Effectiveness and safety of reactive focal mass drug administration (rfMDA) using dihydroartemisinin-piperaquine to reduce malaria transmission in very low-endemic setting of Eswatini: a pragmatic cluster randomised controlled trial

59 60

61 To reduce malari a tr ansmission in very low-endemic settings, scre ening and t rea t ment nea r inde x cases 62 (reactive case de tec tion (RACD)), is widely practiced, but the r apid diagnos tic test s (RDTs ) u sed miss low -63 density infections . Presumptive t rea tme nt near ind ex cas es (reactive focal mass drug administra tion 64 (rfMDA)) may be safe and more effective . 65 66

67 We conduct ed a cluste r-randomis ed con trolle d trial in Eswatini , a very low-ende mic setting. 77 cluste rs 68 were rand omised to rfMDA using dihydr oart emisin-pipera quine (DP) or RACD inv olving RDTs and 69 artem eth er lumefan trine (AL). Int erven ti ons were deliver ed by the loc al program me. An int enti on-to-70 trea t analysis was used to compa re clust er-level cumulative confirme d malaria in cidence among cluste rs 71 with cases. Second ary outcomes includ e d safety and adhe rence . 72 73

74 From Sept 2015– Aug 2017, 220 index cas es from 47 clusters trigger ed 49 RACD events and 68 rfMDA 75 events. R ACD and rfMDA were deliver ed to 1696 and 1932 individuals, resp ectivel y. Index cas e and 76 targe t popula tion in terven tion cover ages for both arms were 75 .6%–81.4% and a dherenc e to DP was 77 98.7%. For rfMDA versus RACD, cumulati ve incidences (per 1000 person-yea rs) of all malaria wer e 2.11 78 (95% CI 1.73–2.59) and 1.97 (1.57–2.47), respectively; and of locally acquired mala ria, th ey were 1.29 79 (95% CI 1.00–1.67) and 0.97 (0.71–1.34), respectively . Adjusting for imbalance in b aseline incide nce, 80 incidence ra te ra tio (aIR R) for rfMDA versus RACD was 0.93 (9 5% CI 0.54–1.60) fo r all malaria and 0 .77 81 (95% CI 0.38–1.56) for locally acquired malaria . No seri ous adverse events occur re d. 82 83

84 In a very low-endemic, rea l-world set ting, this trial is t he first t o evalua te rfMDA u sing DP. rfMDA was 85 safe and resul ted in lower cumula tive inc idence compar ed to R ACD, but we were unable t o confirm its 86 effectiveness, po ten tially due to insuffici ent power . To assess impact of int erventi ons in very low-87 endemic set tings, multi-site , adap tive tri als and use of complement ary inte rventi ons may be needed . 88 89

90 Plasmodium, Swaziland, low transmissio n, malaria elimin ation , active case d etec t ion, reac tive case 91 detec tion, an timala rial, dihydro ar temisin -piperaquin e, cluste r rand omized cont rol led trial 92 93 94 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 4

95 Since 2000, many countries have scal ed up effective malaria con trol in terven tion s, resulting in 96 reductio ns in malaria bur den and a rene wed goal to er adicat e malaria worl dwide by 2050.(1) When the 97 goal is to inte rrup t tra nsmission, it may be necessary to tre at no t only symptomati c malaria but also 98 asymptomatic infections which pe rpe tua te ongoing tr ansmission and re pres ent a n increasing propo rtio n 99 of all infections in low transmission se tti ngs.(2, 3) 100 To address asymptoma tic infections, on e widely practiced str ategy is active case d etec tion in 101 household memb ers and neighb ours of symptomatic cases rec ently rep ort ed fro m health faciliti es, also 102 known as reactive case d etec tion (RACD).(4) Since malaria infections clust er in sp ace and time ,(3) RACD 103 can targe t limited r esourc es to ar eas a t highest risk of infection. I n set tings with substanti al import ed 104 malaria cases th at may seed loc al transm ission, RACD also serves as a focal outbr e ak response .(4) 105 However, th e effectiveness of RACD is limited by the low sensit ivity of currently available poin t-of-care 106 diagnostics to de tec t low-density and no n-falciparum infections. Mol ecular testi n g such as polymerase 107 chain reacti on (PCR) or loop-mediated is otherm al amplification (LAMP) improves sensitivity but is not 108 practical given costs, logistic al challenges of specimen collection and transpo rt , an d turn-arou nd time 109 requir ed for labor ato ry testi ng and re tur n visits to tre at t est-posi tive individuals. I n addition, mass 110 screening and t rea tmen t, which is similar to RACD but deliver ed community-wide, has not sustain ably 111 reduced incid ence in prio r studies .(5) As such, the Wo rld Heal th O rganiza tion (W HO) does not 112 recommend RACD as a stra tegy to r educ e or int errup t tr ansmission. 113 Mass drug administra tion (MDA), or th e trea tmen t all individuals within a specifie d area with an 114 effective antimala rial ir respec tive of infection sta tus,(6, 7) may address some of the challenges of RACD. 115 MDA was a component of many malaria elimination p rogrammes in th e mid-twen tieth c entury bu t fell 116 out of favor due to conce rns rega rding it s effectiveness, sustain ability, cost , and fear of accele rating 117 drug resistanc e. Mo re r ecent evidence s uggests that whe n implement ed in ar eas of low endemicity and 118 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 5 in combination with o the r inte rventio ns, MDA has the po ten tial to sus tainably int errup t tr ansmission.(6, 119 7) Maximizing coverage and adh erenc e may also help to mitiga te risks of drug re sistance.(8) MDA has 120 recen tly been r ecommended by th e WH O in areas a pproaching in ter rupti on of tr ansmission where th ere 121 is good access to tre atme nt, effective im plement ation of vecto r contr ol and surve illance, and minimal 122 risk of re-introduc tion of infection .(9) However, a dea rth of definitive evidence on its effectiveness, 123 safety, and feasibili ty remains.(10) 124 Eswatini (formerly Swaziland) is among 21 countries worldwide tha t were id entifi ed by WHO as 125 the most likely to reach ze ro indigen ous cases by 2020.(11) However, several of these count ries 126 including Eswatini continue to ex peri enc e persist ent local transmission an d resur gence. As a mala ria 127 elimination-sp ecific strat egy, the Eswatin i Nation al Mala ria Programme (NMP) has implemented R ACD 128 since 2009. Prior studies have confirm ed that asympt omatic infectio ns cluster a ro und passively detect ed 129 index cases, wit h the high est risk within 200 meters of the in dex cas e.(12) However, in Eswatini RACD 130 using RDTs missed two-thirds of infections and 40% of hotspots compared to mo re sensitive mol ecular 131 methods.(12) Due to logistical chall enges , att empts to us e more sensi tive molecul ar methods to dir ectly 132 inform trea tmen t have be en unsuccessful (N. Dlamini, personal communica tion). 133 Reactive focal MDA (rfMDA) is an altern a tive inte rventio n tha t builds on RACD for targe ting 134 high-risk populations r esiding near ind ex cases. rfMDA ent ails mass drug administr ation withou t t esting 135 in household membe rs and neighbo urs o f recent inde x cases.(13) A recen t tr ial of rfMDA using 136 artem eth er-lumefant rine (AL) from a low transmission set ting (infection pr evalenc e 1–10%(14)) with 137 minimal importa tion in Namibi a repo rt e d safe administra tion and rfMDA r educe d locally acquired 138 malaria incidenc e by approxim ately 50% compared t o RACD.(13) However, ther e are no t rials of rfMDA 139 from very low transmission settings (infe ction preval ence >0 but <1%(14)) with a high level of 140 importa tion, which charac teri zes most n ear-elimina tion se ttings. rfMDA may be more appr opria te th an 141 blanket MDA in low-end emic settings, si nce it targ ets popul ations whe re mala ria has been r ecently 142 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 6 introduc ed. Also , the re a re no t rials of rfMDA using dihydroart emisinin-piper aqui ne (DP), which 143 compared t o standa rd ar temisinin-combi nation t her apies such as AL, has favorabl e charact eristics for 144 MDA (less frequent dosing and long er pe riod of prot ection), bu t safety concerns a bout ra re QT-inte rval 145 prolongati on leadi ng to arry thmia and sudden dea th e xist.(15) 146 Our objective in this trial was t o evalua te the effectiven ess of rfMDA using DP, co mpared to 147 RACD, for reducing malaria t ransmission in the very low transmission se tting of Eswatini. Bo th th e 148 rfMDA and RACD interven tions were embedded within the Eswatini Nati onal Mal aria Programme; as 149 such, this pragmatic t rial assess ed real-w orld effectiveness of th ese int erven tions when delivered wit hin 150 an existi ng surveillance and respons e pro gramme. 151 152

153 Study design and participants 154 We conduct ed a pragma tic open-lab el, cl uster-ra ndomised con troll ed tria l(16) between Sept ember 2015 155 and June 2017 in th e Kingdom of Eswatini, a low middle-income country in sout he rn Africa. 156 Appro ximat ely 30% of the populatio n lives in the e aste rn malaria- endemic ar ea, which borders 157 Mozambique . Plasmodium falciparum is responsible for ove r 99% of malaria case s in Eswatini. Malari a 158 transmission is unstab le and occurs main ly between Oc tobe r and May.(12) Annua l case loads are 159 repor ted from July to Ju n e each year . 160 After major d eclines in malari a transmiss ion from annual par asite incid ence (API) of 3.9 to 0.07 161 per 1000 popula tion from 1999 to 2009, the NMP r eorie nte d its str ategy from control t o elimina tion of 162 transmission by 2020. Since implemen ta t ion of the elimina tion pr ogramme and until ju st prior to this 163 trial, API h as remain ed <1 pe r 1000 population . In 2014–2015, th e tr ansmission se ason prior to th e trial , 164 ther e were 604 r epor ted cases, of which 50% were classified as imported .(17) 165 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 7 This pragmatic cluster-r andomised con tr olled tri al was designed to compa re rfM DA and RACD 166 effectiveness as implemen ted by th e Eswatini Na tiona l Malari a Programme (NMP) and in the cont ex t of 167 other ongoing inte rventio ns including case managemen t, vector con trol , surveilla nce, and informa tion, 168 communication and ed ucation . 77 malari a endemic localiti es or clust ers, with a to tal of 209085 169 individuals residing in 431 enumer ation a reas, wer e eligible for inclusion . Of th e 77 clusters, 63 had 170 malaria cases in th e thr ee yea rs prior to t he trial ; th e remaind er did no t have case s but had prior 171 historical risk of malaria transmission . W e randomised clus ters with a 1:1 allocati on ratio to rec eive 172 RACD, including rapid diagnostic t esting with Pf- specific First Response (Premier Medical Corpor atio n 173 Ltd, Mumbai, In dia) and tr eatm ent of po sitives with AL (Coartem, Nova rtis Pharm aceuticals, Kemp ton 174 Park, South Africa) or rfMDA with pr esu mptive tre atmen t using DP (Eurartesim, Sigma Tau, Italy) (Figure 175 1). Inclusion and exclusi on crite ria ar e sh own in Appendix 1 . Briefly, microscopy- or RDT-confirmed index 176 cases repor ted from any he alth facility in Eswatini were classified as local, impor t ed, or unknown base d 177 on travel hist ory. RACD or rfMDA was tri ggered if the inde x case r esided within a study cluster. If RACD 178 was conducted in th e prior 5 weeks of th e index cas e repo rt, i t was not re pea ted . Following the 179 manufacture r’s recommen dati on tha t DP not be rep eat ed within 8 weeks, nor ta ken more tha n twice in 180 a year, rfMDA was not r epea ted if th ese criteri a were me t. O the r exclusio n crite ri a for DP included: age 181 < 9 months; weight < 7 kg; pr egnancy and breastfe eding, all ergy to DP, acute illne ss including severe 182 malaria, und erlying kidney or hepa tic pro blems, person al or family history of QT prolongati on, or r ecen t 183 trea tmen t with QT-prolonga ting medicat ions. 184 Randomisation and masking 185 To ensure th at t he baseli ne risk of malari a was balanced be tween in terven tion a r ms, we utilized block-186 stratifie d randomisa tion . We assigned th e 77 localities or clus ters to rand omisatio n blocks by separating 187 them into thre e risk groups based on incidence in th e thr ee years p rior t o the trial and prior hist orical 188 risk according to NMP. W e furth er str atif ied each block by whethe r the siz e of the population at risk was 189 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 8 above or below 650 individuals. A statis ti cian at UC San Francisco (MKD) generate d the rand om 190 allocation s equenc e using SAS (version 9. 4m2) to randomly assign 0 or 1 to each cluster within each 191 block and strat um, and th e NMP flipped a coin to de termin e which interven tion c orrespo nded t o 0 and 192 1. The inte rventio n delivery t eam and st udy investigators wer e not blind ed to in t erventi on assignment 193 due to th e na ture of int erven tions. 194 195 Procedur es 196 Prior to th e study, individuals r esiding in endemic are as receive d indoor r esidual s praying (IRS) per 197 standar d procedu res by the NMP. During the study, mala ria cases who pres ent ed at surveillanc e sites 198 were confirmed using RDT or microscopy. Inde x cases trigge red in terven tions if th ey lived within the 199 study area . The surveillanc e te am att em pted t o visit index cas es’ homes within 4 8 hours to administ er a 200 questionn aire a bout travel his tory and vector con trol me asures . 201 In the R ACD arm, consistent wit h NMP st andard pr actices, all cons enting individuals residing 202 within 500m of the index cas e (the “targ et popul ation” for RACD) received RDT te sting, and a dri ed 203 blood spot (DBS) was collected for subse quent molecul ar t esting. RDT-positive in dividuals were 204 transpo rt ed to t he ne ares t heal th facility for trea tment . The study aimed to delive r interve ntions within 205 7 days of index case prese nta tion, but all owed u p to 5 weeks. 206 In the rfMDA a rm, individuals residing wi thin 200m of the inde x case, bu t ex tendi ng beyond 207 200m to reach a minimum of 30 individuals (the “targe t popula tion” for rfMDA) were ta rget ed for drug 208 administra tion using DP. A radius of 200 meters with a minimum 30 individuals was chosen because 209 prior RACD studies showed tha t the maj ority of infections nea r an inde x case cou ld be captur ed within 210 this targ et popul ation .(12) Field staff assessed wheth er it was safe to administe r DP to enrolled eligibl e 211 individuals. Individuals inel igible to r ecei ve DP received RDT testing, and a DBS was collected for 212 subsequen t molecular testi ng. RDT-positi ve individuals were tr anspor ted t o the n eares t heal th facility 213 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 9 for trea tmen t. Eligible individuals r eceive d the first dose of DP under directly obse rved ther apy and 214 doses for day 2 and day 3 for self-administrati on. In dividuals ineligible for DP were screen ed using RDTs 215 and transp ort ed to the n eares t heal th fa cility for trea tmen t if they test ed posi tive . Participan ts were 216 instruct ed to go to the n ear est he alth fac ility if they expe rienced any illness after t aking DP, and they 217 were instruc ted to cont act an on-call stu dy nurse that was availa ble at all hours. T o assess adher ence, 218 the study t eam re turn ed to a subsampl e of participan ts (all participan ts of the firs t inte rvention for each 219 rfMDA cluster) 7-10 days after enrollm en t to conduct pill cou nts. 220 In both a rms, study te ams retu rned a sec ond and thir d day to rec ruit individuals w ho were 221 initially absen t. The study aimed to achi e ve at least 80% int erven tion coverage of index cases and 80% 222 coverage of the targe t popula tion . 223 224 Laboratory m ethods 225 RDT testing was performed using th e Firs t Response P. falciparum HRP-2 Detecti o n Test (Premier 226 Medical Corpor atio n Ltd.). DNA e xt ractio n from DBS for LAMP testing was conducted as pr eviously 227 described (Loopamp Mal aria Pan and Pf Detectio n Kits, Eiken Chemical Co., Ltd.).( 12) LAMP results were 228 used for rese arch purpos es only. 229 230 Outcomes 231 The primary outcome of th e tri al was the cumulative incidence of malari a cases b y study cluster over 232 two-years of follow-up. Secondary ou tco mes repor ted h ere includ e safety and ad herenc e (acceptabili ty 233 has been r epor ted els ewher e(18)). Infection prevalenc e and ser opreval ence a t two-year follow-up were 234 originally also seconda ry outcomes but t he endline cr oss-sectional su rvey was not conducted du e to a 235 shift in prioriti es within th e Minist ry of Health . 236 237 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 10 Statistical Analysis 238 We estima ted the minimum det ecta ble d ifference in cumulative incide nce per p o pulation a t-risk 239 betwee n arms. Bas ed on surveillanc e dat a from 2012–2015 in areas wher e RACD was conducted, we 240 assumed an annual incid ence of 4 per 10 00 individuals, coefficient of variation of 0.9, and type I err or of 241 0.05. We assum ed the p opula tion at-risk (the tot al popula tion of census enum era t ion areas that 242 repor ted incid ent cases within each clust er) was 55 928 individuals in at least 63 o f the 77 total clus ter . 243 There was 80% statistica l power t o det ec t a 50% percent reducti on be tween a rms if at least 63 of 77 244 clusters had a t leas t one ind ex case .(19) 245 The cumulative incidence in each cluste r was calculated as th e number of passivel y detect ed 246 malaria cases divided by th e produc t of populatio n and follow-up time in each clu ster, st ar ting on the 247 date of first ind ex case d etec tion. Th e first index case in each cluste r was exclude d from incidence 248 calculations since in terven tions wer e deli vered afte r initial ind ex cas e det ection in each cluster . Mala ria-249 free survival was compared, and the assu mption of propor tion al hazar ds was assessed using Schoenfeld 250 residuals t esting.(20) 251 To estimat e inte rventio n effects, we use d an inten tion-to- tre at (ITT) approach th at exclud ed 252 localities with no incid ent cases during th e stu dy period since thes e localiti es did n ot receive 253 interven tions . The primary analysis used negative binomial regressio n models wit h an offset for 254 populati on size to estima te incidenc e rat e ratios in each cluste r over th e study pe riod. Mod els adjuste d 255 for baseline covari ates that we re associa ted with th e outcom e using a likelihood r atio t est (p-value < 256 0.2) and that ha d a Pearson cor rela tion c oefficient with th e outcom e ≥ 0.3.(21) Ba selines covaria tes 257 included: incidenc e (2014–2015), propor tion of import ed cases, pr opor tion of ho uses receiving IRS in 258 the pas t year, mon thly average e nhance d vegetati on inde x, monthly aver age rain fall, monthly average 259 land surface temp era tur e, and el evation . 260 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 11 Malaria transmission is highly het erogen eous in lower t ransmission set tings(22), and tre nds in 261 monthly incidence differed b etwe en arm s in the thr ee yea rs prior t o the trial . To a ccount for thes e pre-262 trial differe nces, we used a synth etic con trol analysis to minimiz e pre-t rial differe nces in incidence 263 betwee n arms (Appendi x 2).(23) We the n estimat ed th e difference-in-differenc es for RACD vs rfM DA 264 and the synth etic RACD vs rfMDA. This analysis was not pre-specified. 265 To assess poten tial cont aminati on due t o a lack of buffer zones between clus ters, we identified 266 all clusters with con tiguous neighbou ring clusters and plo tt ed the incid ence in eac h cluster agains t 267 incidence in th e neighbour ing cluster . The small number of contiguous clust ers pr ecluded th e use of 268 formal statis tical tes ting to assess cor rela tions betw een incidenc es in contiguous c lusters . 269 270

271 Betwe en Sep tembe r 2015 and June 2017 , 22 of the 38 clusters ran domly assigned to RACD had 99 272 repor ted cas es; 56 of these cases we re c overed by forty-nine RACD events . Twenty-five of the 39 273 clusters ran domly assigned to rfMDA ha d 121 report ed cases; 89 of th ese cases were covered by 68 274 rfMDA events. The r emaining cases did n ot receive reac tive inte rventio ns due to s taff limitations, fuel 275 shortages , or wea ther condi tions compli cating transp ort (Figure 2). Of th e 2134 individuals eligible to 276 receive RACD, 1696 (79%) were tes ted b y RDTs. Five RDT -positive cases, of which three we re LAMP 277 positive), were r eferr ed for tr eatm ent wi th AL. The most common re ason for non- receip t of RACD was 278 not pres ent (n=398, 18.7%); only 1.5% (n=33) refused. Of th e 2623 individuals eligible to r eceive rfMDA, 279 1932 (74% ) received DP. The most comm on reasons for non-r eceip t of rfMDA wer e not pr esent (n=302 , 280 11.5%) and ineligibility of receive DP (n=313, 11.9%) mainly due to repor ted po te ntial for medica tion 281 inter action . Seventy-six (2.9%) of eligible individuals refused to p articip at e. Data o n medication typ e 282 were incomple te as nurs es repo rte d sens itivities ar ound par ticipan ts disclosing use of antire trovir als 283 (ARVs). No RDT nor LAMP-positive individuals were iden tified among rfMDA ineli gibles. In to tal, 117 284 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 12 interven tion eve nts were impl ement ed i n the tri al and 3941 individuals in 47 clusters wher e 285 interven tions wer e conduct ed were inclu ded in primary outcom e analyses. Adhe r ence to in terven tion 286 assignment was incomplet e: 20 RACD int erventi ons were delive red in rfMDA a rea s (14 clusters), and 5 287 rfMDA interven tions wer e deliver ed in R ACD areas (four clusters). 288 Not t aking into accoun t clusteri ng, index cases had a similar distri bution of age, se x, case origin 289 (e.g. local, impor ted, o r unknown), occupation, a nd bed ne t ownership b etwe en st udy arms (Appendix 290 3). The percent age of index cas es tha t re porte d having had thei r home spraye d in the past yea r was 291 higher in rfMDA clusters than RACD clust ers (28.6% vs 5.3%). For target popul atio n receiving study 292 interven tions, ther e was a similar distrib ution of age, occupa tion, a nd vector con t rol coverage . A higher 293 propor tion in th e rfMDA arm (1.4%) worked in manufacturing compar ed to R ACD (0.1%). In all study 294 clusters, an ave rage of 35.7% of index ca ses and 2.8% of the ta rget po pulati on re porte d inte rna tional 295 travel in th e prio r 8 weeks during the st u dy period. 296 Taking into account clust ering, th ere was imbalance in baselin e tra nsmission inten sity. 297 Cumulative incidence of all malari a in th e thre e years pr eceding th e tri al was higher in th e rfMDA arm 298 compared t o the R ACD arm (6.30 vs 4.17 per 1000, resp ectively) with a similar tre nd seen for local cases 299 only, and for all and local cases only in 2 014–2015, the yea r prec eding the trial (Table 1, App endix 4a). 300 The percen tage of cases classified as imp orted in each cluste r in the ye ars prio r to the tri al was higher in 301 the RACD arm compare d to th e rfMDA ar m (35.8% v s 28.1% for 2012–2015, and 48.7% vs 32.2% for 302 2014–2015). Mean popul ation siz e and e cological factors including rainfall, e nhan ced vegeta tive index , 303 elevatio n, and daytim e land surface t em pera ture we re bala nced be tween a rms at baseline (Table 1). 304 Index cas e and t arget p opula tion int erve ntion coverag e in the R ACD arm was 80.1% and 75.6%, 305 respectively, compa red t o 77.0% and 81. 4%, respectively, in th e rfMDA arm (Table 2). Total coverage 306 (including both index cases a nd the targe t popula tion ar ound each ind ex case) wa s 60.0% in RACD and 307 68.8% in rfMDA. For all coverage me asur es, 95% confidence inte rvals for each ar m overlapped 308 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 13 substanti ally. The median numb er of days between ind ex case repo rt and in terve ntion resp onse was 7 309 (range: 2, 27) in the RACD arm and 11 (range: 3, 52) in the rfMDA arm. I n the rfM DA arm, two clusters 310 had respons e times of 40 and 52 days; ex cluding those cluste rs, th e range was 3 t o 21. 311 During follow-up, the cumulative inciden ce decreas ed from baselin e levels in bot h the RACD and 312 rfMDA arms, with rfMDA having fewer cases during the final mon ths (Janua ry to May) of the second 313 transmission seas on (Figure 3, Appe ndix 4b). The cumulative incidence from 2015 –2017 was 2.11 per 314 1000 in the rfMDA arm compare d to 1.9 7 in the RACD arm (Table 3) (N = 47 clusters in both a rms). In the 315 inten tion-to-t rea t analysis, crud e and adj usted incidenc e rat e ra tios (IRRs) were 1. 01 (95% CI 0.58, 1.73) 316 and 0.93 (95% CI 0.54, 1.60), respectively (Table 3). Restricting to loc al cases only, the adjust ed IRR was 317 0.77 (95% CI 0.38, 1.56). 318 Cumulative malaria-fre e survival for all cases was similar betwe en arms (Figure 4 a). Restric ting 319 to local cases only, cumulative su rvival was higher in the rfMDA arm until app roxi mately 9 months afte r 320 study initiati on, and subs equen tly it was higher in the R ACD arm throughou t th e second high 321 transmission seas on (13–18 months afte r study initia tion) (Figure 4b). The Schoe nfeld residual t est 322 indicated that su rvival was proportio nal betwee n arms for all cases (rho = -0.04, p-value = 0.308) but not 323 for local cases (rho = 0.12, p-value = 0 .032). 324 In the synth etic cont rol ana lysis accounti ng for pre-trial differ ences in incidence b etween arms, 325 ther e was no difference in incidenc e of all malaria cases be tween the rfMDA arm and the synth etic 326 RACD arm (Appendix 5 and 6). When co mparing incidence in each clus ter t o inci dence in contiguous 327 neighbouring clust ers, clust er-level incid ence was not associa ted with incid ence i n contiguous 328 neighbouring clust ers, suggesting th at th e risk of contaminati on in this tri al was minimal (Appendix 7). 329 Field staff conducted 1114 pill coun ts an d recorde d complet e adhe rence to th e 3-day DP 330 regimen in 1099 (98.7%) individuals. Adverse even ts were e xpe rienc ed by 68 individuals in the rfMDA 331 arm (49 in year 1, 19 in year 2). Based on the WHO s everity scale , 54 (80%) events were mild and 14 332 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 14 (20.6%) were as moderate . The most co mmon complaints were he adache , nause a/vomiting, and 333 abdominal pain . Of five individuals with adverse even ts who did not comple te t h e course of DP, all 334 recovere d. On e had difficulty brea thing a nd chest tightn ess tha t could be consist e nt with DP-associated 335 arrythmia bu t the accompanying diarroh ea is less consisten t (Appendi x 8 and 9). During the study 336 period, ther e was only one reco rded mal aria dea th in th e study ar ea. The infec tio n was locally acquired 337 and the pa tien t lived in an rfMDA cluste r , though rfMDA had no t previously be en conducted in th e 338 targe t are a. N o AEs were r epor ted in t he RACD arm. 339 340

341 In this pragma tic, cluste r-randomised tri al conducted in a very low tr ansmission malaria elimina tion 342 setting, rfMDA clust ers had lowe r locally acquired mala ria incidenc e during th e whole study peri od 343 compared t o RACD clusters, pa rticula rly during the secon d high transmission se ason of the study peri od, 344 but overall , evidence was weak. I nte rven tion coverage was lowe r than e xpec ted, and malaria occur red 345 in fewer clusters th an plann ed in the sa mple size calculatio n. Adh ere nce to p res umptive tre atme nt with 346 DP was high, and as reported els ewher e, acceptabili ty was high.(18) Importantly, ther e were no s erious 347 adverse even ts (SAEs). 348 Progress towards th e 2030 eliminat ion goal in southe rn Africa has slowed d espit e coordinat ed 349 regional effor ts and delivery of stand ard interven tions, including pr e-season ind o or residual sp raying, 350 symptomatic case manageme nt, an d RA CD.(11) While RACD aims in part to add r ess asymptomatic 351 reservoirs of t ransmission, ra pid diagnos tic tests use d in low transmission se tting s have poor sensitivity 352 and miss many low-density infections.(24) While blanket MDA would r each all as ymptomatic infections, 353 it is logistically difficult to implement a t s cale, inefficient in pop ulati ons with few, highly clustered 354 infections, and i t may not be safe or acc e ptable .(7) 355 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 15 A few trials have evalua ted focal MDA de livered to ho tspo ts at th e village or sub-village level in 356 low transmission set tings and resul ts ar e mixed.(25-28) In Zanzibar, which most re sembles our sit e due 357 to high coverage of stand ard in terven tio ns, very low transmission intensi ty, and h igh rates of 358 importa tion, n egative findings of focal MDA effectiveness were hypot hesize d to b e rela ted t o 359 suboptimal timing and the numbe r of MDA rounds, and r e-introd uction of malari a through 360 importa tion.(27) The reac tive appro ach e mployed in our tri al sought to address th ese issues by targe ting 361 the focal MDA to a time and plac e when transmission risk was highest (e.g. whe re there we re r ecent 362 imported o r local cases). 363 This trial is one of thr ee t hat evaluat ed rfMDA. Resu lts from a low t ransmission se tting in Zambia 364 trial ar e forthcoming .(25) A trial in a low transmission se tting in Namibi a evalua te d rfMDA alone and in 365 combination with r eactive vect or cont rol in comparison to RACD.(13) Compared t o RACD, rfMDA 366 reduced local mal aria incid ence by 48%, and rfMDA with additi onal re active vecto r control r educed 367 incidence by 74%. There ar e sever al key differences be tween t he Namibi a tri al an d this trial . First , the 368 Namibia t rial had a highe r baselin e annual malaria incidenc e (30 per 1000 compared to 3 pe r 1000 in 369 this trial) and a lowe r propo rtio n of imported mala ria (3% compared to 40% in thi s trial), both of which 370 may facilitate higher impac t of focal MDA.(26) Second, the Namibia t rial was larg ely implemente d by a 371 resea rch team , while the Eswatini trial w as pragmatic and la rgely implement ed b y the local malaria 372 control pr ogramme.(16) Coverage in the Namibia t rial was also higher compa red t o this trial (study ar ea 373 index and targe t popula tion cover age we re >84% and >85%, compar ed to 78.5% a nd 78.7% in this trial). 374 This trial faced several cha llenges uniqu e to very low incidence set tings including strong 375 spatiot empor al clusteri ng and import ed malaria.(22) The number of cluste rs with at least o ne inde x case 376 during follow-up was lower than ex pect e d (we expecte d 63 but observe d 47). Thus, the t rial was not 377 powered t o det ect t he hypoth esized inci dence red uction of ≥ 50%, no r smaller r e ductions, with 378 precision. S econd, t hough the s tudy was cluster-ran domised, bas eline mala ria inci dence and t he 379 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 16 percen tage of import ed cases was higher in the rfMDA arm th an the R ACD arm. The lower th an 380 expec ted numb er of clusters p er arm like ly contribut ed to t hese imbal ances. Thou gh analyses adjusted 381 for these fact ors, it r emains possible that unmeasured fact ors affecting malaria tr ansmission differed 382 betwee n arms. To furth er add ress baseli ne imbalance, we conduc ted a synth etic control an alysis, which 383 produced similar r esults to th e primary a nalysis. However, th e synthe tic contr ol a nalysis did not 384 completely accoun t for pre-int erven tion differences in incidence b etwe en arms, li kely because few 385 clusters (<25 per a rm) had incident cases during follow-up, and cluster-level incid ence varied . When 386 outcomes ar e rar e and clust ered , tri als r equire very la rge cluste r numbers t o have sufficient statis tical 387 power and basel ine bala nce.(27, 28) 388 Implemen tatio n factors may have influe nced effect estima tes . First, tot al covera ge was lower 389 than th e tri al’s goal of 80%, and imbalan ced across arms (60% for RACD and 68.8% for rfMDA). 390

r elat ed to st affing and transp ort compromise d index cas e level covera ge and particip ants no t 391 being presen t compromised targe t popul ation cover age. Of no te, 12% of the rfM DA target p opula tion 392 was ineligible to r eceive DP, with the mo st common reason b eing poten tial medic ation int erac tion with 393 ARVs. Alth ough saquinavir, t he only ARV contraindic ate d for use with DP, is not available in Eswatini, 394 nurses exp ressed conc ern th at adve rse e vents could be int erpr et ed by the pa rtici pant as due to ARV, 395 and thus compromise A RV adher ence. W here A RV use is common, such as Eswatini which has the 396 highest worldwide incide nce of human immunodeficiency virus (HIV),(29 ) better strat egies to a ddress 397 safety concerns rega rding drug-drug int e raction will be ne eded . The use of less st ringent e xclusion 398 criteri a (e.g. inclusion of pregn ant wome n, young children, individuals with cer tai n morbidities) as has 399 been safely prac ticed by othe rs(30) could also improve coverage . However, cove rage was not associa ted 400 with incidence, suggesting t hat differ enc es in coverage be tween a rms were unlik ely to affect study 401 findings. Second, int erven tion resp onse t ime was substantially highe r for two rfM DA clusters compare d 402 to the R ACD arm. It is possible tha t the re was greater mal aria t ransmission be twe en index cas e 403 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 17 detec tion and in terve ntion d elivery in th e rfMDA arm than the RACD arm. Third , study clusters wer e not 404 separa ted by geograp hic buffer zones to minimize contamina tion, which can occu r due to vect ors or 405 human movement. H owever, clust er-lev el incidence was not co rrela ted b etwe en contiguous cluste rs, 406 suggesting that the chance of con tamina tion in this t rial was low. 407 Import antly, our s tudy is the first t o sho w the safety of rfMDA using DP. The Namibia trial used 408 AL and in comparison, DP may be prefer able for MDA due t o ease of use and lon ger prot ective pe riod 409 (once versus twice daily, and 4–6 weeks versus a few days, due to th e half-life). Rarely, DP-associated 410 QT-interval prolo ngation may lea d to sud den dea th (1/~200 000), and in very low- endemic set tings the 411 risk-to-benefit r atio may not favor DP.(15) Here, one pa rticipa nt had symptoms t hat could be consis tent 412 with arrythmia , and tr ea tment was stop ped. Pharmacovigilance pr ovided by nurs es through follow-up 413 visits and their on-call avail ability likely helped t o preven t SAEs. 414 415

416 This study is the first trial to compar e rfMDA and RACD in a very low malaria-endemic setting . As 417 interven tions wer e embedd ed within an existing na tional mal aria pr ogramme, i t provides such evidence 418 in realistic implem enta tion condi tions. W e found tha t rfMDA was safe. Alth ough rfMDA clusters had 419 lower cumulative incidenc e during th e st udy period, we wer e unable to confirm e ffectiveness of rfMDA 420 compared t o RACD, poten tially due to in sufficient power. Fo r rfMDA to be mor e effective than R ACD, 421 improved coverage and/o r the a ddition of complementary int erven tions, such as IRS, may need t o be 422 delivered in tandem .(13) To improve stat istical power t o det ect impact of int erve ntions in very low-423 endemic set tings, futur e trials may re qui re multi-site d esigns, large r sample sizes, or alte rnatively, 424 smaller units of rand omisation (e .g. a nei ghborhood), or ad aptive d esigns that a djust featur es such as 425 the sample siz e and alloca tion ra tio .(31, 32) Such evidence will be critical to guide countries in th eir 426 quest to move from very low to no trans mission. 427 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 18 428 List of abbreviations 429 430 aITT adjusted int enti on to t rea t 431 ARV antir etr ovirals 432 DP dihydroar temisinin-pipe raquin e 433 HIV human immunodeficiency virus 434 ITT inten tion to tre at 435 IRS indoor resi dual spraying 436 LAMP loop mediat ed isoth ermal amplifica tion 437 MDA mass drug administrat ion 438 NMP Natio nal Mala ria Programme 439 PCR polymerase chain r eacti on 440 RACD reactive cas e det ection 441 rfMDA reactive focal mass drug administ rati on 442 RDT rapid diagnostic test 443 SAE serious advers e events 444 WHO World He alth O rganiza tion 445 446 447 448 449 450 Declarations 451 452 Ethics approval an d co nsent to partici pate 453 Ethics approval was given by Eswatini Ministry of Health (MH/599C) and by University of California San 454 Francisco Human Rese arch Prot ection Pr ogram & IRB (Formerly Committe e on Hu man Research) (14-455 15226). Writt en informed conse nt was o btained from individual p articip ants . For children less th an 18 456 years, writ ten informed cons ent from a p aren t or guardi an was requir ed, as was writt en assen t for 457 children 12–17 years . 458 Trial registration 459 ClinicalTrials.gov, NCT02315690 (registra tion dat e: Decemb er 8, 2014) 460 Consent for publication 461 Not applic able 462 463 Availability of data and materials 464 The data t hat supp ort the findings of this study are availabl e from Eswatini Minist ry of Health but 465 restric tions apply to the availa bility of th ese dat a, which were used und er licens e for the curr ent st udy, 466 and so are no t publicly available . Data ar e however availabl e from the au tho rs upon reason able r eques t 467 and with permission of Eswatini Minist ry of Health. 468 469 Competing inter ests 470 The autho rs declar e tha t th ey have no competing inte res ts. 471 472 Funding 473 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 19 This study was supported by the Bill & M elinda G ates Fo undati on (A122394) and the Horchow Family 474 Fund (5300375400). The funders of the study had no rol e in study design, dat a col lection, d ata an alysis, 475 data in terp ret atio n, or writi ng of the rep ort. 476 477 Patient and Public Involvem ent 478 As assessed formally (through Knowledg es Atti tudes a nd Practices surveys) and informally (during 479 malaria progr amme activiti es including r eactive case d etec tion), th e public’s conc erns abou t malaria and 480 their eagern ess for the coun try to achi eve its goal of malaria elimin ation info rme d the res earch qu estion 481 and study design. As incide nt malari a cas es were th e trigge r for recrui tmen t (targ eting househ old 482 members and neighb ors of index cas es) and the prim ary outcome , thes e aspec ts of the study reli ed on 483 patien ts seeking car e when ill and rec eiving malaria tes ting. To elicit o ngoing feed back regarding th e 484 conduct and burd en of the s tudy interve ntion, pa tien ts and th e public were enga ged in an ongoing basis 485 through focus group discussions, th e res ults of which are published els ewher e.(1 8) 486 487 Authors’ co ntributions 488 MSH, SK, and RG conc eptua lised and d esigned the stu dy. NN , ND, and KB contr ib uted t o study design. 489 NM led th e tri al coordin ation . KB, BD, DH, LMP, and CM additionally suppor ted tri al coordina tion. ND led 490 the field implemen ta tion. SV led the da ta collection. MK and GT oversaw clinical a nd safety aspects of 491 the t rial. KB overs aw data collec tion and analysis of acceptability assessmen t. NN led the la bora tory 492 activities with oversigh t from DH, BG, an d GM. SV and BW l ed dat a managemen t and support ed da ta 493 analyses. MSKD, JBC, and MSH l ed th e da ta analysis. R G and DP advised on the da t a analyses. J BC and 494 MSH wrote the manuscrip t. N N and SK p rovided oversight of local implem enta tio n. MSH provided 495 overall oversight of th e study. All auth ors read and app roved th e final manuscript . 496

497 The autho rs would like to th ank the r esid ents of Eswatini who support ed th e stud y through thei r 498 participa tion and in puts. We th ank the fi eld and labo rato ry staff. We t hank Alema yehu for collecting 499 ecological dat a. W e thank Ad am Soble, Manik Saini, Charlo tt e Lejeune, a nd Tho mas How at CHAI for 500 their supp ort in adminis tra tion and loc al coordina tion. W e th ank Justin Cohe n, Ar naud LeMenach , Hugh 501 Sturrock, Joell e Nadl e, Immo Kleinschmi dt, and Rob er t Haley for th eir inpu ts on t rial design. W e thank 502 the Minis try of Health , Eswatini Pharmac ovigilance committe e, and th e Eswatini Malaria Elimina tion 503 Advisory Group for their supp ort a nd ove rsight. 504 505 506 Figure lege nds 507 Figure 1. Map of th e study ar ea 508 Abbrevia tions: rfMDA, react ive focal mass drug administra tion; R ACD, reactive cas e det ection . 509 510 Figure 2. Trial pr ofile showing randomisa tion and en rolmen t 511 Abbrevia tions: rfMDA, react ive focal mass drug administra tion; R ACD, reactive cas e det ection ; RDT, 512 rapid diagnostic test ; LAMP, loop-media t ed isothe rmal amplification ; AL, ar teme t her-lumefant rine ; DP, 513 dihydroar temisinin-pipe raquin e 514 *not cover ed due to staff limitati ons, fue l shortages, o r weath er condi tions compl icating transp ort 515 †RDT testing conduct ed in 262 of DP ineligibles. As none test ed positive , none we re refe rred for 516 trea tmen t with AL 517 518 Figure 3. Mon thly incidence in e ach stud y arm prior to and du ring the in terve ntio n period. 519 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 20 RACD, reactive case d etec tion; rfMDA, re active focal mass drug administra tion 520 521 Figure 4. Mala ria-fre e survival curves for the ou tcomes of a) all incident mala ria cases, and b) local 522 incident mala ria cases. R ACD, reactive ca se detec tion; rfMDA, reac tive focal mass drug administra tion . 523 High transmission seasons occur red from follow-up months 1-5 and 13-18. 524 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 21

525 1. Feachem RG A, Chen I, Akbari O , Ber tozzi- Villa A, Bhat t S, Binka F, e t al. Malari a era dication 526 within a gener ation : ambitio us, achievab le, and nec essary. Lance t. 2019;394(10203):1056-112. 527 2. Okell LC, Bousema T, Griffin JT, Ou edra o go AL, Ghani AC, Drakel ey CJ. Factors d et ermining the 528 occurrence of submicroscopic mala ria infections and their r elevance for con trol . N at Commun. 529 2012;3:1237. 530 3. Sturrock HJ , Hsiang MS, Cohen JM, Smith DL, Greenhous e B, Bous ema T, et al . Tar geting 531 asymptomatic malari a infections: active surveillance in cont rol and elimin ation . PLoS Med. 532 2013;10(6):e1001467. 533 4. Moonen B , Cohen JM , Snow RW, Slutske r L, Drakeley C, Smith DL, et al. Op era tion al stra tegies to 534 achieve and maint ain malari a eliminat io n. Lancet . 2010;376(9752):1592-603. 535 5. WHO. W HO Evidence R eview Group on mass drug administrat ion, mass scree nin g and 536 trea tmen t and focal scree ning and tr eat ment for malari a. G eneva, Swit zerl and: World He alth 537 Organiza tion; 2015 . 538 6. Poirot E, Skarbinski J, Sincl air D, Kachur S P, Slutsker L, Hwang J. Mass drug admini strati on for 539 malaria. Cochr ane Data base Syst Rev. 20 13(12):C D008846. 540 7. Newby G, Hwang J, Koit a K, Chen I, G ree nwood B, von Seidlein L, e t al. R eview of mass drug 541 administra tion for mala ria and its op era ti onal challenges . Am J Trop M ed Hyg. 2015;93(1):125-34. 542 8. White NJ . Does antimala rial mass drug a dministrati on incre ase or d ecreas e the ri sk of 543 resistanc e? Lancet Infect Dis. 2017;17(1):e15-e20. 544 9. WHO. M ass drug administr ation for falci parum malaria . G eneva, Switz erlan d: Wo rld Healt h 545 Organiza tion; 2017 . 546 10. Eisele TP, Benne tt A, Silumb e K, Finn TP, Chalwe V, Kamuliwo M, et al. Shor t-t erm Impact of 547 Mass Drug Administra tion Wi th Dihydroa rtemisinin Plus Piperaquin e on Mal aria in Souther n Province 548 Zambia: A Cluster-Ran domized Cont rolle d Trial. The Jou rnal of infectious dis eases . 2016;214(12):1831-9. 549 11. Programme WGM . Wo rld Mala ria Rep ort 2019. Geneva : World H ealt h Organiz atio n; 2019. 550 12. Hsiang MS, Ntsha lintshali N, Kang Dufour MS, Dlamini N, Nhlab athi N , Vilakati S, e t al. Active 551 case-finding for malaria: A t hre e-year na t ional evalua tion of optimal a pproach es t o detec t infections an d 552 hotspots through r eactive cas e det ection in the low transmission se tting of Eswati ni. Clin Infect Dis. 553 2019. 554 13. Hsiang MS, Ntuku H, R ober ts KW, Dufour MK, Whitt emore B , Tambo M, e t al. Effe ctiveness of 555 reactive focal mass drug administ rati on a nd reactive focal vecto r cont rol to r educ e malaria t ransmission 556 in the low malaria- endemic set ting of Na mibia: a cluster- randomised con troll ed, open-label , two-by-two 557 factorial design t rial. Lanc et. 2020 ;395(10233):1361-73. 558 14. WHO. A fr amework for malaria eliminati on. G eneva: W orld He alth O rganiza tion; 2017. 559 15. WHO. The ca rdiot oxicity of antim alarials . WHO Evidence Review Gr oup Me eting; Octobe r 13 to 560 14; Gen eva, Switze rland: World H ealth O rganizati on; 2016. 561 16. Ford I, N orri e J. Pragma tic Trials. N Engl J Med. 2016;375(5):454-63. 562 17. Programme SNMC. N ation al Mala ria Con trol Programme A nnual Re por t 2014-2015. In: He alth 563 Mo, edi tor. Mbaban e, Swaziland2015 . 564 18. Baltz ell KA, Maglior A, Bangu K, Mngadi N, Prach LM, Whit temo re B, e t al. "W e were afrai d of 565 the lion t hat has r oar ed ne xt t o us"; com munity response to re active focal mass drug administra tion for 566 malaria in Eswatini (formerly Swaziland). Malaria jour nal. 2019;18(1):238. 567 19. Hayes R, Moult on L. Cluster Ra ndomised Trials: CRC Press; 2009. 568 20. Grambsch PM, Thern eau TM. Propo rtio n al hazards t ests and di agnostics based o n weighted 569 residuals. B iomet rika. 1994;81(3):515–26 . 570 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 22 21. Pocock SJ, Assmann SE, Enos LE, Kasten LE. Subgroup analysis, covaria te adjus tme nt and 571 baseline compa risons in clinical trial repo rting: curr ent p ractice a nd probl ems. Sta t Med. 572 2002;21(19):2917-30. 573 22. Bousema T, Griffin JT, Sau erwein R W, Sm ith DL, Churcher TS, Takken W, et al . Hit ti ng hotspots: 574 spatial t arge ting of malaria for cont rol an d elimination . PLoS Med. 2012;9(1):e100 1165. 575 23. Rehkopf DH, Basu S. A N ew Tool for Case Studies in Epidemiology-the Syn the tic Control M eth od. 576 Epidemiology. 2018;29(4):503-5. 577 24. Newby G, Harva rd K, Cotte r C, Roh M, Be nnet t A, Chen I, et al . Scree n and tr ea t st rategi es for 578 malaria elimina tion : a review of evidence , A Background Paper commissione d by the Bill and Meli nda 579 Gat es Founda tion . San Francisco: Th e Gl obal Heal th Gr oup, Ins titu te for Global H ealth Scie nces, 580 University of California, San Fr ancisco. ht tp://www.shrinkingthemalari amap.o rg/resources-581 publications/scre en-and-tr eat-s tra tegies- malaria-elimina tion-r eview-evidence ; 20 18. 582 25. Bridges DJ, Mille r JM, Chalwe V, Mo onga H, Hamainza B, S teke tee R , et al . Community-led 583 Responses for Eliminati on (CoRE): a study protocol for a community rand omized controll ed tri al 584 assessing the effectiven ess of community-level, reactive focal drug adminis tra tio n for reducing 585 Plasmodium falciparum infection preval e nce and incidence in Sou the rn Province, Zambia. Trials. 586 2017;18(1):511. 587 26. von Seidlein L, Peto TJ, Landi er J , Nguyen TN, Tripura R, Phommasone K, e t al. Th e impact of 588 targe ted mala ria elimina tion with mass d rug administra tions on falciparum mala ri a in Southe ast Asia : A 589 cluster ra ndomised t rial. PLoS Med . 2019;16(2):e1002745. 590 27. Lipsitch M, Eyal N. Improving vaccine tri a ls in infectious disease em ergencies . Scie nce. 591 2017;357(6347):153-6. 592 28. Lang T. Adaptive t rial design: could we us e this appr oach to improve clinical trials i n the field of 593 global healt h? Am J Trop M ed Hyg. 2011; 85(6):967-70. 594 29. Justman J , Ree d JB, Bic ego G, Donnel l D, Li K, Bock N, et al. Swazilan d HIV Incidenc e 595 Measur ement Su rvey (SHIMS): a prospec tive nation al cohor t study. Lance t HIV. 2 017;4(2):e83-e92. 596 30. Gutman J, Kovacs S, Dorsey G, St ergachis A, Ter Kuile FO . Safety, t oler ability, and efficacy of 597 repea ted d oses of dihydroar temisinin-pi peraquin e for preve ntion a nd tr eatm ent of malaria: a 598 systematic review and me ta-analysis. La ncet Infect Dis. 2017;17(2):184-93. 599 600 601 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 23 Tables 602 Table 1. Baseline cha ract eristics of cluste rs (localities) included in the trial 603 Cluster-level cha racteristic Overall n=77 RACD n=38 rfMDA n=39 Transmission intensity and cont rol measures, mea n (95% CI) Sept 201 2–June 2 015 Cumulative inc idence of all cases 5.25 (3.7 6 – 6.74) 4.17 (2.9 1 – 5.43) 6.30 (3.5 9 – 9.01) Cumulative inc idence of loca l cases 4.02 (3.2 1 – 4.82) 3.31 (2.1 5 – 4.47) 4.70 (3.5 7 – 5.83) Propo rtio n of cases classified as impo rted a 31.9 (2 5.0 – 38.7) 35.8 (2 4.5 – 47.2) 28.1 (1 9.9 – 36.3) July 20 14–June 20 15 Cumulative inc idence of all cases 2.99 (2.0 9 – 3.88) 2.59 (1.3 6 – 3.82) 3.38 (2.0 4 – 4.72) Cumulative inc idence of loca l cases 2.44 (1.5 6 – 3.32) 1.80 (0.6 1 – 2.99) 3.06 (1.7 5 – 4.38) Propo rtio n of cases classified as impo rted b 40.1 (2 9.2 – 51.0) 48.7 (3 1.3 – 66.2) 32.2 (1 8.3 – 46.0) Populati on c haracter istics, mean ( 95% CI) Size 2715 (22 75 – 31 56) 2752 (20 86 – 34 18) 2680 (20 70 – 32 89) Ecol ogical factors , media n (ra nge) Rainfa ll, mm c 65.9 (3 6.9 – 92.6) 64.8 (3 9.6 – 92.6) 66.7 (3 6.9 – 89.0) EVI c 0.29 (0.1 9 – 0.44) 0.28 (0.2 1 – 0.39) 0.29 (0.1 9 – 0.44) Elevation , m 368 ( 147 – 8 52) 377 ( 170 – 5 89) 355 ( 147 – 8 52) Daytime LST, °C c 31.2 (2 8.3 – 35.7) 31.4 (2 8.4 – 35.2) 31.1 (2 8.3 – 35.7) Incidences are cases per 10 00 p opu latio n 604 Abbreviatio ns: RACD, reactive case detection; rfM DA, reactive fo cal mass dr ug admin istration ; E VI, e nhanced vegetative index; 605 LST, land su rface temperatu re 606 a Sample size (n ) fo r Overa ll , RACD, an d r fMDA were 74 , 3 6, and 38 c lusters, respectively 607 b Samp le size (n) fo r Overal l, RACD, and rfMDA we re 52 , 2 5, and 27 c lusters, respectively 608 c Mean month ly values Sept 20 15–June 20 17 609 610 611 612 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 24 Table 2. Inte rvention cove rage and r espo nse time 613 Overall n=47 a RACD n=22 a rfMDA n=25 a p-value Index case coverage b , mean % ( 95% CI) 78.5 (6 9.9 – 87.0) 80.1 (6 7.3 – 92.9) 77.0 (6 4.8 – 89.2) 0.72 Target popu latio n coverage c , mean % (95% CI) 78.7 (7 1.1 – 86.3) 75.6 (6 1.8 – 89.5) 81.4 (7 2.6 – 90.1) 0.45 Total c overage d , mean % (95% CI) 64.6 (5 6.4-7 2.9) 60.0 (4 6.0-7 4.0) 68.8 (5 8.6-7 9.1) 0.86 Response time, med ian ( range) Days between index case report and i ntervention response 8 (3 – 52) 7 (2 – 27) 11 (3 – 5 2) 0.37 Abbreviatio ns: RACD, reactive case detection; rfM DA, reactive fo cal mass dr ug admin istration 614 a Sample size (n ) fo r Overa ll , RACD, an d r fMDA were 41 , 2 0, and 21 c lusters, respectively for target populat ion coverage, total 615 coverage, and t ime to intervention , an d respo nse time 616 b Index case coverage was defined as the perce ntage of eligi ble index cases that received a n i ntervention averaged acr oss study 617 arm clusters. 618 c Target pop ulati on c overage was defined as the p ercentage of the target po pulati on w ithin 20 0m z ones arou nd each in dex case 619 that received an interventio n averaged acr oss study arm cl usters. 620 d T otal c overage was defined as the p rod uct o f i nd ex case coverage and target popu lation coverage 621 622 623 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 25 Table 3. Adjusted incidenc e rat e ra tios (I RRs) in 2015–2017 comparing clusters as signed to RACD versus 624 rfMDA 625 Study Arm N clusters Incidence (cases per 1 000 person - years) Crude IRR p- value Adjusted IRR* p- value All cases RACD 22 1.97 (1.57 – 2.47) 1 (Ref) 0.99 1 (Ref) 0.80 rfMDA 25 2.11 (1.73 – 2.59) 1.01 (0.58 – 1.73) 0.93 (0.54 – 1.60) Local cases on ly RACD 22 0.97 (0.71 – 1.34) 1 (Ref) 0.85 1 (Ref) 0.47 rfMDA 25 1.29 (1.00 – 1.67) 1.06 (0.57 – 1.98) 0.77 (0.38 – 1.56) Abbreviatio ns: RACD, reactive case detection; rfM DA, reactive fo cal mass dr ug admin istration 626 95% Conf idence intervals f or inc idence were estimated using the W ilson method. Inci dence rate ra tios (IRRs) compa red l ocal ity-627 level inc idence in the r fMDA arm to the RACD arm using an intentio n-to-t reat appr oach a nd negativ e bin omial mode ls. 628 *Adjusted for baseline covariates that were associ ated with the outc ome: inc idence in 201 4–15 (al l cases model on ly), loca l 629 inci dence in 2 014– 15 ( loca l cases model on ly). 630 . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint 77 Localities in study area (total population 209 085) RACD 38 localities (total population 104 584) rfMDA 39 localities (total population 104 501) 121 cases reported in 25 localities (population 97 142)99 cases reported in 22 localities (population 66 575) 2 134 eligible individuals (44/event) 33 refusals (1.5%) 398 not present (18.7%) 7 not done (0.33%) 1 696 RDTs done (34.6 individuals/event) 76 refusals (2.9%) 302 not present (11.5%) 313 DP ineligible (11.9%)† 87 potential medication interaction 51 reported pregnancy 13 possibly pregnant 35 with heart, kidney, or liver problems 14 less than 9 months 8 weight <7 kg 7 with fever or feeling ill 8 family history of heart problems or sudden death 4 allergic to DP 86 unknown reason1 932 received DP (28.4 individuals/event) RANDOMISATIONTRIGGERING INDEX CASESCOMMUNITY INTERVENTION RESPONSE 49 reactive intervention events covering 56 cases in 22 localities 68 rfMDA intervention events covering 89 cases in 25 localities 43 cases not covered* 32 cases not covered* 2,623 eligible individuals (39/event) 1 691 RDT negative (10/1644 tested LAMP positive) 5 RDT positive (3/5 tested LAMP positive) 5 referred for treatment with AL 355 localities in Eswatini 775 cases reported from 78 health facilities 537 cases reside outside study area 15 found in RACD 3 false positives by microscopy . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint A B 0.995 0.996 0.997 0.998 0.999 1.000 0 4 8 12 16 20 Months since study start Proportion without malaria infectionRACD fMDA RACD 95% CI fMDA 95% CI High Low A) All cases 24474 24469 24461 24455 24449 24417 31927 31920 31910 31901 31886 31854 RACD fMDA 0 4 8 12 16 20 Months since study start Number at risk 0.997 0.998 0.999 1.000 0 4 8 12 16 20 Months since study start Proportion without malaria infectionRACD fMDA RACD 95% CI fMDA 95% CI High Low B) Local cases only 21791 21785 21785 21783 21780 21763 31464 31463 31458 31451 31440 31422 RACD fMDA 0 4 8 12 16 20 Months since study start Number at risk 0.995 0.996 0.997 0.998 0.999 1.000 0 4 8 12 16 20 Months since study start Proportion without malaria infectionRACD fMDA RACD 95% CI fMDA 95% CI High Low A) All cases 24474 24469 24461 24455 24449 24417 31927 31920 31910 31901 31886 31854 RACD fMDA 0 4 8 12 16 20 Months since study start Number at risk 0.997 0.998 0.999 1.000 0 4 8 12 16 20 Months since study start Proportion without malaria infectionRACD fMDA RACD 95% CI fMDA 95% CI High Low B) Local cases only 21791 21785 21785 21783 21780 21763 31464 31463 31458 31451 31440 31422 RACD fMDA 0 4 8 12 16 20 Months since study start Number at risk . CC-BY 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 12, 2021. ; https://doi.org/10.1101/2021.03.12.21252721doi: medRxiv preprint