A review of typhoid fever transmission dynamic models and economic evaluations of vaccination

Despite a recommendation by the World Health Organization (WHO) that typhoid vaccines be considered for the control of endemic disease and outbreaks, programmatic use remains limited. Transmission models and economic evaluation may be informative in decision making about vaccine programme introductions and their role alongside other control measures. A literature search found few typhoid transmission models or economic evaluations relative to analyses of other infectious diseases of similar or lower health burden. Modelling suggests vaccines alone are unlikely to eliminate endemic disease in the short to medium term without measures to reduce transmission from asymptomatic carriage. The single identified data-fitted transmission model of typhoid vaccination suggests vaccines can reduce disease burden substantially when introduced programmatically but that indirect protection depends on the relative contribution of carriage to transmission in a given setting. This is an important source of epidemiological uncertainty, alongside the extent and nature of natural immunity. Economic evaluations suggest that typhoid vaccination can be cost-saving to health services if incidence is extremely high and cost-effective in other high-incidence situations, when compared to WHO norms. Targeting vaccination to the highest incidence age-groups is likely to improve cost-effectiveness substantially. Economic perspective and vaccine costs substantially affect estimates, with disease incidence, case-fatality rates, and vaccine efficacy over time also important determinants of cost-effectiveness and sources of uncertainty. Static economic models may under-estimate benefits of typhoid vaccination by omitting indirect protection. Typhoid fever transmission models currently require per-setting epidemiological parameterisation to inform their use in economic evaluation, which may limit their generalisability. We found no economic evaluation based on transmission dynamic modelling, and no economic evaluation of typhoid vaccination against interventions such as improvements in sanitation or hygiene. © 2015 Published by Elsevier Ltd.

Modelling suggests vaccines alone are unlikely to eliminate endemic disease in the short to medium term without measures to reduce transmission from asymptomatic carriage. The single identified data-fitted transmission model of typhoid vaccination suggests vaccines can reduce disease burden substantially when introduced programmatically but that indirect protection depends on the relative contribution of carriage to transmission in a given setting. This is an important source of epidemiological uncertainty, alongside the extent and nature of natural immunity.
Economic evaluations suggest that typhoid vaccination can be cost-saving to health services if incidence is extremely high and cost-effective in other high-incidence situations, when compared to WHO norms. Targeting vaccination to the highest incidence age-groups is likely to improve cost-effectiveness substantially. Economic perspective and vaccine costs substantially affect estimates, with disease incidence, case-fatality rates, and vaccine efficacy over time also important determinants of cost-effectiveness and sources of uncertainty. Static economic models may under-estimate benefits of typhoid vaccination by omitting indirect protection.
Typhoid fever transmission models currently require per-setting epidemiological parameterisation to inform their use in economic evaluation, which may limit their generalisability. We found no economic evaluation based on transmission dynamic modelling, and no economic evaluation of typhoid vaccination against interventions such as improvements in sanitation or hygiene. © 2015 Published by Elsevier Ltd.

Introduction
Typhoid fever is an exclusively human enterically transmitted systemic disease caused by infection with the bacterium Salmonella enterica enterica serovar Typhi. Although largely controlled in Europe and North America, typhoid remains endemic in many parts of the world, notably Asia, where it is an important cause of febrile illness in crowded, low-income settings [1]. A notable feature of typhoid is the carrier state -asymptomatically infected individuals who continue to shed Salmonella Typhi in their stool or urine for many years, thereby sustaining transmission [2].
Despite a recommendation by the World Health Organization in 2008 that typhoid vaccination be considered for the control of endemic disease and outbreaks, programmatic use remains limited [3].
In the early twentieth-century, public health officials were debating the best methods of evaluating typhoid vaccine effectiveness, and whether vaccination was a distraction from improvements in sanitation and hygiene [4]. These remain contemporary policy issues for ministries of health and other health partners who may be considering programmatic anti-typhoid vaccination as a counterpart to other anti-typhoid measures such as improvements to income distributions, sanitation, water supplies and hand washing with soap (post-defecation and before the preparation of food in the home or sold in the street) as well as identification and management of carriage [5][6][7][8]. Transmission dynamic modelling and economic evaluation are two informative tools to support such decisions [9,10].
Where health budgets are limited, allocation of resources to activities which generate the best value for money maximises the population's health (not withstanding other health programme criteria such as equity). To compare between and across health states, cost utility analysis (CUA) can be employed using a common metric of health, such as disability-adjusted life-year (DALY). The World Health Organization's Choosing Interventions that are Cost-Effective project (WHO-CHOICE) describes interventions as "cost-effective" if they add a DALY at a cost of less than three times Gross Domestic Product (GDP) per capita, and "highly costeffective" if each DALY costs less than GDP per capita. These are arbitrary thresholds and meeting them does not necessarily lead to the intervention being adopted, as health decision-makers are often required to make choices between multiple interventions that fall below these thresholds. Furthermore, even highly costeffective activities may be too expensive overall for a health service to provide within budget: a hypothetical drug adding a year of life and costing GDP per capita for each person treated would require the entire national economy to be spent giving the drug to every member of the population [11].
By building on the germ theory of disease, and mass-action principles from the physical sciences [12], mechanistic mathematical modelling enables extrapolation beyond observed data, and can be used to project the expected trends of disease in a population or the potential impact of control strategies such as vaccination. Through capturing indirect effects of immunisation -the reduced incidence of disease in members of a population not themselves immunised, commonly described as "herd immunity" -these transmission dynamic models capture the impact of such interventions more completely than static economic models measuring only the direct effects in vaccinees [13].
In this review, typhoid transmission dynamic models and typhoid vaccine economic evaluations are examined for their potential contributions to informing disease control, identification of gaps in knowledge and indication of directions for further research.

Methods
PubMed was searched on 23 October 2014 without date restriction using the following terms: ("Typhoid Fever" [ Studies were included if they modelled typhoid transmission and/or analysed the cost-effectiveness of vaccination in endemic settings. Endemic settings were identified using recent high-quality reviews [14,15]. We included cost of illness (COI) studies if they were linked to an analytical study, and willingness-to pay (WTP) studies if they included an economic evaluation or were linked to an analytical study. Studies were excluded if they used geographical or statistical modelling, including time-series analysis, without transmission dynamics, or if they addressed transmission or cost-effectiveness in non-endemic populations, such as international travellers.
Transmission models were assessed for their model structure, data sources, parameter estimates, use of fitting methods, sensitivity analysis and the contribution of their approach to epidemiological understanding of typhoid. Economic studies were evaluated by data sources, economic evaluation approach, perspective, comparator programmes, use of sensitivity analysis and capture of indirect effects of vaccination.

Results
Seventy-nine titles were retrieved. Ten modelling papers were selected for review based on title or abstract. One was discarded as a non-mechanistic time series study [16], one as it modelled outbreaks in a non-endemic setting [17], while two papers were of the same model and considered together [18,19]. These are summarised in Tables 1 and 2. A further, as yet unpublished, transmission model has been developed by the International Vaccine Institute (personal communication, Jin Kyung Park) and is not reviewed here.
Seven titles were identified as economic evaluation and obtained for full-text review (Tables 3 and 4) alongside two underpinning COI studies and one underpinning WTP study (Table 5). A further COI study was identified but excluded as not linked to a published economic evaluation [40].
There was minimal overlap found between transmission modelling and economic evaluation. Of the transmission dynamic models, only those by the Cvjetanović group also had costeffectiveness components [18][19][20]. One economic study included quantitative consideration of indirect protection [27].

Transmission dynamic models
The seven typhoid models identified range from two-state analytical tools to complex individual-based simulation or multi-state compartmental models (see Table 1). Only two models were formally fitted to data [24,25].
The structures of models (Tables 2a and 2b) are based on different assessments or representations of the natural history of typhoid fever, particularly in how immunity to Salmonella Typhi is    Primary or subclinical infection C; Carrier; Ct; C l Temporary; long-term R; Removed/resistant/refractory/recovered; Not able to be infected, immune. Rt; R l ; Temporary immunity; long-term immunity; Clinical immunity is against disease but allows infection and onward transmission. Rc; Rs Natural immunity to clinical disease; natural sterile immunity) Sterile immunity is against any infection.

V Vaccinated Vc; Vs
Vaccine-induced immunity to clinical disease; Vaccine-induced sterile immunity W 'Water' Long-cycle transmission from water or environmental contamination, contributed to by all infectious or carrier classes. considered. González-Guzmán suggests that natural partial immunity is likely to arise but simplifies to a model with vaccine immunity only, noting that sufficiently high infectious doses can overcome immunity [21]. Pitzer uses population compartments to separately represent immunity against typhoid infection ('sterile immunity'), and immunity against typhoid disease ('clinical immunity'), allowing transition from the latter to either full susceptibility or to a subclinical infection that in turn restores full sterile immunity in the individual. This corresponds to immunity boosting repeated infection cycles without overt disease, particularly in adults after recovery from clinical disease in childhood, and allows bacterial shedding during subclinical infection to contribute to sustained transmission [24]. Saul similarly models both sterile and clinical immunity, with infection resulting in sterile immunity that wanes to clinical immunity (potentially after zero time) and then to susceptibility, and explores a range of hypothetical state-transition scenarios based on multiple infections, though he does not clearly resolve a most-likely scenario [25]. Despite asymptomatic boosting being a longstanding hypothesis, or perhaps because of it, there is a paucity of data from microbiological, immunological or epidemiological studies to parameterise models or to validate assumptions [41].
While noting leaky immunity in those naturally infected (each individual has a reduced probability of further infection), González-Guzmán models Ty21a oral vaccine protection as all-or-none, giving each vaccinee a probability of developing immunity or not. In this model, those who develop immunity following vaccination have 100% protection against typhoid, until vaccine wanes and they return to full susceptibility, an approach also applied by Cvjetanović [18][19][20][21]. Pitzer handles injected Vi vaccination the same, noting results were similar in a sensitivity analysis assuming leaky vaccine immunity. Pitzer represents oral Ty21a vaccination as akin to natural immunity, transitioning vaccinees to clinical immunity after full immunity wanes [24].
While vaccination programmes are predicted to reduce typhoid incidence, uncertainty around carriage prevalence, duration and contribution to the force of infection substantially affects vaccines' projected impact [21,24,25]. In reviewing Cvjetanović's 1978 model [19], Anderson and May observe that the implicit assumption that the effective contact rate for carriers is equal to that of acute cases, combined with other fixed parameter estimates, gives carriers a contribution to transmission ten times that of other cases [42]. While illustrative of the potential contribution of carriage in sustaining disease, for policymaking it has been recommended that such assumptions should be tested against data [43,44]. An approach might be to conduct systematic, detailed investigation of incidence cases to identify potential sources, using suitable screening methods to look for carriers as well as acute cases, and combine this with population-level carriage surveys and water quality studies. While labour-intensive, such investigations could be integrated into wider control efforts [45].
Chronic Salmonella Typhi carriage can be treated with antibiotics and/or cholecystectomy for gallstone-associated infection, but there is no demonstrated role for vaccination in clearance of carriage [46,47]. Premised on this, Cvjetanović's and Pitzer's models demonstrate that where carriage contributes substantially to transmission in endemic settings, vaccination is unlikely to result in elimination in the short-to-medium term [18,24]. Similarly, where carriage rates are high the indirect protective effects of vaccination are reduced, as the risk to the unvaccinated of acquiring disease from carriers is not diminished [24,42]. The contribution of carriage, however, requires further epidemiological assessment, as does the role of short-cycle and long-cycle (environmental) transmission [21,24,25].
If immunisation of the susceptible population does not bring about typhoid elimination, then measures to reduce per case or  CBA, cost-benefit analysis; CE, cost-effective(ness); CEA, cost-effectiveness analysis; CFR, case-fatality rate (proportion of cases that result in death); CUA, cost-utility analysis; COI, cost of illness; DALY, Disability adjusted life-year; DALY weight, a scale from 0 (perfect health) to 1 (death). DOMI, Diseases of the most impoverished programme [39]; PAHO, Pan-American Health Organization; SIREVA, Sistema Regional de Vacunas (Regional Vaccine System); U5MR, under-five mortality rate; USD, United States Dollars; VE, vaccine effectiveness; WTP, willingness to pay. per carrier infectivity, such as improved sanitation or hand washing with soap, might be considered instead of or in conjunction with vaccination [48,49]. The multi-compartment models suggest such a reduction in effective contact rates could lead to important reduction in prevalence [21,23,24]. This is consistent with Briscoe's analytical model [22]. Another feature of transmission dynamic studies is that the average age of infection increases as the force of infection decreases, for example, with the introduction of vaccine [20]. This is consistent with burden of disease studies which find earlier average age of infection in settings with higher disease incidence [14,35].

Economic evaluation
Our literature search found seven papers evaluating typhoid vaccine cost effectiveness. The earliest two of which were based on values derived from expert opinion and are less informative to current policy considerations than the most recent five which were based on field studies, as outlined in Tables 3 and 4. Two supporting COI studies and one supporting WTP study are outlined in Table 5. These field-informed analyses share multiple common authorships with collaboration through the Diseases of the Most Impoverished (DOMI) programme. Of the seven economic evaluations, four included a cost-benefit analysis (CBA), two a cost-effectiveness analysis (CEA) and two a cost-utility analysis (CUA). Four of these used a societal perspective [28,30,31,33]. Only one evaluation considers indirect protection quantitatively, but uses hypothetical values for herd immunity from different coverage levels rather than estimates from dynamical modelling [27]. No studies considered improvements in sanitation, hygiene or water supply as an alternative or adjunct to typhoid vaccination.
In a one-year study of a very-high typhoid incidence area -Kalkaji slum, Delhi -Bahl found average costs per episode of illness were high to both the health sector and families, excluding intangible costs such as pain, with hospitalisation an important component of health service costs [35]. A CBA by Poulos on this data reported that a vaccination programme for children under five years of age would be cost-saving to the public sector. Analysis from a societal perspective, incorporating private costs, indicated that vaccination in high incidence settings with modestly priced vaccines could also have net benefits in other age groups [31].
Cook and colleagues conducted a CUA based on field data from multiple Asian sites [30,38], and found that while typhoid vaccination using the Vi-polysaccharide across adults and children would be unlikely to be cost saving to the public sector, in high incidence settings it was likely such a programme would meet the standard for "very cost-effective" health interventions of less than per-capita gross domestic product (GDP) per DALY gained. In these settings, targeting vaccination to the highest incidence age-groups improved cost-effectiveness substantially. Through sensitivity analysis, the main determinants of cost-effectiveness identified were vaccine cost, case-fatality rates (CFR), vaccine duration of protection, (baseline) incidence and vaccine efficacy. Cost-effectiveness was insensitive to vaccine coverage as no indirect protection was assumed [30].
WTP has also been used in economic evaluation of typhoid vaccines as an alternative approach to COI in valuing private or societal benefits. Such an approach is considered to demonstrate the value individuals place on the total benefit of the vaccine, though is confounded by ability to pay, and ability to value public sector activities foregone if vaccines are supplied through the state [50]. One study from Hue, Vietnam suggested that typhoid vaccination would pass a social cost-benefit test (total costs less than total societal benefit), based on demand estimates from a contingent valuation survey addressing hypothetical vaccine purchases for householders and their children [29]. Analysis by Cook of a similar WTP study done in Kolkata, India, by Whittington and colleagues, suggested that vaccination of children or all-ages would not pass a social cost-benefit test using COI, but costing benefits using WTP plus public costs would likely find that such programmes pass such a test [28,36].
WTP studies are also informative to vaccine uptake, with 9% of respondents in the Kolkata survey stating they would not accept a free vaccine, with data suggesting these individuals are more likely to be older, have lower income and never boil drinking water [36]. In the Hue survey, Canh suggests a number of issues affect validity of economic evaluation using WTP estimates, noting that householders were most sensitive to price, with proposed vaccine efficacy making no detectable difference to individual or household demand at a given price [29].
Observing that typhoid vaccines are equivalent to one-sixth of per-capita public sector health spending in India, Cook notes the potential for user fees in financing a state-administered programme [28]. Whittington's Kolkata WTP survey suggested vaccine protection for children was given greater value than vaccinating adults [36]. Drawing on data from this study, a price optimisation model by Lauria of different vaccine prices for children and adults did not find a strong case for differential pricing, but in a non-dynamic sensitivity analysis of potential indirect protection scenarios found herd immunity to be a significant influence on incidence and costeffectiveness [27].

Discussion
This review found a relatively sparse literature on typhoid modelling and vaccine economic evaluations. Of seven transmission models found, only two were published in the last 25 years and use contemporary data-fitting methods. All five field-based economic evaluations we found shared multiple common authorships around the DOMI collaboration.
Although the Global Burden of Disease is a much-criticised ranking, it provides some comparator infectious diseases, with measles and syphilis ranked close to typhoid and paratyphoid fever, and cholera attributed around one-third of the annual number of DALYs [51]. Repeating our (non-comprehensive) PubMed typhoid search strategy for these returned over six times as many measles papers, three times as many syphilis papers, and twice as many cholera papers. Typhoid fever transmission and economic evaluation appear relatively under-studied. Typhoid's low profile-to-burden ratio has variously been attributed to ubiquity in developing countries, inadequate diagnostic tools, the absence of champions in health agencies, and affecting mostly the poor and the underclasses [52]. Effective antibiotics and previous non-availability of a long-acting infant vaccine have also been cited as reasons for the absence of internationally funded typhoid vaccination programmes [52]. For budget-constrained national health agencies, typhoid vaccination programmes may appear unattractive unless public sector cost saving can be demonstrated, for example, if the incidence rate is very high [31,35].
Considering only a health services perspective -of treating typhoid and providing vaccine -omits private costs associated with the disease and therefore underestimates the societal impact of typhoid. Various approaches have been taken to address these costs and more completely capture the benefit of vaccination, from cost of illness studies for private expenses, estimates of health utility forgone due to illness, or by determining the extent to which people value vaccination in willingness-to-pay studies. While user fees and WTP are controversial [10,50], economic evaluations that consider these have been consistent with typhoid economic analysis using more widely accepted cost-utility analysis and private cost of illness. These CUAs suggest typhoid vaccine programmes are likely to be highly cost-effective (against international norms) where disease is highly endemic, particularly when targeted to the age-groups at highest risk of disease. The economic analyses, which do not include mechanistic components for transmission, emphasise as key drivers of cost-effectiveness the vaccine cost, case fatality rates, baseline incidence, vaccine efficacy and duration of protection.
We did not find any economic evaluation comparing typhoid vaccination programmes against other potential means of typhoid control, such as measures to improve sanitation, hygiene or water supplies. It is still not possible to answer the century-old question of whether, in a given setting and within a limited budget, vaccination should be adopted over improved sanitation and hygiene, or what combinations are optimal for control under what circumstances. Such analysis would need to be based on a transmission dynamic model, with extensive epidemiological surveillance for detailed burden of illness measurement, and comprehensive costing for CUA or CBA approaches that allow other diseases to be included in the evaluation.
The only published data-fitted transmission model of typhoid vaccination suggests that while vaccination is effective in reducing disease incidence, if other measures are not enacted to reduce the ongoing force of infection, particularly from asymptomatic carriage, short or medium-term vaccination campaigns are unlikely to result in elimination and would see disease rebound if vaccination stopped [24]. The authors of recent dynamic models emphasise our lack of understanding of certain aspects of the natural history of typhoid (particularly around acquisition of immunity, the role of carriers, and the contribution of short-and long-transmission cycles) [21,24,25]. A model intended to examine what role the putative different forms of Salmonella Typhi immunity have in determining typhoid incidence rates found an absence of suitable immuno-epidemiological data on which to fit parameters and make strong inferences, a challenge further compounded by the absence of age structure in the model [25].
Transmission dynamic modelling and a non-mechanistic economic analysis have shown that the level of indirect protection may have an important impact on vaccine effectiveness and cost-effectiveness respectively [24,27]. None of the economic models mechanistically consider disease dynamics and so cannot scientifically appraise the indirect effects of vaccination in costeffectiveness calculations. Indeed, while a number of economic analyses readily acknowledged indirect effects as an important phenomenon, they specifically excluded them, citing the absence of evidence for Vi polysaccharide vaccine herd immunity pending the publication of cluster randomised controlled trials [28,30,53,54]. Early work by Cvjetanović is the only meeting point we found of mechanistic typhoid transmission modelling and economic analysis, but the complexity of this model and absence of fitting make it difficult to apply findings to contemporary disease control problems [18,19].
Using a static economic model premised only on direct protection in vaccinees may be a reasonable approximation in some situations, such as if vaccine-preventable new typhoid cases (symptomatic or otherwise) make a relatively small contribution to the force of infection compared with carriers and the unimmunised. Even in such circumstances, prior assessment with mechanistic modelling of field epidemiological data would be appropriate in estimating the relative contributions of each group to transmission.
While there seems limited inter-disciplinary dialogue between typhoid modellers and economists, a unifying concern is the importance of accurately determining age-based incidence rates, which can be highly variable within-country or between otherwise similar settings, and are central to estimates of vaccine impact and cost-effectiveness. Heterogeneity in disease rates, transmission mechanisms and health service provision may limit external validity of both typhoid modelling studies and economic evaluations.
Accurate assessment of disease burden could be done with large, population-based studies to inform incidence, complications and case-fatality rates, using blood culture confirmation of cases, or altogether improved diagnostics [30,[55][56][57]. It should be noted that even well-conducted studies are unlikely to provide unbiased estimates, due to the positive health consequences of introducing disease surveillance. Bahl notes that active surveillance with early treatment gave rise to disease that is less severe, and less expensive, than disease detected through passive surveillance [35]. This is echoed in the DOMI disease burden study, which rather than the 1% case-fatality rate widely cited in literature, had a zero percent CFR amongst the 475 cases detected (which gives an upper 95% confidence interval of around 0.63%) [30,37,58].
Other field epidemiology and laboratory investigations could further inform typhoid transmission dynamics [59]: • Large, population-based, Salmonella Typhi carriage studies, similar to those done in Chile in the early 1980s [5], to determine prevalence in a range of endemic settings, potentially with serological surveys [60], alongside investigation of potential sources of infection amongst new cases. • Serological assay development and population-based serological surveys to determine past infection to Salmonella Typhi, natural immunity and waning of this immunity. Seroprevalence could be linked to surveillance records to estimate the proportion of infections that are clinically apparent and notified to national authorities. • Epidemiological time series with consistent, transparent methodology and/or cross-referencing between methods [61]. • Age-based social contact pattern surveys, which may inform short-cycle transmission [62].
For models to assess vaccination against other enteric fever control measures, findings could be incorporated from interventional field studies on the role of improvements to sanitation, hygiene and water supply in changing disease incidence and transmission. Ongoing scrutiny of vaccine efficacy and duration of vaccine protection may also be informative. Estimates of efficacy of Ty21a vaccines in recent systematic reviews are less than in the data sources for early modelling, with the reviews focusing on individual RCTs rather than cluster field studies [20,[63][64][65][66]. Estimates for Vipolysaccharide effectiveness have also been modified downward [31,64]. Analysis of differences between cluster and individual randomised trials may be informative on indirect protection. This review has a number of limitations. In the absence of licensed human vaccines, the review does not cover paratyphoid fever or non-typhoidal salmonelloses. It covers only material in the English language, limited searches to a non-systematic enquiry of a single database and does not attempt to synthesis qualitatively or quantitatively any of the studies reviewed. Comparator studies have not been sought that consider investments in water, sanitation and hygiene as alternatives to typhoid vaccination.
One possible feature observed in the course of this review is a less pessimistic assessment of disease burden, perhaps reflecting true decline, as well as a more sceptical perspective on vaccine efficacy estimates, with fewer inputs based on expert opinion alone. Any such trend towards assessment of vaccine costs and benefit firmly grounded in data is beneficial to equitable, scientifically informed health-policy setting.
It has been suggested that for a model to have sufficient complexity to enable robust cost-effectiveness analysis, substantial data collection may be required [67]. When data is in short supply, theoretically informed modelling may still be a particularly appropriate tool to support decision-making [68,69]. Transmission modelling using existing data explains patterns seen in average age of infection, demonstrates the importance of carriage, suggests optimal strategies for vaccination, and appraises the potential role of other interventions to reduce transmissibility.

Conclusion
Transmission dynamics have not yet been integrated into a comprehensive cost-utility analysis of typhoid vaccination and, as such, there is no economic evaluation that would meet contemporary gold standards [9,70]. Given the costs and time involved in further field study, constructive efforts could be made to integrate existing transmission modelling and cost-effectiveness analyses, such as utilising the extensive collation of typhoid epidemiological and clinical parameters by Saul [25] with the transparent, reproducible modelling approach of Pitzer [24], and DOMI project economic data [30,38]. While such endeavours would not address the fundamental limitations on health service budget in endemic areas, an analysis of typhoid vaccination that enables economic comparison across health arenas could help bring into the public gaze the full potential of measures to control enteric fever, and improve the prospects of protection from typhoid for people living with daily risks from a disease eliminated from most of the affluent world.

Ethics
Literature review, ethics approval not required.

Conflict of interest
CHW and WJE have had travel and expenses paid for by the Coalition against Typhoid to attend meetings on the modelling of typhoid vaccination programmes. WJE has undertaken consultancy for the Coalition against Typhoid, which was paid to a fund held the London School of Hygiene & Tropical Medicine.