Honey contamination from plant protection products approved for cocoa cultivation: a 2 systematic review of existing research and methods

Abstract

2 25 quantify existing studies on honey contamination from plant protection products approved for 26 cocoa cultivation. We observed that one hundred and sixty-nine different compounds, 27 comprising some recommended and other unapproved compounds for cocoa cultivation, were 28 detected in 81% of the reviewed 104 publications. Our results further point to the 29 neonicotinoids as the most detected class of pesticides, with imidacloprid particularly being 30 the single most detected compound. However, the most remarkable observation made from this 31 study points to disproportionate studies of honey contamination from pesticides conducted in 32 cocoa and non-cocoa producing countries with only 19% of the publications taking place in the 33 latter. To bridge the gap, we suggest prioritising increased research in cocoa growing countries 34 to ameliorate the significant gaps in knowledge owing to limited studies emanating from these 35 geographic regions. countries, especially those from the developing world, heavily rely on the income generated 42 from its export to support their economies [3,4]. The economic value of cocoa is estimated in 43 the region of US$11.8 billion from global annual production of 4.2 million metric tons of cocoa 44 beans [2]. The overall cocoa confectionary market generates about US$80 billion worldwide 45 [5] with West Africa being its main production hub. The cocoa plant is, however, vulnerable 46 [6, 7] to attacks from the cocoa swollen shoot virus, beetles and capsids (miridae) and 47 100 following Joint Meetings on Pesticide Residues [35]. As a natural food produced by Appis 101 mellifera, honey is deemed a food substance of animal origin under Directive 2001/110/EC 102 and therefore needs to meet specified requirements. The EU has established MRLs for 103 pesticides in honey which range from 0.05 mg/kg to 0.2 mg/kg [36]. Where an MRL is not 104 specified, the default limit of 0.05 mg/kg is applied in honey [37]. However, different national 105 or regional bodies also tend to set different upper pesticide residue concentration limits but this 106 has been cited as a source of confusion in international markets [38]. Harmonization and 107 standardisation of MRLs is therefore a necessity.

109
Honey bees collect water, nectar pollen and other resources from the environment into the 110 beehive for the production of honey [39]. The foraging behaviour of bees throughout the 111 general environment means that bees can transport foreign bodies which may compromise 112 colony products. Honey, as a hive product, is susceptible to potential contamination from the 113 use of applied pesticides for crop production [40] and can be used as a proxy to evaluate the 114 general state and health of the environment because of its ability to reveal the chemical 115 condition of the environment [41,42]. The levels of contaminants in honeybees, honey and 116 pollen have been used to assess the level of heavy metals in the environment in both natural 117 and human disturbed landscapes [43,44]. Similarly, Perugini et al.,[45] used contaminations 118 in honey bees, honey and pollen to access the presence of polycyclic aromatic hydrocarbons in 119 the environment. It is clear therefore that studies of pesticide contamination in bee products 120 such as honey can be used to assess the extent of pesticide use in a particular environment. This 121 review utilises this relationship between pesticide contamination of honey and pesticide use in study. Prior to developing the search strings, a preliminary study was undertaken in Google 138 Scholar to identify key publications that report on the pesticides recommended for cocoa 139 production to guide the construction of the search strings. Eight publications (seven peer 140 reviews and one international report) were found that reported on approved pesticides for cocoa 141 cultivation in key cocoa producing countries, namely Ghana,Nigeria,Cameroun and Ivory 142 Coast, which account for 70% of the world's cocoa production [46]. These publications 143 facilitated the compilation of a list of the key pesticides comprising 23 insecticides, 17 144 fungicides and 2 herbicides approved for cocoa cultivation (S1 Table). Following the 145 successful compilation of this list, three sets of search strings (S2 Table) were formulated for 146 literature search. Search strings were broken into two to reduce the length of the string for each 147 category of search engine.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in using the developed search strings which resulted in 524 retrieved articles which were also 164 exported into Endnotes bringing the total dataset of retrieved article to 1,884. The author lists 165 and titles of exported dataset were compiled and listed in Microsoft Excel to identify and 166 remove duplicates. Duplicates based on the combined list were then removed using conditional 167 formatting in Excel resulting in a dataset of 1,282 articles. These were then screened based on 168 the titles and the abstracts to assess the studies which report residues of active ingredients of 169 pesticides in honey. However, no restriction on study design, date, or geographical zones were 170 applied during the literature search for the relevant studies for this review. The screened data 171 yielded a total of 91 papers. However, one article was not accessible. An email was sent to the 8 172 corresponding author with a request for the paper, but no positive response was received.

173
Subsequently this paper too was excluded. The shortlisted papers for inclusion at the beginning 174 of the study comprised 90 articles. These were subjected to quality assessment prior to data 175 extraction. This resulted in the further exclusion of 2 studies (see section 2.2). The flow chart 176 in (Fig 1)

195
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 27, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022 2.2 Quality assessment 197 Before conducting data extraction, each selected study was appraised to assess it in terms of 198 quality and suitability (S3 Table). This allowed us to evaluate the strengths and limitations of 199 each study [51]. The appraisal process enabled us to include only studies which were of 200 sufficient quality for this review. Our study was designed to address multiple study objectives 201 by evaluating a broad range of multiple issues that included differences between countries of 202 study, analytical techniques, extraction techniques, detected pesticide residues in the context 203 of limit of detection and quantification, among others. In order to have reproducible criteria for 204 the critical assessment of the quality of all selected studies and also taking into consideration 205 the aims of our review, we applied a checklist of eleven customized questions (S8 Appendix) 206 based on proposed checklists [52] for evaluating quantitative studies. The included studies were 207 appraised by two reviewers using the scoring system previously applied by [52]. In this grading 208 system, selected studies are scored to what extent they satisfied the criteria ("yes"=2, 209 "partial"=1, "no"=0, NA=not applicable). The scoring systems developed by [52] had been 210 designed using the guidelines previously developed by [53] and [54]. In our study, the first 211 reviewer appraised each selected study which was subsequently validated by a second 212 reviewer. The overall scores agreed by reviewers ranged from 45% to 100%. Based on the 213 outcome of the quality assessment conducted, reviewers agreed for eight studies (two from the 214 previous search and six from the up-to-date search) to be excluded. Of these eight studies, three 215 analysed pesticides residues in honey samples were collected from multiple countries and 216 therefore the study could not be assigned to one specific country for analysis. The other five 217 studies only utilised blank honey as a sample matrix exclusively to demonstrate the robustness 218 of the analytical method but did not quantify concentrations of pesticide residues in those 219 samples. They were therefore deemed ineligible for further assessment. Overall, one hundred . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in Pesticides were recorded in three categories, namely insecticides, fungicides, and herbicides.

Results
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in  244 The results reveal that the first of the 104 papers included in this study was published in 1997.

245
However, publications were not sustained immediately after 1997 as the next published paper 246 was recorded five years later (Fig 2A). Even though the trend of publication persisted thereafter  (Fig 2A & B). The aim and key findings from each study are summarised in the data set (S4 252   Table).  CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 27, 2022. ; https://doi.org/10.1101/2022.12.26.521958 doi: bioRxiv preprint 266 cultivated. Further observation made revealed that at least one study was conducted on each 267 continent except Antarctica where no studies took place (Fig 3). Out of the 20 studies which took place in cocoa producing countries results show that 8 studies 275 were conducted in Brazil which is ranked sixth highest cocoa producing country in the world 276 and accounts for 5% of global production of cocoa beans (S5 Table). A further 4 studies were 277 carried out in India ranked as the sixteenth cocoa producing country with production levels 278 being less than 1% of global production. Only one study was conducted in each of Ivory Coast

289
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in  is the smallest concentration of a substance which can successfully be quantified [57,58]. Both 295 LOD and LOQ were evaluated for all pesticide residues that were detected in the included 296 studies (S3 Table & S4 Table). Our result showed that 80% of included studies applied LODs 297 which were below the specified EU MRLs for the studied compounds. However little  Similarly, most studies (77%) clearly indicated the applied LOQs which were found to be lower 301 than the MRL. In 17 others, limited information was provided on applied LOQ. From the 302 clearly stated LOQs, three were found to exceed the MRL. Furthermore, in two other studies, 303 the LOQ did not exceed the MRL, but they were not sufficiently low enough for the 304 quantification of trace elements.  Overall, only 4 publications studied insecticides, fungicides, and herbicides together in the 311 same study. It was observed that pesticide residues were detected in 80% of the 104 studies 312 reviewed. Our study found that a total of 169 different compounds, comprising of some of . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in  Table). Eleven 330 approved insecticides for cocoa cultivation namely capsaicin, chlorantraniliprole, 331 thiamethoxam, acetaprimid, etofenprox, indoxacarb, pirimiphosmethyl, promecarb, 332 pyrethrum, sulfoxaflor, and teflubenzuron and one herbicide (i.e., paraquat) were not detected 333 in any of the studies conducted in the cocoa growing countries. Additionally, our findings 334 showed that only 2 of the 18 recommended fungicides for cocoa production (S1 Table), namely . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in  The outcome from this study shows that multiple studies took place in some countries (i.e., 347 54% of included 35 countries) with one study taking place in the rest. Spain and China, 348 accounted for the highest number with fifteen studies each. Among the cocoa producing 349 countries, our results show that Brazil, India, and Mexico were the only the countries where 350 more than one study was conducted. Our results show that some pesticides were separately 351 detected in different studies within in the same country (seven countries in all). An assessment 352 of the concentrations of these pesticides, where these pesticides were detected in the same 353 jurisdiction, sowed that they were usually found to be at varying concentrations (Table 2).

Exceedance of EU MRLs
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in Overall, 13% of the included studies in ten countries recorded concentrations of pesticide 388 residues which exceeded the maximum residue limit (MRL) set by the European Union in 389 honey (Table 3). Of note, EU MRLs are occasionally revised in light of additional scientific 390 data becoming available to the European Food Safety Authority, and during the time period of 391 this study, these revisions resulted in an increase in MRLs for certain pesticides, It was 392 additionally observed that there was one study where the malathion concentration was found 393 to have exceeded the MRL set in India, but the upper limit was found to be below that of the  . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in Outcomes from such multiple sample collection recorded varied outcomes. No pesticide 438 residues were detected in the study conducted in Uganda. It was observed that in two 439 independent studies conducted in Chile, no pesticide residues were detected in one study, but

459
The findings from our study showed that a vast array of both traditional and other novel 460 extraction techniques was applied for trace analysis. However, the "Quick, Easy, Cheap, . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in  involving either a technique being applied independently or in combination with others, were 508 applied for analysis. However, we observed that the most predominantly applied technique was 509 Liquid Chromatography-tandem mass spectrometry (LC-MS/MS), which was applied in 40 % 510 of the total studies. It was further applied in conjunction with 13 other techniques for analysis.
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 27, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022 Like the extraction, the performances of the different analytical techniques applied in the 512 various studies were not a subject matter of this study. However, the higher frequency of use  Pesticides can be classed according to their uses and in this regard can be categorised as 575 fungicides, weedicide/herbicides, nematicides, rodenticides and insecticides [88]. Moreover, 576 on the basis of chemical structure, further grouping of major pesticides into organochlorines, 577 organophosphates, carbamates, pyrethroids, triazines, and neonicotinoids also exists [42,88].

578
The findings from this study provide evidence of these different classes of pesticides receiving 579 attention in the included studies even though the compounds or classes of pesticides studied 580 were not given equal attention across different geographic locations (Fig 5). We observed that 581 the neonicotinoids were the most detected of all compounds (Fig S6). These findings agree CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 27, 2022. ; https://doi.org/10.1101/2022.12.26.521958 doi: bioRxiv preprint 586 very much likely even after several months or years after application. Moreover, ever since 587 neonicotinoids were developed from the 1980s to replace the more persistent organochlorines 588 in the environment [91,93], there has been a great demand for them, and particularly for 589 imidacloprid after it was introduced to the market [94]. It was therefore not surprising that 590 imidacloprid was found to be the most detected compound in our study. Interestingly, this 591 observation was also confirmed by Mitchell et al [95] and Kavanagh et. al. [96]. Imidacloprid   Table). Among 603 the plausible reasons for this finding are that these pesticides have been found to be inexpensive countries has witnessed remarkable successes [101] but such success has not been witnessed . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 27, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022   MRLs have been set for honeybees and hive matrices including honey by the European 634 Commission [34]. In the present study, we found that 13% of included studies detected 635 pesticides whose concentrations exceeded allowable limits required for human consumption . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 27, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022 (Table 3), of which three cocoa producing countries accounted for a quarter of the studies where 637 pesticides exceeded the MRL. One further important observation from our study shows that 638 some detected pesticides residues, which exceeded the previously specified MRLs set by the 639 EU at the time of the study, are presently below the revised MRLs that have since been 640 implemented in the EU. This is significant as it implies that products that were previously where specified guidelines including detailed information on labels, are followed for the safe 652 and sustainable production of crops and livestock to maximise profit with minimal impacts to 653 the environment, has been touted as capable of preventing excessive leaking of plant protection 654 products into the environment [88,110,111]. While we do not have detailed information of 655 agricultural practices in the areas where MRL was exceeded, it should be considered that 656 generally a significant proportion of applied pesticides also find their way to the general 657 ecosystem with just about 0.01% of applied pesticides reaching its target and the rest filtering 658 into the into general ecosystem [18,19]. Though strict enforcement of compliance to specified 659 MRLs in honey will help promote health safety measures, a globally standardised MRLs will 660 ensure clarity and prevent potential confusion. We observed that the concentration of malathion . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in India but found to be below upper limits of the MRL set by the EU. This can be a potential 663 source of confusion for which reason standardisation across borders will be essential. It should 664 also be considered that while the revision of MRL can impact the assessment of honey as a 665 food product, it does not alter the assessment of pesticide contamination levels as an indicator 666 of potential contamination of the environment surrounding the hive. suggests that studies included in this review largely applied analytical methods with good 674 sensitivity. It must, however, be noted that LODs for three studies were not suitable for 675 detecting pesticide residues below EU MRLs, compromising the extent to which their results 676 could be considered within this study. Specifically, even though no pesticide residues were 677 detected by [114] in a study conducted in Brazil, their reported LODs and LOQs which ranged 678 from 0.07 mg/kg to 0.25 mg/kg and 0.02 mg/kg to 0.08 mg/kg respectively were higher than 679 EU MRLs, and so the study's reports of no pesticides detected cannot be interpreted to mean 680 that there were no pesticides present at concentrations with the potential to cause harm.

681
Similarly, although pesticides were detected by [115] and [116] in studies conducted in Brazil, 682 the LOQs achieved for the method were at concentrations so high that their findings cannot be 683 interpreted to mean that the pesticides detected were the only ones which were a cause for 684 concern.
685 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in also belong to. Toxicity levels of imidaclorpid, clothianidin and thiamethoxam in the ranges 696 of of 0.004 to 0.075 µg/bee have been established to be lethal to bees [122][123][124]. Although this 697 is below the known LD 50 s for these compounds for bees [125], they are within the range of 698 concentrations known to have sub-lethal effects. For instance bumble bees were found to 699 reduce learning capability and their short-term memeory impaired drastically when exposed to 700 field realistic levels to 0.0024 mg/kg of thiamathoxam [126], which is 10-fold lower impacted by exposure to 0.0024 mg/kg of thiamathoxam [127]. In another study, Straub,704 Villamar-Bouza [128] confirmed that the survival of honeybees was reduced by 51% as well 705 as reduced flight activities when exposed to 0.0043 mg/kg and .0011 mg/kg concentrations of 706 thiamethoxam and clothianidin respectively. Therefore, there may be potential sub-lethal 707 effects of the detected pesticide residues in honey on bees which deserves further consideration.

709
It was observed from this study that pesticide residues were detected in 80% of both 710 commercial and raw honeys analysed in the included studies. This finding reflects a similar . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in  The current state of knowledge of studies of honey contamination from pesticides approved for 731 cocoa cultivation has been evaluated through a systematic literature review. The studies 732 conducted over the period have been disproportionately focused on non-cocoa growing areas 733 leaving a huge gap of knowledge of how pesticides approved for cocoa cultivation affects bee 734 products particularly honey and, by proxy, how prevalent these pesticides abound in the 735 localities of cocoa production. As a crop whose production hinges on the intense application . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 27, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022 of large volumes of pesticides to prevent huge losses, continuous monitoring and strict 737 compliance of pesticide application will ensure the correct use of pesticides, thereby ensuring 738 that pesticide residues are kept below tolerable levels. Residue analysis in honey could serve 739 as a proxy for monitoring the extent to which pesticides are imputed for cocoa cultivation. It is 740 recommended that cocoa producing countries are prioritised for such studies especially in the 741 two leading cocoa producing countries namely Ivory Coast and Ghana, which account for 70% 742 of the world cocoa but where studies were rarely conducted. These studies could form the basis 743 for policy formulation for sustainable and effective beekeeping and pesticide application in 744 cocoa producing countries, especially in cocoa growing landscapes.    CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 27, 2022. ;https://doi.org/10.1101https://doi.org/10. /2022