- 1 Fluorine is the first element of the halogen group in the periodic table (atomic number 9, relative (...)
1Fluoride, the ion-derived form of the element fluorine,1 is an essential element for human health, and specifically for the formation of bones and teeth. Good health depends on a delicate balance of this element in the body: on the one hand, too little fluoride can lead to poor bone mineralisation and a high prevalence of dental caries. On the other hand, too much fluoride can result in pathological conditions such dental and skeletal fluorosis (WHO, 1992; 2019; Zohoori and Duckworth, 2017). The most frequent signs of dental fluorosis include developmental hypoplasia defects, staining, and surface irregularities in the enamel (Littleton, 1999). In contrast, skeletal fluorosis is a chronic metabolic bone disease that is caused by prolonged and excessive ingestion of fluoride. It affects a variety of cells involved in bone turnover, resulting in abnormal bone densification (Teotia et al., 1971; Boivin et al., 1989; Littleton, 1999; Petrone et al., 2013; Wei et al., 2019). Clinically, early symptoms can include stiffness and pain in the joints. Severe cases can be crippling and are associated with osteosclerosis, calcification of tendons and ligaments and bone deformities (Teotia et al., 1971; 1986; Teotia and Teotia, 1988; Boivin et al., 1989; Littleton, 1999; Petrone et al., 2013; Nelson, 2015; Nelson et al., 2016; 2019; Zohoori and Duckworth, 2017; Zuo et al., 2018; Wei et al., 2019; WHO, 2019; Walser et al., 2020). The condition also causes extensive sub-periosteal and endosteal bony accretion, often accompanied by increased resorption of the old cortex on the long bones, making the bones weaker than normal and increasing the risk of fractures (Ortner, 2003; Brickley and Ives, 2008b; Petrone et al., 2013; Nelson et al., 2016). The condition is cumulative and gradual (WHO, 2019), and the pathological bone modifications it causes may decrease overall life expectancy (Walser et al., 2020). Prolonged fluoride intake (exceeding ten years) can lead to mild skeletal changes, with fluoride accumulation slowing and reaching a plateau after approximately 55 years of age (Ayoob and Gupta, 2006; Ponikvar, 2008). If fluoride intake ceases, the reduction of fluoride levels in bone occurs gradually (Khairnar et al., 2015). However, there are still few published data on the relationship between bone fluoride levels, age and fluoride concentrations in water.
2In the environment, fluoride is naturally present in soils, water, air and dust. It can also be present as a result of human activities such as industrial pollution (e.g. aluminium industry) or as an artificial additive in foodstuffs, water or dental care products (WHO, 1992; Ayoob and Gupta, 2006; Ghosh et al., 2013; Zohoori and Duckworth, 2017). Despite this wide variety of sources, water is considered to be the main source of fluoride, with concentrations varying between geographic regions and influenced by parameters such as pH, temperature and water hardness (WHO, 1984; Ayoob and Gupta, 2006; Barbier et al., 2010; Zohoori and Duckworth, 2017; Chowdhury et al., 2019). Groundwater in volcanic regions is particularly rich in fluoride. Hydrogen fluoride is one of the most soluble gases in magma, and is partially released into the environment during eruptive activity. It is also released with fumarole gases and by degassing from volcanic systems, then dissolving when it comes into contact with water (D’Alessandro, 2006; Yeşilnacar et al., 2016; Sahu, 2019). As a consequence, while fluoride concentrations in unpolluted surface waters are usually between 0.01-0.3 mg/l, groundwater in volcanic areas and in areas with geological deposits of marine origin usually contain high fluoride concentrations, up to 30-50 mg/l (Barbier et al., 2010; Shamsollahi et al., 2015; Zohoori and Duckworth, 2017). These figures are higher than World Health Organization (WHO) recommendations for fluoride intake in drinking water, which should not exceed 1.5 mg/l, because higher doses could lead to poisoning (WHO, 1992; 2019).
3In modern times, high levels of fluoride in groundwater, and the resulting fluorosis, have been raising significant environmental and public health challenges (Ghosh et al., 2013; WHO, 2019; Srivastava and Flora, 2020; Shaji et al., 2024). Skeletal fluorosis affects millions of people worldwide and is prevalent in many volcanic areas (WHO, 1992; D’Alessandro, 2006; Barbier et al., 2010), particularly those associated with the five fluoride belts. These belts include, for example, Turkey and the East African Rift in Belt 1; Northern Africa in Belt 2; India, Iran, Iraq, and China in Belt 3; Southern Europe, the United States, and Latin America in Belt 4; Japan and Indonesia in Belt 5 (for more information see Chowdhury et al., 2019).
4As for past populations, a few studies have demonstrated the presence of skeletal fluorosis in different geographic regions and in a variety of chrono-cultural and geological contexts, including areas of volcanic activity (Callaghan, 1986; Littleton, 1999; Weinstein, 2005; Yoshimura et al., 2006; Nelson, 2015; Nelson et al., 2016; 2019; Walser et al., 2020; Zhou et al., 2023). It should be noted, however, that these studies have mainly focused on non-cremated skeletal remains. As a result, no proper protocol has been proposed to diagnose the condition in cremated human remains, although they are the main source of data for some chrono-geographical contexts (notably in southern Europe during a large part of the Roman period until the end of the 2nd century AD).
- 2 A "secondary cremation burial" should be understood as follows: "a cremation burial is considered t (...)
5Our study addresses this issue by investigating the possible evidence of fluoride poisoning in the Campanian region of southern Italy during the Roman period (2nd century BC – 1st century AD) through palaeopathological observations and chemical analyses of osteological remains from secondary cremation burials.2 Even though Italy is often not considered to be part of the above-mentioned fluoride belts, there is modern evidence of endemic fluorosis in southern locations, such as Sicily (D’Alessandro, 2006) and Naples (Eager, 1901; Petrone et al., 2013). Moreover, ancient cases of fluorosis have already been identified in buried human remains in this region, both in the Somma-Vesuvius area at Herculaneum (Torino et al., 1995; Petrone et al., 2011; 2013; 2019) and at the site of Cumae in the Phlegraean Fields (Torino et al., 2012) (see the locations of the cited sites in figure 1). Recently, a few researchers have highlighted the possible presence of the condition in cremated remains preserved in the necropolis of Porta Nocera in Pompeii (Van Andringa et al., 2013; 2021) and the Necro- polis of Porta Mediana at Cumae (Duday, 2018; 2019), mainly based on observations of increased bone density in the remains of single burials. This study aims to examine this hypothesis for the second of these sites and to develop a detailed protocol to assess the presence of skeletal fluorosis in cremated human remains.
Figure 1
A) Map of Italy highlighting the Campanian region, where Cumae is located; B) Cumae, shown by a star, is situated amidst the volcanic area of the Phlegraean Fields to the north of the Gulf of Naples; to the south, Herculaneum and Pompeii are situated within the Somma-Vesuvius volcanic area (adapted from https://tessadem.com/) |
A) Carte de l’Italie mettant en évidence la région de Campanie, où se trouve Cumes; B) Au nord du golfe de Naples, Cumes, représentée par une étoile, est située au sein de la zone volcanique des Champs Phlégréens; Herculanum et Pompéi, au sud, sont situés dans la zone volcanique de Somma-Vésuve (modifié d’après https://tessadem.com/)
- 3 The pre-calderic period is defined as a phase of submarine volcanism, with less frequent eruptions (...)
6The Cumae archaeological site is located in the Phlegraean Fields, a volcanic region to the west of Naples in southern Italy (figure 1). Numerous eruptions have occurred over the history of this super-volcano, which is still active. These eruptive events have caused the formation of mostly single-edifice structures and the deposition of significant amounts of pyroclastic rocks along with occasional small-scale lava flows. Researchers have divided its volcanic activity into a pre-calderic period3 (ca. 80 to 60 Ky) characterised by the formation of multiple separate volcanic centres followed by two major volcanic events: the Campanian Ignimbrite eruption (39 Ky BP) and the Neapolitan Yellow Tuff (15 Ky BP). Contrary to the Somma-Vesuvio, also located in the Campanian volcanic arc, the eruptive activity of the Phlegraean Fields is spread over a very large area, in which numerous small volcanic systems have formed (Rosi et al., 1983; Bousquet, 2018; Forni et al., 2018; Stellato et al., 2020; Sbrana et al., 2021; Cappelletti et al., 2022). The Phlegraean Fields caldera also presents a unique phenomenon known as bradyseism, involving periodic vertical ground deformations over the centuries that have varied in scale from millimetres to metres. Explosive activity in this region is also characterised by water/magma interactions within a vast hydrothermal system (Adinolfi, 1978; Rosi et al., 1983; Di Vito et al., 1999; Orsi et al., 1999; Ciaramella et al., 2011; Calò and Tramelli, 2018). Due to this significant volcanic activity, the groundwater in the Phlegraean Fields is likely to contain elevated levels of fluoride, raising potential risks for both the environment and human health.
7Historically, Cumae is one of the oldest Magna Graecia settlements around the Western Mediterranean Sea. The site has seen more than 20 centuries of human occupation, from indigenous occupations during the Bronze Age, followed by the Greeks’ foundation in 740 BC and through to the abandon of the site between the 13th and 14th century AD. Around the second half of the 4th century BC, the Romans took control of the Campanian coasts. Cumae was then incorporated into Roman territory (civitas sine suffragio) and gained municipal status towards the end of the 1st century BC (Caputo et al., 1996; Munzi and Brun, 2011). Cumae’s boundaries are marked by a partially defined wall circuit, encompassing Monte di Cuma on the north-western side and Monte Grillo on the eastern side (Caputo et al., 1996; Munzi and Brun, 2011).
8The "Porta Mediana" necropolis, investigated by the Centre Jean Bérard since 2001, is located to the north-east of Monte di Cuma, just outside one of the northern city gates, also called the Porta Mediana. The excavations have revealed funerary practices attributed to different groups including Cumaeans, Greeks, Italics and Romans, thus confirming the continuous funerary function of this area from the period of indigenous settlements and the Greek colony until the early Middle Ages. The typology of the tomb architecture and the sepulchral characteristics display differences based on the chronology, cultures and religious beliefs of the communities that inhabited the settlement (Brun and Munzi, 2009; Munzi and Brun, 2011; Munzi, 2022).
- 4 Between the 2nd and 1st centuries BC, secondary cremation burials were deposited either in parallel (...)
9By the second half of the 2nd century BC and the final years of the Republic, hypogean monuments marking the burial places of Cumae elites were mainly surrounded by secondary cremation burials of lower-class individuals,4 arranged in multiple rows along roadsides. With the beginning of the Augustan age, mausoleums for the upper classes persisted while "cippo" tombs were gradually replaced by small "a dado" monuments with an aedicule façade and an underground funerary chamber (figure 2). The secondary cremation burials preserved in these monuments are mostly those of employees or freedmen of notable Cumae families. After the Augustan period, both cremation and inhumation continued, the former being slightly more prevalent. Most tombs built during the 1st century AD were columbaria intended to contain cremated remains in urns inserted into the walls. It was only after the 1st century AD that inhumation became the main funerary practice, in monuments as well as individual graves (Brun and Munzi, 2009; Brun et al., 2010; 2013; Munzi and Brun, 2011; Munzi et al., 2018; Munzi, 2022).
Figure 2
Tomb ENF34009, a small "a dado" funerary monument from the Augustan age, and an urn containing cremated remains CU34242 (photographs: CJB archives) |
Tombe ENF34009, petit monument funéraire "a dado" d’époque augustéenne et sépulture secondaire à crémation CU34242 (clichés : archives CJB)
10This study covers forty-five individual burials of cremated remains in the Necropolis of Porta Mediana at Cumae. Only adults (n=43) and older adolescents (n=2) were included in our sample (table 1), as the completion of development and ossification makes bones less porous and fluoride uptake therefore proceeds at a more regular pace than in children (Teotia and Teotia, 1988; Whitford, 1994; 1999; Ozsvath, 2009). All these cremated remains were placed in sealed ceramic urns. Previous chemical analyses conducted by Loeff (2018) on a sample from Cumae, which included pottery fragments of the urn, its cover, and the plaster used to seal the funerary urn, found no traces of fluoride in these components of the funerary structure. These findings rule out the possibility of post-depositional contamination of the cremated remains inside the urns.
Table 1
Sample information and types of analyses undertaken. Legend: Treatment (funerary treatment): B = burnt, Ub = unburnt. Origin: V = volcanic area, nV = non volcanic area. Sex estimation: Ind = indeterminate, F = female, F? = Probably female, M = male, M? = probably male, - = not identified. Final* = Sex estimation after taking the two criteria into consideration (coxae morphology + epitaph). Group of age: J = juvenile (16-20 years), yA = young adult (20-29 years), mA = middle-aged adult (30-59 years), oA = elderly adult (> 60 years), A = adult (when no further detail was available). Analysis: Macro. = macroscopic analysis (palaeopathological protocol), LIBS = Laser-induced breakdown spectroscopy |
Information de l’échantillon étudié et les types d’analyses entreprises. Légende : Traitement (traitement funéraire) : B = brûlé, Ub = non brûlé. Origine : V = zone volcanique, nV = zone non volcanique. Diagnose sexuelle : Ind= indéterminé, F = féminin, F? = féminin probable, M = masculin, M? = masculin probable, - = non identifié. Final* = Diagnose sexuelle après prise en compte deux critères (morphologie coxale + épitaphe). Classe d’âge : J = juvénile (16-20 ans), yA = jeune adulte (20-29 ans), mA = adulte moyen (30-59 ans), oA = adulte âgé (> 60 ans), A = adulte (lorsqu’il n’était pas possible de donner une meilleure précision). Analyse : Macro. = analyse macroscopique (protocole paléopathologique), LIBS = Spectroscopie d’émission atomique de plasma induit par laser
11Sex and age assessment was conducted previously by one of the authors of this study (HD). Sex estimation was based on sexually dimorphic traits of the os coxae, following Bruzek (2002). To aid sex estimation, information from epitaphs was also used when available. The sex distribution of our sample was determined from these two criteria, resulting in 31% male individuals (n=14), 4% probably male (n=2), 47% female (n=21), 9% probably female (n=4), and 9% of indeterminate sex (n=4). Age estimation was based on the stage of bone maturation and development of the teeth, when these observations were possible. For the juveniles (between 16 and 20, n=2), age was estimated according to observations of secondary ossification centres, following Scheuer and Black (2000). For the adults, age was estimated on the basis of age-related changes to the auricular surface, following Meindl and Lovejoy (1989). Adults were classified into three age groups: young adults (20-29 years) (n=11), middle-aged adults (30-59 years) (n=23), and elderly adults (> 60 years) (n=9) (figure 3).
Figure 3
Demography of the Cumae sample. For each age group, sex estimation is specified along with the percentage relative to the whole sample (n=45) |
Démographie de l’échantillon de Cumes. Pour chaque classe d’âge, la diagnose sexuelle est précisée ainsi que le pourcentage par rapport à l’échantillon complet (n=45)
12The palaeopathological observations involved a thorough study of all the remains from each individual. Following the recommendations of Turkekul and colleagues (2020), a femur fragment was selected from each individual for archaeometric analysis. Preference was given to fragments from the femur’s proximal third, particularly those containing the gluteal tuberosity when possible.
13In order to better assess fluoride enrichment in the Cumae sample, a control sample of six individuals from non-volcanic contexts was also analysed (table 1). Four of the samples were from a collective inhumation assemblage from the osteological Neolithic-Chalcolithic collection of La Caouno (Moux, Aude, southern France) (Gutherz, 1986). The other two individuals are cremated remains from an Augustan Roman necropolis at Classe (Ravenna, northeast Italy) (Scheid, 2008). These samples enabled us to assess whether there is a relevant difference in fluoride, firstly between volcanic and non-volcanic contexts, and secondly between non-cremated and cremated remains. Additionally, the comparison between the Cumae sample and the control sample provided a better understanding of variations in fluoride concentrations within the Cumae sample.
14Previous research (Duday 2018; 2019) revealed the possible presence of skeletal fluorosis in cremated individuals
15from the Cumae sample, and highlighted the need to build up a specific protocol for this kind of material. The approach proposed in this research is based on skeletal fluorosis lesions documented in previous palaeopathological studies (Littleton, 1999; Brickley and Ives, 2008a; Petrone et al., 2011; 2013; Nelson et al., 2016; 2019; Walser et al., 2020; Zhou et al., 2023) and modern cases reported in the clinical literature (Jolly et al., 1968; Teotia et al., 1971; Teotia and Teotia, 1988; Buchancová et al., 2008; Qin et al., 2009; Izuora et al., 2011; Gupta and Ayoob, 2016; Kurdi, 2016; Zuo et al., 2018; Fabreau et al., 2019; Guan et al., 2019; Cook et al., 2021; Joseph et al., 2022). Our protocol takes into account the most frequent of these pathological bone changes, all of which can also result from other pathological conditions (table 2).
- 5 Including the trochlear surface and capitulum of the humerus, the head of the radius, the trochlear (...)
- 6 Including the distal articular surface and ulna notch of the radius, and the radial circumferential (...)
16Pathological lesions were observed macroscopically in 12 different anatomical areas (figure 4): the cranial vault, vertebrae, ribs, sternum, glenohumeral joint, elbow joint,5 wrist,6 pelvic girdle, femur, patella, tibia and calcaneus. Teeth were excluded from our analysis since they are frequently broken and enamel tends to shatter due to the high temperatures and rapid desiccation that occur during cremation. Each of the regions considered can show one or more types of bone lesions (table 2). Each lesion was recorded as "present", "absent" or "non-observable". Given the fragmented and incomplete nature of cremated remains, at least 50% of the joint area or ligament insertion needed to be visible, even if spread across multiple fragments, in order to meet the criteria for observability.
Table 2
Differential diagnoses of pathological signs analysed for this study (modified from Nelson 2015 and Nelson et al. 2016; 2019) |
Diagnostic différentiel des signes pathologiques analysés pour cette recherche (modifié d’après Nelson 2015 et Nelson et al. 2016 ; 2019)
Figure 4
Anatomical zones considered in our protocol and examples of bone lesions observed in the Cumae sample. Arrows indicate bone modifications. A) Cranial vault fragment from CU34283; detail showing abnormal diploe thickness; B) Rib fragment from CU37044 exhibiting excessive bone formation, ossification of intercostal muscles on the superior and inferior margins, and porous surface alteration (a: exothoracic side, b: endothoracic side); C) Sternum fragment from CU39306 displaying cartilage ossification, particularly in the visceral surface and the articular facets; D) Cervicothoracic portion of the spinal column from CU37044 illustrating osteophytic development and excessive cartilage ossification on the articular processes; E) Iliac crest fragment from CU34220 with enthesophyte formation; F) Patella from CU34220 showing enthesophyte development; G) Calcaneal tuberosity from CU2001 displays enthesophyte formation and bony spicules; H) Posterior view of tibia fragment from CU37044 showing enthesophyte development on the soleal line; I) Femur fragment portion showing the linea aspera from CU37044 with excessive dense cortical development (a: section view, b: posterior view); J) Distal portion of radius fragment from CU35417 showing evidence of enthesophyte formation; K) Ulna fragment from CU29050 with enthesophyte growth on the trochlear notch; L) Glenoid fossa of scapula fragment from CU34220 with enthesophyte formation |
Zones anatomiques considérées dans notre protocole et exemples de lésions osseuses observées dans l’échantillon de Cumes. Les flèches indiquent les modifications osseuses. A) Fragment de voûte crânienne provenant du CU34283, le détail montre un épaississement anormal du diploë ; B) Fragment de côte du CU37044 avec formation osseuse excessive, ossification des muscles intercostaux sur les bords supérieur et inférieur et altération de la surface sous la forme de porosité (a : côté exothoracique, b : côté endothoracique) ; C) Fragment de sternum du CU39306 avec ossification du cartilage visible au niveau de la surface viscérale et des facettes articulaires ; D) Portion cervicothoracique de la colonne vertébrale du CU37044 illustrant le développement ostéophytique et l’ossification excessive du cartilage sur les processus articulaires ; E) Fragment de crête iliaque du CU34220 avec formation d’enthésophytes ; F) Patella du CU34220 montrant le développement d’enthésophytes ; G) Tubérosité calcanéenne du CU2001 présente la formation d’enthésophytes et de spicules osseux ; H) Vue postérieure d’un fragment de tibia du CU37044 montrant le développement d’enthésophytes sur la crête soléaire ; I) Fragment de fémur montrant la ligne âpre du CU37044 notamment l’épaisseur excessive de la corticale (a : vue en coupe, b : vue postérieure) ; J) Partie distale du fragment de radius du CU35417 avec la formation d’enthésophytes ; K) Fragment d’ulna du CU29050 avec formation d’enthésophytes sur l’incisure trochléaire ; L) Cavité glénoïdale du fragment de scapula du CU34220 avec formation d’enthésophytes
17We examined the cranial vault for indications of osteosclerosis and increased thickness, particularly on the frontal, parietal, and occipital bones. The presence of osteophytes and enthesophytes on the vertebrae was documented based on large anatomical regions (cervical, thoracic and lumbar). In addition to observing widening and irregular edges on the ribs, we considered coarse, diffuse porosity of the rib surface as possibly indicative of the mixed osteoblastic/osteoclastic activity associated with the disease. Another criterion added to our protocol was the presence of cartilage ossification on the sternum. Furthermore, ossification or calcifications in limb anatomical areas were noted in terms of presence or absence according to Villotte’s classification of entheseal changes (Villotte et al., 2010; 2016; Villotte and Knüsel, 2013). Additionally, we documented fractures as well as bone lesions that are indicative of other conditions, such as Diffuse Idiopathic Skeletal Hyperostosis (DISH).
18Because none of the pathological lesions recorded are pathognomonic of fluorosis, individuals were classified into three levels of suspicion for fluorosis, based on the percentage of observable anatomical areas showing bone changes across the twelve regions considered. "Strong suspicion" was assigned when more than 70% of preserved anatomical areas had lesions. "Moderate suspicion" encompassed cases where 30 to 70% of preserved anatomical areas had lesions. When less than 30% of the preserved anatomical areas showed lesions, the individual was recorded as having "no suspicion" of fluorosis. Our sample consisted mainly of fragmented and incomplete burnt remains that limited observations in many areas. To avoid extremely positive or negative diagnoses resulting from preservation biases, a category of fluorosis suspicion was assigned only when at least five anatomical regions could be observed. Thus, individuals with only four or fewer observable anatomical zones were classified as "non-attributable".
19As well as biological characteristics, total weight and weight index were also considered as pointers to suspected fluorosis, since values exceeding the theoretical weight for a single individual can be related to pathological processes on the skeleton (Van Andringa et al., 2013; Duday, 2018; 2019). McKinley’s (1993) research on modern crematoria aimed to establish comparisons with archaeological cremations. The results show a total average weight for individuals of adult size of 1,627.2 g (σ=426.4 g) (male individuals: x̄=1,864 g, σ=333.5 g; female individuals: x̄=1,271.9 g, σ=280.7 g). Despite the differences, there is a substantial overlap between the sexes, also demonstrated by other studies (see Duday et al., 2000:8). Moreover, there seem to be differences between modern and archaeological data that can be attributed to several factors. Most archaeological remains studied are from secondary cremation burials, often with no complete skeleton. Additionally, sediment in the porous bones may increase the weight of the skeletal remains studied (Duday et al., 2000; Depierre, 2014). Duday (in Van Andringa et al., 2013) suggests that the average total weight of a cremated archaeological individual is between 1,250 g and 1,499 g, regardless of sex. These theoretical values apply to adults and adolescents, which is consistent with our sample. Based on Duday’s (in Van Andringa et al., 2013) proposed categories for the cremated remains at Porta Nocera (Pompeii), in this study, if the total weight exceeds 1,500 g the secondary cremation burial is considered to have been of an individual with a high total weight overall.
20It is important to note that the overall weight of cremated remains is influenced not only by the actual weight of the skeleton, but also by various funeral-related factors such as temperature and collection type (total or partial) (Duday, 2019; Duday et al., 2000). Factors like the state of conservation and rates of fragmentation can also have an impact on the total weight analysed, so that identifying specific anatomical areas is challenging. Therefore, differential identifications based on anatomical regions and theoretical weight indexes becomes crucial. For example, diaphyseal fragments of long bones are often classified as indeterminate when they cannot be attributed to a specific bone (Duday et al., 2000). In our study, we focused on not just the total cremation weight but also prioritized the trunk weight index, which has a theoretical reference value of 17%. This specific region was chosen because it has higher identification rates than other anatomical regions (Duday et al., 2000), and due to the higher frequency of bone alterations potentially related to fluorosis based on the protocol proposed in this study.
- 7 Hydroxyapatite is the main inorganic content of bone. Bone is a biocomposite of inorganic component (...)
21As a consequence of its pronounced electronegativity, fluoride has a natural affinity for calcium hydroxyapatite (Ca10(PO4)6(OH)2)7 (Miller and Phillips, 1953; 1956). After intake, fluorine passes through blood or the gastrointestinal tract and is then rapidly incorporated into calcified tissues, which contain up to 99% of body fluorine (Whitford, 1994; Zohoori and Duckworth, 2017). The hydroxyl ions of hydro-xyapatite crystals are then replaced by fluoride, converting the hydroxyapatite into fluorapatite and consequently producing changes in the physical characteristics of the crystals (WHO, 1992; Barbier et al., 2010; Ghosh et al., 2013).
22Fluoride is a halogen gas with a low atomic number, which means that its detection requires very accurate and reliable analytical methods. In previous palaeopathological studies of bone fragments from inhumations, fluoride has been detected using methods such as atomic absorption spectrometry (AAS) (Littleton, 1999), instrumental neutron activation analysis (INAA) (Petrone et al., 2011; 2013), ion-selective electrodes (ISE) (Yoshimura et al., 2006; Walser et al., 2020) or electron microprobes (EMP) (Zhou et al., 2023). All these methods allow the fluoride content to be measured, but they also require the extraction and destruction of the samples to perform the analysis. To overcome this limitation, we used laser-induced breakdown spectroscopy (LIBS), a method for elemental detection recently introduced in archaeometry. This method is sensitive enough to detect elements qualitatively, particularly halogens or light elements (Gaft et al., 2014; Pořízka et al., 2017). In addition, it can measure quantitative concentrations at around the part-per-million (ppm) level by combining LIBS with other methods (Rehse et al., 2012; Spizzichino and Fantoni, 2014). Furthermore, its potential for use with unprepared samples (Rehse et al., 2012; Álvarez et al., 2014; Alvarez-Llamas et al., 2017), its portability and its nano-destructive nature ensure a minimally invasive approach to data collection (Syvilay et al., 2019).
23The LIBS method relies on a brief laser pulse (lasting nanoseconds - Ns - or less) directed at a small area of the sample (approximately 100-200 μm in diameter). This process leads to sample ablation and the formation of a dynamic laser-induced plasma. An optical emission spectrometer is then used to measure the radiation emitted from the excited plasma species (Alvarez-Llamas et al., 2016). The most intense emission lines of fluoride lie at wavelengths shorter than 100 nm, which standard LIBS spectrometer set-ups using atmospheric air cannot detect (Alvarez-Llamas et al., 2016). Therefore, for this study, we used CaF I molecular emission bands at 535 nm (Álvarez et al., 2014).
24The portable instrument used is an EasyLIBS from IVEA Solution, equipped with a Nd:YAG laser at 1064 nm with a frequency of 1 Hz. The pulses delivered are 5 ns and 30 mJ. The plasma is collected via optical fibres connected to three separate compact Czerny-Turner spectrometers (Ocean Optics), which cover a spectral range between 200 and 1013 nm, with a spectral resolution of 0.2 nm. The analysis distance was set at 10 cm.
25LIBS models for analysis of non-prepared samples are based on solid homogeneous specimens (Jantzi et al., 2016). Femurs have wider and more homogeneous cortical sections, a significantly higher fluoride content compared to other bones (Turkekul et al., 2020) and a slow cortical turnover rate (5%/year) (Parfitt, 2002; Lerebours et al., 2020). Therefore, one femur fragment from each individual in the Cumae sample and the control sample were analysed using the following protocol: 6 measurement points per sample, with 50 laser shots per point. The spectra processing to obtain an average spectrum for each sample was then carried out with LIBStick software, developed by one of us (YL) (https://github.com/crp2a/LIBStick). The initial ten spectra were excluded to remove pollution and alterations from the surface. Subsequently, the spectral data were processed based on the simple measurement of CaF I molecular emission within the 525-562 nm range and standardised, resulting in intensities expressed in arbitrary units (a.u.). These baseline-subtracted signals are therefore semi-quantitative data that allow comparisons of intensities within the sample analysed.
26A comparative multivariate analysis was conducted using R v.4.1.3, incorporating the following categories as variables: "total cremation weight", "trunk weight index", "LIBS measurements" and "suspicion". The aim of this analysis was to explore the relationships between the study sample and the various criteria used for diagnosing fluorosis. The multivariate techniques used, including cluster analysis and principal component analysis (PCA), were applied to identify systematic relationships among the variables, as well as to examine the nature of these relationships.
27Applying the macroscopic observation protocol to the Cumae sample enabled us to identify various pathological changes, some of which are illustrated in figure 4. The specific details of lesions observed for each individual are given in table 3. The twelve anatomical regions observed were unequally affected by pathological changes, with the prevalence of lesions as follows (from the lowest value to the highest): 13% for the wrist (2/15), 22% for both the cranial vault (9/41) and the glenohumeral joint (7/32), 40% for the ribs (17/42), 45% for both the patella (9/20) and the tibia (5/11), 46% for the pelvic girdle (17/37), 48% for the elbow joint (10/21), 53% for the sternum (9/17), 67% for the femur (26/39), 70% for the vertebrae (30/43) and 82% for the calcaneus (14/17).
Table 3
Palaeopathological observations. Legend: Pathological alterations in the twelve anatomical zones observed: A = absent, P = present, NO = non-observable, n Obs = total number observable, n P = total number present, % P = relative frequencies of present within the observable total. Fluorosis suspicion: 1 = no suspicion, 2 = moderate, 3 = strong, NA = non-attributable |
Observations paléopathologiques. Légende : Altérations pathologiques dans les douze zones anatomiques observées : A = absente, P = présente, NO = non observable, n Obs = nombre total observable, n P = nombre total présent, % P = fréquences relatives des présentes dans le total observable. Suspicion de fluorose : 1 = sans suspicion, 2 = modérée, 3 = forte, NA = non attribuable
28Nine individuals were excluded from category attribution because observations could only be made in less than five anatomical areas. These account for 21.7% of the total sample. Among the remaining individuals (n=36), 39% are males (n=12) or probably males (n=2), and 58% are females (n=19) or probably females (n=2). In 51% of the Cumae sample, observations could be made in eight or more anatomical zones, i.e. at least two thirds of the areas considered in our protocol.
29Among the samples to which a category of fluorosis suspicion could be attributed, the results are as follows: 13 individuals (36%) have "no suspicion", 14 (39%) were placed in the "moderate suspicion" group and 9 (25%) in the "strong suspicion" group (figure 5A). All sex groups are represented among the three categories of fluorosis suspicion in this study. However, as shown in figure 5, the representation of female individuals in the "strong suspicion" category is notably lower (n=2).
30Figure 5B shows the correlation between age group and categories of fluorosis suspicion. These results indicate that no juveniles and only one young adult present bone signs suggestive of fluorosis. In contrast, bone lesions suggesting a moderate or strong suspicion of fluorosis are common in middle-aged and elderly individuals. A detailed view of the "strong suspicion" category indicates that it is more frequent among male individuals over 30 years old. Additionally, the ratio between strong and moderate suspicion is higher in the middle-aged adult group (7:8) than in the elderly adult group (2:4).
Figure 5
Fluorosis suspicion in the Cumae sample. A) Fluorosis suspicion and sex distribution; B) Sex and age distribution correlated with fluorosis suspicion |
Suspicion de fluorose dans l’échantillon de Cumes. A) Distribution de suspicion de fluorose par sexe ; B) Répartition par classe d’âge et sexe corrélé à la suspicion de fluorose
31When we analyse the distribution and relative frequencies of pathological bone changes in the "moderate suspicion" and "strong suspicion" groups in our sample (n=22), our observations show similar results to those from previously published studies (Littleton, 1999; Petrone et al., 2013; Nelson et al., 2019). The thoracic anatomical zones are more affected overall than the cranial vault and upper limbs. After changes observed in the vertebrae, the most characteristic alterations were seen in the lower limb anatomical zones, particularly on the femur and pelvic girdle. The higher frequency for the calcaneus is also comparable to the results of Littleton (1999); however, our higher value can also be attributed to the large and well-preserved fragments, where calcaneal tuberosities can be very clearly distinguished (table 4).
Table 4
Relative frequency of pathological bone alterations in the "moderate and strong suspicion" groups. Legend: n P = number of individuals with pathological alterations present, n Obs = number of individuals where the anatomical zone can be observed, % (relative Obs) = relative frequency of the number of individuals with pathological alterations |
Fréquence relative des lésions ostéologiques pour les groupes de suspicion modérée et forte. Légende : n P = nombre d’individus présentant des lésions ostéologiques, n Obs = nombre d’individus où la zone anatomique peut être observée, % (Obs relative) = fréquence relative du nombre d’individus présentant des lésions ostéologiques
32In the graphs shown in figure 6, each category of fluorosis suspicion is represented within the correlation of total cremation weight (x-axis) and trunk weight index (y-axis). In the "no suspicion" (figure 6A) and "moderate suspicion" (figure 6B) groups, it is notable that some individuals exceed the theoretical value for the trunk weight index and/or average value for the total weight. Among the individuals in the "no suspicion" group, the highest total weight recorded is 1,847.2 g, belonging to an older adolescent female (CU34241). In the "moderate suspicion" group, there is slightly greater dispersion, with four individuals showing a higher trunk index and total weight overall. The highest weights recorded are around 2,000 g, corresponding respectively to a young adult male (CU35370) and two middle-aged adult females (CU39396 and CU34283). The data dispersion in the "strong suspicion" group (figure 6C) is clearly different, with seven male individuals above the theoretical values for both trunk index and total weight. The two individuals with a lower total weight are females (CU37045 and CU35482), but they both have a significantly higher value for trunk weight index. CU37045 is a noteworthy case, as this individual is an older adult female whose lower total weight could suggest a natural demineralisation process specific to the menopause (see Duday in Van Andringa et al., 2013).
Figure 6
Total trunk weight index and total burnt bones weight correlated with the different categories of fluorosis suspicion. A) Correlation for the "no suspicion" group; B) Correlation for the individuals with "moderate suspicion"; C) Correlation for the "strong suspicion" group. The dotted line on the x-axis shows the average normal total weight (1,500 g) and the dotted line on the y-axis shows the theoretical trunk weight index (17%) |
Indice pondéral du tronc et masse totale des os humains brûlés corrélés aux différentes catégories de suspicion de fluorose. A) Corrélation pour le groupe sans suspicion ; B) Corrélation pour les individus avec suspicion modérée ; C) Corrélation pour le groupe de suspicion forte. La ligne pointillée sur l’axe x montre la masse totale moyenne (1 500 g) et la ligne pointillée sur l’axe y montre l’indice pondéral théorique du tronc (17 %)
33The control sample analyses were crucial to identify average fluoride values for individuals from non-volcanic contexts with no suspicion of fluorosis, whether they were cremated or not. Results from LIBS analyses have confirmed the potential of this method for studies of cremated human remains, including fluoride detection. The results obtained for the Cumae sample reinforce our palaeopathogical protocol and our proposed division into categories of suspicion.
34Figure 7 illustrates the correlation of the LIBS results with the three categories of suspicion outlined in this study. The results indicate that our control sample has lower LIBS values compared to the Cumae sample, whose overall values are significantly higher. The control sample values range from 0.0160 to 0.0258 a.u. (x̄=0.0220, σ=0.0041); the Cumae sample values are higher and range from 0.0372 to 0.1616 a.u. (x̄=0.0878, σ=0.0366) (table 5). The elevated levels of fluoride in the Cumae population may be attributed to environmental contributions associated with the underlying volcanic system and therefore with nutrient intake, most probably through the consumption groundwater, which is a primary source of fluoride for humans (Barbier et al., 2010; Baxter and Horwell, 2015; Yeşilnacar et al., 2016).
Figure 7
Cross-analysis of data from palaeopathological observations by category of suspicion and LIBS measurements. The dotted line shows the proposed threshold for a strong suspicion of fluorosis |
Analyse croisée des données d’observations paléopathologiques par catégorie de suspicion et mesures LIBS. La ligne pointillée montre le seuil proposé pour une forte suspicion de fluorose
Table 5
LIBS measurement results. Legend: Suspicion of fluorosis: 0 = control sample, 1 = no suspicion, 2 = moderate, 3 = strong, NA = non-attributable. Sex: Ind = indeterminate, F = female, F? = Probably female, M = male, M? = probably male, - = not identified |
Résultats des mesures LIBS. Légende : Suspicion de fluorose : 0 = échantillon contrôle, 1 = sans suspicion, 2 = modérée, 3 = forte, NA = non attribuable. Diagnose sexuelle : Ind= indéterminé, F = féminin, F? = féminin probable, M = masculin, M ? = masculin probable, - = non identifié
35A close examination of the LIBS values for each suspicion category (no suspicion, moderate, strong) shows that our palaeopathological protocol classifying subjects into groups indeed correlates with fluoride detection (table 5) and is therefore relevant when attempting to diagnose fluoride poisoning. The "no suspicion" group has the lowest values with an average of 0.0706 a.u. (σ=0.0220), followed by the moderate and strong suspicion groups, with averages of 0.0966 a.u. (σ=0.0384) and 0.1274 a.u. (σ=0.0228) respectively.
36Since our sample consists only of older adolescents and adults, the results can be analysed by sex (see details in table 5). Hereafter, we include both "female" and "probably female" in the female group, and likewise for male individuals. The mean values for females are 0.0698 a.u. (σ=0.0233) for "no suspicion", 0.0934 a.u. (σ=0.0277) and 0.1082 a.u. (σ=0.0084) for "moderate" and "strong suspicion", whereas for males they are respectively as follows: 0.0750 a.u. (σ=0.0177), 0.1102 a.u. (σ=0.0501), 0.1329 a.u. (σ=0.0228). From the data analysis, it is evident that the average values for male individuals consistently exceed those for female individuals across all categories. However, this difference between the sexes appears to widen with increasing levels of suspicion. Specifically, the difference between male and female averages is three times greater in the "moderate suspicion" group (0.0168) than in the "no suspicion" group (0.0051), and roughly one-and-a-half times greater in the "strong suspicion" group (0.0248) than in the "moderate suspicion" group.
37A Mann-Whitney non-parametric test (U test) was performed to evaluate the differences between sample groups. There is a highly significant difference between the Cumae sample and the control sample (p=0.00038). When comparing two contiguous groups (no suspicion vs. moderate; moderate vs. strong) there is no significant difference (p=0.120 and p=0.153, respectively). However, there is a significant difference when comparing the "no suspicion" group with the "moderate" and "strong suspicion" groups together, and also when comparing the "strong suspicion" group with the rest of the Cumae sample (p=0.00705 and p=0.00729, respectively). Furthermore, there is also a significant difference in LIBS values between males and females (p=0.02996).
38Among the three categories of suspicion, there is higher variability in the "moderate" category. The LIBS values for this group overlap with both the "no suspicion" and "strong suspicion" categories (figure 7). These results do not offer a definitive differentiation of this category based solely on LIBS, but they do emphasise the importance of incorporating anthropological observations with this method. They also raise the question of the pertinence of this category in terms of its osteological reality, suggesting that it may need further evaluation with additional skeletal and clinical evidence.
39Upon thorough analysis of the findings, it becomes evident that the LIBS values allow for a clear separation between the "no suspicion" and the "strong suspicion" groups (figure 7). We propose that the cut-off point for identifying definite cases of fluorosis can be established by using the maximum measurement possible in the sample with the lowest intensity value in the "strong suspicion" category (CU34243: x̄=0.0879, σ=0.0067, max.=0.0946) and rounding it to three decimal points (i.e. 0.095 a.u.; figure 7). This threshold could serve as a baseline for future studies of populations from the same geological context.
40Figure 8 shows the results of the multivariate analysis. The hierarchical tree (figure 8A) reveals three main groups. The Principal Component Analysis (PCA) scatter-plot (figure 8B), with the first two components, explains 80.53% of the variance (63.86% and 16.67%). The trunk weight index is strongly correlated with the suspicion variable (r=0.562, p < 0.01).
Figure 8
A) Hierarchical clustering tree illustrating the classification of individuals into three distinct clusters; B) PCA scatter-plot depicting the distribution of individuals within each cluster, with categories of suspicion indicated for each individual. In brackets, the individual number corresponding to those indicated in the hierarchical clustering tree |
A) Arbre de classification hiérarchique illustrant la répartition des individus en trois clusters distincts ; B) Diagramme de dispersion en ACP montrant la répartition des individus au sein de chaque cluster, avec les catégories de suspicion indiquées pour chaque individu. Entre parenthèses, le numéro de l’individu correspondant à ceux indiqués dans l’arbre de classification hiérarchique
41Cluster 1 comprises all "no suspicion" individuals and eight "moderate suspicion" individuals. This cluster is characterised by low values for the "LIBS measurements", "trunk weight index", and "total cremation weight" variables, in order from the lowest to the highest. Cluster 2 is distinguished by high values for the "LIBS measurements" and "trunk weight index" variables, with individuals in this cluster exhibiting moderate (n=5) and strong (n=4) suspicion. Cluster 3 is made up of "strong suspicion" individuals and is marked by high values for the "total cremation weight", "trunk weight index" and "LIBS measurements" variables, in order from the highest to the lowest.
42This analysis demonstrates that the bone changes noted in the Cumae sample, the correlation with cremation data (total weight and trunk weight index) and the LIBS measurements are probably indicative of skeletal fluorosis, particularly for those categorised under "strong suspicion". It should be noted that some of the bone lesions recorded can also be associated with other pathological conditions (see table 2). Most of these disorders, such as myositis ossificans progressiva, Paget’s disease, hypoparathyroidism, osteopetrosis, treponematoses and hematogenous osteomyelitis can be readily ruled out. These conditions are rare and/or result in bone changes that are not present in the Cumae individuals (see Supplementary Information for a detailed differential diagnosis). Diffuse idiopathic skeletal hyperostosis and ankylosing spondylitis can cause a mosaic of bone changes resembling those in skeletal fluorosis, so they cannot be definitely excluded. However, these latter conditions are unlikely to be found in such a large proportion of individuals and would most certainly not result in elevated fluoride levels. Based on this differential diagnosis, the most probable cause of the bone modifications recorded throughout the Cumae sample is skeletal fluorosis. However, since the changes caused by skeletal fluorosis are gradual, we cannot rule out the possibility that some of the bone modifications observed may have arisen from osteoarthritis or been exacerbated by normal mechanical demands and age-related degenerative processes.
43Although the protocol proposed was applied to cremated remains, our results for bone lesions are consistent with those from other archaeological contexts where the remains do not show signs of burning (Callaghan, 1986; Littleton, 1999; Weinstein, 2005; Yoshimura et al., 2006; Petrone et al., 2013; Nelson, 2015; Nelson et al., 2016; 2019; Walser et al., 2020; Zhou et al., 2023), as well as with those from clinical studies (e.g. Jolly et al., 1968; Brickley and Ives, 2008a; Datta and Datta, 2013; Sellami et al., 2020).
44It seems evident that one of the anatomical areas most affected by bone modifications is the trunk. As seen on the Cumae sample, the relative frequencies of bone alterations observed are 100%, 69% and 64% for the vertebrae, sternum and ribs, respectively. These results, besides being similar to those from previous palaeopathological studies (Littleton, 1999; Petrone et al., 2013; Nelson et al., 2019), are also consistent with radiological data for the diagnosis of skeletal fluorosis in living patients (Sellami et al., 2020). We can also infer that the higher prevalence in this specific anatomical region is consistent with bone biology, as this area contains multiple zones with connective tissue and is therefore prone to ossification (de Vlam et al., 2006).
45The fact that the second most affected area is the lower limb, more specifically the femur (86%), is likewise consistent with typical skeletal fluorosis bone alterations (Littleton, 1999; Nelson et al., 2019). Furthermore, in addition to the observations proposed in our protocol, we were able to observe sections of femoral diaphysis and note that the cortical bone was abnormally thick (figure 4I). Abnormal bone formation is related to osteosclerotic changes (de Vlam et al., 2006), and is therefore responsible for the abnormal weight or "heaviness" characteristics that we observed among the individuals of the Cumae sample.
46Although our sample is small, our results show that bone alterations are higher amongst males than females. This observed sex difference in fluorosis may be linked to various factors in the variables studied for diagnosing this condition. Some lesions attributed to fluorosis could be due to age-related degenerative changes or traits associated with robustness, especially when the diagnosis relies solely on osteological features.
47Additionally, data on total cremation weight and trunk weight index also suggest a higher prevalence in males. However, these findings may be biased when considered separately. The robustness often attributed to males could lead to an overestimation of the prevalence rate in males, which is a significant limitation in sex estimations of cremated remains (Duday et al., 2000). From a statistical perspective, using fixed thresholds based on theoretical references might underestimate the number of females.
48It is important to note that LIBS results are not influenced by the aforementioned variables and exhibit the same trend, as shown by the multivariate analysis. Consequently, the demographic data appears to be reliable, as it is supported by evidence unaffected by these biases. However, the underrepresentation of older adult females in the sample is a significant limitation that could influence the results.
49While many clinical and palaeopathological studies support a higher prevalence of skeletal fluorosis in males, the underlying causes of this observed difference require further investigation (e.g. sexually dimorphic biology or behaviour, Walser et al., 2020). For instance, clinical research by Mohammadi and collaborators (2017) on an Iranian population found a higher prevalence in females. Moreover, hormonal changes during menopause in adult females might also be a factor to consider. Duday (in Van Andringa et al., 2013) has already discussed that this female specificity can affect the total weight of cremated remains, but there is also the possibility that natural demineralisation can mask some bone modifications and an overall diagnosis of fluorosis, whether in cremated or non-cremated remains or even in clinical samples. Although the limited number of females in our sample reduces support for this hypothesis at present, further research is needed to fully explore this potential explanation. Before doing so, additional research on the mineral density (Yildiz and Oral, 2003) and micro- architecture of cremated remains should be carried out (by microtomographic analyses for example).
50When considering the data by age group, it seems that the rarity of strong suspicion of fluorosis among the youngest individuals in the sample (juvenile and young adults) may be linked to the duration of exposure and the time necessary to develop bone signs. A higher occurrence of strong suspicion of fluorosis was observed in the individuals over 30 years old (especially in the middle-aged adult group); these results also correspond to clinical studies that report a higher occurrence of skeletal fluorosis from around 30 to 40 years of age and beyond (Datta and Datta, 2013; Sellami et al., 2020). Similarly, our findings align with Littleton’s results (1999), which indicate that few bone lesions are observed before the age of 30. However, it is important to stress that there is a potential bias: because skeletal fluorosis is an age-related condition, it is difficult to isolate bone lesions caused by aging from those caused by fluorosis, particularly with fragmented remains. Consequently, the risk of false positives increases with age, as older individuals are more likely to develop osteophytes, enthesophytes and similar conditions. This difficulty underscores the importance of combining macroscopic observations with archaeometric analysis. Furthermore, while robusticity and age-related changes do not account for all the features observed (e.g. thickened diploe and certain forms of bone densification), they do play a significant role in many of the lesions or features detected, including porosity, bone densification and the development of enthesophytes and osteophytes. These age-related changes are most commonly seen in the spine and lower limb areas, which are particularly susceptible to fluorosis. However, it is important to recognise that some features observed in individuals are not attributable to aging.
51The main challenges for this research were to study a pathological condition with no pathognomonic lesions, and dealing with the nature of our sample. Cremated remains, as stated in previous sections, can limit observations due to their fragmentary state and incomplete representations of anatomical areas. However, by incorporating the average total weight and weight indexes for the different anatomical areas into the proposed protocol, more comprehensive interpretations can be achieved. The findings in this study, including the multivariate analysis, reveal a significant correlation between the trunk weight index and the cases of strong suspicion of fluorosis. This anatomical area not only has higher rates of identification among cremated remains (Duday et al., 2000), but also a higher relative frequency of the pathological changes identified in this study (see table 4). The correlation results demonstrate a clear association between these two sets of cremation data, when weights exceed both the overall average normal total weight (1,500 g) and the trunk weight index (17%), supporting a strong suspicion of fluorosis. In contrast, there is greater variation in the results for the "no suspicion" and "moderate suspicion" categories, which makes the correlations established less conclusive. Further research is necessary to determine whether higher weight values in these categories should be attributed to factors other than pathology.
52Our findings demonstrate that fluoride can be effectively detected by laser-induced breakdown spectroscopy (LIBS). The use of LIBS in our study is pivotal, as it produces a more accurate determination of the origins of the skeletal lesions observed. Without this elemental analysis, drawing conclusions about their aetiology would have been difficult. Additionally, our application of this innovative methodology represents a significant contribution to the field.
53As proposed by previous researchers, emissions from CaF bands (around 535 nm) offer a viable alternative for accurately detecting and assessing fluoride levels (Álvarez et al., 2014; Alvarez-Llamas et al., 2016; Pořízka et al., 2017; Foucaud et al., 2019), and this method can also be applied to cremated human remains. Our current research shows that LIBS has promising potential as a method for elemental analysis. Its portability enables analyses to be conducted both in the field and in the laboratory. Our findings also show its potential for application to archaeometric studies relating to questions in biological anthropology, complementing its existing uses in areas such as art history (Spizzichino and Fantoni, 2014) and biomedicine (Samek et al., 2001). However, it must be noted that LIBS is a relatively new technique which is still undergoing development and improvement, specifically for quantitative analysis.
54The control sample in this study was set up to establish a standard set of data for individuals who show no signs of fluorosis and come from non-volcanic areas with different burial practices (inhumation and burial after cremation). Apart from providing important data on the threshold for individuals not exposed to high levels of fluoride, the control group prompts an inquiry into the interaction between fluoride and fire, particularly in the context of cremation as a funerary practice. However, due to the limitations of our sample size, we cannot further explore how exposure to fire might influence fluoride fixation or loss. Nevertheless, it is apparent that in comparison to the control sample, all the Cumae individuals exhibit higher levels of fluoride, probably influenced by their environment.
55Even though our results are semi-quantitative, they still allow us to identify different groups in the Cumae sample. The correlation with anthropological observations has enabled us to validate the potential of our macro-observation protocol and the proposed categories, and to establish a threshold for identifying strong suspicions of fluorosis. Further research on larger skeletal samples from the same geological context presented in this study, as well as from other volcanic areas, would allow confirmation or adjustment of the proposed threshold, which in turn would improve the identification of suspected fluorosis using LIBS measurements.
56Unfortunately, direct comparisons between our findings on fluoride detection and those of other researchers (Littleton, 1999; Petrone et al., 2013; Walser et al., 2020; Zhou et al., 2023) are not feasible due to the substantial differences in the methods used. However, we concur with their recognition of higher values for populations in volcanic environments. The higher values for the Cumae sample, as opposed to our control sample, are probably associated with fluoride poisoning.
57As stated earlier, LIBS is a qualitative method with which quantitative data could be acquired by using reference samples or coupling with another quantitative method to allow the construction of calibration curves (Samek et al., 2001; Rehse et al., 2012; Spizzichino and Fantoni, 2014). Further research using a quantitative method (e.g. PIXE-PIGE) would allow us to explore the possibility of developing a calibration curve for detecting and quantifying fluoride in human remains. Conducting further research with a larger sample size and integrating LIBS with other quantitative methods will facilitate future investigations, allowing the Cumae values to be compared with those from other archaeological populations in the same geological context (Campanian volcanic arc), as well as in volcanic areas elsewhere in the world.
58The environmental health impacts of historical water consumption are still underexplored. While bioarchaeological studies have investigated cases of poisoning due to fluoride (e.g. Littleton, 1999; Yoshimura et al., 2006; Petrone et al., 2011; 2013; 2019; Nelson, 2015; Nelson et al., 2016; 2019; Walser et al., 2020), arsenic (e.g. Özdemir et al., 2010; Fresnais et al., 2015; Galaz-Mandakovic and Rivera, 2022) and heavy metals (e.g. Emslie et al., 2015; Moore, 2019; Moore et al., 2021; Proctor, 2021), research into these issues is still limited. Concretion analyses have sometimes been used to investigate environmental pollution (e.g. Carlut et al., 2009; Delile, 2014; Benjelloun et al., 2019). However, more interdisciplinary research is essential to fully assess historical natural risks, particularly those related to water. Addressing this gap in knowledge would provide a critical entry point for the development of historical ecotoxicology.
59Toxins, whether from natural or human sources, can be harmful even in small amounts (Sánchez-Bayo, 2011). Fluoride is a cumulative toxin that entails risks depending on the duration of exposure, amounts and sources (Ayoob and Gupta, 2006; Brickley and Ives, 2008b). Despite its benefits for dental health, the potential hazards of fluoride intake were under-appreciated until recent studies highlighted its adverse effects even at low doses (Barbier et al., 2010). The World Health Organization (WHO) has set the maximum safe fluoride concentration in drinking water at 1.5 mg/l to prevent poisoning (WHO, 1992; 2019). However, excessive fluoride intake can also occur through water used to produce crops, raise livestock or prepare foodstuffs (Chowdhury et al., 2019).
60In Antiquity, groundwater consumption was common (Tölle-Kastenbein, 1993; Mays et al., 2007) and probably the primary source of fluoride intake. In Greek-era Cumae, water was collected from cisterns and wells (D’Acunto, 2020), a practice that most likely continued into the Roman period. The Serino aqueduct, constructed during the Augustan period (also known as the Aqua Augusta, ca. 38 BC- 11 AD), supplied water to Pompeii and Herculaneum (Caputo, 2004; Dessales, 2008; 2013; Keenan-Jones, 2010a; 2010b; 2013; Döring, 2012; Linoli, 2012; Ferrari and Lamagna, 2013; Potenza, 2016). Although indirect evidence suggests that Cumae was connected to this aqueduct (e.g. inscriptions, lead pipes, decorative fountains) (Camodeca, 1997; Capaldi, 2007a; 2007b; 2008; Guardascione, 2007; Zevi et al., 2008; Brun and Gasparri, 2009; Gasparri, 2009; Gallo, 2015; Ferrari and Lamagna, 2016; Potenza, 2016; Ferrari et al., 2018), questions remain about whether the aqueduct actually reached Cumae and, if so, how it influenced fluoride intake. Further research with a larger sample is necessary to address these questions.
61Fluoride exposure in the Phlegraean Fields has probably been a health risk since ancient times due to the high fluoride concentrations in the groundwater, driven by continuous volcanic activity (Rosi et al., 1983; Stellato et al., 2020; Cappelletti et al., 2022). Current levels (3.6-15 mg/l) (Ducci and Sellerino, 2012) far exceed safe drinking standards, suggesting that Roman-era water had similarly high fluoride levels, raising a risk of fluorosis, although exact historical intake levels are difficult to assess.
62This study applies an innovative multidisciplinary approach, using new methods to examine skeletal fluorosis in cremated human remains. To the best of the authors’ knowledge, it is the first to address this issue from a population perspective and thus contributes to the development of historical ecotoxicology. Although cremated remains raise many difficulties for bioarchaeological studies, the correlation observed between palaeopathological macro- observations and LIBS measurements demonstrates the potential for providing greater scientific certainty in diagnosing pathological conditions such as fluorosis. Further research with a bigger sample, including a larger sample from Cumae and other sites in volcanic regions such as Pompeii, should enable us to enrich and reinforce our results and assess differences between various environments. For this study, we examined only femur fragments, but to conduct a more comprehensive assessment of intoxication stages, future research should involve analysing other bones (e.g. ribs) in order to evaluate fluoride levels in bones with varying turnover rates.
63The supplementary information is available in .pdf from the BMSAP website (https://journals.openedition.org/bmsap/15711?file=1).
Acknowledgments: This research was conducted as part of a PhD funded by the French Ministry of Higher Education and Research, through the Réseau des Écoles Françaises à l’étranger, in collaboration with the École française de Rome. It also benefited from the scientific framework provided by the University of Bordeaux’s IdEx "Investments for the Future" program/GPR "Human Past". The authors are grateful for their collaboration to Professor Jean-Pierre Brun (Collège de France) and to the entire team involved in the "Cumes : Aux marges de la ville" research programme. Sincere thanks also to Anne Le Maître, editor-in-chief of the BMSAP, and the anonymous reviewers for their invaluable feedback and constructive comments, which have significantly enhanced the quality of this manuscript.