1. IntroductionBotulism is an infectious disease caused by the activity of the neurotoxin produced mainly by the anaerobic, Gram-positive and spore-forming bacterium Clostridium botulinum, which can be found in the environment, such as the soil and dust. These exotoxins consist of two protomers (an active part in the light chain and a host cell binding part in the heavy chain) linked by a disulfide bridge [1,2]. Botulinum toxins bind to the presynaptic membrane by a double receptor mechanism [3], blocking acetylcholine release at the neuromuscular synapse, causing a progressive symmetrical and descending flaccid paralysis that can be fatal by asphyxia due to diaphragmatic arrest [4]. To date, several different isoforms, grouped into seven toxin serotypes, have been described; from A to G, but only A, B, E and, less frequently, F can cause human botulism [5]. In addition, some new chimeric toxins (C/D, D/C and F5A) have been described recently [6,7]. Although C. botulinum is the most common species producing botulinum toxins, toxin type E is also released by C. butyricum and toxin type F is usually related to C. baratii [8,9]. In addition, neurotoxin-like botulinum has been found in Chryseobacterium piperi, Weisella oryzae and Enterococcus spp. strains [10,11,12,13,14]. Botulism can be acquired by ingestion of improperly processed food containing pre-formed toxin (food-borne botulism); by toxin release at the intestinal level when microbiome and/or intestinal peristalsis cannot prevent spore germination and gut colonization, as happens mainly in children below 1 year of age (infant or intestinal botulism); by contamination of wounds by spores which develop to vegetative forms and then produce the toxin (wound botulism); or, less frequently, due to iatrogenic injection [15] or bioterrorism. Severe cases of all ages usually require ventilation for respiratory failure and admission to the Intensive Care Unit (ICU). Food-borne botulism is found worldwide, related to canned/preserved food, especially home-canned food [16,17], in which the anaerobic environment allows C. botulinum to develop. Since C. botulinum does not grow in food products with high acidity (pH 16]. Regarding infant botulism, its relation to honey, where spores can be preserved, is well documented, but there are several possible sources, such as dust [17]. Consequently, national and regional public health primary prevention actions aim to reduce these known risk factors by promoting good practices during home-canning, reducing the production of home-canned food that is not reheated before consumption, and avoiding the consumption or use of honey by children under 1 year of age [18,19,20]. Feces, serum and/or gastric juice are required for laboratory diagnosis; taking into account that toxin detection in serum is less probable than in feces, and that the serum sample must be collected before antitoxin administration [21,22].The treatment of choice consists of the administration of antitoxin as soon as possible. Different equine antitoxin formulations including two (AB) or three (ABE, available from 2008 to 2018) serotypes have been used, and, since 2018, despeciated heptavalent Botulism Antitoxin (BAT) [23,24], which includes all serotypes, is used in Spanish hospitals for patients of all ages. From 2007, human antitoxin including serotypes A and B (BabyBIG), which has a longer half-life and requires a lower dose than BAT, has been used as the treatment of choice for infant botulism [25,26].According to the latest ECDC annual report [27], with data from 2015, and the ECDC Atlas with 2021 data [28], botulism in Europe has remained stable over the last decade. The reporting rate was 29]. Case information is stored on an electronic platform managed by the National Centre for Epidemiology.

This study aims to answer questions related to the epidemiological characteristics of botulism in Spain, including those associated with the outcomes of antitoxin administration, in order to increase early medical investigation, improve patient management and review the main risk factors identified and its trend to reduce or avoid them.

3. DiscussionThe epidemiological data show similarities with those previously published for the European Region [27] in terms of the notification rate, 21], or around 5% in the US, where type A toxin is more frequent [32,33]. This low fatality rate makes it—fortunately—difficult to assess its relationship with the delay in antitoxin administration. Spanish botulism cases are mostly food-borne, similar to what happens in France and Italy [21,34], but contrary to what happens in the US [35]. In our series, most of the cases occurred in males aged 25 to 80 years and females aged 45 to 80 years, and infants 27] but with differences in the 15 to 24 year age group, in which the data for both sexes were similar in our series.Homemade preserved foods are the main source of food-borne disease, similarly to the previously published data [36]. However, it is noteworthy that fish products were implicated in almost one-third of the cases in which the most suspected food was commercial and were the commercial foodstuff with the highest number of positive cases.Though the toxin type was missing in more than 50% of the food-borne cases (typing can only be performed on feces) in Spain, as in Europe in general, the type B neurotoxin predominates [37], followed, by far, by toxin type A. The only outbreak caused by type E neurotoxin was linked to salted fish products [38,39,40], and all the cases in the type F neurotoxin outbreak required ICU admission due to rapid clinical progression [16,31], as seen in the literature previously.In our series of infant botulism cases, the source was found positive only in one case; it was honey, which is the main food item related to infant botulism described in the literature [41,42,43]. The absence of infant botulism cases related to honey from 2010 may show the success of information campaigns carried out by local public health authorities and pediatricians [44]. However, we should note that honey was among the components of an enema that was the main suspected source of infection in a recent case in Spain (out of the study period), and in a previous case included in this paper, which resembles another possible case reported in Malaysia and published recently, in which the author considers the topical application of medical-grade honey on a wound as the most likely cause of infection [45]. In our series, the suspected food item most mentioned was herbal tea, even if C. botulinum was not detected in this food vehicle. Due to the nature of herbal tea, the presence of C. botulinum spores is not surprising, although this finding is not very frequent, and cross-contamination once the package is opened cannot be excluded easily. Bianco et al. from Argentina detected spores of C. botulinum in 7.5% of chamomile samples, being higher in products sold by weight than in tea bags [46]. More studies are needed to determine the role of herbal tea in infant botulism but, for caution, the Spanish Food Safety and Nutrition Agency recommends avoiding its consumption by children under one year of age [18].According to our results among food-borne cases, early antitoxin administration from the onset of symptoms (≤48 h) is associated with a higher probability of a short stay in hospital, while late administration or no antitoxin administration do not differ substantially in terms of the length of hospitalization. It should be noted that IQR in the late administration group is almost double that in the early administration group, so we can expect some kind of benefit in those patients who received antitoxin closer to the cut-off point. As in our case, recent USA studies have shown a relationship between early antitoxin administration and a shorter hospitalization compared to late administration [23,47]. Even if the general recommendation is to use antitoxin in all cases of botulism, the latest CDC review concludes that patients with mild symptoms and no clinical progression, who have not been treated within the first two days, are unlikely to improve with antitoxin administration, particularly after seven days from the onset of symptoms [36], linking the lack of clinical progression to the absence of the toxin in the bloodstream. From an economic perspective, Anderson et al. quantified the mean extra costs derived from the delay in the administration (>2 days) of antitoxin as 2.5 times higher than early administration [48]. Considering that, in our series, more than half of the patients were admitted around 48 h from the onset of symptoms or later, the window of opportunity is very short, and clinicians from emergency wards must be aware of the need to act fast.It was not possible to develop a specific analysis comparing the length of ICU stays, due to the frequent lack of data on the dates of admission and discharge in these units. However, it should be mentioned that some clinical guidelines used in Spain recommend admission to the ICU in any case with a high suspicion of botulism [49,50]. This could detract from the association between ICU admission and severity since, in some cases, ICU admission responds to the need for constant monitoring for the early detection of symptoms of worsening respiratory function, explaining the lack of differences in the mean and median length of hospitalization between cases with severity criteria and those without them. Furthermore, it is well known that the severity of symptoms is proportional to the dose of toxin [36], and this information exceeds the data routinely collected for case studies, being a major limitation.Regarding the classification of cases, the low level of confirmed cases may be associated with the limitations of the bioassay [14], especially with serum samples, in which botulinum toxin is detected for a shorter time than in feces, particularly in infants [14,51], as well as delays in sample collection, especially feces. For this reason, the confirmed cases of food-borne botulism could be those with more suggestive symptoms and those more serious cases, in which the botulinum toxin is detectable in the serum for a longer time than those with negative results, even if they ate the same contaminated food, explaining a prolonged hospitalization of the confirmed cases. We could not analyze the percentage of positivity among the different samples because, in many cases, only one sample was tested, which is usually serum because it is easier to take and constipation can complicate feces collection. Moreover, in some cases, the information about the nature of samples sent to the laboratory was not collected in the survey or just reflected that with a positive result.On the other hand, the higher frequency of outbreaks during the summer months explains the higher proportion of probable cases in this season. We decided not to perform a statistical analysis to assess seasonality due to the low number of events, so the test does not have the required statistical power; however, the frequency of cases shows a high difference between seasons. Similarly, in the USA, the three months with the highest number of reported cases are June to August [33]. As this is a retrospective study, we cannot explain the reason for this higher incidence during the third quarter of the year, but we can propose some ideas, such as the difficulty of keeping unrefrigerated preserved food in a cool place and a probable increase in the consumption of non-heated or non-reheated food. After consulting the monthly data for several years on food consumption habits in Spain [52], we could not find an increase in the consumption of home-canned food in summer compared to the rest of the year.All the commercialized antitoxins for botulism treatment in adults are of an equine origin, which increases the risk of anaphylactic reaction compared to non-equine products [26]. In Spain, no adverse events have been reported through the official pharmacovigilance system, but the online reporting program does not recognize botulinum antitoxin as a drug, forcing manual reporting, with all the classification difficulties and mistakes that this kind of notification entails. None of the cases in our series had diagnostic codes for adverse events to any antitoxin or drug (note that there is not a specific ICD code for botulism antitoxin), but the 72-year-old male case detected in our analysis was also mentioned in a short, previously published article [53]. Therefore, we conclude that, to date, there are no data regarding the frequency of adverse events related to botulism antitoxin in Spain. In our series, we found four cases with confirmed or probable adverse events according to epidemiological surveys and/or an ICD code analysis, giving an anaphylaxis rate of 1.5% and an overall adverse events rate of 3%. These data are in agreement with those published for the current antitoxin (BAT), at around 2.8% and 54,55], although all these cases were treated with a previous antitoxin, thus no adverse events have been detected with BAT. However, we did not include some possible adverse events, such as nausea, dizziness or hemodynamic alterations, as it was impossible to distinguish them from the botulism symptoms or the patients’ comorbidities.The first case of infant botulism treated with human antitoxin (BabyBIG) in the European Region took place in Spain in 2007 [56], and, until 2019, it was administered to four more cases without incident. Two other infants were treated with the equine-derived antitoxin with no adverse events reported. To the best of our knowledge, there are few recommendations to prioritize one antitoxin over the other, including in Spain, and we were not able to find any study comparing these two antitoxins. While the CDC recommends human antitoxin because of its safety profile and longer half-life (note that intestinal colonization can last up to three months [57]), the Norwegian Institute of Public Health recommends BAT because it is stocked and less expensive [42,58]. Due to the difficulties in performing a case–control study, we consider it important to share all experiences with both antitoxins. Previous studies detected a reduction in the length of hospitalization in early treated patients [59]. In our series, a delay in antitoxin administration (all the cases were treated >48 h after the onset of symptoms, including one case treated on day +20 [60]), along with a higher level of severity in the treated cases and the low number of total cases, could explain the lack of difference in the duration of hospitalization versus the no-treatment cases. In addition, the absence of infant botulism cases reported in some years is notable and could reflect an under-diagnosis or under-notification.The absence of wound botulism may suggest a low level of clinical awareness among medical staff. During this period, several wound botulism cases were detected in neighboring countries, including an international outbreak among injecting drug users [27].

The main limitation of the study is the nature of the data collected, which are useful information for the epidemiological and clinical management of each specific case, but which were not collected for a specific study. There are also differences in data collection from different regional public health services and hospitals. As previously mentioned, delays in sample collection notably diminish the chances of getting a confirmed diagnosis, as well as the lack of feces samples, where the toxin can be detected over a longer period than in serum, and typing can be performed. As the majority of the “not specified” toxin type cases probably belong to subtype B, and only 13 cases were related to other toxin types, we did not analyze the differences according to the toxin type, which is a relevant limitation. Another limitation arises from the assumption that survival probabilities are the same for all patients throughout the study period, although it is known that healthcare systems have improved over this time. However, as antitoxin effectiveness is high for all the antitoxins used in this period, is among the untreated cases where differences due to better management in the more recent cases could be significant.

5. Material and MethodsThis retrospective study was conducted using epidemiological surveillance data from the RENAVE database. To increase the sensitivity and completeness of the data, they were supplemented with the Minimum Basic Data Set (MBDS) of hospital discharge records. Thus, all the cases between 1997 and 2019, registered at hospital discharge and according to the ninth and tenth editions of the International Classification of Diseases (ICD-9 and ICD-10), with any diagnostic of botulism, were reviewed. Due to the low incidence of botulism, it was possible to link each case from the RENAVE database to the MBDS database using basic demographic information (date of birth, sex and place of residence) and the date and place of hospitalization. All the non-coincident data were requested from the regional public health services for review and, in cases of a mismatch, the information from epidemiological surveys (RENAVE) was prioritized. The cases without epidemiological survey data (absence in the RENAVE database) that could not be confirmed by regional public health authorities were excluded. The latest European case definition [61] (the same as the national one) was used to classify the confirmed and probable cases, and the national definition for the possible cases [25] was included (Table 1). Only cases with a positive result in the clinical samples were considered confirmed; for this purpose, serum, prior to antitoxin treatment, and feces samples were collected and analyzed using the standard methods (Table 1). The cases that did not meet the criteria for classification were excluded. The cases requiring ICU admission, ventilation or death were considered severe. The length of hospitalization was calculated as days between admission and discharge dates, collapsing consecutive stays (≤2 day between the first discharge and the next admission date) and transfers into a single episode. For food-borne cases, the incubation period was established as days between the date of consumption of the suspected food and the onset of symptoms.

The ICD codes of other pathologies compatible by clinical symptoms, and often part of the differential diagnosis, were searched to assess the specificity of the included cases in the final dataset. These differential diagnoses included myasthenia gravis, Guillain–Barré and Miller Fisher syndrome, Lamber–Eaton syndrome, stroke-related pathology, poliomyelitis, Wernicke encephalopathy, methanol poisoning and tick-borne encephalitis.

An epidemiological description of the cases was performed, assessing the differences in case classification related to categorical variables using the chi-square test. Logistic regression was used to assess severity by sex, age, whether or not they received antitoxin and the delay until its administration and, for the outbreak-related cases, whether it was the first case detected or a subsequent case. For time series analysis, the frequency of cases per year and the quarter of year were calculated.

To assess the differences in the length of hospitalization based on the time of antitoxin administration, the Cox proportional hazards test and Kruskal–Wallis test were used after checking for non-normality of the data (skewness and kurtosis tests for normality of the length of hospitalization p-value= 0.0051). The time to antitoxin administration was classified according to the clinical guideline recommendations [36] in three groups: “early administration” when used within the first 48 h after symptom onset, “later administration” when administered more than 48 h after symptoms onset, and “non-administration” when it was clear that the patient did not receive the antitoxin treatment. Those cases who received antitoxin, but the date was not specified, and were admitted ≥72 h after the symptoms onset were classified in the “later administration” group. We used 72 h instead of 48 h as the breakpoint in order to avoid the misclassification of cases that may have been treated in the emergency ward before getting a hospital admission, even if using 48 h as the breakpoint found similar results. We consider planned discharge as the endpoint and right-censored the observations from the rest (including death, discharge against medical advice, transfer to other unit or center for further management and unknown reason for discharge). The cases receiving an antitoxin that did not match the toxin detected in the clinical samples were considered untreated [36]. When no toxin was detected in the clinical samples or typing was not performed, but patients received an antitoxin, they were considered treated, as the likelihood of matching is higher.

The following symptoms and syndromes related to probable adverse events were searched: urticarial, anaphylaxis, iatrogenic hypotension, allergic angioedema and any antitoxin poisoning. This information was complemented with the epidemiological surveys and public data from the Spanish Medicines and Medical Products Agency (AEMPS, by its acronym in Spanish) which manages the system for drug adverse events notification in Spain. Long-term adverse events were not assessed due to the nature of the available information.

The epidemiological analyses were performed using Stata®BE17 and Excell 2016. The results with two-sided p-values < 0.05 were considered statistically significant.