomeasure the differences in NDMA formation
under different disinfection conditions. However, Shen and
Andrews (2011b) demonstrated that the NDMA formation ki-
netics in a real water matrix was relatively independent of the
initial pharmaceutical concentration because chloramine was
present in large excess relative to the pharmaceutical con-
centrations; therefore similar reaction kinetics are expected
for selected pharmaceuticals at their environmental levels.
Findings from this work could be of particular concern for
water reuse processes where much higher concentrations of
pharmaceuticals might be subjected to chloramination.
2. Materials and methods
Chemical structures of the eight selected pharmaceuticals are
illustrated in Fig. 1. The pharmaceuticals (25 nM of each) were
dosed into selected water matrices (raw, not filtered) individ-
ually and subjected to different disinfection strategies. The
chloramination experiments (preformed chloramine) were
carried out under the same simulated distribution system
(SDS) conditions as applied in Shen and Andrews (2011b). The
sequential disinfection experiments (prechlorination fol-
lowed by chloramination) employedmodified SDS conditions,
where a sodiumhypochlorite (NaClO) solutionwas first added,
followed by the addition of ammonia chloride (NH4Cl) to form
chloramine after a range of target free chlorine contact times
(0.5e120 min). The resulting chloramine concentration at the
point of NH4Cl addition was the same as that which was
applied in the preformed chloramination experiments for
each matrix (i.e., 2.5 � 0.2 mg/L plus the 24 h chloramine de-
mand for eachmatrix). The initial free chlorine dosage and the
NH4Cl dosage for each matrix were determined via pre-
liminary chlorine/chloramine demand tests (see details in
Supporting Information). All of the other conditions remained
Fig. 1 e Structures of selected pharmaceuticals.
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 72448
the same as the SDS conditions (i.e., 21 � 1 �C; pH ¼ 7.0 � 0.1;
Cl2: N mass ratio ¼ 4.2:1). Further details concerning the
preparation of stock solutions, working solutions, and the
NDMA analysis have been described in Shen and Andrews
(2011a,b).
NDMA-FPs and kinetics evaluations were both conducted
under the (modified) SDS conditions described above. The 24 h
NDMA-FPwas determined for all eight pharmaceuticals in lab-
gradeMQ (Milli-Q�, MilliPore, Etobicoke, Ontario) water, where
samples (dosed with pharmaceuticals) and blanks (water ma-
trix only) were both prepared in triplicate. The NDMA forma-
tion kinetics was monitored for ranitidine and sumatriptan,
where samples were prepared in triplicate along with one
blank control (using the respective water matrix) at each time
point. Error bars in all graphs demonstrate the variability due
to multiple formation potential tests (n ¼ 3 unless otherwise
specified) under the same reaction conditions.
The kinetic experiments were performed in three water
matrices; the basic water quality parameters for each source
are summarized in Table 1. Lake Ontario water and Otonabee
River water samples were taken from the influent of two
drinking water treatment plants in September 2011 and
January 2012, respectively. The methods for measuring these
basic water quality parameters have been described in Shen
and Andrews (2011b). Use of raw water instead of partially-
treated water as the experimental matrix provided the study
Table 1 e Water matrix source and basic water quality measu
Water matrix Water source TOC (mg/L)
Milli-Q� (MQ) MilliPore; Etobicoke, Ontario 0 7.5
Lake Lake Ontario Ajax, Ontario 2.3 � 0.1a 8.2
River Otonabee River, Peterborough,
Ontario
6.2 � 0.1 7.8
a These basic water quality parameters (except for turbidity) were mea
months) and the values here are the average and standard deviation of 5
b Turbidity was recorded at the water treatment plant on the day of sam
with the maximum amount of unmodified NOM for the cur-
rent investigations. Future studies could employ partially-
treated water to account for the removal or modification of
someNOM fromvarious treatment processes or process trains.
In order to better characterize the water matrix, NOM
components in each raw water source were analysed using
size-exclusion liquid chromatography-organic carbon detec-
tion (LC-OCD) at the University of Waterloo (ON, Canada). The
LC-OCD uses a size-exclusion column to separate the hydro-
philic NOM fractions. The hydrophobic matter does not elute
within the limited measuring time of 120 min, and thus is
calculated as “DOCminus hydrophilic DOC”. A typical LC-OCD
chromatograph for surface water is shown in Fig. S1 in
Supporting Information. The major hydrophilic NOM frac-
tions are eluted in the order of biopolymers, humics, building
blocks, low-molecular-weight (LMW) acids, and LMW neu-
trals. Further details about the LC-OCD system are described
by Huber et al. (2011). Proprietary software was used for data
acquisition and processing (ChromCalc, DOC-LABOR, Karls-
ruhe, Germany). The LC-OCD results for the selected water
matrices are summarized in Table 2. In addition, several
chlorine- and/or chloramine-treated water samples (lake and
river; from the blank controls) were collected along with the
kinetic experiments to investigate the potential change of
NOM upon the disinfection treatment. The chlorine/chlora-
mine residual was quenched with sodium thiosulfate
rements.
pH Alkalinity(mg/L)
UV254 (cm�1) SUVA
(L/mg-m))Turbidityb
(NTU)
� 0.1 1.8 � 0.3 0.000 0.0 0.0
� 0.1 95.4 � 1.8 0.023 � 0.002 1.0 � 0.1 0.67
� 0.1 94.9 � 2.9 0.162 � 0.002 2.6 � 0.1 0.46
sured every week during the sample storage period (no more than 2
e10 measurements.
pling.
Table 2 e LC-OCD results for the selected water matrices (Unit: mg/L carbon).
Watermatrix
DOC HydrophobicDOC
HydrophilicDOC
Hydrophilic DOC fractions
Biopolymers Humics Building blocks LMW neutrals LMW acids
MQ 0.07 0.03 0.04 0.01 0.000 0.003 0.02 0.004
Lake 2.20 0.20 2.00 0.34 0.90 0.46 0.19 0.10
River 5.90 0.20 5.70 0.34 3.82 0.72 0.62 0.20
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 7 2449
(Na2S2O3) at the dosage of twice the stoichiometric ratio of
chlorine (personal communication, University of Waterloo);
the samples were then filtered through a 0.45 mm filter paper
and shipped to University of Waterloo for LC-OCD analysis.
Samples were stored at 4 �C for no more than one week prior
to the analysis.
3. Results and discussion
3.1. Prechlorination impacts in MQ water
3.1.1. 24 h NDMA-FP upon prechlorinationTo begin investigating the impacts of prechlorination, 24 h
NDMA-FPs following sequential chlorine and chloramine
Fig. 2 e 24 h NDMA-FP from eight pharmaceuticals upon seque
water).
disinfection were determined for eight selected pharmaceu-
ticals in MQ water (Fig. 2). The 0 min pre-Cl condition repre-
sents the use of preformed chloramine. In general, the NDMA
formation from most of the pharmaceuticals (except for su-
matriptan and diltiazem) decreased with increasing pre-
chlorination contact time ( p-value < 0.0001 except for
doxylamine ( p-value ¼ 0.0034); ANOVA analysis, 95% confi-
dence level). However, the rate and degree of reduction in
NDMA formation varied among the different compounds. For
example, ranitidine and nizatidine responded very quickly to
chlorine, with a 50% reduction in NDMA formation achieved
by 0.5 min and 3 min of prechlorination, respectively. For
tetracycline, a similar amount of reduction required 30 min of
prechlorination. For the three structurally similar H1-antihis-
tamines (carbinoxamine, chlorphenamine, and doxylamine),
ntial chlorination and chloramination disinfection (MQ
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 72450
no significant reduction was observed until the chlorine con-
tact time increased to 60e120 min. Sumatriptan was the only
pharmaceutical showing consistently increasing NDMA for-
mation as the prechlorination contact time increased ( p-
value < 0.0001). For diltiazem, a very short chlorine contact
time (30 s) seemed to reduce the NDMA formation, but further
increases in chlorine contact time increased the NDMA for-
mation to levels approaching those observed when preformed
monochloramine was used. Detailed calculations for the sta-
tistical analyses are provided in Supporting Information
(Table S2).
These initial tests showed that not all of the NDMA for-
mation potential of the target compounds was destroyed by
prechlorination. Further testingwould focus on ranitidine and
sumatriptan as representatives of pharmaceuticals that
showed either decreases or increases in NDMA formation,
respectively, following prechlorination.
3.1.2. NDMA formation kinetics upon prechlorinationIn order to further investigate the impact of prechlorination
on NDMA reaction kinetics, the NDMA formation was moni-
tored over time for ranitidine and sumatriptan following
sequential chlorination and chloramination (Fig. 3). Markers
in the figures represent the measured NDMA molar conver-
sion values and the lines represent model-predictions.
The three-parameter kinetic model used was of the form:
Y ¼ q
1þ 10k�ðLag�tÞ
where Y is the NDMAmolar conversion at given reaction time
(t); q is the ultimate NDMA molar conversion; Lag is the time
required to achieve 50% of the ultimate molar conversion and
thus associated with the length of the initial lag phase; k is the
pseudo-first order reaction rate constant. Further details
about the model have been discussed in Shen and Andrews
(2011b). The 0 min prechlorination time represents the use
of preformed chloramine.
The formation kinetics results were in good agreement
with the 24 h formation potential results (Fig. 2). NDMA for-
mation from ranitidine was significantly reduced upon pre-
chlorination even with a short free chlorine contact time at
3 min, and no further reduction was achieved as the free
chlorine contact time increased. In contrast, the ultimate
NDMA conversion from sumatriptan increased with free
Fig. 3 e NDMA formation kinetics from ranitidine and sumatript
MQ water (Error bars represent the maximum and minimum v
chlorine contact time, and the initial lag phase was reduced
upon prechlorination. The estimated model parameters for
both compounds are shown in Table S3 in Supporting
Information, and the associated analytical analyses on the
estimated model parameters are summarized in Tables S4
and S5.
3.2. Prechlorination impacts in real water matrices
Previously, the authors demonstrated that water matrix
components could slow down the initial NDMA formation
from pharmaceuticals (i.e., much longer initial lag phase with
the presence of NOM), while they had less impact on the ul-
timate NDMAmolar conversion (Shen andAndrews, 2011b). In
the present work, the impact of prechlorination on the NDMA
formation kinetics was also investigated in natural water
samples dosed with selected pharmaceuticals. Some studies
have reported the possible impact of bromide ion on NDMA
formation (Chen et al., 2010; Le Roux et al., 2012a; Valentine
et al., 2005); however, the bromide levels in the waters
selected for this work (Lake Ontario, w40 mg/L (Comerton
et al., 2006); Otonabee River, <11.4 mg/L (Woodbeck, 2007))
aremuch lower comparedwith the concentrations reported to
be significant in literature (mg/L). Therefore the role of bro-
mide ion in the NDMA formation from selected pharmaceu-
ticals is not investigated in this study. The results for
ranitidine and sumatriptan in water from Lake Ontario and
the Otonabee River are summarized in Figs. 4 and 5, respec-
tively. Similar to Fig. 3, markers represent the measured
NDMA molar conversion values and the lines represent
model-predicted conversions. The estimated model parame-
ters in different water matrices are included in Table S3 for
easier comparison with similar data determined using MQ
water; similarly, the associated statistical analyses on the
model parameters are provided in Tables S4 and S5.
For ranitidine, prechlorination impacts observed in the
lake water (Fig. 4) showed similar trends to those that were
observed in MQ water (Fig. 3); the ultimate NDMA formation
was significantly reduced upon prechlorination, and increases
in free chlorine contact time did not cause further reductions
in NDMA formation.Moreover, the initial lag period in the lake
water was significantly shortened by more than half upon
prechlorination, and the estimated model parameters in MQ
an upon sequential chlorine and chloramine disinfection in
alues under the same reaction conditions (n [ 2)).
Fig. 4 e NDMA formation kinetics from ranitidine upon sequential chlorine and chloramine disinfection in Lake Ontario and
Otonabee River water.
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 7 2451
and lake water were comparable with each other when pre-
chlorination was applied at the same contact time (Table S3).
Previously, a NOMepharmaceutical binding hypothesis was
proposed to explain the initial lag phase observed in realwater
matrices (Shen and Andrews, 2011b). The results here suggest
that chlorination may destroy some of the binding between
NOM and ranitidine and thus enhance the initial reaction
between ranitidine and chloramine to form NDMA.
However, the results in the river water were quite different
for ranitidine. Upon 3 min prechlorination, the ultimate
NDMA molar conversion was reduced by 23% and the initial
lag period was shortened from 31 h to 4 h. When the chlorine
contact time further increased to 10 min, the NDMA conver-
sion was significantly inhibited to 12.5%, and very little NDMA
(molar conversion less than 3%) was detected at longer chlo-
rine contact times. It is not likely that prolonged chlorination
simply destroyed the NDMA precursor (ranitidine in this case)
because similar results would have been obtained in MQ and
the lake water. Considering the NOMeranitidine binding
theory, the results suggest that prolonged chlorination might
further break down the NOM to form smaller products which
may rebind with ranitidine and thus inhibit the NDMA for-
mation. However, it is still unclear why this “rebinding” was
Fig. 5 e NDMA formation kinetics from sumatriptan upon seque
and Otonabee River water.
not observed in the lake water. One possible explanation
could be the much higher TOC level in the river water, thus
more NOMeranitidine complexes were formed.
Sumatriptan’s response to prechlorination in naturalwater
matrices (Fig. 5) was quite different compared with that
observed in MQ water (Fig. 3), except for the common trend in
all of the matrices of a reduced initial lag phase upon pre-
chlorination. However, unlike the consistently increasing
NDMA conversion from sumatriptan in MQwater, there was a
maximum NDMA molar conversion observed upon 30 min of
prechlorination. Further increases in chlorine contact time
reduced the ultimate molar conversion. This phenomenon
might also be explained by the proposed rebinding theory that
prolonged chlorination could breakdown NOM molecules to
smaller products and thus enhance their rebinding with su-
matriptan, as will be discussed further in Section 3.3.3.
In summary, prechlorination was shown to affect both the
ultimate NDMA conversion (q) and the reaction kinetics (Lag
and k). As compared in Fig. 6, the impact was both compound-
specific and matrix-specific, thus requiring case-by-case
investigation. Currently, most studies have found that pre-
chlorination with enough CT (concentration � contact time)
has, in general, been able to reduce the NDMA formation.
ntial chlorine and chloramine disinfection in Lake Ontario
Fig. 6 e Comparison of the kinetic model parameters upon sequential chlorine and chloramine disinfection in three water
matrices (Error bars represent the 95% confidence interval).
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 72452
However, only single-precursor scenarios have been consid-
ered in most studies, i.e., NOM or DMA as the only precursor.
This work suggests that the potential interactions between
NOM and other precursors could also affect NDMA formation.
Moreover, an enhanced NDMA formation rate upon prechlori-
nation has not been reported previously, which is a new
contribution to the field.
3.3. NOMepharmaceuticaleCl2 interactions
Both free chlorine and NOM have been shown to individually
affect the conversion of selected pharmaceuticals into NDMA.
Since systems involving free chlorine, NOM, and the target
pharmaceuticals are more complicated, it is important to
investigate the interactions in between them to further
understand the impact from matrix components as well as
prechlorination.
3.3.1. NOMepharmaceutical interactionsIn the absence of prechlorination, the matrix components
were found to slow down the initial NDMA formation (i.e.,
longer initial lag phases were associated with higher TOC and
SUVA), while they had less impact on the ultimate NDMA
molar conversion (Fig. 6, pre-Cl ¼ 0 min). The results are in
good agreement with a previous study by the authors (Shen
and Andrews, 2011b), where the potential interactions be-
tween NOM and selected pharmaceuticals have been dis-
cussed. The formation of NOMepharmaceutical complexes is
most likely due to the electrostatic attraction between the
negatively charged NOM surface (De Ridder et al., 2011) and
the positively charged amines. Although there is no direct
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 7 2453
spectroscopic evidence to confirm such binding in the
aqueous phase, the NOMepharmaceutical binding hypothesis
has been suggested by other researchers to explain the
enhanced pharmaceutical removal during treatment pro-
cesses like coagulation and membrane filtration (Ballard and
Mackay, 2005; Diemert, 2012; Stackelberg et al., 2007; Vieno
et al., 2006; Westerhoff et al., 2005). A recent study by Ding
et al. (2012) has developed an efficient solid phase extraction
method to determine the association of a variety of pharma-
ceuticals with dissolved humic acids, in which two of the
pharmaceuticals studied had the DMA functional group in
their structures. The difference in their affinity with NOMwas
well explained by the different amounts of positively charged
sites in the pharmaceuticals (e.g., amine groups) and nega-
tively charged sites in the humic acids (i.e., carboxylic and
phenolic moieties). Although the concentrations applied in
Ding et al. (2012) were relatively high (i.e., 100e2500 mg/L of
pharmaceuticals; 40e80 mg/L of NOM) compared with the
levels of pharmaceuticals and NOM in natural waters, their
study proves the possibility of binding between amine-based
pharmaceuticals and NOM in the aqueous phase.
3.3.2. Cl2epharmaceutical interactionsResults from this work showed that prechlorination can either
increase or decrease NDMA formation from selected phar-
maceuticals. Because the NDMA formation can be described
as an electrophilic reaction on the nitrogen atom of the amine
group (Mitch and Schreiber, 2008), theoretically chlorination
could enhance the reaction if it enhances the electron density
on the amine group, and vice versa. Therefore, it is important
to evaluate the different reactivity of selected pharmaceuti-
cals towards free chlorine in order to determine how the final
NDMA conversion will be affected by prechlorination.
Table 3 e Comparison of chlorine reactivity towards ranitidine
Ranitidine
DMA group protonated on the parent
compound at given pH (7.0 � 0.1)?
w94%
Primary chlorine attack site of the
parent compound
Sulphur
Possible fragmentation upon chlorination? Yes
DMA group protonated on the breakdown
product(s) at given pH (7.0 � 0.1)?
w92%
Primary chlorine attack site(s) of the
breakdown product(s)
DMA
Electron density of DMA group increased? No
Ultimate NDMA molar conversion
upon prechlorination
Reduced
Moreover, the DMA group is not the only functional group on
the pharmaceuticals that could react with chlorine. If the
groups close to the DMA group are modified upon chlorina-
tion, they may also impact the electron density of the amine
group, and thus affect the NDMA conversion. Among the eight
pharmaceuticals being tested, ranitidine and sumatriptan had
relatively highNDMA conversions but showed opposite trends
when being subjected to prechlorination. Thus the following
discussion mainly focuses on these two pharmaceuticals.
Deborde and von Gunten (2008) have reviewed the kinetics
and mechanisms for the chlorination of organic compounds.
Generally, rate constants for the reaction of chlorine with
sulphur-moieties are typically 1e2 orders of magnitude higher
thanwith amines. According to this review, themost potential
attack site of chlorine on ranitidine would be the sulphur,
forming ranitidine sulfoxide (S]O). Then the ranitidine sulf-
oxide may go through CeS fragmentation upon further chlo-
rination, resulting in the formation of smallermolecules (Table
3). The CeS fragmentation upon chlorination was confirmed
by Le Roux et al. (2012b) for ranitidine, and the breakdown
product (5-(dimethylaminomethyl)-2-furanmethanol (DFUR))
is still a potential NDMA precursor, with NDMA molar con-
version being reported from 50% to 80% upon preformed
chloramination (Schmidt et al., 2006; Le Roux et al., 2012a).
However, the DMA group on DFUR may now become the pri-
mary chlorine attack site. Abia et al. (1998) also proposed a
reaction mechanism for the chlorination of tertiary amines,
including an elementary step in which a positive charge is
developed on theN-atom via a chlorine transfer. This supports
the results of a previous study which has suggested that pre-
chlorination can significantly reduce the NDMA formation
fromDMA due to the partially formed Cl-DMA that inhibits the
electrophilic attack on the amine group (Schreiber and Mitch,
and sumatriptan.
Sumatriptan
w100%
C2 site of the indole ring
No
/
/
Yes
Enhanced
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 72454
2005). Le Roux et al. (2012b) have also suggested that the
chlorine attack on the amine group could reduce the NDMA
conversion from ranitidine because it limited the direct sub-
stitution on the DMA group.
Unlike ranitidine, the NDMA conversion from sumatriptan
was enhanced upon prechlorination. As indicated in Table 3,
the sulphur moiety of sumatriptan is already fully oxidized
(eSO2), therefore the proposed CeS fragmentation for raniti-
dine may not occur. Moreover, the DMA group is almost
completely protonated at pH 7 (pKa ¼ 9.6; 99.8% protonated),
while the chlorine reaction is only significant with neutral
amines (reaction rate constants in the range of
101e102 M�1 s�1 at pH 7 (Deborde and von Gunten, 2008)). As a
result, the DMA group may not be the primary chlorine attack
site of sumatriptan at the given reaction conditions. In addi-
tion, Xu et al. (2001) have suggested that the C2 position of
indole can be activated for oxidationwhen the C3 site has an R
group. In the case of chlorination, the C2 position could be
substituted with chlorine (Gilow and Burton, 1981) and the
reactionwould occur very fast, with a conservative estimation
of the reaction rate constants to be at least in the range of
101e102 M�1 s�1 at pH 7 (Lin and Carlson, 1984). The
substituted chlorine on the C2 position may then increase the
electron density at C3 position due to the conjugation effect
between the double bond and the lone electron pair on the Cl
(Carey, 2008), thus enhancing NDMA formation from suma-
triptan. Although this reaction is plausible from the literature
cited above, it would benefit from mass spectral confirmation
of the intermediate species which, unfortunately, was not
possible as part of the current study.
In summary, twomajor factors should be considered when
evaluating the effect of prechlorination on the NDMA con-
version from amine-based pharmaceuticals: i) if the DMA
group on these pharmaceuticals is the primary chlorine attack
site; ii) if the DMA group is protonated (pHepKa relevant).
Table 3 compares the differences in the chlorine reactivity
towards ranitidine and sumatriptan.
Usually chlorine would cause a small modification on the
parent compound rather than a complete oxidation. The
Fig. 7 e Potential interactions in between NO
modification may change the pharmaceutical reactivity to-
wards the subsequent chloramine, but it could also change
their binding potential with NOM which would affect the
availability of the pharmaceuticals as well. For example, if
chlorination increases the electron density on the amine
group of sumatriptan (Table 3), it will not only enhance the
NDMA conversion from sumatriptan (i.e., favouring the elec-
trophilic attack), but also reduce the binding with the nega-
tively charged NOM surface thus improving the initial
reaction. The latter might also contribute to the much short-
ened initial lag phase for sumatriptan upon prechlorination in
lake and river water (Fig. 5).
3.3.3. Cl2eNOM interactionsChlorine can also react with NOM and modify its surface
properties, which would then affect the binding potential
between NOM and pharmaceuticals. The potential in-
teractions between chlorine and NOM were investigated via
LC-OCD analysis. Water samples were collected along with
the chloramination and sequential disinfection experiments
(from the blank control samples in each set of kinetic exper-
iments) to investigate the potential change of NOM compo-
nents upon the treatment.
For the rawwater samples used in this study (lake and river
water), the majority of NOM was hydrophilic (more than 90%
of the total DOC), and the major fractions were humic sub-
stances (HS) and building blocks (accounting for 70%e80% of
the total hydrophilic NOM). In general, DOC was not removed
by chlorination or chloramination at the applied reaction
conditions, but there were changes in the major NOM frac-
tions (i.e., HS and building blocks). Typically, degradation of
HS was observed to lead to the formation of building blocks,
and further breakdown of building blocks would lead to the
increase in LMW fractions (data not shown). At the employed
reaction conditions, free chlorine itself did not change any
NOM fractions significantly; however, it facilitated the trans-
formation of HS into building blocks when NH4Cl was added
subsequently to form chloramine. For example, following
120 min of prechlorination, the HS in lake water started to
M, pharmaceuticals, and free chlorine.
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 7 2455
degrade after 2 h of NH4Cl addition. However, without pre-
chlorination, the reduction in HS occurred only after 4 h of
chloramination (Fig. S2, top left).
This trendwas further confirmed in the river water (Fig. S2,
bottom two). The application of prechlorination enhanced the
degradation of HS, and the reduction was improved as the
chlorine contact time increased. Compared with preformed
chloramine (i.e., pre-Cl ¼ 0 min), there were more building
blocks formed when prechlorination was applied; as chlor-
amination contact time increased, the building blocks first
increased (from the breakdown of HS) and then decreased
subsequently (further breakdown to LMW fractions). Specif-
ically, the highest increase in building blocks occurred at
30 min of prechlorination followed by chloramination; this is
in agreement with the kinetic experiments for sumatriptan in
lake and river water, where the maximum NDMA molar
conversion was also observed upon 30 min of prechlorination
(Fig. 5). It was proposed that prolonged prechlorination may
destroy some NOM fractions and the breakdown products
may rebind with the pharmaceutical and thus inhibit its
conversion into NDMA. The LC-OCD results suggest that pro-
longed chlorine contact time does lead to early breakdown of
building blocks, yielding more LMW molecules which may
rebind with pharmaceuticals and inhibit their further con-
version into NDMA.
Fig. 7 summarizes the possible interactions in between
NOM, pharmaceuticals, and free chlorine. Without the appli-
cation of prechlorination, the presence of NOM usually does
not affect the ultimate NDMA conversion, but will affect the
reaction rate due to the formation of NOMepharmaceutical
complexes. However, free chlorine could affect both the
NDMA conversion and the reaction rate by altering the prop-
erties of the pharmaceuticals and/or NOM, as well as the
binding potential between the two.
4. Conclusions
Previous research that looked into prechlorination effects
mostly considered the single-precursor scenarios, while very
few took into consideration the interactions between matrix
components and the precursors. Moreover, little is known in
terms of the prechlorination impact on the NDMA formation
kinetics. This work demonstrates that prechlorination affects
both the ultimateNDMA conversion and the reaction rate, and
investigates how the interactions in between NOM, pharma-
ceuticals, and free chlorine could affect the NDMA formation.
The prechlorination impact is compound-specific and
matrix-specific, requiring case-by-case investigation. In gen-
eral, prechlorination tends to reduce the NDMA conversion
from ranitidine while it increases that from sumatriptan. In
the presence of NOM, prechlorination may enhance the initial
reaction by breaking the binding between NOM and pharma-
ceuticals, but prolonged prechlorinationmay also break down
NOM into smaller products which may rebind with pharma-
ceuticals and thus inhibit its further conversion into NDMA. In
addition, an enhanced NDMA formation rate upon prechlori-
nation has not been reported previously.
Inmost cases, it would be expected that applying sufficient
prechlorination may help reducing the maximum NDMA
concentration at the farther end of the distribution system.
However, for certain precursors like the amine-based phar-
maceuticals studied in this work, short prechlorination
sometimes might break the binding between NOM and pre-
cursors and thus enhance the initial reaction, resulting higher
level of NDMA within the sections with shorter water age.
Therefore, a good NDMA control strategy would require
knowledge about the sourcewater, the precursor property, the
treatment applied at the plant, the size of the distribution
system, as well as the location of households and water age.
Acknowledgement
This researchwasfinancially supportedby theCanadianWater
Network, the Natural Sciences and Engineering Research
Council of Canada, and the Ontario Research Fund. Special
thanks are dedicated to Richard Jones and John Armour in the
water treatment plants for their assistance in water sampling.
Appendix A. Supplementary data
Supplementary data related to this article can be found at
http://dx.doi.org/10.1016/j.watres.2013.02.017.
r e f e r e n c e s
Abia, L., Armesto, X.L., Canle, L.M., Garcia, M.V., Santaballa, J.A.,1998. Oxidation of aliphatic amines by aqueous chlorine.Tetradedron 54, 521e530.
Andrzejewski, P., Kasprzyk-Hordern, B., Nawrocki, J., 2008. N-nitrosodimethylamine (NDMA) formation during ozonation ofdimethylamine-containing waters. Water Research 42,863e870.
Ballard, B.D., Mackay, A.A., 2005. Estimating the removal ofanthropogenic organic chemicals from raw drinking water bycoagulation flocculation. Journal of EnvironmentalEngineering 131, 108e118.
Carey, F.A., 2008. Organic Chemistry, seventh ed. McGraw-Hill,New York, Chapter 12. Reactions of Arenes: ElectrophilicAromatic Substitution. pp. 496e498.
Charrois, J.W.A., Hrudey, S.E., 2007. Breakpoint chlorination andfree-chlorine contact time: implications for drinking water N-nitrosodimethylamine concentrations. Water Research 41,674e682.
Chen, Z., Valentine, R.L., 2007. Formation of N-Nitrosodimethylamine (NDMA) from humic substances innatural water. Environmental Science and Technology 41,6059e6065.
Chen, Z., Valentine, R.L., 2008. The influence of the pre-oxidationof natural organic matter on the formation of N-Nitrosodimethylamine (NDMA). Environmental Science andTechnology 42, 5062e5067.
Chen, Z., Yang, L., Zhai, X., Zhao, S., Li, A., Shen, J., 2010. N-nitrosamine formation during chlorination/chloramination ofbromide-containing water. Water Science and Technology:Water Supply 10, 462e471.
Choi, J.H., Valentine, R.L., 2003. N-Nitrosodimethylamineformation by free-chlorine-enhanced nitrosation ofdimethylamine. Environmental Science and Technology 37,4871e4876.
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 72456
Comerton, A.M., Andrews, R.C., Bagley, D.M., 2006. Impact ofblending reuse and lake water on treated water quality.Journal of Environmental Engineering and Science 5,359e363.
De Ridder, D.J., Verliefde, A.R.D., Heijman, S.G.J., Verberk, Q.J.C.,Rietveld, L.C., van der Aa, L.T.J., Amy, G.L., van Dijk, J.C., 2011.Influence of natural organic matter on equilibrium adsorptionof neutral and charged pharmaceuticals onto activatedcarbon. Water Science and Technology 63, 416e423.
Deborde, M., von Gunten, U., 2008. Reactions of chlorine withinorganic and organic compounds during water treatment ekinetics and mechanisms: a critical review. Water Research42, 13e51.
Diemert, S.A., 2012. The Impact of Coagulation on EndocrineDisrupting Compounds, Pharmaceutically Active Compoundsand Natural Organic Matter. Master’s thesis, University ofToronto, Ontario.
Ding, Y., Teppen, B.J., Boyd, S.A., Li, H., 2012. Measurement ofassociations of pharmaceuticals with dissolved humicsubstances using solid phase extraction. Chemosphere. http://dx.doi.org/10.1016/j.chemosphere.2012.11.039.
Dotson, A., Westerhoff, P., Krasner, S.W., 2007. Nitrosamineformation from natural organic matter isolates and sunlightphotolysis of nitrosamines. In: Proceedings of AWWA AnnualConference and Exposition. Toronto, ON, Canada, June 24e28,2007.
EPA Integrated Risk Information System (IRIS), 1993. N-Nitrosodimethylamine; CASRN 62-75-9. www.epa.gov/iris/subst/0045.htm.
Gerecke, A.C., Sedlak, D.L., 2003. Precursors of N-Nitrosodimethylamine in natural waters. EnvironmentalScience and Technology 37, 1331e1336.
Gilow, H.M., Burton, D.E., 1981. Bromination and chlorination ofpyrrole and some reactive 1-substituted pyrroles. Journal ofOrganic Chemistry 46, 2221e2225.
Harvey, M, 2009. The Impacts of UV Direct Photolysis and UV/H2O2
Advanced Oxidation Processes on the Formation ofNitrosamines and Organic Chloramines from SubsequentChlor(am)ination. Master’s thesis, University of Toronto,Ontario.
Health Canada, 2010. Guideline Technical Document on N-nitrosodimethylamine (NDMA) in Drinking Water for PublicComment. http://www.hc-sc.gc.ca/ewh-semt/consult/_2010/ndma/draft-ebauche-eng.php#a3.
Huber, S.A., Balz, A., Abert, M., Pronk, W., 2011. Characterisation ofaquatic humic and non-humic matter with size-exclusionchromatographye organic carbon detectione organic nitrogendetection (LC-OCD-OND). Water Research 45, 879e885.
Kemper, J.M., Walse, S.S., Mitch, W.A., 2010. Quaternary aminesas nitrosamine precursors: a role for consumer products?Environmental Science and Technology 44, 1224e1231.
Kohut, K.D., Andrews, S.A., 2003. Polyelectrolyte age and N-nitrosodimethylamine formation in drinking water treatment.Water Quality Research Journal of Canada 38 (4), 719e735.
Krasner, S.W., 2009. The formation and control of emergingdisinfection by-products of health concern. PhilosophicalTransactions of the Royal Society A Mathematical andPhysical Sciences 367, 4077e4095.
Le Roux, J., Gallard, H., Croue, J., 2011. Chloramination ofnitrogenous contaminants (pharmaceuticals and pesticides):NDMA and halogenated DBPs formation. Water Research 45,3164e3174.
Le Roux, J., Gallard, H., Croue, J., 2012a. Formation of NDMA andhalogenated DBPs by chloramination of tertiary amines: theinfluence of bromide ion. Environmental Science andTechnology 46, 1581e1589.
Le Roux, J., Gallard, H., Croue, J., Papot, S., Deborde, M., 2012b.NDMA formation by chloramination of ranitidine: kinetics and
mechanism. Environmental Science and Technology 46,11095e11103.
Lee, C., Schimidt, C., Yoon, J., Von Gunten, U., 2007. Oxidation ofN-Nitrosodimethylamine (NDMA) precursors with ozone andchlorine dioxide: kinetics and effect on NDMA formationpotential. Environmental Science and Technology 41,2056e2063.
Lee,C., Lee,Y., Schmidt, C.,Yoon, J.,VonGunten,U., 2008.Oxidationof suspected N-nitrosodimethylamine (NDMA) precursors byferrate (VI): kineticsandeffect on theNDMAformationpotentialof natural waters. Water Research 42, 433e441.
Lin, S., Carlson, R.M., 1984. Susceptibility of environmentallyimportant heterocycles to chemical disinfection: reactionswith aqueous chlorine, chlorine dioxide, and chloramine.Environmental Science and Technology 18, 743e748.
Massachusetts Department of Environmental Protection(MassDEP), 2004. Current Regulatory Limit: n-Nitrosodimethylamine (NDMA). CASRN 62759. http://www.mass.gov/dep/water/drinking/standards/ndma.htm.
Mitch, W.A., Krasner, S.W., Lee, C.F.T., Wong, T.M., 2010.Tradeoffs in DBP formation during chloramination followingpre-oxidation for nitrosamine control. In: Proceedings of theAWWA Water Quality Technology Conference, Savannah, GA,November 14e18, 2010.
Mitch, W.A., Schreiber, I.M., 2008. Degradation of tertiaryalkylamines during chlorination/chloramination: implicationsfor formation of aldehydes, nitriles, halonitroalkanes, andnitrosamines. Environmental Science and Technology 42,4811e4817.
Mitch, W.A., Sedlak, D.L., 2004. Characterization and fate of N-nitrosodimethylamine precursors in municipal wastewatertreatment plants. Environmental Science and Technology 38,1445e1454.
MOE, 2003. Ontario Regulation 268/03 Made under the SafeDrinking Water Act, 2002. http://www.e-laws.gov.on.ca/html/source/regs/english/2003/elaws_src_regs_r03268_e.htm.
Najm, I., Trussell, R.R., 2001. NDMA formation in water andwastewater. Journal of the American Water WorksAssociation 93 (2), 92e99.
Office of Environmental Health Hazard Assessment (OEHHA),2006. Public Health Goal for N-nitrosodimethylamine andCadmium in Drinking Water. http://www.oehha.org/water/phg/cadndma122206.html.
Oya, M., Kosaka, K., Asami, M., Kunikane, S., 2008. Formation ofN-nitrosodimethylamine (NDMA) by ozonation of dyes andrelated compounds. Chemosphere 73, 1724e1730.
Pisarenko, A.N., Stanford, B.D., Yan, D., Gerrity, D., Snyder, S.A.,2012. Effects of ozone and ozone/peroxide on trace organiccontaminants and NDMA in drinking water and water reuseapplications. Water Research 46, 316e326.
Radjenovic, J., Farre, M.J., Gernjak, W., 2012. Effect of UV and UV/H2O2 in the presence of chloramines on NDMA formationpotential of tramadol. Environmental Science and Technology46, 8356e8364.
Schmidt, C.K., Brauch, H.J., 2008. N, N-Dimethylsulfamide asprecursor for N-Nitrosodimethylamine (NDMA) formationupon ozonation and its fate during drinking water treatment.Environmental Science and Technology 42, 6340e6346.
Schmidt, C.K., Sacher, F., Brauch, H.J., 2006. Strategies forminimizing formation of NDMA and other nitrosaminesduring disinfection of drinking water. In: Proceedings of theAWWA Water Quality Technology Conference, Denver, C.O.,November 5e9, 2006.
Schreiber, I.M., Mitch, W.A., 2005. Influence of the order ofreagent addition on NDMA formation during chloramination.Environmental Science and Technology 39, 3811e3818.
Schreiber, I.M., Mitch, W.A., 2006a. Nitrosamine formationpathway revisited: the importance of chloramines speciation
wat e r r e s e a r c h 4 7 ( 2 0 1 3 ) 2 4 4 6e2 4 5 7 2457
and dissolved oxygen. Environmental Science and Technology40, 6007e6014.
Schreiber, I.M., Mitch, W.A., 2006b. Occurrence and fate ofnitrosamines and nitrosamine precursor in wastewater-impacted surface waters using boron as a conservativetracer. Environmental Science and Technology 40,3203e3210.
Shah, A.D., Krasner, S.W., Lee, C.F.T., von Gunten, U., Mitch, W.A.,2012. Trade-offs in disinfection byproduct formationassociated with precursor preoxidation for control of N-Nitrosodimethylamine formation. Environmental Science andTechnology 46, 4809e4818.
Shen, R., Andrews, S.A., 2011a. Demonstration of 20pharmaceuticals and personal care products (PPCPs) asnitrosamine precursors during chloramine disinfection. WaterResearch 45, 944e952.
Shen, R., Andrews, S.A., 2011b. NDMA formation kinetics fromthree pharmaceuticals in four water matrices. Water Research45, 5687e5694.
Stackelberg, P.E., Gibs, J., Furlong, E.T., Meyer, M.T., Zaugg, S.D.,Lippincott, R.L., 2007. Efficiency of conventional drinking-water-treatment processes in removal of pharmaceuticals andother organic compounds. Science of the Total Environment377, 255e272.
USEPA, 2009. Contaminant Candidate List 3 (CCL3). http://water.epa.gov/scitech/drinkingwater/dws/ccl/ccl3.cfm.
Valentine, R.L., Choi, J., Chen, Z., Barrett, S.E., Hwang, C.,Guo, Y.C., Wehner, M., Fitzsimmons, S., Andrews, S.A.,Werker, A.G., Brubacher, C., Kohut, K., 2005. Factors Affectingthe Formation of NDMA inWater and Occurrence, Denver, CO,pp. 81e90.
Vieno, N., Tuhkanen, T., Kronberg, L., 2006. Removal ofpharmaceuticals in drinking water treatment: effect ofchemical coagulation. Environmental Technology 27, 183e192.
Westerhoff, P., Yoon, Y., Snyder, S., Wert, E., 2005. Fate ofendocrine-disruptor, pharmaceutical, and personal careproduct chemicals during simulated drinking water treatmentprocesses. Environmental Science and Technology 39,6649e6663.
Wilczak, A., Assadi-Rad, A., Lai, H.H., Hoover, L.L., Smith, J.F.,Berger, R., Rodigari, F., Beland, J.W., Lazzelle, L.J.,Kincannon, E.G., Baker, H., Heaney, C.T., 2003. Formation ofNDMA in chloraminated water coagulated with DADMACcationic polymer. Journal of the American Water WorksAssociation 95 (9), 94e106.
Woodbeck, M., 2007. Evaluating the Potential Genotoxicity ofDisinfected Source Waters. Master’s thesis, University ofToronto, Ontario.
Xu, X., Bartlett, M.G., Stewart, J.T., 2001. Determination ofdegradation products of sumatriptan succinate using LC-MSand LC-MS-MS. Journal of Pharmaceutical and BiomedicalAnalysis 26, 367e377.