The article was considered relevant for the meta-analysis study if it met the following inclusion criteria:

The information sources are websites gathering research papers in agriculture, ecology, and related fields. The above-mentioned sources included Web of knowledge, Google Scholar, Scopus, IEEExpore, and the website Zenodo. We searched on these sources to find primarily relevant articles. Books, book chapters, review papers, and notes were not considered in the primary selection. We subscribed to publication notifications in Google Scholar and Academia to receive continuous updates about the articles on the topic of interest. The terms related to the topic were then introduced in the above-mentioned websites as follows: For Cactus, (prickly pear cactus OR Cactus OR Opuntia spp OR Figue de barbarie) AND for the pest (Dactylopius opuntiae OR prickly pear cochineal OR Cactus white cochineal) AND for the control, (management OR control OR eradication). Based on the results from the search, a separate dataset was built from each source prior to analysis with artificial intelligence tools.

The selection process of relevant articles to be included in the meta-analysis was performed in the ASReview software (Van De Schoot et al., 2021) tool after charging its library through python 3.13.2 version. After removing duplicates, a single file combining data from different sources was uploaded in ASReview for machine learning screening. A total of 2,049 articles were processed in the oracle mode. The following keywords were used as prior knowledge: “Dactylopius opuntiae, Cochineal, Control, Prickly pear cactus, Opuntia spp, Resistance”. We carefully checked titles and abstracts and set the training model with 61 papers identified as relevant and 60 as irrelevant labeled records. The active learning model (computer programs that help in the identification of eligible studies) was the TfidfVectorizer (TF-IDF) algorithm, which calculates the product of Term Frequency and Inverse Document Frequency (Lazuardi et al., 2023), while the machine learning model used as a classifier to compute relevant scores was naive Bayes. The screening was stopped after 162 continuous irrelevant papers out of 810 screened (Figure 1).

Once the papers were selected as relevant for the study, the following key information was extracted and arranged in an Excel file: author name, date of publication, the duration of the experimentation, the study location and country, and the control mean used against D. opuntiae. All control means used were classified into four categories, including chemical, biological, genetic, and organic. The data were likewise collected both for nymphs and female adult insects. If the author tested the control method at different levels of insect attacks, the result for the high level was calculated.

The publication bias was qualitatively screened through the funnel plot, a funnel plot of effect sizes plotted against the weight of each. The publication bias was next evaluated through the Egger’s regression test for funnel plot asymmetry using the weighted regression model “lm” (Egger et al., 1997). This was followed by the Duval and Tweedie trim-and-fill method to correct possible bias.

The systematic review yielded 2,638 publications related to the control of D. opuntiae in cactus production. After removing duplicates, 2,049 were selected for quality check in ASReview. The Web of Knowledge produced the highest number of publications (n = 879), followed by Google Scholar (n = 649), Scopus (n = 269), Zenodo (n = 235), and IEEExplore (n = 17). The screening process in the ASReview platform with machine learning algorithms identified 131 relevant publications to be considered for further selection analysis. After a deep check, 54 publications and seven from citations and the website Zenodo fall into the eligibility criteria for a total of 61 papers selected to conduct the meta-analysis. The PRISMA flow diagram delineating the selection process is shown in Figure 2.

The metafor library yielded 221 effect sizes from 43 publications on nymphs and 192 effect sizes from 38 publications on adult insects. The number of effects sizes per paper ranged from one to 20 on nymphs and one to 17 on adult insects. Organic products were the most common control method used, and a large number of these methods were evaluated in laboratory conditions for many days (Figure 3). Regarding the field location, trials were recorded in arid and semi-arid regions (Figure 4), and most of the tests were conducted in Morocco (202) and in Brazil (96).

The meta-regression with the robust variance estimation of the log-transformed D. opuntia mortality yielded an estimate of 4.19 ± 0.08 (95% CI: 4.03–4.35) and 3.99 ± 0.09 (95% CI: 3.81–4.18) for the nymphs and for the adult insects, respectively. The meta-regression robustness was confirmed in sensitivity analysis in both D. opuntiae development stages, where the estimate and resulting standard error were insensitive to different values of the correlation (Figure 5). A high heterogeneity among nymphs (I² = 99.99%, τ² = 0.09) and among adult insects (I² = 99.99%, τ² = 0.05) was observed. Obviously, the two groups of insects were considered in subgroup proportional meta-analysis. Supplementary Tables S1, S2 report the statistics of the RVE with correlated models on nymphs and adult insects, respectively.

The mortality percentage of nymphs pooled in proportional meta-analysis yielded 0.74 (95% CI: 0.71–0.78, p < 0.001) for nymphs and 0.63 (95% CI: 0.6013–0.6739, p < 0.0001) for adult insects. The high heterogeneity was confirmed in nymphs (I² = 98.6%, τ² = 0.06, H = 8.43) and for adult insects (I² = 99.1%, τ² = 0.06, H = 10.74). Organic products showed the most efficient control on nymphs (0.79 ± 0.03), while the genetic approach was the most efficient on adult insects (0.91 ± 0.07). The meta-proportion of the nymph and adult insect mortality variation are reported in Figures 6 and 7, respectively. Between-groups tests for difference showed significance only on adult insects (Q = 18.31, df = 4, p < 0.001). Figures 8 and 9 show how the experiment conditions affect the proportion of nymphs and adult mortality, respectively. Studies conducted in field conditions yielded the highest proportions in both development stages of Dactylopius (0.79 ± 0.03 for nymphs and 0.65 ± 0.03 for adults), but no significant difference was observed among experimental conditions. The trials conducted in Morocco obtained the highest result (0.83 ± 0.02) on nymphs compared to other countries (Figure 10), while trials conducted in Lebanon exhibited the highest proportion on adults (0.84 ± 0.04), as reported in Figure 11. Finally, the longer the test lasted, the greater the proportion of mortality. Evaluation conducted at least during a month yielded the highest proportion of nymphs (0.96 ± 0.02) and adults (0.93 ± 0.03) killed, as displayed in Figures 12, 13. Interestingly, among countries that hosted the experimentation (Q = 35.06, df = 4, p < 0.0001 for nymphs; Q = 31.16, df = 5, p < 0.0001 for adults) and among the durations of these trials (Q = 116.07, df = 5, p < 0.0001 for nymphs; Q = 61.78, df = 5, p < 0.0001 for adults), a significant difference was observed. Accordingly, the experimentation condition was not included in moderator analysis due to the homogeneity observed both in nymphs and adults. Combined effect of moderators yielded significant variation of mortality for nymphs’ mortality (estimate = 3.70 ± 0.43, 95% CI: 2.83–4.58, p < 0.0001). The interaction between nature, country, and the duration of experimentation likewise showed significant variation for adult mortality (estimate = 3.84 ± 0.26, 95% CI: 3.29–4.39, p < 0.0001).

The symmetry of the funnel plot indicates less publication bias, although many publications did not fall into the funnel triangle (Supplementary Figures S1, S2). The regression test for the funnel plot asymmetry result proved the absence of significant bias in the publication included to evaluate the efficacy of the control method against D. opuntiae nymphs (t = -1.8082, df = 219, p = 0.0720) and adult insects (t = -0.6857, df = 190, p = 0.4938). This result was confirmed by the trim-and-fill analysis using restricted (or maximum likelihood (REML) estimator, which showed that zero (0) additional articles were necessary to validate the mortality of nymphs and adults used in the meta-analysis.

After machine learning screening and identification of new studies via other methods, 199 articles passed the eligibility criteria (Figure 2). Most of these studies assessed diverse control methods against the pest, with 221 and 192 results reporting the mortality of D. opuntiae nymphs and adults, respectively. Despite the few single studies gathering many factors studied in this work, a large diversity of control measures against D. opuntiae was demonstrated in the present work. Trends in publication on D. opuntiae were also observed in a previous report, which concluded that this can help to build a collaboration network between scientists and contribute to the development of an efficient control of the threat (Méndez-Gallegos and Bravo-Vinaja, 2022). In this research, Brazil, Morocco, and Mexico are countries with the highest number of research activities on D. opuntiae control, while organic methods are the most commonly used against the pest.

The trend in publication reported that most of the studies are conducted in laboratory conditions, converging with similar observations on other insect families (Bernardes et al., 2022). Nevertheless, field trials yielded the highest D. opuntiae mortality compared to laboratory and greenhouse tests, bringing a promising route to control the pest under real conditions. More interestingly, non-chemical products performed well in both nymphs and adults, assuming a sustainable management of the pest. In fact, much attention has been given to biological and organic control approaches (Vigueras et al., 2007; Flores et al., 2013; Lopes et al., 2018; Idris et al., 2021). The proportional mortality of D. opuntiae showed a significant difference among location and trial duration, where Morocco (0.83 ± 0.02) and Lebanon (0.84 ± 0.04) exhibited the highest performances on nymphs and adults, respectively. The importance of location while choosing a niche for cactus cultivation was demonstrated by Acharya et al. (2019). Many factors, including climate and edaphic components, were identified as key components that can affect cactus growth. When considering that these components considerably vary throughout the environment, the present study result shows the evidence of location influence on the efficacy of pest management, including D. opuntiae control. About 15 species were found associated with D. opuntiae as natural enemies in two different zones of Morocco (El Aalaoui et al., 2019a). This shows that this latter pest does not act alone and should be managed based on the location. Despite the lack of a significant difference among experimental conditions, the proportion of insects killed in field conditions surpassed that observed in a greenhouse. The controlled conditions do not represent the real environment (Cornell et al., 2021), and the conditions in the laboratory and greenhouse may be favorable to the pest against the control method. In addition, the ecosystem type was reported as an important moderator in meta-analyses including environmental conditions in trials (Breza and Grandy, 2025).

In D. opuntiae female insects, the insecticidal effect leads to the destruction of the barrier. However, very few articles included in this work have reported the biological mechanisms specifically attributed to each method used to control the pest. The use of insecticides in general has a toxic effect on insect pests. This toxicity not only reduces reproduction in female insects but can also affect their nervous system and growth. Some organic products, such as vegetable oils, have ovicidal activity and can thus prevent pest oviposition. According to El Aalaoui et al. (2019b), mineral oils reduce the number of eggs as well as the survival rate in Choristoneura rosaceana. Genetic control can likewise promote the development of defense mechanisms in the resistant variety. These mechanisms include the production of biochemical compounds that can hinder the reproduction, development, and survival of the insect on the plant. Work conducted with resistant varieties of Dactylopius coccus concluded that the maintenance of insects in the stage nymph I on the plant was probably due to the production of phyto-ecdysones (terpenoids) by the plant, which hindered molting in the larva (Berhe et al., 2022).

The pooled effect size in meta-analysis is a convincing result, as it contributes to solving the contradiction observed between studies and to confirming or not an assumption in a specific situation based on relevant studies and not on a single research work (Hak et al., 2018). Accordingly, this study found that efforts to control the cochineal insect population of cactus are promising, in general, when we see the robustness of the meta-regression in the present study. Although meta-analysis has become an important source of information for decision-makers, one observed the increase of contradiction in many publications due to the locating, selecting, and combining of studies (Egger et al., 1997b). To keep the reliability of insights provided by meta-analysis, the identification of the risk of bias in included studies is now mandatory in PRISMA and other scientific consortiums (Page et al., 2021a, 2021b). The biases in publications that were included in the present study were not significant, as confirmed by the regression and adjusted test, showing that insights provided in this work are useful for any stakeholder engaged in the control of the D. opuntiae pests. Relative asymmetry observed in this case may be due to many other reasons, and the output from the bias test must be interpreted with some caution as advised by Viechtbauer (2010). Interestingly, the adjusted mortality of nymphs (4.2205 ± 0.03, p < 0.001, 95% CI: 4.15–4.28) and adult insects (4.0337 ± 0.04, p < 0.001, 95% CI: 3.95–4.11) was very close to the estimated ones.

Many studies reported a large range of time to evaluate the efficacy of the product. The highest mortality reported after 1 month may suggest that farmers should apply products for this period instead the single or punctual applications that are usually observed. The proportional meta-regression revealed that, although it was showing that all groups of control methods yielded a good mortality of D. opuntiae (proportion > 0.5), the highest proportions observed for organic methods on nymphs (0.79 ± 0.03) and genetic method on adult insects (0.91 ± 0.07) are promising for the control of this pest without chemical products. This also confirms the importance of promoting resistant material as a sustainable option, as proven in many studies (Vasconcelos et al., 2009; Borges et al., 2013; Falcão et al., 2013).

The control of pests and diseases in conventional agriculture uses chemical products with harmful effects on the environment and biodiversity (Nicolopoulou-Stamati et al., 2016; Von Cossel et al., 2025). Many reports show the alarming situation led by the use of chemical products and the presence of contaminants such as arsenic (As), cadmium, fluorine, and lead reduce the biodiversity (Zhang et al., 2018). In the US, 97% of stream water samples contain pesticides, according to a survey of 51 major rivers (Zhang et al., 2018). The reduction of biodiversity due to chemical products affects the equilibrium created by this biodiversity, which allows crops to be resilient in the face of biotic and abiotic constraints (El Ouali, 2021). The integrated approaches promote the use of environmentally friendly products for the eradication of plant biotic constraints while conserving the equilibrium of the ecosystems. In the present study, proportional subgroup analysis has shown how non-chemical control means, such as organic, genetic, and biological, have demonstrated a strong efficacy against Dactylopius in field conditions across many regions (Vigueras et al., 2007; Flores et al., 2013; Idris et al., 2021; El Aalaoui et al., 2022; Naboulsi et al., 2022). This result is a good signal toward the adoption of these methods, which remain marginal in both conventional and peasant agricultures. Most of the authors reporting these products suggest biological and organic controls as alternatives to chemical products in management programs of D. opuntiae (Gonçalves Diniz et al., 2020; El Finti et al., 2022; Abadi et al., 2024). In addition, cultivars that demonstrated resistance against the pest could be used in breeding programs as parents to introduce other agronomic traits if absent or directly shared with farmers for direct use.