key: cord-0795274-dddx26uv
authors: Laydon, D. J.; Cauchemez, S. J.; Hinsley, W. R.; Bhatt, S.; Ferguson, N. M.
title: Prophylactic and reactive vaccination strategies for healthcare workers against MERS-CoV
date: 2022-04-06
journal: nan
DOI: 10.1101/2022.04.06.22273497
sha: 47b005f668bed3c61fa09381ad1499d2c94e5477
doc_id: 795274
cord_uid: dddx26uv

Several vaccines candidates are in development against Middle East respiratory syndrome-related coronavirus (MERS-CoV), which remains a major public health concern. Using individual-level data on the 2013-2014 Kingdom of Saudi Arabia epidemic, we employ counterfactual analysis on inferred transmission trees ("who-infected-whom") to assess potential vaccine impact. We investigate the conditions under which prophylactic "proactive" campaigns would outperform "reactive" campaigns (i.e. vaccinating either before or in response to the next outbreak), focussing on healthcare workers. Spatial scale is crucial: if vaccinating healthcare workers in response to outbreaks at their hospital only, proactive campaigns perform better, unless efficacy has waned significantly. However, campaigns that react at regional or national level consistently outperform proactive campaigns. Measures targeting the animal reservoir reduce transmission linearly, albeit with wide uncertainty. Substantial reduction of MERS-CoV morbidity and mortality is possible when vaccinating healthcare workers, underlining the need for at-risk countries to stockpile vaccines when available.

As a sensitivity analysis, we also consider slower waning of immunity using the 159 sigmoidal Hill function, and so in this case 160

where Y is the efficacy half-life and a governs the speed of decline. We set a = 4 as a 162 balance between allowing the vaccine to maintain its efficacy for longer than 163 exponential waning, while also having a reasonably gradual decline [ Figure 1 ]. 164 165 We consider mean durations (or half-lives if waning is sigmoidal) of 1, 2, 5, 10, 15, and 166 20 years, as well as no vaccine waning. We simulate values of 6 months, and 1 to 10 167 years for the time between vaccination and the next outbreak (which we term the "wait" 168 for brevity). If S denotes the set of people to be vaccinated, and Pc is the coverage achieved in a 176 campaign, then under a proactive strategy the probability Pv,k that a given case k will 177 be vaccinated is given by 178 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. where tk * is the time post vaccination for case k. In this work, S is the set of healthcare 185 workers, and we assume full coverage (i.e. Pc = 1) of this group, although the effects 186 of reduced coverage can simply be obtained through scaling. For example, 45% 187 efficacy with 100% coverage is equivalent to 90% efficacy with 50% coverage. 188 189 Under a reactive campaign, a delay must be incorporated to account for first the react 190 time τI between the first case in a hospital (or region or country), and second the lag 191 τP between vaccination and protection. Hence the probability that case k will be 192 vaccinated is given by the following product (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 of symptom onset of the first case in hospital h. However, if reacting at regional or 198 national level, T0,h is defined as the time of symptom onset of the first case in region h 199 or the entire country. 200 201 For reactive campaigns, only the vaccine's initial efficacy is relevant (and not its 202 duration of protection) as vaccine waning will be negligible in the time frames we 203 consider between an outbreak and implementation. Here then the probability that case 204 k will be protected and deleted from the transmission tree, is 205 ChAdOx1 MERS peaked at 14 days, and while for antibodies the peak was observed 210 at 28 days, antibody titres were still high at 14 days [19] , and therefore we set τP = 14 211 days. It is important to note that if the vaccine takes longer than 14 days-post-dose to 212 confer protection, this is effectively already included in our analysis, as it is really the 213 react time plus the time to protection that is important and so a longer time to protection 214 is essentially a relabelling. 215 216 We investigate the effect of control measures aimed at limiting animal reservoir 217

transmission, which we model as a simple proportion γ of reservoir infections that are 218

In the inference of transmission trees and parameter posteriors, all priors are uniform 232 and fitted on a log scale. We performed 55,000 iterations with a burn-in period of 5,000, 233 thinning every 5 iterations, resulting in 10,000 posterior samples per model run. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 In the 18 months from the start of 2013, there were 681 MERS cases, where date of 242 symptom onset and patient's hospital was reported, of which 534 (78%) were 243 symptomatic at presentation and 276 (41%) were fatal. 187 (28%) of cases were in 244 healthcare workers (HCWs), among whom there were 15 deaths, giving an 8% case-245 fatality ratio among healthcare workers, and comprising only 5% of all 276 deaths. The 246 case-fatality ratio among non-HCWs was 53% [ Figure 2 ]. 247 248 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

Example model output 255 Figure 3 shows a series of transmission trees ("who-infected-whom" plots) from 256 example model runs with and without vaccination. The trees show the contribution to 257 transmission from the animal reservoir, as well as transmission within hospitals, 258 between hospitals but within regions, and between regions. 259 260 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 14 Under a proactive strategy, healthcare workers are vaccinated in anticipation of the 269 next outbreak, and therefore all vaccinees have at least some protection from its 270 outset. However, a proactive strategy depends on the extent of vaccine waning. 271 Therefore success is a function of initial efficacy, duration and the wait time until the 272 next outbreak, where the latter cannot be known in advance. 273 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 In the absence of camel control measures, under an optimistic scenario of 90% 275 efficacy with a 20-year mean duration and only 6 months until the next outbreak, 64% 276 (95% CrI: 54% -74%) of cases and 51% (95% CrI: 39% -64%) of deaths would be 277 averted. However, if the next outbreak occurred 8 years after vaccination, then only 278 54% (95% CrI: 41% -67%) of cases and 41% (95% CrI: 28% -58%) of deaths would 279 be averted. The 2013-2014 KSA MERS-CoV outbreak has been the only one of its 280 scale [31] , which suggests that the wait until next large outbreak (i.e. an outbreak that 281 will most require vaccination) will be long. 282 283 Figure 4 shows the proportion of cases averted as a function of efficacy, duration, wait 284 between vaccination and the next outbreak, and the effectiveness of camel control 285 measures. Figure S2 shows the equivalent plots for the proportion of deaths averted. 286

The success of a proactive campaign increases with the vaccine's efficacy and 287 duration of protection, and decreases with the wait until next outbreak. The wait 288 between vaccination and outbreak is largely irrelevant if duration is long (e.g. 20 289 years), whereas the duration matters far less for short waits, and so here success is a 290 function primarily of efficacy. In any case, low efficacies (e.g. ≤25%) struggle to make 291 any impact, achieving at most a 31% (95% CrI: 18% -51%) reduction in cases and 292 23% (95% CrI: 10% -42%) reduction in deaths. Short durations (≤ 2 years) similarly 293 struggle unless waits are short (<1 year) and efficacy is at least moderate (e.g. ≥50%). 294 295 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

Adding measures targeting the animal reservoir can make large differences to 304 proactive campaigns [ Figure 4 ]. For example, 30% effective camel controls would 305 improve the above optimistic scenario (90% efficacy, 20-year mean duration and 6 306 months wait) to 75% (95% CrI: 63% -87%) of cases and 66% (95% CrI: 49% -83%) 307 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint of deaths. 50% effective camel control measures would improve this further to 82% 308 (95% CrI: 69% -92%) of cases and 75% (95% CrI: 57% -90%) of deaths. Trends with 309 efficacy, duration and wait time hold with the addition of camels, and the uncertainty 310 in modelling camels in isolation is also present in combination with proactive 311 campaigns. Lower credible intervals [ Figure S3 ] show substantially less effective 312 campaigns, whereas upper credible intervals [ Figure S4 ] practically eliminate the 313 epidemic for most values of efficacy, duration, wait time and camel control 314 effectiveness that we considered. 315 316

In the likely event that vaccine efficacy wanes over time, even for a high efficacy and 318 duration, a proactive campaign is still dependent on there being a sufficiently short 319 wait until the next outbreak, at least in the absence of widespread and effective camel 320 control measures. 321 322 Under a reactive campaign, an outbreak is already underway and so neither the wait 323 until the next outbreak, nor the vaccine's duration of protection are relevant. However, 324 the react time between the first case of an outbreak and the implementation of a 325 vaccination campaign will determine how a vaccine will fare. We model react times in 326 2-day intervals between 0 and 28 days and assume it takes 14 days for the vaccine to 327 elicit an immune response. 328 329 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ;

In the ideal reactive scenario, with a 100% efficacious vaccine with instant 330 implementation, vaccinating all healthcare workers in response to the first case at 331 hospital level would avert 59% (95% CrI: 51% -68%) of cases and 48% (95% CrI: 332 38% -58%) of deaths. Since healthcare workers constituted only 28% of cases, this 333 discrepancy illustrates the disproportionate effect of removing downstream cases 334 [ Figure 5 ]. 335 336 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint 

If a react time of 14 days is assumed, impact falls to 53% (95% CrI: 43% -62%) of 344 cases and 42% (95% CrI: 32% -51%) of deaths, and for a 28-day react time falls 345 further to 51% (95% CrI: 41% -61%) of cases and 41% (95% CrI: 30% -51%) of 346 deaths. A vaccine efficacy of 50% with react times zero, 14 and 28 days would 347 respectively reduce cases by 41% (95% CrI: 28% -55%), 36% (95% CrI: 24% -51%), 348 20 and 35% (95% CrI: 22% -50%) [ Figure 5 ], and deaths by 32% (95% CrI: 18% -46%), 349 28% (95% CrI: 16% -42%), and 27% (95% CrI: 15% -42%) [ Figure Greater impact can be achieved where a campaign reacts at regional level [ Figure 5 ], 356 where react times matter less than for hospital-level reactions. A perfect vaccine 357 deployed instantaneously to healthcare workers would achieve a 69% (95% CrI: 61% 358 -77%) reduction in cases and a 55% (95% CrI: 44% -67%) reduction in deaths. 359

Assuming a 14-day react time, this falls to 66% (95% CrI: 58% -74%) of cases and 360 51% (95% CrI: 41% -62%) of deaths, and a 28-day react time reduces 65% (95% CrI: 361 58% -73%) of cases and 50% (95% CrI: 41% -62%) of deaths. Therefore, the react 362 time makes less difference than at hospital level. Reacting at national level offers little 363 further improvement [ Figure 5 ], although interestingly the impact is the same 364 regardless of whether the react time takes zero, 14 or 28 days, reducing cases by 69% 365 (95% CrI: 61% -77%) and deaths by 55% (95% CrI: 45% -67%) in each instance. 366

The above national-level reductions are the best that can be achieved from reactive 368 campaigns without camel control measures. For 30% effective camel controls, this 369 maximum impact increases to 78% (95% CrI: 68% -88%) of cases and 68% (95% CrI: 370 54% -84%) of deaths averted, whereas with 50% effective camel controls, 84% (95% 371

CrI: 74% -93%) of cases and 78% (95% CrI: 62% -91%) of deaths can be averted. 372 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The gains that can be achieved from reacting at regional or national level versus 376 hospital level vary by efficacy and react time [ Figure S8 ]. On average, across all 377 efficacies and react times considered, regional and national level reactions offer a 31% 378 and 34% improvement over hospital level reactions. The level at which a reactive 379 campaign takes place is more important for longer react times. For example, there are 380 greater differences (and ratios) between regional and hospital level reactions for an 381 implementation that takes 28-day react time than for an instantaneous implementation. 382

Broadly, the ratio of cases averted between reactive campaign levels decreases with 383 efficacy, although this is mostly due to the limited impact of hospital-based reactive 384 campaigns with low efficacy. However, the absolute difference increases with efficacy. 385

At very long react times, national offers a slight improvement over regional and is 386 approximately 5% better for efficacies above e.g. 60% [ Figure S8 ]. Trends are largely 387 the same for the number of deaths averted. 388 389 Proactive vs. reactive campaigns 390 Figure 6 shows the ratios of cases averted between proactive and reactive campaigns. 391

If vaccine efficacy does not wane, then the wait time until the next outbreak is 392 irrelevant, and so proactive campaigns will always outperform reactive campaigns. 393

Otherwise, a regional level reactive campaign with a 28-day react time is far superior 394 to a proactive campaign, except where both vaccine duration is long (≥15 years) and 395 the wait is short (≤1 year). National-level reactive campaigns were superior to 396 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Figure 6 ], and more often if the react time is reduced to say 8 days [ Figure S9 ]. 400 401 Figure 6 . Each plot shows the ratio of cases averted between proactive and reactive (regional) 408 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint relatively insensitive to camel control measure effectiveness [ Figure S10 ], although as 411 their effectiveness increases, there is less proportional difference between proactive 412 and reactive campaigns. Trends further hold when considering the ratio of deaths 413 averted [ Figure S11 ], although hospital-and regional-level campaigns outperform 414 proactive slightly less often. 415

There is an asymmetry in the relative performance of proactive and reactive 417 campaigns (if the vaccine does not retain its efficacy permanently). Assuming 10% 418 vaccine efficacy, proactive campaigns can at best avert 39% more cases than 419 hospital-level reactive campaigns, or 31% more cases assuming 90% efficacy. 420 However, regional and national reactive campaigns of a 10% efficacious vaccine can 421 respectively avert approximately 97 and 100 times more cases than a proactive 422 campaign, or 421 and 444 times more if vaccine efficacy of 90% is assumed, albeit 423 with the important caveat than the very poor performance of proactive campaigns with 424 short duration and long wait results in very small numbers being compared. transmission type (nosocomial, regional, national and reservoir). Otherwise, the 432 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint 24 change in transmission proportions with efficacy is highly similar between strategies.

between approximately 45% and 50%, depending on the specific strategy. 435

Transmission within hospital, region and nationally goes down and therefore relative 436 contribution of animal reservoir goes up (approximately 12% of cases without 437 vaccination rising to as much as 30% depending on the strategy) [ Figure S12 ]. Trends 438 in absolute case numbers also do not differ markedly between vaccination strategies 439 [ Figure S13 ], although regional reactive campaigns avert more nosocomial, regional 440 and national transmission than hospital level reactive campaigns. 441 442 Sensitivity analyses

We investigated the sensitivity of our results to the choice of vaccine waning model. 444

Our main analysis considers waning of immunity to be exponential. However it may 445 be that a slower decline with a sigmoidal function, would be more appropriate. We 446 therefore reran our analysis of proactive campaigns using a Hill function [ Figure 1 ], 447 considering the vaccine's half-life as opposed to its mean duration. 448

The relationships between the proportion of cases averted and efficacy, duration (in 450 this instance half-life), and the wait until the next outbreak are largely the same as for 451

our default exponential waning model [ Figure S14 ], although longer half-lives (>5 452 years) perform slightly better and shorter half-lives (<5 years) perform slightly worse. 453

In general though, the predicted impact of proactive campaigns is marginally greater 454 when considering sigmoidal waning. 455 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint campaigns ostensibly avert many times the cases than reactive campaigns, there are 482 very few cases to avert in the first place, particularly among healthcare workers [ Figure  483 2]. Therefore, in situations where there are most cases where a vaccine is most 484 needed, reactive policies at national level are best. 485 486

No vaccine against MERS-CoV has yet been licensed in humans. If and when such a 488 vaccine becomes available, determining its optimal deployment is nontrivial. In this 489 study, we analysed multiple vaccine campaign strategies as a function of efficacy and 490 duration. Each strategy was evaluated by estimating multiple transmission trees (who-491 infected-whom), and then "pruning" them to generate counterfactual epidemics to 492 determine the number of cases and deaths that a vaccine would prevent. Our intention 493 is that all strategies considered could at least in principle be implemented, and 494 therefore that our analysis will be relevant for policymakers. We considered 495 vaccination of healthcare workers only, as they will be most easily vaccinated and 496 most exposed, and therefore more cost-effective. 497

We considered the relative merits of proactive and reactive campaigns, where the 499 fundamental difference between the two approaches is whether to vaccinate in 500 anticipation of the next outbreak, or in response to the current outbreak. The success 501 of a proactive campaign is a function of vaccine efficacy, duration and the wait time 502 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint between vaccination and the next outbreak. Whereas for reactive campaigns, success 503 is dependent on efficacy, react time and the spatial level at which a vaccination 504 campaign reacts: in response to a hospital, regional or national outbreak. In all 505 scenarios examined, vaccination has a greater effect on cases than deaths, likely 506 because healthcare workers firstly constitute only 27% of cases, and secondly 507 because they are younger than non-healthcare workers (mean age 39 vs. 51 years) 508 and so are probably healthier. 509 510 Short durations or long waits (or both) greatly diminish the impact of a proactive 511 campaign. While the wait time until the next major outbreak cannot be known in 512 advance, still less its magnitude, given that the 2013-2014 KSA MERS-CoV outbreak 513 was the only one of its scale [31], it is reasonable to think that the wait until the next 514 large outbreak will be long. In contrast, we have not modelled reactive campaigns to 515 depend on the vaccine's duration, and so vaccinees are conferred maximum possible 516 benefit, provided that vaccines can be administered and elicit an immune response 517 before people would otherwise be infected. 518 519 Therefore, the spatial scale at which a campaign reacts is crucial. If each hospital 520 reacts individually to its own outbreak, many cases (and their secondary cases) occur 521 before vaccination or immunity. However, campaigns that react at regional or national 522 level suffer far less from these delays and therefore can avert many more cases than 523 their proactive equivalents, even where the time until the next outbreak is short and 524 durations are long (although proactive campaigns are always better than reactive if 525 vaccine efficacy does not wane). Interestingly, the relative performance of reactive and 526 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint 28 proactive campaigns does not depend on efficacy, and the introduction of measures 527 to limit transmission from the animal reservoir does not affect the rank order of 528 campaigns. Essentially, regionally and nationally reactive campaigns offer an 529 opportunity to get ahead of the epidemic, and can be viewed as a proactive campaign 530 but with a more certain and shorter wait time. 531 532 Our analysis is reasonably robust to whether vaccine waning is exponential or 533 sigmoidal, but more sensitive to the choice of timeframe, in that hospital-level reactive 534 campaigns are rarely superior to proactive. However, our main conclusions firstly that 535 the spatial scale at which a vaccination campaign reacts is crucial, and secondly that 536 nationally reactive campaigns are campaigns are the most effective way to reduce 537 MERS-CoV case numbers and deaths, are strengthened. Further, focussing on the 538 impact on a single smaller outbreak slightly misses the point: we are ultimately most 539 interested in the maximum possible morbidity and mortality reductions over the widest 540 possible timeframe. In effect this blurs the distinction between reactive and proactive 541 campaigns: a reactive campaign against one outbreak can also be considered as a 542 proactive campaign against a subsequent outbreak. 543

We are aware of some limitations in our analysis. We assume that all downstream 545 cases of a successfully inoculated person are deleted, and this is unlikely to be 546 completely true for two reasons. First, if downstream cases had not been infected by 547 their index case, they may still have been infected through another route. Second, the 548 vaccine may have differing efficacies against disease and transmission. A vaccine that 549 inoculates against disease may not stop transmission or vice-versa. Additionally, we 550 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 have any data to inform such analysis, and so any attempt to do so would be overly 554 speculative. 555 556 It is also possible that some sub-clinical infections were not detected, and are therefore 557 missing from our line list. If such cases contributed meaningfully to transmission, then 558 our results could be biased upwards, and arguably more so for reactive campaigns, 559 as reactive campaigns might not react to index case in a hospital or region or country. 560

We have also not accounted for any behavioural change or risk compensation in 561 response to an available vaccine. 562

In considering camel control measures, we have assumed only a simple proportional 564 reduction in contribution from animal reservoir, without specifying what this would 565 entail (e.g. vaccination, better hygiene, or reduced physical contact between humans 566 and camels), and clearly additional data to inform more precise analysis would be 567 helpful. 568 569 To reduce the otherwise prohibitively large number of simulations, we have assumed 570 no vaccine waning under reactive campaigns. However, unless duration was very 571

short the effects of waning would be negligible, and in this instance, waning would still 572 affect reactive campaigns less than proactive campaigns. On the other hand, we 573 assume zero immunity until 14 days post vaccination, whereas in practice there would 574 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ;  30 be at least some protection prior to this. Greater delays between vaccination and 575 immunity would affect our results, but not the trends we describe, and would 576 essentially constitute a relabelling (e.g. if the time was 10 or 18 days, the react times 577 we list must be decreased or increased respectively by 4 days). 578 579 Because MERS-CoV outbreaks are relatively infrequent, traditional randomised 580 controlled trials may not be feasible [35] , and therefore vaccine efficacy or its wider 581 effectiveness may be difficult to measure empirically, and this is even more so with the 582 vaccine's duration of protection. It is therefore useful to have an indication of the most 583 effective strategies even if values of efficacy and duration are unknown. Unless the 584 vaccine maintains its efficacy for a long time (>20 years) a reactive campaign at 585 regional or national level will usually be superior to a proactive campaign. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint 

Initial vaccine efficacy (VE) Between 5% and 100% in intervals of 5% 1 year to 10 years, and 6 months.

Spatial reaction level for reactive campaign i) hospital; ii) regional; iii) national React time (τI) 0 to 28 days in 2-day intervals, between start of (hospital, regional or national) outbreak and vaccination Time between vaccination and immunity (τP) 14 days (zero immunity assumed between 0 and 13 days post vaccination) 615 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 Gamma distributed and its mean and standard deviation are inferred from the data. where E0 is the initial number of reservoir infections at the beginning of the study period 729 (i.e. January 1, 2013), and α is the (positive or negative) growth rate. 730 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 , : 

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; 

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. show the ratios of and differences between the proportions of cases averted, by react 799 time and efficacy. Left, middle and right columns show: i) regional vs. hospital; ii) 800 national vs. hospital; iii) national vs. regional. 801 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint Figure S9 . As per Figure 6 , but assuming an 8-day react time. 803 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint Figure S10 . As per Figure 6 , but assuming 30% effective camel control measures in 805 tandem with both proactive and reactive campaigns. 806 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint Figure S11 . As per Figure 6 , but showing ratio of deaths averted, as opposed to 808

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. proactive campaign with 30% effective camel control measures. Reactive campaigns 819 assume 28-day react time. Proactive campaigns assume 5-year mean duration and 820 6 months wait until next outbreak. 821 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint Figure S13 . Change in absolute contributions of transmission types with efficacy. As 823 per Figure S12 but showing absolute case numbers by transmission type and 824 efficacy / effectiveness, not relative contributions. 825 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint Figure S15 . As per Figure 6 , but where vaccine efficacy wanes sigmoidally, and not 832 exponentially as in our main analysis. Note that the y-axes denote vaccine efficacy 833 half-life, as opposed to mean duration. 834 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted April 6, 2022. ; https://doi.org/10.1101/2022.04.06.22273497 doi: medRxiv preprint

MERS transmission and risk factors: a systematic review

Progress 635 on the global response, remaining challenges and the way forward

Preliminary epidemiological 638 assessment of MERS-CoV outbreak in South Korea

Infection in the Kingdom of Saudi Arabia

Estimates of the severity of 645 coronavirus disease 2019: a model-based analysis

Asymptomatic and presymptomatic transmission of SARS-CoV-2: A

Waning 675 of BNT162b2 vaccine protection against SARS-CoV-2 infection in Qatar

Duration of 677 effectiveness of vaccines against SARS-CoV-2 infection and COVID-19 Disease: results of a 678 systematic review and meta-regression

Duration of Protection 680 against Mild and Severe Disease by Covid-19 Vaccines

Waning 31. WHO Middle East respiratory syndrome coronavirus (MERS-CoV) 2022