Journal of Chromatography A, 1333 (2014) 124–133

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

A chemometric approach to elucidate the parameter impact in the hyphenation of evaporative light scattering detector to supercritical fluid chromatography Marie Lecoeur a,∗ , Nicolas Simon a,b , Valérie Sautou c,d , Bertrand Decaudin a,b , Claude Vaccher a , for the ARMED study group a

Univ Lille Nord de France, UDSL, EA 4481, UFR Pharmacy, 59006 Lille, France Department of Pharmacy, University Hospital, Lille, France c Department of Pharmacy, University Hospital, Clermont-Ferrand, France d EA 4676 C-Biosenss, Auvergne University, Clermont-Ferrand, France b

a r t i c l e

i n f o

Article history: Received 21 October 2013 Received in revised form 16 January 2014 Accepted 19 January 2014 Available online 29 January 2014 Keywords: Evaporative light scattering detector Experimental design Plasticizers Synergi Polar-RP Supercritical fluid chromatography

a b s t r a c t The aim of this work was to elucidate the effects of parameters influencing the evaporative light scattering detector (ELSD) response when it was coupled to supercritical fluid chromatography (SFC). Phthalates, currently used as plasticizers in medical devices, were selected as model compounds. The configuration of the hyphenation setup was firstly optimized and shown that both peak efficiency and sensitivity were improved by connecting the ELSD to the SFC before the back pressure regulator (BPR). By using a tee-junction which splits the flow after the PDA towards the collect fraction (or waste) and the ELSD, this instrument configuration has the advantage to be applicable for small-scale preparative SFC. The impacts of other parameters such as mobile phase composition and flow rate, outlet pressure, column oven temperature and ELSD drift tube temperature on the ELSD signal were evaluated using a chemometric approach. First, it was demonstrated that a classical mobile phase composed of CO2 –methanol 90:10 (v/v) was suitable to obtain great nebulization efficiency. The flow rate of the eluent was the second main effect factor. The setting must be as low as possible to avoid the loss of large particle size in the drift tube resulting in a loss of signal intensity. Concerning the outlet pressure, the configuration of the setup between SFC and ELSD requires a setting as high as possible to limit the partial liquid–vapor separation of the mobile phase in the restrictor tube. Finally, due to the low quantity of solvent which must be evaporated in the detector, a drift tube temperature of 25 ◦ C is suitable for the hyphenation of ELSD to SFC. In the optimized conditions, the proposed SFC/ELSD method could be suitable to quantify plasticizers in medical devices. © 2014 Elsevier B.V. All rights reserved.

1. Introduction For several years, supercritical fluid chromatography (SFC) is gaining interest because of the typical properties of the mobile phase, the low cost of carbon dioxide (CO2 ) compared to organic solvents, the fast analysis speed, the wide polarity compatibility and also the current development in SFC instrumentation, both for analytical [1,2] and preparative scale [3] applications. Due to its versatility, SFC could be considered either as an alternative to normal or reverse phase liquid chromatography, depending on the nature of the stationary phase used [4,5]. Several recent reviews reported

∗ Corresponding author. Tel.: +33 3 62 28 30 27; fax: +33 3 20 95 90 09. E-mail address: [email protected] (M. Lecoeur). 0021-9673/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.chroma.2014.01.054

various applications in the field of enantiomeric separations [6], drug discovery [7] and food analysis [8]. SFC system is commonly equipped with a UV–visible detector located before the back pressure regulator (BPR). However, for several classes of molecules (for example: fatty acids, saccharides, PEG-based polymers), the use of UV detection is very limited. That is why the development of SFC methods using evaporative light scattering detector (ELSD), which do not require heavy instrument modification, is gaining interest. Although the ELSD suffers from a dramatic lack of sensitivity, it presents the advantages to be robust, inexpensive [9] and ability to be used under solvent–gradient conditions [10]. ELSD is considered as a mass detector with a poor range of linearity which requires the use of double logarithm coordinates for calibration [11]. The hyphenation of ELSD to liquid chromatography has been largely investigated [9,12–14]. Regarding SFC, only few studies

M. Lecoeur et al. / J. Chromatogr. A 1333 (2014) 124–133

were achieved to describe the effect of various parameters acting on ELSD detection [15–19]. So far, optimization was systematically conducted by varying one experimental parameter at a time, all the others being fixed, whereas they could present interaction effects. Moreover, when a chemometric approach is implemented [17], only ELSD settings (nitrogen flow rate, drift tube temperature, orifice size) was studied without taking into account the influence of SFC parameters. Indeed, both outlet pressure and eluent flow rate are often optimized during the method development but only two papers deals about their effect on the ELSD signal [15,19]. In regard to the published papers, authors did not agree with the setting of detection parameters. The first discrepancy is about the configuration of the SFC-ELSD interface. For example, some authors connected the ELSD after the back pressure regulator (BPR) [15], whereas in some other papers, the flow rate was split by using a tee-junction between the detector and the BPR [20]. Likewise, the choice of the ELSD drift tube temperature was not systematically explained. Hence, in very similar chromatographic conditions (outlet pressure, eluent flow rate and composition), drift tube temperature was set at 30–40 ◦ C [20,21] or at 60–70 ◦ C [17,22–25] without any discussion about the impaction on the ELSD response. To our knowledge, only E. Lesellier et al. showed that the increase in the drift tube temperature leads to a decrease in response of a semivolatile compound [19]. Consequently, there is a real need for understanding the impact of the experimental parameters on the SFC/ELSD performances. For this purpose, a chemometric approach based on a fractional factorial design was used for the first time, to our knowledge, for the hyphenation of ELSD to SFC. Phthalic acid esters usually called phthalates are widely used in industrial applications (vinyl flooring, ink, glue, cosmetics, textile. . .). In polymer industry, phthalates act as plasticizers to improve flexibility and workability of polyvinylchloride (PVC). Among applications, di(2-ethylhexyl)phthalate (DEHP) was very useful for producing medical devices such as blood bags and tubing used for blood transfusion, drugs infusion, dialysis, feeding, cardiopulmonary bypass and endotracheal intubation. As phthalates are not chemically bonded in the PVC, they can be released from the medical device during contact with blood, enteral or total parenteral nutrition admixtures or lipophilic drugs, or drugs that contain surfactant then penetrate into the human fluid. DEHP is considered as class-1B carcinogen, mutagen or toxic for reproducing chemical compound due to its reproductive toxicity in animal studies [26]. Hence, since 2010, European authorities have challenged the use of DEHP in medical devices destined to the administration or removal of drugs, biological liquids or other substances into or from the human body. This action has forced the manufacturers to replace DEHP by alternative plasticizers (ATBC, DEHA, DEHT, DINP, DINCH, TOTM) whose impact on health has not yet been reported. HPLC-UV and GC–MS are both techniques usually developed to quantify phthalates in environmental matrices or intravenous injection solutions [27]. However, UV detection is not suitable for non UV-absorbent plasticizers such as ATBC, DEHA and DINCH and the implementation of GC seems to be difficult for the less volatile compound (TOTM). Hence, SFC–ELSD could be a good alternative technique to analyze new plasticizers added to medical devices. The aim of this study was to better understand the impact of factors influencing the ELSD response in the hyphenation to SFC. For this purpose, a series of five UV detectable and undetectable plasticizers with a wide range of log P values (from 4.3 to 11.6) and boiling points (from 173 to 414 ◦ C) were selected as model compounds. The results obtained contribute to elucidate the impact of experimental parameters on ELSD responses and allow reduction of the number of parameters to be optimized.

125

2. Materials and methods 2.1. Chromatographic apparatus Chromatographic separations were carried out using an SFC–PICLAB hybrid 10–20 apparatus equipped with an autosampler, three 40P pumps, a column oven with a 10-column selection valve and a 6-solvent switching valves (PIC solution, Avignon, France). The proportion of the co-solvent in the mobile phase was adjusted by a piston pump. It was then directly added in the carbon dioxide feeding, and the mixture of co-solvent and carbon dioxide was pumped by another piston pump at the total flow-rate. The pump head used for CO2 was cooled to −8 ◦ C by a cryostat (Huber Minichiller, Offenburg, Germany). The injection valve was supplied with a 10 ␮L sample loop. The unit was also composed of a Smartline2600 DAD detector with a high-pressure resistant cell (Knauer, Berlin, Germany). Detection wavelength was set at 225 nm. After the detector, the outlet pressure was controlled by a back-pressure regulator (BPR) with a void volume of 250 ␮L. The outlet tube was heated at 55 ◦ C to avoid ice formation during the carbon dioxide depressurization. Data were recorded with SFC PicLab Analytic Online 3.1.2 and processed with Analytic Offline 3.2.0. An ELSD model Sedex 85 (Sedere, Alfortville, France) was used in this study. It was either plumbed between the PDA and the back pressure regulator using a 0.010 in. i.d. stainless steel tee-junction and a 65 ␮m × 160 cm Peek tubing or to the outlet capillary after the BPR using a 100 ␮m × 160 cm Peek tubing. During the separation optimization, the pressure of the nebulizer gas (N2 ) was set at 3 bar, the drift tube temperature and the gain were 30 ◦ C and 7, respectively. The columns tested were a Viridis 2-Ethylpyridine (250 mm × 4.6 mm, 5 ␮m) and a Viridis C18 (same dimensions, carbon load: 16%) from Waters Europe (Guyancourt, France), a Synergi Polar RP (250 mm × 4.6 mm, 5 ␮m) from Phenomenex (Le Pecq, France), an Uptishere Strategy C18-3 (carbon load: 22%) with the dimensions (250 mm × 4.6 mm, 3 ␮m) from Interchim (Montluc¸on, France). During the separation optimization, the column temperature was 35 ◦ C and the outlet pressure was 150 bar. The mobile phase was delivered at a flow rate of 3 ml/min. The flow rate of the fluid entering into the nebulizer of ELSD was measured with an ADM 1000 flowmeter (Agilent, Waldbronn, Germany). 2.2. Chemicals and reagents Acetyl tri-n-butyl citrate (ATBC), benzylbutylphthalate (BBP), di(2-ethylhexyl phthalate) (DEHP), di(2-ethylhexyl) terephthalate (DEHT), trioctyltrimellilate (TOTM) were purchased from Sigma–Aldrich (Steinheim, Germany). All chemical structures were presented in Fig. 1. All solvents used (2-PrOH, EtOH, MeOH, ACN) were of analytical grade and were provided by VWR (Val de Fontenay, France). The carbon dioxide of N45 quality was purchased from Air Liquide (Puteaux, France). 2.3. Sample preparation Stock solutions of each plasticizer (1000 ␮g/ml) were prepared in ACN. Experiments were carried out using a test solution composed of each plasticizer at a concentration of 100 ␮g/ml. BBP was added to the test solution as a potential internal standard for quantitative analysis. 2.4. Design of experiments 2.4.1. Choice of factors, factors levels and responses Considering previous works reported in the literature [14,15,19,28] and our preliminary experiments, five potentially

126

M. Lecoeur et al. / J. Chromatogr. A 1333 (2014) 124–133

O

O

O

O

O

O

O

O

O O

O

O

ATBC (log P = 4.3; b.p. = 173°C)

DEHP (log P = 7.6; b.p. = 384°C)

O

O

O

O

O

O

O

O

DEHT (log P = 8.4; b.p. = 383°C)

BBP (log P = 4.8; b.p. = 370°C)

O O O

O O

O

TOTM (log P = 11.6; b.p. = 414°C) Fig. 1. Chemical structure of plasticizers investigated. Values in parentheses refer to the logarithm of octanol–water partitioning factor (log P) and to the boiling point (b.p.). They were obtained from the database “PhysProp” (http://www.srcinc.com). Table 1 Values of the different studied factors at factorial (−1; 1) and central (0) levels. Coded levels

FR MeOH Pout Toven TELSD

Flow rate of the mobile phase (ml/min) MeOH content in the mobile phase (%) Outlet pressure (bar) Oven temperature (◦ C) Evaporative tubing temperature of ELSD

−1

0

+1

2.0 1.0 100 30 25

3.5 4.5 150 40 42.5

5.0 8.0 200 50 60

influent factors were selected to build the design of experiments: flow rate of the mobile phase (FR), methanol content in the mobile phase (MeOH), outlet pressure (Pout ), oven temperature (Toven ) and drift tube temperature of the ELSD (TELSD ). The flow rates of the mobile phase were from 2.0 to 5.0 ml/min. The percentage of MeOH in the mobile phase varied from 1 to 8%. The outlet pressure was chosen between 100 and 200 bar. The drift tube temperatures of ELSD were selected from 25 to 60 ◦ C. Factor levels, in initial and coded values, of the fractional design are given in Table 1. The responses chosen to evaluate the influence of each factor were the ELSD peak areas of each analyte. 2.4.2. Choice of the experimental design and matrix of experiments The chemometric approach was a two-level fractional design (25−1 ) with three centre points. All experiments were repeated twice; hence, the design was composed of 40 experiments which were randomized to avoid being affected by uncontrolled factor variations (Table 2). With this design, main factor effects and first order interactions (two-factor interactions) were estimated. Indeed, with this design of resolution V, the two-factor interactions

themselves are unconfounded meaning that 25−1 design is almost as good as full factorial designs [29]. Between each temperature change (column oven or ELSD), the system was equilibrated for 30 min before injection. 2.4.3. Determination of the effects Factor effects were calculated by least-squares multiple linear regression. It was depicted as follows: Y = b + i Ei Xi + i j Eij Xi Xj , where Y is the predictive response; b, the intercept, is the constant term; Ei is the coefficient of factor Xi (namely the effect of the factor); and Eij is the coefficient of the interaction between the factors Table 2 Matrix of experiments for the 25−1 design in coded values. Run order

Flow rate

MeOH

Pout

Toven

TELSD

24–28 2–19 5–15 31–34 23–30 18–27 32–36 26–20 9–33 4–6 11–29 16–35 17–22 12–38 7–37 10–25 3–21 1–13 8–14

− + − + − + − + − + − + − + − + 0 0 0

− − + + − − + + − − + + − − + + 0 0 0

− − − − + + + + − − − − + + + + 0 0 0

− − − − − − − − + + + + + + + + 0 0 0

+ − − + − + + − − + + − + − − + 0 0 0

M. Lecoeur et al. / J. Chromatogr. A 1333 (2014) 124–133

127

AU

AU 1200

1200

ATBC + DEHT + DEHP + TOTM

A) Silica

1000

B) 2-EP 1000

800

800

600

600

400

400

ATBC + DEHT + TOTM

BBP DEHP BBP 200

200

0

0

0

0,5

1

1,5

2

2,5

3

3,5

0

0,5

1

1,5

min AU

2,5

3

3,5

AU DEHT

200

C) C18

180

2

min 800

D) SP

DEHP 700

ATBC

160

BBP

600

TOTM

140

BBP DEHT DEHP TOTM

500

120 100

400

ATBC

80

300

60

200

40

100

20 0

0 0

0,5

1

1,5

2

2,5

3

3,5

4

min

0

1

2

3

4

5

min

Fig. 2. Chromatograms of the five plasticizers on various bonded stationary phases obtained with the split mode. Conditions: Pout : 150 bar; CO2 /MeOH 95:5 (v/v); T◦ : 35 ◦ C; flow rate: 3 mL min−1 ;T◦ ELSD drift tube : 30 ◦ C, gain: 7. (A) Silica, (B) 2-ethylpyridine, (C) C18–22% carbon load, (D) Synergi Polar RP stationary phases.

Xi and Xj . Usually coefficients of interest are the main effects Ei and the first order interaction Eij . If the interaction coefficient is not significant, the effect of factor Xi can be examined individually. If the first order interaction coefficient bij is significant, both effects of Xi and Xj factors have to be studied together with the use of an interaction diagram [29]. Analysis of variance (ANOVA) was used for the evaluation of multiple linear regression model. ANOVA is based on partitioning the total variation of a selected response into one part due to the regression model and another part due to the residuals. From the repeated measurements, ANOVA also decomposes the residual variation into one part related to the model error and another part linked to the replicate error. Subsequently, the numerical sizes of these variance estimates are compared by means of F-tests with a type I risk of 5%. Main effects, Ei , and the first order interaction, Eij , significance were then evaluated by comparing the corresponding regression coefficient to its confidence interval (risk of 5%). Factor and two-factor interaction are statistically insignificant if the confidence interval includes zero. [29]. Coefficients were expressed as relative effects (effect normalized by the average value of the response) to visualize the impact of each factor better. 3. Results and discussion

plasticizers were retained less on silica and 2EP stationary phases (Fig. 2A–B) even if the mobile phase was composed of 100% CO2 . With log P values ranging from 4.3 and 11.6, the lipophilicity of plasticizers investigated is too strong to interact with these polar stationary phases. Both C18 bonded stationary phases with 16 and 22% carbon load, respectively, displayed a lack of retention and selectivity for the plasticizers. The decrease in MeOH from 5 to 0% increased the retention time but without any improvement in selectivity (Fig. 2C). The Synergi Polar-RP was the last bonded stationary phase investigated. Despite its name, this stationary phase belongs to the group of stationary phases of intermediate polarity [30] due to the inclusion of an oxygen atom in the spacer arm of the phenyloxypropyl-bonded silica phase. In the literature, triterpenoids [23] and barbiturates [31] with close related structures were successfully resolved using this column. First separation of plasticizers using the Synergi Polar-RP was very promising (Fig. 2D). The retention order is ATBC, DEHT, DEHP, BBP and TOTM. As expected, ATBC is the less retained plasticizer on this support because no ␲–␲ interaction can be generated between this non aromatic compound and the chromatographic support. Great selectivity was obtained between the two isomers (DEHT and DEHP). Hence, the Synergi Polar-RP was selected for this study.

3.1. Setup of the separation

3.2. Hyphenation of ELSD to SFC

The separation optimization was studied, first with a mobile phase composed of CO2 –MeOH 95:5 (v/v), at 35 ◦ C and a backpressure of 150 bar. The results were presented in Fig. 2. As expected,

Two instrumental setups were evaluated for the hyphenation of ELSD to SFC. In the first case (Fig. 3A), the ELSD was plumbed to the outlet capillary after the BPR using an 0.010 in. i.d. stainless

128

M. Lecoeur et al. / J. Chromatogr. A 1333 (2014) 124–133

A

Tee-juncon

Column

BPR T, mV

λ, A

CO2 + modifier

PDA Column oven

ELSD

B

Tee-juncon

Column

BPR

λ, A

CO2 + modifier

PDA Column oven T, mV

Waste or collector

ELSD Fig. 3. Schematic diagrams of the on-line SFC/ELSD interface. Void volume of the BPR and of the capillary from the tee-junction to the BPR: 250 and 27 ␮L, respectively; (A) the ELSD was connected after the BPR using a similar tee and a 100 ␮m × 160 cm Peek tubing (void volume: 12.5 ␮L) (B) the ELSD was plumbed between the PDA and the back pressure regulator (BPR) using a 0.010 in. i.d. stainless steel tee-junction and a 65 ␮m × 160 cm Peek tubing (void volume: 5.3 ␮L).

steel tee and a standard capillary with an internal diameter equal to 100 ␮m (160 cm total length). The total eluent flow is introduced into the ELSD. The effect of the organic modifier nature (MeOH, EtOH, iPrOH, ACN) was investigated in order to improve both separation quality and ELSD response. The chromatograms of the five plasticizers were presented in Fig. 4A. The chromatographic behavior of analytes is still the same. Indeed, Synergi Polar belongs to the group of polar phases, so the selectivity is affected quite less by the change of the mobile phase composition [31]. Concerning the detection sensitivity, methanol, ethanol and isopropanol provide similar results, even if few artifacts appear on the chromatogram using isopropanol. With acetonitrile, the ELSD signals decrease in half, indicating that few particles are going to the light beam. As deeply explained by Lesellier et al. [15], an insufficient solvation of analytes by the low quantity of this aprotic solvent in the mobile phase may be responsible for the weak ELSD response. In the second configuration (Fig. 3B), the ELSD was connected to the PDA outlet and before the back pressure regulator using an 0.010 inch i.d. stainless steel tee and a 65 ␮m Peek tubing (same length as before). This configuration splits the flow after the PDA, such that most of the column flow goes to the collect fraction (or waste) and only a portion of interest goes to the ELSD. This is the configuration that is mostly found with instrument manufacturers and for coupling SFC to mass spectrometry detection [32]. The length of the tubing has to be long enough to allow the decompression of the supercritical fluid before entering into the nebulizing tube of the ELSD. Small i.d. tube (65 ␮m) is also required to avoid any variations of the outlet pressure value. It was noticed that the BPR cannot correctly regulate the outlet pressure if 100 ␮m i.d. tubing is implemented. This instrument configuration has the advantage to be applicable for small-scale preparative SFC [28], as shown in Fig. 3B. Similar experiments were made as previously described and chromatograms were presented in Fig. 4B. When the ELSD is plumbed between the PDA and the BPR, a substantial increase in peak height and peak area is obtained, despite the majority of the flow rate was split to the waste. As a result, a 2–7-fold gain in peak efficiency is achieved, depending on the modifier (Fig. 5A). This

series of experiments reveals that the hyphenation of ELSD to SFC after the BPR is not suitable. This may be due, first, to the liquidvapor separation of the CO2 -methanol phase occurring after the BPR because of the sudden decrease in pressure from 150 bar to the atmospheric pressure [33]. As a result, the analytes are not completely dissolved in the mobile phase, leading to a non homogeneity of the particle size entering into the detector and then leading to a decrease in the ELSD signal. The second and the most likely explanation is the too large dead volume of the BPR (250 ␮L) and of the capillary from the tee-junction to the BPR (27 ␮L) which cannot maintain good chromatographic profiles. Indeed, experiments have been performed on hybrid SFC apparatus able to work until semipreparative scale. Hence, BPR design of the system is not optimized for analytical purpose and all extra-column volumes are too large to enable the connection of the ELSD after the BPR. A dramatic improvement in the detector response is also observed when the ELSD is coupled to the SFC according to the configuration depicted in Fig. 3B. Indeed, the sensitivity is improved by a factor 2 with methanol, ethanol and isopropanol and can achieve a12-fold gain when using ACN (Fig. 5B). It is interesting to notice that similar ELSD responses are obtained, whatever the organic modifier is, by selecting the suitable setup between SFC and ELSD. Although spike peaks are found on the chromatogram (Fig. 4) with ACN, this result opens the way of specific separations which require the addition of ACN in the mobile phase and an ELSD detection. 3.3. Influence of the experimental factors 3.3.1. Choice of the factor levels Although the chromatograms of plasticizers (Fig. 4B) seem to be similar in term of sensitivity, best ELSD responses were obtained with methanol. Hence, several factors i.e. methanol content, eluent flow-rate (FR), outlet pressure (Pout ), column oven temperature (Toven ) and evaporation temperature in the drift tube of the ELSD (TELSD ) have been selected in order to evaluate their effect on the ELSD response, using a multivariate approach. Concerning methanol, this factor could not only have some effect on the ELSD

M. Lecoeur et al. / J. Chromatogr. A 1333 (2014) 124–133

A - ELSD connected aer BPR

129

B - ELSD connected before BPR 800

800

700

700

MeOH

600

MeOH

600 500

500

400

400

1

300

3 4

2

300

5

200

200 100

100

0

0 0

1

2

3

4

0

5

800

1

2

3

4

5

2

3

4

5

2

3

4

5

2

3

4

5

800

700

700

EtOH

600

EtOH

600

500

500

400

400

300

1

200

2 3

300

4

200

5

100

100

0

0 1

0

2

3

4

0

5

1

800

800 700

700

iPrOH

600

600

500

500

iPrOH

400

400

2

300

3

1

200

4

300 200

5

100

100

0

0 0

1

2

3

4

0

5

800

1

800

ACN

700

ACN

700

600

600

500

500

400

400

300

300

200

1 100

4 2 3

200

5

100

0

0 0

1

2

3

4

5

Time (min)

0

1

Time (min)

Fig. 4. Effects of SFC/ELSD setup and organic modifier nature on the separation and on the ELSD responses of the five standards. Conditions: column: Synergi Polar RP; mobile phase: CO2 /organic modifier 95:5 (v/v); flow rate: 3 mL min−1 ;T: 35 ◦ C; Pout : 150 bar;TELSD : 30 ◦ C; gain: 7.1-ATBC, 2-DEHT, 3-DEHP, 4-BBP, 5-TOTM.

response but also has a strong influence on the separation quality. In order to obtain baseline resolution between all plasticizers as a function of the level of MeOH selected in the experimental design (Table 1), BBP was deleted from the test mixture which was also composed of four plasticizers (ATBC, DEHT, DEHP, TOTM). 3.3.1.1. Influence of MeOH content in the mobile phase. MeOH contents added to the mobile phase were selected in the 1 and 8% range to have a baseline separation of plasticizers in the most

eluent conditions, while maintaining reasonable runtime in the less eluent ones. From a chromatographic point of view, an increase in MeOH (from 1 to 8%) favors the elution of plasticizers (runtime of 4.3 min instead of 8.8 min), increases the peak efficiency twofold but at the expense of a loss of resolution (20%). However, plasticizers are still separated with a minimum resolution of 2.62 obtained between DEHT and DEHP. Fig. 6 shows that MeOH exerts a strong influence on peak areas using ELSD detection. Upon increasing the percentage of MeOH

130

M. Lecoeur et al. / J. Chromatogr. A 1333 (2014) 124–133

3.3.1.2. Influence of mobile phase flow rate. The flow rate of the mobile phase was set from 2.0 to 5.0 mL min−1 . As shown in Fig. 6, the flow rate of the mobile phase has a negative effect on the peak areas (70–80%). These results are in agreement with those obtained previously by other authors [14,15,34,35]. ELSD response intensity (A) is related both to the quantity of non volatile microparticules which scattered the light in the detection chamber (a) and to the particle size of droplets (b) produced by the nebulizer. These two phenomena which are strongly dependant can be materialized in the following equation:

7

A Efficiency improvement

6 5

ATBC 4

DEHT DEHP

3

BBP

2

TOTM 1

A = a × mb

0 MeOH

EtOH

iPrOH

where m represents the analyte mass. A recent study [15] has shown that the coefficient b is not impacted by increasing the flow rate of the mobile phase meaning that particle size distribution seems rather constant to whatever the flow rate is. However, the same authors have noticed that the value of the coefficient is divided by a factor 10 by increasing the flow rate from 1.5 to 3.0 ml/min, suggesting that a less amount of particles crosses the light beam. Indeed, after depressurization, the speed of the gaseous fluid entering into the nozzle can be improved until a factor 1000 in relation to the supercritical fluid velocity. Hence, working at 5 ml/min leads to a too high fluid velocity which could explain the loss of compounds onto the walls of the nebulizing chamber. Another explanation for the reduction of signal intensity, often mentioned using ELSD hyphenated to HPLC, is the incomplete solvent vaporization occurring at higher eluent flow rate. This second reason is unlikely in SFC due to the small percentage of organic modifier (maximum 8%) added to the mobile phase.

12

Sensivity improvement

10

B

8

ATBC DEHT

6

DEHP 4

BBP TOTM

2

0 MeOH

EtOH

(1)

ACN

iPrOH

ACN

Fig. 5. Efficiency (A) and sensitivity (B) improvement by connecting the ELSD before the BPR compared to a connection after the BPR. Conditions: same as Fig. 4.

in the mobile phase, peak areas increase as well. This result is contradictory with regard to previous publications dealing with the analysis of various compounds (pentacyclic triterpenoids, corticoids, xanthine) by SFC-ELSD [15,19,23]. Indeed, whatever the nature of organic modifier (MeOH or EtOH), a strong increase in ELSD response is classically observed by decreasing the percentage of modifier in the mobile phase. In our case, the dramatic loss of signal intensity observed with a weak percentage of co-solvent is probably due to some adsorption of plasticizers onto the internal surface of the capillary. Increasing the methanol percentage could improve plasticizer’s solubility and then favors their introduction into the ELSD.

3.3.1.3. Influence of outlet pressure. Outlet pressure was ranged between 100 and 200 bar. Fig. 6 shows that the outlet pressure has a positive impact on peak areas (50%). The flow rate of the fluid entering into the nebulizer was measured working with a mobile phase composed of 100% CO2 delivered at a flow rate of 3.5 ml/min. By applying a higher back pressure to the BPR, the flow rate of the fluid entering into the nebulizer was increased (280 ml/min at 100 bar vs. 520 ml/min at 200 bar) because of the greater difficulty for the fluid to go through the BPR. This enhancement of fluid velocity in the nebulizer seems to improve the ELSD signal.

1.00

ATBC

0.80

DEHT

0.60

Relave effect

DEHP 0.40 TOTM 0.20 0.00 -0.20 -0.40 -0.60

Toven*TELSD

Pout*TELSD

Pout*Toven

MeOH*TELSD

MeOH*Toven

MeOH*Pout

FR*TELSD

FR*Toven

FR*Pout

FR * MeOH

TELSD

Toven

Pout

MeOH

-1.00

FR

-0.80

Fig. 6. Relative effects of each experimental factor and factor interactions on peak areas recorded by ELSD. FR: flow rate; Pout : outlet pressure; Toven : oven temperature; TELSD : ELSD drift tube temperature. Bar stands for a standard deviation. Only significant effects and interactions are reported

M. Lecoeur et al. / J. Chromatogr. A 1333 (2014) 124–133

1400

MeOH (high)

1000

1000

A DEHT (mV)

A DEHT (mV)

B - FR x Pout

1100

1200

800 600 MeOH (high) 400

Pout (high)

1200

A - FR x MeOH

131

900 800 700 Pout (low)

600 500

MeOH (low)

Pout (high)

400 300

200

MeOH (low) 2.0

2.2

2.4

2.6

2.8

3.0

3.2

3.4

3.6

3.8

4.0

4.2

4.4

4.6

4.8

Pout (low)

200 2.0

5.0

2.2

2.4

2.6

2.8

3.0

-1

3.4

3.6

3.8

4.0

4.2

4.4

4.6

4.8

5.0

-1

Flow rate (mL.min ) 1000

3.2

Flow rate (mL.min )

TELSD (low)

1100

C - FR x TELSD

900

Pout (high)

D - MeOH x Pout

1000 900

A DEHT (mV)

800

TELSD (high)

Pout (low)

800 700

700

600 Pout (high)

500

600

400

500

300 200

400 300 2.0

2.2

2.4

2.6

2.8

3.0

3.2

3.4

3.6

3.8

4.0

4.2

4.4

TELSD (low)

100

TELSD (high)

0

4.6

4.8

Pout (low) 1.0

5.0

1.5

2.0

2.5

3.0

3.5

-1

4.0

4.5

5.0

5.5

6.0

6.5

7.0

7.5

8.0

MeOH (%)

Flow rate (mL.min ) TELSD (low)

1100

E - MeOH x TELSD

1000

A DEHT (mV)

900 TELSD (high)

800 700 600 500 400 300

TELSD (high) TELSD (low) 1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

5.5

6.0

6.5

7.0

7.5

8.0

MeOH (%) Fig. 7. Interaction plot between (A) flow rate × MeOH, (B) flow rate × outlet pressure, (C) flow rate × ELSD drift tube temperature, (D) MeOH × ELSD drift tube temperature. High and low mentioned in parentheses refer to values of the corresponding factor at factorial levels: MeOH (high): 8%; MeOH (low): 1%; TELSD (high): 60 ◦ C;TELSD (low): 25 ◦ C; Pout (high): 200 bar; Pout (low): 100 bar.

3.3.1.4. Influence of column oven temperature. An increase in column temperature does not significantly affect the ELSD response. 3.3.1.5. Influence of drift tube temperature. The effect of the drift tube temperature was then studied with temperature ranging from 25 to 60 ◦ C. Fig. 6 clearly shows that the drift tube temperature has negative effect on the ELSD response. However, the relative impact depends on the plasticizer. It is well-known that the drift tube temperature only influences the evaporation process of solvent and eventually those of the analyte. TOTM and DEHT are slightly affected by the temperature variation (c.a. >−20%). Since the boiling points of TOTM, DEHP and DEHT are close to 380 ◦ C, partial vaporization of these solutes could be suspected during the residence time in the detector leading to non-uniform particle sizing. This effect (−90%) becomes more marked for ATBC, a volatile compound (b.p. = 173 ◦ C). Similar results have already been obtained

during the SFC–ELSD analysis of semi volatile compounds such a lipids [12], ginkgolides [17] and caffeine [19]. As a conclusion, a temperature of 25 ◦ C is sufficient to evaporate the low quantity of organic modifier added to the CO2 . 3.3.1.6. Interactions between the factors. Fig. 6 reveals the existence of several significant interactions between factors which has a strong impact on the ELSD response of most of the plasticizers. The interaction plots, displayed in Fig. 7 as an example, allow to explore the nature of interactions and to study their effects on the DEHT response. First interaction occurs between the mobile phase flow rate and the methanol content. Indeed, at a flow rate of 5 ml/min, the detector response is double when the MeOH percentage increases from 1 to 8%. In contrast, increasing the MeOH percentage in the same manner as previously, a 4-factor improvement of the response is observed at 2 ml/min (Fig. 7A). As

132

M. Lecoeur et al. / J. Chromatogr. A 1333 (2014) 124–133

already explained, adsorption phenomenon of plasticizers occurs with low MeOH percentage in the mobile phase which seems to be accentuate at low flow rate. Although, Fig. 6 shows a significant interaction between the flow rate and the outlet pressure, the corresponding interaction plot depicted in Fig. 7B reveals a weak interaction between these factors. In other words, the effect of outlet pressure on the ELSD response is almost independent on the level of the flow rate. For all compounds, the influence of flow rate and drift tube temperature seems to be correlated (Fig. 7C); however, no explanation was proposed. Fig. 7D shows that MeOH is moderately correlated with outlet pressure whereas strong interaction occurs with drift tube temperature (Fig. 7E). In the latter, ELSD signal of DEHT is poor and quite similar to whatever the drift tube temperature, using a weak percentage of MeOH (1%). This result is in agreement with the fact that plasticizers’ lack of solubility using 1% MeOH and adsorption of them onto the internal surface of the capillary leads to a poor ELSD signal. At the opposite, the higher the MeOH content is (8%), the lower the drift tube temperature has to be selected to avoid poor ELSD response due to the partial vaporization of the analytes. All these observed interactions confirmed the relevance of the chemometric approach used for this study. Fig. 8 displays chromatograms of the five plasticizers using ELSD and UV detector. The most noteworthy interest of ELSD is for the analysis of ATBC which cannot be detected by UV even at low wavelength. However, strong ELSD response (Fig. 8B, upper trace) is obtained after sensible selection of both chromatographic and ELSD parameters.

mV 180

A

160

DEHT

DEHP

TOTM

120 100 80 60 40 20 0

min 1

0

2

3

4

5

mV 900

B

800

DEHT

DEHP

ATBC

700

The purpose of this work was to optimize the hyphenation setup between SFC and ELSD using plasticizers, currently used in medical devices, as model compounds. First, their separation was optimized on different stationary phases and best resolutions were obtained using Synergi Polar. A substantial gain in peak efficiency and sensitivity were obtained by connecting the ELSD to the SFC before the back-pressure regulator. To our knowledge for the first time, a chemometric strategy was then used to determine the behavior of both chromatographic and detection parameters influencing the ELSD response. From these results, an optimization guideline could be proposed. First of all, the mobile phase flow rate must be set at 2 ml/min to reach strong ELSD response while keeping reasonable runtime and satisfied peak efficiency. The setting of MeOH in the mobile phase at 8% is preferred to lower values. In regard to plasticizers, it seems to be possible to generalize this statement for organic modifiers (EtOH, iPrOH, ACN) classically implemented in SFC. However, this study did not take into account the behavior of ELSD detector while selecting organic modifier content higher than 10%. Concerning the outlet pressure, the configuration of the setup requires its setting at 200 bar to increase the flow rate of the fluid entering into the nebulizer and to improve the ELSD response. As a consequence, attention must be paid during the development of SFC–ELSD methods to select the stationary phase which leads to good retention and selectivity in the operating constraints previously cited. Finally, due to the low quantity of solvent (maximum 8%) which must be evaporated in the detector and to avoid the partial volatilization of semi-volatile plasticizers, a drift tube temperature of 25 ◦ C is suitable for the hyphenation of ELSD to SFC. This work also demonstrated that SFC/ELSD can be a suitable method to analyze plasticizers in medical devices. Acknowledgements

ATBC

140

4. Conclusions

TOTM

600 500 400 300

This work was a part of the project ARMED (Assessment and Risk Management of Medical Devices in plasticized polyvinylchloride) which has received the financial support of the French Medicine Agency (ANSM: Agence Nationale de Sécurité des Médicaments et des Produits de Santé). The authors also wish to thank collaborators of the ARMED study group in its task 1 “Characterization of plasticizers in medical devices”: Lise Bernard, Daniel Bourdeaux, Philip Chennell, Damien Richard, Bruno Pereira,Valérie Sautou (University Hospital, Clermont-Ferrand, France); Nathalie Azaroual, Christine Barthelémy, Bertrand Décaudin, Thierry Dine, Frédéric Feutry, Stéphanie Genay, Nicolas Kambia, Marie Lecoeur, Pascal Odou, Nicolas Simon, Claude Vaccher (EA 4481, University of Lille 2, France); Régis Cueff, Emmanuelle Feschet (EA 4676 C-Biosenss, Auvergne University, France); Colette Breysse (Technology Research Centre CASIMIR, Aubière). The authors are grateful to Vincent Desfontaine (Phenomenex) and Julien Lefebvre (Interchim) for the loan of stationary phases for this study.

200

100

References min

0 0

1

2

3

4

5

Fig. 8. Chromatograms of the five plasticizers. (A): Synergi Polar RP; Pout : 100 bar; CO2 /MeOH 99:1 (v/v); T◦ : 35 ◦ C; flow rate: 5 ml/min; lower trace: UV detection at 225 nm; upper trace ELSD response (T◦ ELSD ◦ drift tube : 50 C, Gain: (7); ATBC + DEHT + BBP + TOTM both at 100 ␮g/ml. (B) Synergi polar RP; Pout : 200 bar; CO2 /MeOH 92:8 (v/v); T◦ : 35 ◦ C; flow rate: 2 ml/min; lower trace: UV detection at 225 nm; upper trace ELSD response (T◦ ELSD ◦ drift tube : 25 C, gain: (7); ATBC + DEHT + BBP + TOTM both at 100 ␮g/ml.

[1] [2] [3] [4] [5] [6]

T.A. Berger, B. Berger, R.E. Majors, LC-GC N. Am. 28 (2010) 344. M. Saito, J. Biosci. Bioeng. 115 (2013) 590. G. Guiochon, A. Tarafder, J. Chromatogr. A 1218 (2011) 1037. E. Lesellier, J. Chromatogr. (1216) (2009) 1881. E. Lesellier, J. Sep. Sci. 31 (2008) 1238. K. De Klerck, D. Mangelings, Y. Vander Heyden, J. Pharm. Biomed. Anal. 69 (2012) 77. [7] W. Ren-Qi, O. Teng-Teng, T. Weihua, N. Siu-Choon, Trends Anal. Chem. 37 (2012) 83.

M. Lecoeur et al. / J. Chromatogr. A 1333 (2014) 124–133 [8] J.L. Bernal, M.T. Martin, L. Toribio, J. Chromatogr. A 1313 (2013) 24. [9] R. Lucena, S. Cardenas, M. Valcarcel, Anal. Bioanal. Chem. 388 (2007) 1663. [10] B. Herbreteau, M. Lafosse, L. Morin-Allaury, M. Dreux, J. Chromatogr. A (505) (1990) 299. [11] J.M. Charlesworth, Anal. Chem. 50 (1978) 1414. [12] F.S. Deschamps, A. Baillet, P. Chaminade, Analyst 127 (2002) 35. [13] A.M. Van Nederkassel, V. Vijverman, D.L. Massart, Y. Vander Heyden, J. Chromatogr. A 1085 (2005) 230. [14] S. Almeling, U. Holzgrabe, J. Chromatogr. A 1217 (2010) 2163. [15] E. Lesellier, A. Valarché, C. West, M. Dreux, J. Chromatogr. A 1250 (2012) 220. [16] S. Hoffmann, T. Greibrokk, J. Microcolumn Sep. 1 (1989) 35. [17] J. Thompson, B. Strode, L.T. Taylor, J. Chromatogr. Sci. 34 (1996) 261. [18] G. Guiochon, A. Moysan, C. Holley, J. Liq. Chromatogr. 11 (1988) 2547. [19] E. Lesellier, T. Bizet, A. Valarche, C. West, M. Dreux, LC-GC Eur. 25 (2012) 360. [20] S. Staskiewicz, A. Jones, D. Melillo, J. Labelled Compd. Radiopharm. 50 (2007) 629. [21] M. Köhler, W. Haerdi, P. Christen, J.L. Veuthey, J. High Resolut. Chromatogr. 20 (1997) 62. [22] K. Takahashi, S. Matsuyama, T. Saito, H. Kato, S. Kinugasa, T. Yarita, T. Maeda, H. Kitazawa, M. Bounoshita, J. Chromatogr. A 1193 (2008) 146. [23] E. Lesellier, E. Destandau, C. Grigoras, L. Fougère, C. Elfakir, J. Chromatogr. A 1268 (2012) 157.

133

[24] R.M. Smith, M. Hyytiainen, A.V. Felipe, P.M. Morris, J. Sep. Sci. 24 (2001) 208. [25] Y. Zhao, J. McCauley, X. Pang, L. Kang, H. Yu, J. Zhang, C. Xiong, R. Chen, B. Ma, J. Sep. Sci 36 (2013) 3270. [26] A. Ahlbom, J., Bridges, W. De Jong, J. Hajslová, P. Hartemann, T., Jung, M.O.Mattsson, J.-M.Pagès, K., Rydzynski, D., Stahl, M., Thomsen, D. Williams, Scientific Committee on Emerging and Newly-Identified Health Risks (SCENIHR) Preliminary report on the safety of medical devices containing DEHP-plasticized PVC or other plasticizers on neonates and other groups possibly at risk, June 2007, Brussels. [27] A.D. LaFleur, K.A. Schug, Anal. Chim. Acta 696 (2011) 6. [28] J. Cole, R. Chen, LC-GC Chromatogr. 2 (2009) 1, Online- The application Notebook. [29] L. Eriksson, E. Johansson, N. Kettaneh-Wold, C. Wikström, S. Wold, Design of Experiments–Principles and Applications, 3rd edition, Umetrics, Sweden, 2008. [30] C. West, E. Lesellier, J. Chromatogr. A 1191 (2008) 21. [31] C. West, E. Lesellier, J. Chromatogr. A 1302 (2013) 152. [32] J.D. Pinkston, Eur. J. Mass. Strectrom. 11 (2005) 189. [33] T.L. Chester, J.D. Pinkston, J. Chromatogr. (807) (1998) 265. [34] A. Stolywho, H. Colin, M. Martin, G. Guiochon, J. Chromatogr. 228 (1984) 253. [35] L.E. Oppenheimer, T.H. Mourey, J. Chromatogr. 323 (1985) 297.