Surface partitioning studies of N-methylcarbamate-treated post-harvest crops using SFE-HPLC-postcolumn reaction-fluorescence

Iain A. Stuart,† Ray O. Ansell, John MacLachlan and Peter A. Bather

Department of Physical Sciences, Glasgow Caledonian University, Cowcaddens Road, Glasgow, UK G4 0BA

Received 15th October 1998, Accepted 8th January 1999

The partitioning characteristics of selected carbamate insecticides (carbaryl, aldicarb, bendiocarb and pirimicarb) on five fruit and vegetable types were investigated. Post-harvest samples were surface-saturated with a methanolicÐaqueous mixed carbamate spiking solution for a number of time periods. Samples were taken at 3, 7, 10 and 14 d, and extracted using supercritical CO2 at pressure = 300 atm modified with 10% dimethyl sulfoxide. Extracts were analysed by HPLC-postcolumn reaction-fluorescence detection at lex = 330 nm and lem = 450 nm for N-methylcarbamates and at lex = 315 nm and lem = 380 nm for pirimicarb. The relative partitioning of each insecticide between sample skin and flesh was investigated. This included the determination of both half-life and normalised matrix metabolic rate studies with respect to each carbamate. Multilinear regression (MLR) was applied to a number of insecticide and matrix-based variables to develop regression models for carbamate partitioning for each matrix type studied. Experimentally derived carbamate half-lives ranged from 3.6 d (carbaryl in pear flesh) to 8.0 d (bendiocarb in banana skin). Determinations of normalised metabolic rates were based on calculating the time period from the point of sampling through to the point where carbamate concentration was reduced to 5% of its initial value. These values ranged from 16.2 d (bendiocarb in potato skin) to 34.7 d (bendiocarb in banana skin). Although no practicable MLR partitioning models were obtained, it was found that the models created indicated that carbamate solubility in water (and hence log P) and the number of days in contact with the spiking solution were the most important parameters in model construction.

Introduction The compounds for this study (carbaryl, bendiocarb, pir- imicarb and aldicarb) were chosen for their variation in methods As a consequence of the practice of increasing pesticide of crop uptake, crop host and group class. Additional work has loadings onto crops to counteract developing insect resistance, also determined the relative rate of insecticide degradation there exists a potential for significant quantities of material to exhibited by each fruit or vegetable studied through the pass into the foodchain through uptake in roots, leaves and, calculation of both carbamate half-life and carbamate longevity more importantly, the crop itself. It is this latter area that has in each matrix. A Multilinear regression (MLR) technique was been of increasing interest in recent studies as workers have used to develop regression models between relevant parameters attempted to model the partitioning effects of many hydrophilic in the partitioning process. pesticides on fruit and fruit-like vegetables. Yoshida et al.1 have indicated a partitioning ‘order of importance’ for pesticide cross-over into crop flesh by determining the ratio between Experimental [pesticide]flesh to [pesticide]skin. For ratios of the most common classes of pesticides, this order was shown to be: orga- Instrumentation and reagents nochlorides, organophosphates, carbamates and chrysanthemic acid derivatives in ascending order. This is in good agreement All extractions were completed on a SFE-723M Supercritical with the mean hydrophobicities of each pesticide class. Since Fluid Extraction (SFE) system (Dionex, Camberley, UK) using Noble first suggested that pesticides with log P values > 4.0 two 16 cm3 capacity extraction cells (Keystone Scientific, could be classed as being ‘fat-soluble’,2 the above order is also Bellefonte, PA, USA) in parallel (1200 cm3 linear restrictors as might be expected. were used to maintain back-pressure and flow rate) for each The work reported here provides further information on these extraction run. 99.99% SFE-grade CO2 was used as the primary partitioning effects by using the inert extraction medium of solvent, supplied with a 110 bar He over-pressure (BOC 3 modified supercritical CO2 while also utilising the highly Speciality Gases, Guildford, UK). An extracting fluid density of selective carbamate detection method of HPLC-postcolumn 23 0.795 g cm unmodified was used, modified with 10% mol/v reaction-fluorescence.4Ð17 The increase in partitioning of each DMSO using the instrumental conditions as described in compound in individual skin and, subsequently, flesh samples Table 1. has been monitored over a period of 14 d continuous carbamate contact. Samples were taken at 3, 7, 10 and 14 d, and extracted using dimethyl sulfoxide (DMSO)-modified supercritical Procedure CO2. All carbamate standard and spiking solutions were made up † Present address: Strathclyde Institute for Drug Research, University of from certified insecticides: carbaryl, carbofuran, aldicarb, Strathclyde, Royal College, 204 George Street, Glasgow, UK G1 1XW. bendiocarb and pirimicarb (Promochem, UK). Chromato-

Analyst, 1999, 124, 275Ð280 275 graphic solvents used were of HPLC-grade (Sigma-Aldrich, reaction-fluorescence using the conditions as described Poole, Dorset, UK) pumped via a LC 9012 Solvent Delivery earlier. System (Varian, Walnut Creek, CA, USA) with a 10 ml injection loop. Detection was completed by postcolumn reaction- fluorescence on a scanning wavelength detector (model 9070, Study of the metabolic action of crops on carbamates. The sample spiking procedures for this work were identical to those Varian) at lex = 330 nm and lem = 450 nm for carbaryl, bendiocarb, carbofuran (internal standard) and aldicarb iso- used previously. Individual carbamate levels were determined in the skin and flesh of all five samples after a period of 10 d in indole derivatives. Pirimicarb was detected at lex = 315 nm and l = 380 nm without the presence of the reagents. contact with the carbamate spiking solution. These ‘Day 10’ em samples were removed from the spiking solution, washed and A linear gradient of 100 + 0 H2OÐTHF to 30 + 70 H2OÐTHF in 20 min was used to complete carbamate elution at a flow rate refrigerated. Samples were taken on days 3, 7, 10 and 14 post- of 1 cm3 min21. All determinations were completed on a 150 removal from the carbamate spiking solution, mixed directly with Celite and extracted using the conditions as described in mm 3 4.6 mm C18 column at 42 ¡C (Pickering Laboratories, Mountain View, CA, USA) contained in the postcolumn Table 1. Both relative carbamate half-life and longevity values reaction module (PCX 5100, Pickering Laboratories). Fluores- were calculated on the data obtained from this part of the work. cence reagents, o-phthaldehyde (OPA), NaOH hydrolyser Due to the dehydrating effect of refrigeration, these samples (0.3% at 100 ¡C), OPA diluent (0.3% boric acid) and Thiofluor were periodically moistened with a water spray to maintain (N,N-dimethyl-2-mercaptoethylamine hydrochloride) were of biological activity at an, albeit, reduced capacity due to the chromatographic grade (Pickering Laboratories). incapacitating effect of the reduced temperature on crop metabolic rate. The additional benefit of refrigeration is that it is possible to mimic the storage method adopted by many fruit and Crop sample preparation. Washed whole banana, pear, vegetable producers/vendors and so obtain a more ‘real world’ onion, potato and apple samples were immersed in a MeOHÐ carbamate half-life value under these conditions. 23 H2O (20 + 80) mixture of the carbamates (50 mg cm with the exception of aldicarb at 25 mg cm23). All samples were subsequently refrigerated for 3, 7, 10 and 14 d (4 ¡C) in order to simulate any partitioning/contact processes involved between Results and discussion the crop and the carbamates. A 10 cm3 sample of the spiking solution was removed prior to the addition of the fruit/vegetable Carbamate concentration against matrix types (Fig. 1) sample and stored under the same conditions. This aliquot was analysed on each day of sampling with the resultant carbamate When comparing carbamate concentrations in both parts of each response being used to normalise carbamate recovery, inde- crop type, it is possible to rank which crops are susceptible to pendent of carbamate stability in the spiking solution as a greatest pesticide cross-over. Those crops that possess high function of time. partitioning characteristics in both parts of the crop are naturally of greatest concern. It was noticeable that the partitioning trends for the aromatic Sample extraction and determination. On completion of carbamates, e.g., bendiocarb, pirimicarb, and carbaryl, show the required time period of contact, each sample was washed that the magnitude of partitioning across all matrices tends to thoroughly in a deionised water bath to remove any surface- rank in order of decreasing hydrophobicity, e.g., carbamates adhered residues (adsorption). Samples were divided into skin with higher relative log Pcarbamate values possess a higher and flesh portions with the inner surface of each skin portion affinity for partitioning than those of more hydrophilic nature. being scraped to remove the lower dermis layer which was However, this correlation is not maintained in recovery trends bulked with the flesh fraction. In addition, only the outermost for aldicarb (aliphatic) which is absorbed to a greater level in all 0.5 cm thickness of sample flesh was used. Individual flesh and cases than its relative hydrophobicity would suggest. One skin samples were then chopped into individual 2 g samples, possible reason for this apparent anomaly is that the lipophilic air-dried for 24 h in a fan-assisted oven (no heat) and stored in 2-methylthiopropyl tail may orientate the molecule in such a a desiccator on the day of use. The samples were mixed with an way as to maximise absorbed aldicarb onto the dried, lipidic equal amount of Celite to ensure uniform cell packing. In surface. This is further confirmed in Fig. 1(a) as absorbed addition, cells were packed at each end with methanol-cleaned aldicarb is shown to be the second highest in the entire data set glass wool to prevent end-cell frit blockage. On completion of when determined from onion skin (the most lipophilic substrate extraction, each extract was filtered as necessary and spiked investigated). with 1 cm3 of an aqueous 50 ppm carbofuran internal standard. From the viewpoint of matrix type, all carbamates are The extract was then reduced to near dryness and reconstituted strongly absorbed by all matrices studied. One important with 1 cm3 of methanol and analysed by HPLC-postcolumn observation is that in matrices that can be classified as being lipophilic, such as the crop skin samples, residue partitioning is considerably higher in comparison with matrices of high Table 1 Supercritical fluid extraction conditions moisture content (flesh samples). Within this, there is a definite Parameter Level/setting ranking of partitioning in both forms of the crop. It is most noticeable that, for all carbamates, onion skin demonstrates Extraction temperature 70 ¡C significantly greater partitioning kinetics in comparison with Extraction pressure 300 atm either other skin samples or the entire data set as a whole. Total extraction time 60 min Additionally, banana flesh also indicates elevated concentra- Restrictor temperature 70 ¡C Restrictor volume 500 ml tions of carbamates across the data set. Flow rate (gas state)a 860 ml min21 Difficulties in completely drying flesh samples rapidly prior Modifiers used DMSO (10%) to extraction without incurring excessive thermal damage to the Solvent collection Liquid collection in vial (15 ml) matrix studied can cause errors in determining the true Solvent collection temperature 1 ¡C concentration of individual carbamates. Inaccurate carbamate Extraction cell geometry 14 mm 3 100 mm (16 ml capacity) concentrations can also be obtained through the presence of Cell packing Methanol-cleaned glass wool with Celite wet support greater inherent moisture in some matrices as a combination of moisture and heat also maintains enzymatic activity within the a Mean flow rate. host matrix thus decreasing the carbamate concentration more

276 Analyst, 1999, 124, 275Ð280 rapidly. Consequently, future work may focus on using freeze processes of carbamate build-up and decay. In this context, drying at this stage. Equally as important, competitive solvation MLR is used to individually model these concepts with a view between bound water within the matrix and the extracting fluid to quantifying their respective effect on carbamate persistence causes additional limitations in carbamate quantification from in the foodstuffs selected for study. the flesh matrices. Generally, it appears that the aromatic In this study, days of carbamate contact, sample matrix, carbamates studied have a greater affinity to lipophilic sub- carbamate type, log Pcarbamate and Sol.H2Ocarbamate predictors strates, again implying that the attraction between the site of were investigated for their respective relations to [carba- partitioning and the carbamate is directed by the hydrophobic mate]skin, [carbamate]flesh and their ratio (flesh/skin). Statistical property of the carbamate backbone. outputs were obtained using MINITAB v10 software (MINI- In comparing partitioning values obtained for ‘Day 10’ and TAB, PA, USA) with Table 2 illustrating the relations ‘Day 14’, it is apparent, by varying degrees, that the relative investigated in the first instance. concentrations of carbamate have decreased between these It became evident from the models obtained that carbamate times. This implies that the equilibrium between carbamate partitioning/recovery data did not produce models that were partitioning and carbamate metabolic degradation has shifted statistically suitable to be used for partitioning prediction. It was 2 towards carbamate degradation. This indicates that the surface found that due to low r adj values and poor fitting illustrated by has become saturated with carbamate. As no further partitioning low F-values for significance, the models created only partially is possible due to both the build-up of sorbed carbamate explained the data set used in their construction. The most degradation residues and the continuing action of crop enzymes, relevant models produced involved the investigation of the this is as might be expected. variables examined in tests 10Ð13 (Table 3). On consideration of the relative P values obtained, Table 3 also ranks all predictors in order of relevance. Multilinear regression modelling of carbamate From Table 3, it can be concluded that carbamate solubility in partitioning water, carbamate log P and carbamateÐmatrix time of contact are of most importance in surface partitioning of the carbamates The technique of MLR modelling has been applied to the to the matrices studied. This suggests that the models developed recovery data obtained for carbamate partitioning onto both skin in tests 10 and 12 are most relevant and that their respective and flesh samples of selected fruit and vegetable types. models of both skin and flesh concentration are the most Similarly to simple regression, MLR is capable of identifying practicable. This conclusion is further confirmed as these tests 2 the relevance of particular variables considered in the develop- produced the two highest r adj values for all tests conducted. In ment of a model. In addition, it is also possible to determine addition to this, these tests also return the lowest number of which variables, either controlled or independent, have the unusual observations‡ implying a high level of model fit to the greatest effect on the model and which variables can be data. discarded. Fig. 2 displays a typical plot obtained for carbamate partitioning characteristics against both matrix type and time ‡ Unusual observations are determined from large standard residuals (for pear samples) illustrating both partitioning and metabolic calculated and compared to a predetermined level in model construction.

Fig. 1 Matrix type against time. (a) Aldicarb partitioning; (b) pirimicarb partitioning; (c) bendiocarb partitioning; and (d) carbaryl partitioning.

Analyst, 1999, 124, 275Ð280 277 The common factor within these tests is that the ratio term development, [Days of contact], log[Sol. H2O] and the constant ([R]) has been replaced with the individual carbamate concen- terms return acceptably low P values. High P values were trations in skin and flesh. Although Table 3 indicates that less returned for both the sample matrix (0.735) and carbamate type than half of the data set can be explained by the models (0.808) indicating poor relevance to the model. This implies that produced, within this, coefficient relevance was found to be the constructed regression equation is both carbamate type and exceptionally well fitted by contrast. In tests 10 and 12, P values matrix independent. As the carbamate solubility term has been obtained for coefficient relevance to the model were found to be shown to be highly relevant to the model, it can be seen that the zero in all cases; thus the P values obtained for the overall best use of carbamate type is unnecessary. The above also indicates fitted model were also zero, implying high variable sig- that there is no significant difference between carbamate nificance. In testing the larger model involving the additional partitioning in comparing different matrix skin and flesh terms of carbamate type and matrix shown in tests 12 and 13, the absorbed concentrations respectively. coefficient P values obtained varied greatly and thus can be All these observations are mirrored in the test 13 regression interpreted for relevance to the overall model. In test 12, where model where carbamate concentration in sample flesh was used carbamate concentration in sample skin is used in model as the basis in model construction. It is shown that the majority of coefficient weightings produced are significantly smaller in

carbaryl comparison with the previous model in test 12. Of the most Day 14 flesh bendiocarb significant predictors in the model (Table 3), the lower values in pirimicarb Day 14 skin aldicarb both log[Sol. H2O] and [Days of contact] are of most interest in Day 10 flesh determining carbamate partitioning patterns. As the flesh Day 10 skin portions of each matrix possess the greatest moisture content, it 

Sample Day 7 flesh is apparent that at the spiking levels used in this work, relative Day 7 skin carbamate insolubility in each matrix type is not an issue. Day 3 flesh Conversely, the lipidic nature of skin portions from each matrix Day 3 skin will control the rate of the partitioning process as a function of

0 50 100 150 200 250 300 carbamate hydrophobicity. In addition, time of contact with the [Pear]/mg g Ð1 carbamate spiking solution has also been shown to be an

 important factor, resulting in the model differences between Fig. 2 Graphical representation of the build-up of absorbed carbamate as tests 12 and 13. One possible reason for this is that when a function of time of contact for pear samples. samples were placed in the spiking solution and observed over the 14 d time period, considerable sample swelling was Table 2 CarbamateÐmatrix partitioning relations investigated noticed. Log P values were used in this model but, due to their Test Predictor(s) Response correlation with log[Sol. H2O], log P was removed during model construction. As a result, the predictor log[Sol. H2O] 1 [D] * [H] [R] represents this parameter in models created in tests 12 and 13. 2 [D] * [S] [R] 3 [S] [R] 4 [H] [R] 5 [D] * [H] * [S] (for banana extracts) [R] Carbamate half-life determination 6 [D] * [H] * [S] (for potato extracts) [R] 7 [D] * [H] * [S] (for pear extracts) [R] As insecticide degradation in fruit and vegetable samples can be 8 [D] * [H] * [S] (for onion extracts) [R] estimated as being first order with respect to sorbed carbamate 9 [D] * [H] * [S] (for apple extracts) [R] residues, the following can be applied to determine individual 10 [D] * [S] [Sk] 11 [D] * [S] [F] rate constants for each metabolic process: 12 [Ca]* [D] * [M] * [H] * [S] [Sk] d[Carbamate] 13 [Ca]* [D] * [M] * [H] * [S] [F] Metabolic Rate = - = k[Carbamate] dt [D] Days of contact [Sk] log[carbamate]skin [Ca]carbamate [H] log P [F] log[carbamate]flesh [M] matrix rearranged to: [S] log[Sol. H2O] [R] log F/S ratio - d[Carbamate] = kdt [Carbamate] Table 3 Most relevant regression models constructed for carbamate partitioning and on integrating between initial sampling (0) and final sampling (t): 2 Test Constructed regression model R adj (%) [Carbamate]t t d[Carbamate] 10 log[Carbamate]skin = 0.0565 [Days of contact] 2 45.3 - = k dt [Carbamate] 0.418 log[Sol. H2O] + 2.33 Ú Ú [Carbamate] 0 11 log[Carbamate]flesh = 0.0486 [Days of contact] 2 41.9 0 0.38 log[Sol. H2O] + 2.28 12 log[Carbamate]skin = 0.0108 [Matrix] + 0.0565 44.0 On solving for [Carbamate]t and [Carbamate]0 and rearrang- [Days of contact] + 0.039 [Carbamate type] 2 ing: 0.356 log[Sol. H2O] + 2.01 [Carbamate]t - kt 13 log[Carbamate]flesh = 0.003 [Matrix] + 0.0486 40.5 log = [Days of contact] + 0.072 [Carbamate type] 2 [Carbamate]0 2.303 0.265 log[Sol. H2O] + 1.75 From this, k, the metabolic rate constant, was determined. It was Test example Predictor P value also necessary to determine individual carbamate half-lives in each matrix type. As a consequence [Carbamate] was substi- 12,13 log[Sol. H2O] 0.000 t 1 12,13 log P 0.000 tuted by ⁄2[Carbamate]0 in the above to give: 12,13 [Days of contact] 0.000 1 [Carbamate] - kt1 13 [Carbamate type] 0.808 log 2 0 = 2 12 [Matrix type] 0.919 [Carbamate]0 2.303

278 Analyst, 1999, 124, 275Ð280 On cancellation and simplification: level and determine the time period required to reach this level. 0.693 For illustration, only full plots for banana skin and flesh, Fig. t1 = 2 3(a) and (b), respectively, are included. Calculated values for k the above are also included in Table 4. Half-lives were determined by pre-calculation of the rate constant, k1⁄ , for the decay and each constant was used in the 2 Calculated carbamate half-life and longevity values final determination of t1 . Calculated carbamate half-life values ⁄2 in each matrix type are included in Table 4. Data obtained for the half-life calculations for each carbamate (Tables 4 and 5) indicate that, in general, carbamate stability is Carbamate longevity calculation low in each matrix type in comparison with previous insecticide generations, e.g., the organochlorides.¤ In matrices that contain From the carbamate recovery data obtained for the metabolic relatively high moisture values, e.g., the majority of the matrix flesh samples (t1 : 3Ð4 d), the carbamate half-lives are seen to be rate study, exponential decay regression equations were devel- ⁄2 oped for each carbamate from each matrix studied. These lower in comparison with matrices of low inherent moisture, e.g., banana skin (t1 : 4.02Ð8.01 d). It would also appear that the expressions were applied in the determination of carbamate ⁄2 longevity in each matrix type. This was completed by setting an sole aliphatic carbamate, aldicarb, is the most stable under these arbitrary ‘safe level’ of carbamate concentration at 5% of the conditions with carbaryl being the least stable. total incurred carbamate concentration determined from the Values obtained for ‘5%’ levels (Tables 4 and 5) all indicate extraction and quantification of carbamate concentration from that detectable levels of carbamate residue exist in all matrices ‘Day 10’ samples. Using exponential fits for each carbamate over periods of up to 1 month, e.g., bendiocarb and carbaryl in concentration plotted as mg g21 of sample against post-sampling time, it was possible, through algebraic manipulation of each ¤ Kushwaha et al. determined the half-life of aldrin in soil at 49 and 177 d expression, to extrapolate the carbamate decay rate to the 5% (25 and 35 ¡C, respectively).18

Table 4 Observed and calculated recovery data for carbaryl, pirimicarb, bendiocarb and aldicarb from selected fruit and vegetable samples

[Day 0]/ [Day 14]/ 21a 21a b c Matrix Portion Carbamate mg g mg g Exponential fit Half-life (t1 )/d ‘5%’ level/d ⁄2

Banana Skin Aldicarb 48.47 4.34 y = 46.73e20.1743x 4.0 17.0 Pirimicarb 15.90 2.67 y = 14.99e20.1210x 5.4 24.3 Bendiocarb 80.23 23.92 y = 79.41e20.0867x 8.0 34.7 Carbaryl 339.41 95.76 y = 342.45e20.0909x 7.6 33.0 Flesh Aldicarb 131.47 29.60 y = 131.27e20.1079x 6.5 27.7 Pirimicarb 40.51 5.81 y = 35.25e20.1350x 4.9 21.2 Bendiocarb 191.41 38.99 y = 185.38e20.1108x 6.1 26.7 Carbaryl 680.06 86.55 y = 691.84e20.1439x 4.7 20.9 Potato Skin Aldicarb 45.46 7.01 y = 39.02e20.1308x 5.2 21.7 Pirimicarb 25.47 4.46 y = 21.69e20.1218x 5.7 23.3 Bendiocarb 81.07 5.73 y = 87.95e20.1898x 3.6 16.2 Carbaryl 379.92 31.74 y = 404.51e20.1801x 3.9 26.1 Flesh Aldicarb 20.21 3.91 y = 18.07e20.1173x 5.9 24.6 Pirimicarb 16.12 3.14 y = 14.84e20.1124x 5.9 25.9 Bendiocarb 37.13 5.17 y = 37.53e20.1395x 4.9 21.5 Carbaryl 143.77 16.20 y = 140.14e20.1546x 4.4 19.2 Pear Skin Aldicarb 16.16 2.39 y = 12.93e20.1349x 5.1 20.5 Pirimicarb 31.27 4.09 y = 27.36e20.1406x 4.7 20.3 Bendiocarb 30.29 3.78 y = 25.27e20.1451x 4.6 19.4 Carbaryl 115.50 15.07 y = 114.10e20.1474x 4.7 20.2 Flesh Aldicarb 38.54 4.23 y = 35.77e20.1601x 6.7 18.2 Pirimicarb 17.78 2.91 y = 16.08e20.1273x 5.3 22.7 Bendiocarb 48.42 4.06 y = 49.43e20.1738x 3.9 17.3 Carbaryl 235.35 16.23 y = 250.67e20.1912x 3.6 16.0 Onion Skin Aldicarb 233.65 21.13 y = 246.74e20.1701x 4.0 17.9 Pirimicarb 85.42 8.66 y = 72.90e20.1649x 4.2 17.2 Bendiocarb 328.89 30.04 y = 365.75e20.1716x 4.0 18.1 Carbaryl 964.14 76.91 y = 841.36e20.1734x 3.8 16.5 Flesh Aldicarb 19.52 3.74 y = 16.35e20.1133x 5.8 24.9 Pirimicarb 24.32 4.02 y = 21.28e20.1265x 5.3 22.6 Bendiocarb 20.57 3.46 y = 16.68e20.1210x 5.4 23.0 Carbaryl 124.58 11.47 y = 125.40e20.1689x 4.0 17.7 Apple Skin Aldicarb 22.57 3.03 y = 18.81e20.1412x 4.8 31.1 Pirimicarb 16.22 3.92 y = 14.15e20.1008x 6.8 28.4 Bendiocarb 44.33 5.18 y = 41.19e20.1474x 4.5 19.8 Carbaryl 190.03 16.08 y = 170.11e20.1785x 3.9 16.2 Flesh Aldicarb 77.90 10.38 y = 69.00e20.1382x 4.8 20.8 Pirimicarb 38.54 5.18 y = 32.99e20.1391x 4.8 20.4 Bendiocarb 126.90 12.82 y = 124.35e20.1616x 4.2 18.4 Carbaryl 433.53 50.28 y = 409.19e20.1517x 4.5 19.4 a Carbamate concentrations on the first and final day of sampling, respectively. b Best exponential fit for each data set over the five sampling days. c The time taken for each carbamate residue to decrease to within 5% of the respective total day 0 sample value.

Analyst, 1999, 124, 275Ð280 279 banana skin. As expected, matrices of high moisture content regression models using MLR has been limited. This result only metabolise each carbamate at a greater rate, with typical ‘5%’ strengthens the argument that the application of linear model- level time periods of approximately half that of matrices of ling techniques to active biological systems such as matrix lower moisture content, e.g., bendiocarb and carbaryl in pear partitioning could be inappropriate. This is due mainly to the flesh. introduction of large uncertainties from many, normally uncontrolled, non-linear systems active within the experiment. As it may be possible to improve model relevance by increasing Conclusions experimentation replicates, it would be more prudent to focus this solely on one matrix type. Although no practical models In this work it has been shown that it is possible to estimate both were developed, it was possible to determine which predictors pesticide half-life and potential longevity in a variety of had the greatest effect on model construction. It was found that, matrices. The potential of carbamate cross-over into the food- of all predictors used, only three were consistently statistically chain has been demonstrated as a function of resident time relevant in the 13 tests conducted, namely, log[Sol. H2O] (log within the food sample. The success in developing practical P), [Days of contact] and [Carbamate type].

References

1 S. Yoshida, H. Murata and M. Imaida, J. Jpn. Soc. Biosci. Biotech. Agrochem., 1992, 66(6), 1007. 2 A. Noble, J. Chromatogr., 1993, 642, 3. 3 I. A. Stuart, J. Maclachlan and A. McNaughtan, Analyst, 1996, 121, 11R. 4 H. A. Moye, S. J. Scherer and P. A. St. John, Anal. Lett., 1977, 10, 1049. 5 Y. Tsumura, K. Ujita, Y. Nakamura, Y. Tonogai and Y. Ito, J. Food Prot., 1994, 571, 1001. 6 S. Chiron and D. Barcelo, J. Chromatogr., 1993, 645, 125. 7 H. Hiemstra and A. de Kok, J. Chromatogr. A, 1994, 667, 155. 8 R. J. Argauer, K. I. Eller, M. A. Ibrahim and R. T. Brown, J. Agric. Food Chem., 1995, 43, 2774. 9 Y. Tsumura, K. Ujita, Y. Nakamura, Y. Tonogai and Y. Ito, J. Food Prot., 1995, 58, 217. 10 K. M. S. Sundaram and J. Curry, J. Chromatogr. A, 1994, 672, 117. 11 M. S. Ali, J. D. White, R. S. Bakowski, E. T. Phillippo and R. L. Ellis, J. AOAC Int., 1993, 76, 1309. 12 M. S. Ali, J. D. White, R. S. Bakowski, N. K. Stapleton, K. A. Fig. 3 Graphical representation of carbamate decay in (a) banana skin and Williams, R. C. Johnson, E. T. Phillippo, R. W. Woods and R. L. (b) banana flesh as a function of post-sampling time. Ellis, J. AOAC Int., 1993, 76, 907. 13 V. A. Simon, K. S. Pearson and A. Taylar, J. Chromatogr., 1993, 643, Table 5 Reduced summary statistics of Table 4 calculated values 317. 14 H. Frister, H. Meisel and E. Schlimme, Fresenius’ Anal. Chem., 1988, Mean half- Mean ‘5%’ 330, 631. Carbamate life/d RSD (%) level/d RSD (%) 15 S. S. Simons and J. Johnson, J. Am. Chem. Soc., 1976, 98, 7098. 16 R. T. Krause, J. Chromatogr. Sci., 1978, 16, 281. Aldicarb 5.2 0.9 22.4 4.6 17 R. T. Krause, J. Chromatogr., 1979, 185, 615. Pirimicarb 5.3 0.7 22.6 3.1 18 K. S. Kushwaha, H. C. L. Gupta and V. S. Kavadia, Ann. Arid. Zone, Bendiocarb 4.9 1.3 21.5 5.5 1978, 17, 200. Carbaryl 4.5 1.1 20.5 5.3 Paper 8/08013E

280 Analyst, 1999, 124, 275Ð280