Contents lists available at ScienceDirect
`Crop Protection
`
`journal homepage: www.elsevier.com/locate/cropro
`
`Is tank mixing site-specific premixes and multi-site fungicides effective and
`economic for managing soybean rust? a meta-analysis
`Franklin J. Machado a,1, Jhonatan P. Barro a,1, Cl´audia V. Godoy b, Alfredo R. Dias c,
`Carlos A. Forcelini d, Carlos M. Utiamada e, Edson R. Andrade Jr. f, Fernando C. Juliatti g,
`Jos´e Fernando J. Grigolli h, Hercules D. Campos i, Ivan Pedro Araujo Jr. j,
`Jo˜ao Mauricio Trentini Roy k, Jos´e Nunes Jr. l, Luís Henrique C.P. da Silva m,
`Marcelo G. Canteri n, Marina Senger o, Mˆonica A. Müller j, Mˆonica C. Martins p,
`Mˆonica Paula Debortoli q, Silvˆania H. Furlan r, Tiago Madalosso k, Valtemir J. Carlin s,
`Wilson S. Venˆancio t, Emerson Del Ponte a,*
`a Departamento de Fitopatologia, Universidade Federal de Viçosa, Viçosa, MG, 36570-000, Brazil
`b Embrapa Soja, Londrina, PR, 86001-970, Brazil
`c Fundaç˜ao Chapad˜ao, Chapad˜ao do Sul, MS, 79560-000, Brazil
`d Universidade de Passo Fundo, Passo Fundo, RS, 99052-900, Brazil
`e Tagro, Londrina, PR, 86070-460, Brazil
`f Instituto Mato-Grossense do Algod˜ao, Cuiab´a, MT, 78049-015, Brazil
`g Universidade Federal de Uberlˆandia, Uberlˆandia, MG, 38400-902, Brazil
`h Famiva Pesquisa e Soluç˜oes Agrícolas, Ribeir˜ao Preto, SP, 14026-160, Brazil
`i Universidade de Rio Verde, Rio Verde, GO, 75901-970, Brazil
`j Fundaç˜ao Mato Grosso, Rondon´opolis, MT, 78750-000, Brazil
`k Centro de Pesquisa Agrícola Copacol, Cafelˆandia, PR, 85415-000, Brazil
`l Centro Tecnol´ogico para Pesquisas Agropecu´arias, Goiˆania, GO, 74130-012, Brazil
`m Agro Carregal, Rio Verde, GO, 75907-454, Brazil
`n Universidade Estadual de Londrina, Londrina, PR, 86057-970, Brazil
`o 3M Experimentaç˜ao Agrícola, Ponta Grossa, PR, 84046-060, Brazil
`p Círculo Verde Assessoria Agronˆomica e Pesquisa, Luís Eduardo Magalh˜aes, BA, 47850-000, Brazil
`q Instituto Phytus, Santa Maria, Santa Maria, RS, 97111-970, Brazil
`r Instituto Biol´ogico, Campinas, SP, 13012-970, Brazil
`s Agrodinˆamica, Tangar´a da Serra, MT, 78300-000, Brazil
`t Universidade Estadual de Ponta Grossa, Ponta Grossa, PR, 84010-330, Brazil
`
`A R T I C L E I N F O
`
`Keywords:
`Phakopsora pachyrhizi
`Chemical control
`Management
`Profitability
`
`A B S T R A C T
`Soybean rust (SBR), caused by Phakopsora pachyrhizi, is controlled with sequential applications of commercial
`premixes containing two and, more recently, three site-specific fungicides. However, their efficacy has been
`reduced due to the development of fungicide resistance in the fungal population; hence the use of multi-site
`fungicides in tank mixing has been encouraged. In this work we used data from 45 uniform fungicide trials
`conducted across eight Brazilian states during three crop seasons (2014/15, 2015/16, and 2017/18) to identify
`scenarios when the practice of adding multi-site fungicides in commercial premixes was both technical- and cost-
`effective. Premixes of quinone outside inhibitor (QoI) + demethylation inhibitors (DMI) or succinate dehydro-
`genase inhibitors (SDHI) were applied alone, or tank mixed with multi-site fungicides. Three premixes
`(PICOxystrobin + CYPRoconazole, PYRAclostrobin + FLUXapyroxad and AZOXystrobin + BENZovindiflupyr)
`were tank mixed with one of three multi-site fungicides (+MANCozeb, +COPpeR oxychloride, and
`+ChLORothalonil). The benefits from tank mixing a multi-site tended to be highest for the least effective premix.
`Improvements in control efficacy (C, percent point; p.p.) and yield response (D, kg/ha) ranged from 3 to 15 p.p.
`
`* Corresponding author.
`E-mail address: delponte@ufv.br (E. Del Ponte).
`1 These authors contributed equally.
`https://doi.org/10.1016/j.cropro.2021.105839
`Received 16 June 2021; Received in revised form 4 October 2021; Accepted 6 October 2021
`
`Crop Protection 151 (2022) 105839
`
`Available online 7 October 2021
`0261-2194/© 2021 Published by Elsevier Ltd.
`
`SYNGENTA EXHIBIT 1030
`Syngenta v. UPL, PGR2023-00017
`
`

`

`F.J. Machado et al.
`
`and 58–240 kg/ha, respectively. The improvements in C and D were affected by severity in the non-treated
`check; significantly higher improvements in D were determined in trials experiencing high SBR severity levels
`(>80%). The economic analysis for scenarios of soybean price and multi-site costs showed that the addition of
`+MANC, given its lower price, was more likely to be profitable compared with +CLOR and +COPR, particularly
`when tank mixed with the least effective commercial premix.
`
`1. Introduction
`Soybean rust (SBR) is one of the most damaging fungal diseases of
`soybean (Glycine max (L.) Merr.) caused by the obligate biotrophic
`pathogen Phakopsora pachyrhizi Syd. & P. Syd. (Goellner et al., 2010; Li
`et al., 2010). In Brazil, SBR is present in all soybean growing areas and
`where severe epidemics may develop and lead to around 80% yield
`losses under disease-conducive conditions (Dalla Lana et al., 2015). A
`couple years after the first reports of SBR epidemics in Brazil, a
`mandatory soybean-free period was implemented to reduce early-season
`inoculum. The wide adoption of early-maturing cultivars and earlier
`sowing dates, more recently, have greatly helped to reduce the impact of
`the disease (Godoy et al., 2016a). None of the current commercially
`grown soybean cultivars are fully resistant to all P. pachyrhizi races
`(Childs et al., 2018; Hartman et al., 2005) and hence soybean farmers
`should rely on routine fungicide applications for managing SBR (Beruski
`et al., 2020; Dalla Lana et al., 2018).
`Several options of active ingredients have been made available
`during almost twenty years of disease management in Brazil using
`fungicides (Godoy et al., 2016a). Updates in fungicide programs have
`been frequently made based on information generated by field research
`to evaluate the performance of currently used and newly available
`products in the market. For instance, a few years after the sole use of
`demethylation inhibitor (DMI) fungicides, the premixes of DMIs with
`quinone outside inhibitors (QoI) gained attention up until the first ten
`years of chemical control of soybean rust. Following 2013, commercial
`premixes of succinate dehydrogenase inhibitors (SDHIs) and QoIs have
`also become available for SBR control. Recently, SDHIs have also been
`included in triple mixtures with QoIs and DMIs (Godoy et al., 2016a).
`However, a decline in control efficacy of soybean rust has been observed
`in the cooperative trials over the past years (Dalla Lana et al., 2018).
`First the DMI alone, then DMI + QoI mixtures and, lately, QoI + SDHI
`fungicides are no longer as effective as they were during the first years of
`use (Barro et al., 2021). Such decline was not only reported for single-a.i.
`fungicides but also for premixes after at least four years (Barro et al.,
`2021; Dalla Lana et al., 2018). Declines in efficacy have been linked to
`reports of less sensitive populations of Phakopsora pachyrhizi to DMI,
`QoI, and SDHI fungicides (Klosowski et al., 2016; Schmitz et al., 2013;
`Sim˜oes et al., 2018).
`FRAC (Fungicide Resistance Action Committee) published a state-
`ment about the increasing importance of multi-sites for managing
`fungicide resistance, including soybean rust (FRAC 2018). Since 2015,
`multi-site fungicides, e.g. chlorothalonil, copper oxychloride, and
`mancozeb, have been recently evaluated in tank mix with commercial
`premixes to improve SBR control efficacy and reduce the risk of resis-
`tance (Godoy et al. 2015, 2016b, 2017, 2018). Chlorothalonil, a
`multi-site fungicide that belongs to the chloronitriles group, was first
`registered in 1966 and is still in use as a protectant fungicide (Battaglin
`et al., 2011; Miles et al., 2007). Copper oxychloride (3Cu (OH)2.CuCl2) is
`another multi-site fungicide that has been recently evaluated against
`soybean rust (Chechi et al., 2020; Juliatti et al., 2017). Mancozeb, which
`belongs to the dithiocarbamate group (FRAC 2018), was introduced in
`1962 and still has a significant importance in the fungicide market
`worldwide (Gullino et al., 2010; Thind and Hollomon, 2018). As a
`typical cost-effective multi-site protectant-only fungicide, mancozeb
`requires relatively high rate and frequency of application compared to
`modern fungicides to achieve satisfactory protection (Gullino et al.,
`2010; Thind and Hollomon, 2018).
`
`In Brazil, fungicides have been evaluated annually since 2003/2004
`in a network of uniform field trials (UFTs) conducted by the anti-rust
`consortium (CAF) with the goal of monitoring and comparing the
`effectiveness of a range of fungicides (Dalla Lana et al., 2018; Godoy
`et al., 2016a). Considerable variations in control efficacy have been
`observed between fungicide chemistries over the years and across re-
`gions. Meta-analysis, a statistical technique that combines results from
`previous studies selected by defined criteria, have become standard for
`summarizing and comparing fungicide performance in plant disease
`management (Barro et al. 2019, 2020; Edwards-Molina et al., 2019;
`Machado et al., 2017; Paul et al. 2008, 2018) including SBR (Barro et al.,
`2021; Dalla Lana et al., 2018; Delaney et al., 2018). In the present study,
`we used meta-analysis to: 1) estimate the improvement in both efficacy
`and yield return provided by tank mixing multi-site fungicides with
`premixes of QoI plus DMI or SDHI fungicides evaluated from 2014/15 to
`2017/18 across all major soybean regions in Brazil; 2) evaluate whether
`the gains in the responses vary across regions; and 3) estimate the
`probability of breaking-even on costs for a range of scenarios of soybean
`prices and multi-site fungicide costs.
`2. Material and methods
`2.1. Data source and experimental procedures
`The data used in this study were obtained from 45 UFTs, which have
`been published as yearly reports (Godoy et al. 2015, 2016b, 2017,
`2018). The experimental design was a randomized complete block with
`four replications. Replicated plots were at least six rows wide and 5 m
`long. Three fungicide sprays were applied at label-recommended rates
`across crop seasons using a CO2 backpack sprayer with a volume of 120
`L ha(cid:0) 1. Sprays initiated at around 50 days post plant emergence (before
`canopy closure) with subsequent applications approximately 15 days
`apart. Percent disease severity, which takes both percent area affected
`and defoliation into account (Dalla Lana et al. 2015, 2018), was assessed
`at the full-seed growth stage (R6) across four different points of each plot
`where 10 leaves were examined. The central 5 m2 of each plot were
`harvested at the R8 stage (full maturity) and mechanically threshed.
`Soybean yields were calculated as kg/ha at 13% grain moisture. All
`weed and insect control practices followed regional recommendations.
`
`2.2. Criteria for study inclusion and fungicide selection
`To be included in the analysis, a soybean rust fungicide trial should
`have: 1) a non-treated plot; 2) a commercial premix applied alone or
`tank mixed with a multi-site in the same trial, and 3) a measure of dis-
`ease severity at R6 stage (Dalla Lana et al., 2015) and/or grain yield in
`both fungicide-treated and non-treated plots. Among several fungicides
`that matched the criteria, we performed a preliminary exploratory
`analysis and selected three different premixes that represented a low
`(picoxystrobin + cyproconazole [PICO + CYPR]), intermediate (pyr-
`aclostrobin + fluxapyroxad [PYRA + FLUX]) and high (azoxystrobin +
`benzovindiflupyr [AZOX + BENZ]) SBR control efficacy to compare the
`sole application of the premix with its tank mixing with one the three
`multi-site fungicides: chlorothalonil (+CLOR), copper oxychloride
`(+COPR) and mancozeb (+MANC) (Table 1). The concentration of
`active ingredient (a.i.) for + CLOR and +COPR was 1080 g a.i./ha and
`294 g a.i./ha, respectively, in all trials (n = 45), while for + MANC, the
`concentration varied from 1125 (n = 24) to 1500 (n = 21) g a.i./ha. The
`
`Crop Protection 151 (2022) 105839
`
`2
`
`

`

`F.J. Machado et al.
`or non-transformed yield means (D) of treatments was estimated from
`the mean square error (MSE), as described previously (Paul et al., 2008,
`2010). Given the statistical properties of the data (Fig. S1), mean SBR
`severity values were log-transformed, while no transformation was
`required for the mean difference in yield.
`
`Table 1
`Fungicide treatments applied for controlling soybean rust, evaluated in 45
`uniform field trials conducted during three growing seasons (2014/15, 2015/16
`and 2017/18) in 25 municipalities across eight Brazilian states by the Anti-rust
`Consortium (CAF).
`Fungicide
`Active ingredient
`code
`[a.i.]
`
`Trade name
`
`FRAC
`codea
`
`Rate (g
`a. i.
`ha(cid:0) 1)
`
`∑
`
`2.4. Network meta-analytic model
`A network model, also called a two-way unconditional linear mixed
`model, was fitted directly to the treatment means (log-transformed or
`non-transformed) (Machado et al., 2017; Madden et al., 2016; Paul
`et al., 2008). The model can be written as:
`(1)
`​ + ​ Si) ​
`Yi
`​ ∼ ​ N ​ (μ, ​
`where Yi is the vector of L (log of the mean SBR severity) or mean yield
`(D) for the 12 treatments plus the non-treated check for the ith study, μ is
`a vector representing mean of Yi across all studies, Σ is a 13 × 13
`between-study variance-covariance matrix (for the 12 treatments plus
`the non-treated check), and Si is within-study variance-covariance ma-
`trix for the ith study. N indicates a multivariate normal distribution. An
`unstructured Σ matrix was used and the models were fitted to the data
`with a maximum-likelihood parameter. All models were fitted to the
`data using the rma.mv function of the metafor package (Viechtbauer,
`2010) of R (R Core Team, 2019).
`The overall mean log of the response ratio (LSEV) for each fungicide
`treatment (̂μTreat) relative to the non-treated check (̂μCheck) was used as
`effect size in the model and estimated as LSEV = ̂μTreat - ̂μCheck. Overall
`mean percent SBR control and the respective 95% confidence intervals
`(CIs) were obtained by back-transforming LSEV and the upper and lower
`limits of the confidence intervals around LSEV (equation (2)).
`(2)
`C = ​ (1 ​
`​ exp(LSEV)) × 100
`The mean difference in yield (D) was calculated by subtracting the
`mean yield in the non-treated check from the mean yield in the fungicide
`treatment (Madden et al., 2016). To test for network inconsistency, we
`fitted a factorial-type linear model to determine the significance of the
`treatment x design interaction, evaluated based on the Wald test statistic.
`The null hypothesis suggests that the network is consistent (Higgins
`et al., 2012; Madden et al., 2016; Piepho, 2014). Three and five different
`designs (here design refers to the set of treatments in the trial) were
`found in the trials reporting soybean yield and SBR severity, respectively
`
`(cid:0)
`
`Dose
`(g or
`ml
`ha(cid:0) 1)
`300
`
`PICO þ
`CYPR
`PYRA þ
`FLUX
`AZOX þ
`BENZ
`þCLOR
`þCOPR
`þMANC
`
`Picoxystrobin +
`Cyproconazole
`Pyraclostrobin +
`Fluxapyroxad
`Azoxystrobin +
`Benzovindiflupyr
`Chlorothalonil
`Copper
`oxychloride
`Mancozeb
`
`11 + 3
`
`60 + 24
`
`11 + 7
`11 + 7
`M05
`M01
`M03
`
`Aproach
`Prima
`(DuPont)
`350
`116.55
`Orkestra
`+ 58.45
`(BASF)
`200
`60 + 30
`Elatus
`(Syngenta)
`1500
`1080
`Previnil
`(Helm)
`500
`294
`Difere
`(Oxiquímica)
`1500;
`1125;
`Unizeb Gold
`2000
`1500
`(UPL)
`a FRAC = Fungicide Resistance Action Committee; 11 = QoI, quinone-outside
`inhibitors; 7 = SDHI, succinate dehydrogenase inhibitors; 3 = DMI, demethy-
`lation inhibitors; M = multi-site contact activity.
`
`field experiments were conducted during three growing seasons
`(2014/15, 2015/16 and 2017/18) at 25 locations across eight Brazilian
`states (Bahia [BA], Goi´as [GO], Minas Gerais [MG], Mato Grosso do Sul
`[MS], Mato Grosso [MT], S˜ao Paulo [SP], Paran´a [PR], and Rio Grande
`do Sul [RS]) (Fig. 1). The trials were geographically grouped into the
`Northwestern (NW) region (n = 30, MT, MS, GO, BA, and MG states) and
`the Southeastern (SE) region (n = 15, PR, RS and SP states) (Fig. 1). MT
`(n = 12), PR (n = 10), MS (n = 7), and GO (n = 7) were the states with
`the largest number of trials, accounting for 80% of all trials.
`
`2.3. Response variables
`Two response variables were of interest in this study, similarly to a
`previous meta-analysis (Dalla Lana et al., 2018): control efficacy (C),
`and yield difference (D), both relative to a reference treatment (non--
`treated check) for the sole premix treatment or tank mixed with a
`multi-site fungicide (Dalla Lana et al., 2018; Madden and Paul, 2011;
`Paul et al., 2008). Given the availability of data at the plot level, the
`within-study or sampling variance (V) of the log mean SBR severity (L)
`
`Fig. 1. Geolocation of the 25 municipalities across eight Brazilian states where 45 fungicide evaluation trials were conducted during three growing seasons (2014/
`15, 2015/16 and 2017/18). The size of the circle is proportional to disease severity in the non-treated CHECK in each location. The color of the circle represents two
`geographic regions defined in our study: Northwestern (NW) states (MT, MS, GO, BA, MG) and Southeastern (SE) states (PR, RS, SP). (For interpretation of the
`references to color in this figure legend, the reader is referred to the Web version of this article.)
`
`Crop Protection 151 (2022) 105839
`
`3
`
`

`

`F.J. Machado et al.
`(Table S1).
`
`P = Φ[DI (cid:0) (FC / SP)
`
`/ ̅̅̅
`√
`̂τ
`]
`
`(3)
`
`2.5. Effect of moderator variable
`The model was expanded to include categorical moderator variables
`that could explain, at least in part, the heterogeneity of the effects across
`trials (Madden et al., 2016). The first categorical variable was the
`baseline for SBR disease intensity based on the median in the
`non-treated check. The trials were divided into two groups representing
`low (<80% SBR severity) (n = 20) and high (≥80% SBR severity) (n =
`25) disease scenarios. The second categorical variable was based on the
`geographical region where trials were grouped into the Northwestern
`(NW) region (MT, MS, GO, BA, MG) (n = 30) and the Southeastern (SE)
`region (PR, RS, SP) (n = 15) as mentioned previously. Linear contrasts
`were used to estimate the mean effect sizes and their standard errors and
`95% CIs for each level of the categorical moderator (Madden et al.,
`2016).
`
`2.6. Probability of breaking-even fungicide application cost
`The probability (P) of breaking even on the multi-site costs (FC) was
`calculated using the estimates of the mean improvement in yield from
`adding the multi-site fungicide (DI = Dfungicide + multi - Dfungicide alone), and
`the respective between-study variance (̂τ) obtained from meta-analysis.
`This was calculated as the cumulative standard-normal distribution
`function of
`
`where Φ is the cumulative standard-normal function and SP is the soy-
`bean price.
`Average prices of the multi-site fungicides considering: an exchange
`rate of $ 5.3 BRL = 1 U$ during May 2021; fungicide price of 2019/20
`crop season; and three applications were: +CLOR: 36.00 U$/ha,
`+COPR: 27.00 U$/ha and +MANC: 24.00 U$/ha. To obtain the national
`average soybean price, we gathered data from the AGROLINK database
`(AGROLINK, 2020) since 2014. The average soybean price used was 250
`U$/ton. Tile plots of the probability classes of breaking even on
`multi-site costs were produced for each premix.
`3. Results
`3.1. SBR severity and yield
`SBR severity varied across treatments and trials. In the non-treated
`plots, it ranged from 40.3 to 100% and was higher than 98% in one-
`quarter of the studies. As expected, SBR severity was higher in the
`non-treated check (median 80%) compared with premixes alone or with
`the multi-site fungicides (Fig. 2A). Soybean yield in the non-treated
`check ranged from 904 to 3760 kg/ha. In general, mean yield was
`lower in the non-treated check plots (median 2364 kg/ha) than in the
`plots sprayed with either premixes alone or tank mixed with a multi-site
`fungicide, but there was also a considerable variation among treatments
`and studies. Yield in the non-treated check was higher than 2684 kg/ha
`
`Fig. 2. Box plots and individual means of (A) soybean rust severity at R6 stage (%) and (B) soybean grain yield (kg/ha) across 45 uniform field trials conducted
`during three growing seasons (2014/15, 2015/16 and 2017/18) across 25 municipalities in Brazil. The treatments evaluated consisted of a non-treated (CHECK) or
`treated plot with 3 sprays of picoxystrobin + cyproconazole (PICO + CYPR), pyraclostrobin + fluxapyroxad (PYRA + FLUX) and azoxystrobin + benzovindiflupyr
`(AZOX + BENZ), applied alone (without) or amended with multi-site fungicides (chlorothalonil [+CLOR], copper oxychloride [+COPR] and mancozeb [+MANC]).
`The line within the box represents the median, whereas the top and bottom lines of the boxes represent the 75th and 25th percentiles of the data, respectively. The
`circles represent each treatment mean.
`
`Crop Protection 151 (2022) 105839
`
`4
`
`

`

`F.J. Machado et al.
`in one-quarter of all studies (Fig. 2B).
`were very low (~3 p.p.), but in D were the highest for + MANC (190 kg/
`ha) and the lowest for + COPR (58 kg/ha) (Fig. 3).
`
`3.2. Overall improvement in control and yield using the tank mix
`The percent control efficacy (C) obtained from the back-
`transformation of the estimated differences of the log-transformed SBR
`severity in the fungicide-treated without multi-site and non-treated plots
`ranged from 48 to 78% (Table 2). Linear contrasts showed that there was
`a significant increase (P < 0.05) in C when adding any of the multi-site
`fungicides for all the three premixes evaluated. Among the multi-sites,
`efficacy for PICO + CYPR was higher when the premix was amended
`with +MANC (63.8%) or + CLOR (67.6%) which did not differ among
`them (P = 0.2668), but both differed (P < 0.02) from +COPR (59.8%).
`However, there was no statistical difference (P > 0.3) when AZOX +
`BENZ was amended with +MANC (82.1%), +CLOR (82.4%) or + COPR
`(81.6%). Finally, for PYRA + FLUX, higher C was obtained when
`amended with +MANC (79.3%), which differed significantly (P < 0.01)
`from +CLOR (75.1%) and +COPR (72.3%) (Table 2). The Wald test for
`the treatment x design interaction showed that the network was
`consistent (P = 0.9987).
`The mean estimates of yield difference (D) between fungicide-treated
`without multi-site and non-treated plots ranged from 463 to 926 kg/ha
`(Table 3). Similar to control efficacy, linear contrasts showed that D was
`significantly higher (P < 0.05) when mixing any of the multi-site fun-
`gicides to all three premixes evaluated. For PICO + CYPR, D was higher
`when the fungicide was mixed with +MANC (691.6 kg/ha), which was
`different (P = 0.0042) from +COPR (624.3 kg/ha) but not different (P =
`0.7328) from +CLOR (681.8 kg/ha). Similarly, D was higher when
`AZOX + BEN was mixed with +MANC (1110 kg/ha), which was
`different (P < 0.02) from +CLOR (1021 kg/ha) and +COPR (978 kg/ha),
`the latter two not differing among them (P = 0.1172). However, there
`was no statistical difference in D between multi-sites for PYRA + FLUX
`(P > 0.1), and yield difference between the premix applied alone and
`with multi-sites ranged from 105 to 167 kg/ha (Table 3). The Wald test
`for the treatment x design interaction showed that the network was
`consistent (P = 0.8133).
`In general, the pattern of the relationship between the improvement
`in control efficacy and yield from using multi-sites compared with the
`premix applied alone was consistent. For PICO + CYPR, improvements
`in C (percent point; p.p.) and D (kg/ha) were higher for + MANC (15 p.
`p.; 240 kg/ha) and lower for + COPR (11 p.p.; 158 kg/ha). Similarly, for
`PYRA + FLUX, +MANC resulted in the highest improvement in C (12p.
`p.) and D (202 kg/ha). Finally, for AZOX + BENZ, improvements in C
`
`3.3. Effect of moderator variable
`The categorical moderator variable region did not affect control ef-
`ficacy (C) or yield (D) estimates (P > 0.05). However, baseline SBR
`disease intensity significantly affected C and D (P < 0.0001). We found
`that C was generally lower in scenarios with high disease pressure
`(≥80%). The differences in C values between low- and high-disease
`scenarios ranged from 1 to 20.7 percent points among the fungicide
`treatments. When this difference was above 10 percent points, disease
`baseline intensity significantly affected C, as for PICO + CYPR (20.7 p.p.;
`P < 0.0001), PICO + CYPR + MANC (14.1 p.p.; P = 0.0062), PICO +
`CYPR + COPR (12.5 p.p.; P = 0.0104) and PICO + CYPR + CLOR (10.6
`p.p.; P = 0.0359) (Fig. 4; Table S2). In contrast, yield response from the
`use of fungicides was generally greater in the high-disease than in the
`low-disease scenarios, with differences ranging from 83 to 486 kg/ha.
`Differences between low- and high-disease scenarios above 200 kg/ha
`significantly affected D (Fig. 4; Table S2).
`
`3.4. Profitability of multi-site addition to the premix
`The improvement in yield (DI) from adding the multi-site fungicide
`and between-study variance (̂τ) obtained from the meta-analysis were
`used to calculate the probability (P) of breaking even on multi-site costs.
`In general, P was affected by a range of scenarios and the variation in the
`multi-site cost and soybean prices. In most scenarios, probabilities were
`between 45 and 55%, mainly when the premixes were amended with
`+CLOR and +COPR (Fig. 5). A higher number of favorable scenarios (P
`>55%), were identified combining the premixes with +MANC (the less
`expensive multi-site). However, probabilities <45% were observed in
`scenarios of high multi-site costs and low soybean price, especially when
`the premix AZOX + BENZ was combined with +CLOR and +COPR
`(Fig. 5).
`4. Discussion
`This study summarizes the benefits of tank-mixing a few selected
`premixes including QoI + DMI or SDHI fungicide with three different
`multi-site fungicides evaluated during three growing seasons (2014/15,
`2015/16, and 2017/18) across a range of environments in eight Bra-
`zilian states. Most importantly, we found that the benefits tended to be
`
`Table 2
`Overall means and respective confidence intervals of log response ratio (LSev) and calculated percent control (C) of Soybean Rust (SBR) relative to non-treated check
`provided by three fungicide treatments applied alone or amended with multi-site fungicides in 45 uniform field trials conducted during three growing seasons (2014/
`15, 2015/16 and 2017/18) in 25 municipalities across eight Brazilian states by the Anti-rust Consortium (CAF).
`Premixesa
`Multi-sitea
`kb
`Effect Size
`SE(LSev)
`LSev
`0.0604
`(cid:0) 0.6586
`28
`–
`0.0751
`(cid:0) 0.9828
`28
`+CLOR
`0.0686
`(cid:0) 0.9122
`28
`+COPR
`0.0837
`(cid:0) 1.0171
`28
`+MANC
`0.0903
`(cid:0) 1.1273
`24
`–
`0.1061
`(cid:0) 1.3900
`23
`+CLOR
`0.0916
`(cid:0) 1.2831
`24
`+COPR
`0.1330
`(cid:0) 1.5741
`24
`+MANC
`0.1125
`(cid:0) 1.5590
`28
`–
`0.1300
`(cid:0) 1.7410
`28
`+CLOR
`0.1211
`(cid:0) 1.6952
`27
`+COPR
`0.1188
`(cid:0) 1.7239
`27
`+MANC
`a See Table 1 for complete information of the evaluated fungicides.
`b Number of trials that each fungicide was evaluated.
`c Upper (CIU) and lower (CIL) limits of the 95% confidence interval around LSev and C.
`
`SBR control (%)
`CILc
`C
`48.24
`41.74
`62.57
`56.64
`59.84
`54.06
`63.84
`57.39
`67.61
`61.33
`75.09
`69.33
`72.28
`66.83
`79.28
`73.11
`78.97
`73.78
`82.47
`77.38
`81.64
`76.73
`82.16
`77.49
`
`CIUc
`54.02
`67.70
`64.89
`69.31
`72.87
`79.77
`76.84
`84.03
`83.13
`86.41
`85.52
`85.87
`
`PICO + CYPR
`
`PYRA + FLUX
`
`AZOX + BENZ
`
`CILc
`(cid:0) 0.7769
`(cid:0) 1.1300
`(cid:0) 1.0467
`(cid:0) 1.1811
`(cid:0) 1.3044
`(cid:0) 1.5981
`(cid:0) 1.4626
`(cid:0) 1.8347
`(cid:0) 1.7796
`(cid:0) 1.9958
`(cid:0) 1.9324
`(cid:0) 1.9567
`
`CIUc
`(cid:0) 0.5403
`(cid:0) 0.8356
`(cid:0) 0.7777
`(cid:0) 0.8531
`(cid:0) 0.9502
`(cid:0) 1.1820
`(cid:0) 1.1036
`(cid:0) 1.3134
`(cid:0) 1.3385
`(cid:0) 1.4863
`(cid:0) 1.4579
`(cid:0) 1.4911
`
`P value
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`
`Crop Protection 151 (2022) 105839
`
`5
`
`

`

`F.J. Machado et al.
`
`Table 3
`Overall means and respective confidence intervals of difference in soybean yield (D) between fungicide-treated and non-treated plots for the effect of three fungicide
`treatments applied alone or amended with multi-site fungicides on soybean rust severity in 45 uniform field trials conducted during three growing seasons (2014/15,
`2015/16 and 2017/18) in 25 municipalities across eight Brazilian states by the Anti-rust Consortium (CAF).
`Premixesa
`Multi-sitea
`kb
`Yield response (kg/ha)
`D
`564.52
`28
`–
`796.77
`28
`+CLOR
`722.58
`28
`+COPR
`804.80
`28
`+MANC
`803.49
`24
`–
`926.46
`24
`+CLOR
`946.52
`24
`+COPR
`1005.39
`24
`+MANC
`1069.95
`28
`–
`1176.46
`28
`+CLOR
`1128.38
`28
`+COPR
`1260.88
`28
`+MANC
`a See Table 1 for complete information of the evaluated fungicides.
`b Number of trials that each fungicide was evaluated.
`c Upper (CIU) and lower (CIL) limits of the 95% confidence interval around D.
`
`PICO + CYPR
`
`PYRA + FLUX
`
`AZOX + BENZ
`
`SE(D)
`51.54
`58.65
`50.11
`57.72
`57.96
`67.51
`76.91
`75.74
`73.22
`79.22
`76.39
`76.96
`
`CILc
`463.49
`681.80
`624.36
`691.66
`689.89
`794.14
`795.78
`856.93
`926.43
`1021.18
`978.65
`1110.02
`
`CIUc
`665.55
`911.74
`820.79
`917.93
`917.10
`1058.78
`1097.27
`1153.86
`1213.47
`1331.75
`1278.12
`1411.73
`
`P value
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`<0.0001
`
`highest for the least effective dual premix PICO + CYPR, and the use of
`+MANC, given its lower price in the scenario evaluated, was more likely
`to be profitable compared with the other multi-sites. Finally, we found
`generally lower levels of SBR control and greater levels of yield response
`in trials with conditions conducive for severe epidemics.
`The poor performance of the application of PICO + CYPR (48%)
`alone reported herein confirms the trend of significant decline in control
`efficacy determined from a previous meta-analysis (Dalla Lana et al.,
`2018). The previous study found a much higher efficacy (80.2%) for the
`same premix after summarizing the performance of fungicides using
`data collected from 250 UFTs conducted in a 10-year period (2004/05 to
`2013/14) (Dalla Lana et al., 2018). Additionally, our previous
`meta-analysis (Barro et al., 2021), which summarized data from 177
`UFTs conducted during six growing seasons (2014/15–2019/20) (data
`not included in our analysis) reported a significant decline in perfor-
`mance over time for the premix AZOX + BENZ, which was one of the
`most effective fungicides in our analysis (up to 2017/18 season), even
`when not tank mixed with a multi-site (78%).
`Fungicide resistance in P. pachyrhizi populations in Brazil has been
`linked to declines in control efficacy over time (Barro et al., 2021; Dalla
`Lana et al., 2018). Indeed, the continuing use of QoIs, DMIs, and SDHIs
`in the past crop seasons has resulted in the selection of isolates with
`reduced sensitivity to those distinct chemistries (Klosowski et al. 2016;
`Sim˜oes et al. 2018; Schmitz et al. 2013). Multi-site fungicides do not act
`systemically, as they are only surface protectants that affect different
`metabolic pathways of the fungus and have low resistance risk, since
`several mutations in target genes would be required for resistance to
`emerge (Thind and Hollomon, 2018). Hence, the use of premixes
`including sing