AMERICAN JOURNAL OF PHYSICAL ANTHROPOLOGY 115:380–387 (2001) Epidemiological Approach to the Paleopathological Diagnosis of Leprosy Jesper L. Boldsen* Anthropological Data Base Odense University and Danish Center for Demographic Research, SDU-Odense University, DK 5000 Odense, Denmark KEY WORDS sensitivity; specificity; simultaneous estimation; skeletal lesions; lepers institutions; cemeteries; Denmark ABSTRACT In paleopathology it is usually assumed samples were scored for the presence of seven osteological that modern diagnostic criteria can be applied to infectious conditions indicating leprosy. For the osteological conditions, diseases in the past. However, as both the human species sensitivity varied from 0.36–0.80, and specificity from 0.58– and populations of pathogenic microorganisms undergo evo- 0.98. The frequency of leprosy in the three samples was: lutionary changes, this assumption is not always well- Odense (a lepers’ institution), 0.98, 95% CI 0.64–1.00; founded. To get valid estimates of the frequency (the point Malmo¨ (urban cemetery), 0.02, 95% CI 0.00–0.07; and Tirup prevalence at death) of leprosy in skeletal samples, sensitiv- (rural cemetery), 0.36, 95% CI 0.23–0.46. It is concluded that ity, specificity, and sample frequency must be estimated it is indeed possible to estimate disease frequencies without simultaneously. It is shown that more than three symptoms reference to modern standards, and that leprosy occurred must be evaluated in at least three samples in order to reach with widely differing frequencies in different segments of the estimates with well-described properties. The method is ap- Medieval population in southern Scandinavia. Am J Phys plied to three skeletal samples from Medieval Denmark; the Anthropol 115:380–387, 2001. © 2001 Wiley-Liss, Inc. Leprosy was a dreaded disease in the Middle true spread of the disease. The cemetery of Sct. Ages. The clearly visible symptoms of advanced lep- Jørgensgård contains many more lepers than would romatous leprosy stigmatized the diseased. Special be a reflection of the population prevalence, and institutions were created in many places in Europe ordinary cemeteries would perhaps be lacking in to accommodate the victims of leprosy. The name of lepers as the victims had been selected out of the the disease actually reflects this institutionalization general population. However, before one can begin of the sufferers in several languages: German “Aus- to address questions about the processes related to satz” (excluded) and, in the Scandinavian lan- leprosy in the once-living population, it is necessary guages, “spedalskhed” (derived from the word hos- to go beyond the counting of leprous lesions. All pital). The bringing together of lepers in an symptoms (in this case, osteologically identifiable institution and the following burial of dead inmates leprosy-related lesions) have special epidemiological of the institutions in special cemeteries has permit- properties which can be described by specificity and ted research into Medieval leprosy. The first modern sensitivity. Generally, the specificity and sensitivity studies date back to the 19th century (Ehlers, 1898). of leprosy-related osteological conditions are un- This study was archaeologically and historically fol- known, and it is highly probable that they have lowed up by Richards (1960, 1977). It was Møller- changed over centuries with the coevolution of hu- Christensen (1951, 1953, 1961, 1963, 1978) who ini- mans and the bacterium causing leprosy. This paper tiated osteological research into Medieval leprosy. sets out to create methods (osteological and statisti- Andersen (1969, 1991) went on to analyze the diag- cal) to permit the simultaneous estimation of sensi- nostic process for leprosy in Medieval Denmark. Os- tivity and specificity of several osteological symp- teological studies have widened to include estima- toms of leprosy and the frequency of leprosy (the tion of the frequency of lepers in skeletal samples point prevalence of leprosy at death) in several skel- from lepers institutions and ordinary cemeteries etal samples. (Andersen, 1969; Boldsen, 1978; Bennike, 1985; Ar- cini, 1990, 1999; Arentoft, 1999). The frequency of leprous (or any other) lesions Grant sponsor: Danish Foundation for Basic Research. among excavated skeletons has very little bearing *Correspondence to: Jesper L. Boldsen, Anthropological Data Base on the processes of health and disease in the once- Odense University, SDU-Odense University, Campusvej 55, DK 5230 living population (Wood et al., 1992). The selection Odense M, Denmark. E-mail: [email protected] of lepers into lepers’ institutions (in Danish: Sct. Jørgensgård) makes it very difficult to study the Received 12 September 2000; accepted 30 April 2001. © 2001 WILEY-LISS, INC. PALEOPATHOLOGICAL DIAGNOSIS OF LEPROSY 381 TABLE 1. Relationship between disease and symptom in It is obvious from the formulae listed above that relationship to leprosy knowledge about sensitivity and specificity is re- Leprosy quired to estimate the population prevalence of the Sumptom Yes No Total disease, and that the estimation of the probability that a person showing the symptom actually had the Yes Truly positive Falsely positive Positive No Falsely negative Truly negative Negative disease also requires information on population Total Sick Not sick All prevalence. It is likely that human populations have evolved through their interaction with populations of patho- MATERIALS AND METHODS gens like Mycobacterium leprae (the bacterium caus- Epidemiological background ing leprosy), and the pathogens are likely to have changed even more through interaction with specific Paleopathological diagnosis is in most ways simi- human populations. This means that diseases have lar to screening for diseases in modern society. The not retained the same clinical symptomatic image purpose of the exercise is both to estimate the prev- over centuries, and that sensitivities and specifici- alence of the disease in a given (sub)population and ties for specific symptoms have not remained un- to assess if a given individual suffers from the dis- changed. Therefore, in order to study the prevalence ease. In reality, no osteological symptom occurs only of leprosy in Medieval Denmark, specificities and among people with the disease, and no single symp- sensitivities for the symptoms involved must be es- tom is shown by all who suffered from the disease. timated based on Medieval data. Table 1 summarizes the basic relationship between Gold standards do not exist in modern epidemiol- disease (in this case leprosy) and any given symp- ogy, and the situation is even more difficult in pa- tom. leoepidemiology. This means that there is no terra Epidemiologically, a symptom is characterized by firma from which sensitivities and specificities can sensitivity (the probability of having the symptom, be estimated. However, if given that an individual had the disease) and spec- ificity (the probability of not having the symptom, p ⅐ s Ͼ p ϩ 2 ⅐ s, (5) given that an individual did not have the disease). where p is the number of populations studied and s truly positive the number of symptoms used to analyze the dis- sensitivity ϭ sick ease, then it is possible to estimate sensitivities and specificities of the s different symptoms, as well as ϭ p͑positive͉sick͒ (1) the p different population prevalences, and to test truly negative the goodness of fit of the model to the observations. specificity ϭ In the present analyses, p ϭ 3 and s ϭ 7, so the not sick number of degrees of freedom (df) for the goodness of ϭ p͑negative͉not sick͒ (2) fit test becomes four: Following this, an observable (osteological) condi- df ϭ p ⅐ s Ϫ ͑p ϩ 2 ⅐ s͒ ϭ 21 Ϫ 17 ϭ 4 (6) tion is a symptom of a given disease if sensitivity is larger than one minus specificity. In other words, This line of reasoning makes it possible to estimate the probability of having the condition is larger the epidemiological properties of paleopathological among people with the disease than among people symptoms and the population prevalence of the dis- without that particular disease. This does not imply ease (leprosy) under the assumption of indepen- that the majority of people showing the symptom dence among the expression of the different symp- actually have the disease. In paleopathology, and in toms among people without the disease. modern screening programs, the only observable Samples statistic is the frequency of people showing the symptom. However, using sensitivity and specificity Three samples of skeletons from Medieval Den- it is possible to estimate the population prevalence mark have yielded data for the analyses in this (frequency of lepers in the skeletal sample p(sick)) paper: Sct. Jørgen in Odense, Kvarteret St. Jo¨rgen from the frequency of the symptom (p(positive)): in Malmo¨ (Swedish since 1660), and Tirup. Two of the cemeteries (Odense and Malmo¨) have been as- p͑positive͒ Ϫ ͑1 Ϫ specificity͒ p͑sick͒ ϭ (3) sociated with lepers’ hospitals (the so-called Sct. Jør- sensitivity Ϫ ͑1 Ϫ specificity͒ gensgård; Arentoft, 1999), and the third is a rural And once the population prevalence of the disease parish cemetery (Kieffer-Olsen et al., 1986). The (p(sick)) has been estimated from an equation like locations of the three cemeteries are indicated in (3), it is possible to estimate the predictive value for Figure 1. The Odense and Tirup samples are part of the symptom in a given population (P(sick͉symptom)): the collections of the Anthropological Data Base Odense University (ADBOU), and the Malmo¨ sam- ⅐ ͑ ͒ sensitivity p sick ple is temporarily being kept in ADBOU for regis- p͑sick symptom͒ ϭ (4) p͑symptom͒ tration and analysis. 382 J.L. BOLDSEN Fig. 1. Denmark in Middle Ages, including provinces of Skåne, Halland, and Blekinge (now part of Sweden) and of South Slesvig (now in Germany). Sites are indicated by name of town or village where site is situated: Odense, Sct. Jørgen in Odense; Malmo¨, Kvartetet St. Jo¨ rgen in Malmo¨; and Tirup. Sct. Jørgen in Odense was excavated by Odense yielded data on 635 individuals aged 15 or more at Bys Museer in 1979–1981 (Jensen and Tkocz, 1983).