Incubation period of typhoidal salmonellosis: a systematic review and meta-analysis of outbreaks and experimental studies occurring over the last century

Typhoid fever is a systemic infection caused by the bacterium Salmonella enterica subsp. enterica serotype Typhi (S. Typhi). It has been a major human pathogen for thousands of years, and while incidence has greatly reduced in developed countries, it remains a major cause of morbidity and mortality in developing countries [1]. Laboratory diagnosis of typhoid fever involves culture and identification through biochemical and serological means. Widal agglutination test was a common method until its diagnostic capabilities were questioned [2].

The mode of transmission is faecal-oral via ingestion of contaminated food and water or contact with an infected carrier [3]. Although it is a gastrointestinal infection, diarrhoea and vomiting are atypical symptoms of typhoid fever. People suffering from typhoid fever typically complain of systemic symptoms including gradual onset of fever, general malaise, headaches and abdominal pain [4].

Prevention of typhoid fever, like most other gastrointestinal infections, includes good hand hygiene [5] and exclusion of infected food handlers. In addition, typhoid vaccine has been known to offer some protection [6].

Although typhoid fever in developed countries is mostly a sporadic disease associated with travel to endemic countries, outbreaks are still commonly reported [7, 8]. Factors such as improper cooling, inadequate heating of food, contact with contaminated raw products and infected food handlers have been implicated as contributory factors of outbreaks [9].

Investigation of outbreaks contributes to the control and reduction of the burden of disease by identifying and eliminating the source of infection. Information obtained from observational studies of outbreak investigations, particularly outbreaks with well-defined point of exposures, as well as experimental studies provide a means of studying and understanding the disease by studying certain parameters in real-life and under laboratory conditions.

Incubation period of typhoid fever, which is the time between exposure and onset of clinical symptoms, is one of the essential parameters to be studied. It is expected to be a distribution rather than a single estimate as a result of individual variation based on factors such as dose response, food matrix and host characteristics. An accurate knowledge of the incubation period distribution of typhoid fever is relevant for surveillance and implementation of public health interventions. It also contributes towards effective outbreak investigation as incorrect estimations can be misleading if they were used to determine the time of exposure [10]. Accurate knowledge of the incubation period helps in correctly classifying primary and secondary cases and exclusion of travel related cases. The incubation period also offers insights into the pathophysiology of typhoid fever and is important in conducting epidemiological and ecological studies [11].

Despite the importance of knowing the accurate incubation period distribution of typhoid fever, large organisations give a wide range with no clear indication of which durations are common and which are rare. According to the World Health Organisation (WHO), incubation periods ranging from three to sixty days have been reported [3]. The Centres for Disease Control and Prevention (CDC) reported the incubation period as three to thirty days [12].

In this study, we have systematically reviewed the literature for outbreaks of well-defined point source exposures and human experimental studies of typhoid fever with the aim of estimating the distribution of the incubation period and identify factors that may explain any variation observed. We extracted summary estimates and individual incubation period data reported, tested for the presence of heterogeneity, identified factors that may contribute to heterogeneity and defined subgroups of studies than can be combined for analysis.

We have undertaken a systematic review to describe the distribution of incubation period of Salmonella Typhi. We tested for the presence of heterogeneity, attempted to identify factors that could influence the incubation period and classified the studies into subgroups with limited evidence of heterogeneity. Due to the high heterogeneity amongst the studies, estimating a single distribution of incubation period was impossible. However, we defined subgroups of similar studies and found their mean incubation period to range from 9.7 days to 21.2 days. These values are within the range quoted by international organisations reporting 3 to 60 days [3], 3 to 30 days [12] and also scientific literatures reporting 10 to 20 days [18]. Some literature report shorter incubation periods than we observed, like Raffatellu et al. reporting a median of 5 to 9 days [19] or 8 to 14 days [18]. We also identified attack rate and vaccination as possible factors that could affect the incubation period distribution.

The relationship between attack rate and incubation period is inversely proportional, such that higher attack rates result in shorter incubation periods [20, 21]. This has been reported in numerous outbreaks [22‐24] and also observed in our review. Attack rates are associated with factors such as virulence of the organism, host characteristics and infecting dose [22]. When either of these factors are present in a way that increases the attack rate: high virulence, susceptible host and large infecting dose, the incubation period is shortened as the onset of illness is quicker.

Vaccination is one of the preventive measures advised by public health professionals [25, 26] to control or mitigate the burden of disease brought on by S. Typhi [27]. It is recommended to people living in or visiting endemic areas and to control outbreaks [3]. Our study did not aim to review the efficacy of typhoid vaccine in preventing outbreaks, however we identified a couple of studies with outbreaks reported in vaccinated population [28, 29] and similar studies have also been reported [30]. As observed from our review, we may say that in populations where vaccination does not prevent illness, it may delay the onset of symptoms thereby prolonging the incubation period. This could have an adverse effect in outbreaks with continuous source as more people will be exposed and possibly infected before the outbreak is discovered.

The subgroup analysis indicated that outbreaks and experimental studies that took place before the year 1950 belonged to subgroups reporting shorter incubation period (subgroups 3 and 4). In the univariate analysis, post-1950 studies had a significantly longer incubation period, however, in the multivariate analysis, this effect on the incubation period was reversed as we observed a shorter incubation period in post-1950 studies. The change in the direction of effect is indicative of the presence of a confounding factor in the univariate analysis.

A lot of reported outbreaks caused by S. Typhi either had a continuous source of exposure such as municipal water supply [31‐34], or were caused by widely distributed food produce [35, 36]. This means that point source outbreaks, where the time of exposure is definite and incubation period could be accurately calculated, were few. This was a similar finding in the review conducted by Naylor [24]. The author stated the criteria of an outbreak suitable for incubation period analysis and reported that very few reports fulfilled these criteria. However, in their review, studies where the date of exposure was not confirmed were included and date of infection was inferred from date of purchase, date of sale for a batch of food item, date of water contamination and so on. Continuous source outbreaks can be useful if the investigators identify and report the exact dates of exposure and onset but this was not the case in the outbreaks reviewed by Naylor. Hence, some studies identified in Naylor’s review were not included in our study. In our study, we identified only 13 studies with point source exposures or where the time of exposure was known. Some studies however reported point source exposures, although the exposures could have occurred over a few days [37, 38].

Dose response is known to be associated with incubation period [20, 39], and although this was not particularly studied in our review, there were studies were this relationship was observed. In an outbreak associated with ice cream [17], a dose response relationship was observed as those who had a mixture of ice creams had longer incubation periods while those who had only the contaminated ice cream had comparatively shorter incubation periods. Another dose response relationship was observed in the experimental study conducted by Waddington et al. [21]. Two cohorts of volunteers were infected with different doses of inoculum, 10³ and 10⁴. The cohort receiving the lower dose had a longer incubation period compared to those receiving the higher dose.

The case definition used in identifying cases or in including cases as part of an outbreak is very subjective and based on what the authors identify as a case. This is quite essential because it can either prolong or shorten the incubation period of a case based on what is considered as the onset of symptoms. Although there are guidelines on what to consider when defining a case [40], case definitions depend on the authors’ perception. The complex syndrome of typhoid fever [7] makes the process of defining cases arbitrary. In our review, some of the case definitions were broad and syndromic, such as ‘clinical illness’, while others are very specific in terms of fever at a particular temperature [21, 29, 41] or dismissing symptoms such as diarrhoea and vomiting [17]. Neither the broad nor specific definitions are incorrect, however, the varied case definitions may have contributed to the observed heterogeneity.

A limitation we encountered in this review was the difference in the reported incubation period data extracted from the included studies. Although the raw incubation times were reported in days, some studies reported an aggregated date of symptom onset in three or seven-day intervals [42, 43]. Onset times were either rounded up or rounded down to fall into the reported interval. Rounding down the onset time from three days to one will spuriously shorten the incubation period just as rounding up the onset time from one day to seven will spuriously prolong the incubation period. We therefore excluded these studies from the hierarchical cluster analysis and their reported mean incubation periods should be interpreted with caution.

Our study showed that the reported incubation period varied more than could be explained by chance and identified attack rate and previous vaccination as associated factors, although not significantly. This may be due to the limited information available from the small number of studies we reviewed. Analysing individual patient data will provide an opportunity to assess additional patient characteristics such as underlying medical conditions, drug interactions and full vaccination history. In addition, using mathematical models to describe the process of infection will help identify parameters intrinsic to the infection pathway that could influence the distribution of incubation period such as gastric transit times, phagocyte characteristics and cellular bacteria growth.

In order to maximise the use of outbreak reports in studying parameters such as incubation period, authors should endeavour to precisely report exposure times and onset times. Where the outbreak is non-point source, identifying and reporting the exposure time of each case where possible will be useful. Case definitions should be developed in such a way as to ensure that cases are included based on similar thresholds which can also be comparable with other outbreak reports.