A Framework for Anemia Differential Diagnosis in Paleopathology Incorporating Metric Methods

ABSTRACT

Objectives

This paper explores metric manifestations of anemia in crania undergoing growth and development using micro‐CT imaging. It proposes a framework for assigning a most‐likely diagnostic option for anemia, based on evaluating the parameters proposed in this study.

Materials and Methods

Sixty‐eight orbits/frontal bones of individuals aged birth to 15 years from Quebecois and Dutch archaeological collections dating to the 18th and 19th centuries underwent micro‐CT analysis. Individuals were visually assessed for skeletal manifestations of marrow hyperplasia within the internal marrow space using a scoring rubric. Bone microarchitecture measurements were used to calculate T‐scores and identify individuals who displayed potential manifestations of marrow hyperplasia. Relative cortical thickness ratios of the frontal bone were calculated for 16 individuals. Error testing was performed for all evaluations.

Results

Using the micro‐CT analysis and our diagnostic framework, anemia was inferred in 16% (10/61) of the sample that was preserved well enough for the study. Trabecular separation T‐scores were considered the most significant metric for evaluating anemia. Frontal bone ratios were regarded as less insightful due to the imaging technique used. Age had a significant effect on bone measurements, and high repeatability was seen across methods.

Discussion

In this study, recommendations for assigning a diagnostic option prioritize evaluating metric features strongly related to anemia through a biological approach that considers the etiology of marrow hyperplasia. Including a combination of metric and internal visual evaluation criteria provides clearer lines of evidence for the assessment of abnormal bone changes associated with anemia beyond the macroscopic evaluation of porous lesions.

1. Introduction

Although anemia in skeletal remains has been studied by anthropologists since the 1960s (e.g., Angel ¹⁹⁶⁶), confident identification is complex and a subject of ongoing debate, emphasizing the need to standardize and improve diagnostic criteria. Paleopathologists and bioarchaeologists are increasingly highlighting the complexity of recognizing anemia (Brickley 2024; Grauer ²⁰¹⁹; O'Donnell et al. ²⁰²⁰; Panzer et al. ²⁰²³). Issues such as sole reliance on visual assessment of gross porotic abnormalities, which can have many potential and overlapping etiologies (Wapler et al. 2004), minimal clinical data on skeletal manifestations of anemia (Grauer 2019), a lack of clear diagnostic guidelines (Brickley 2024), and high rates of interobserver error when evaluating lesions (Anderson 2023; Buckberry et al. ²⁰²³; Santos et al. In press) have hindered differential diagnosis of anemia in paleopathological research.

As an alternative to qualitative visual observation of porous lesions, quantitative methods have occasionally been used to assess anemia in skeletal remains (Galea 2013; Morgan ²⁰¹⁴; Morgan et al. ²⁰²⁴; Panzer et al. ²⁰²³; Stuart‐Macadam ^{1987a, 1987b}; Zuckerman et al. ²⁰¹⁴). However, there is currently a limited amount of comparative metric data, and the establishment of metric standards is still in the initial stages (Anderson et al. 2021; Brickley ²⁰²⁴). No standardized frameworks have yet been presented for incorporating metric methods and visual evaluation for anemia diagnosis in skeletal remains. Establishing a systematic framework with consistent terminology based on an understanding of the underlying biological mechanisms that affect the expression of skeletal changes (Klaus and Lynnerup 2019) enables comparisons across diverse contexts and ensures that anemia differential diagnosis based on identifying marrow hyperplasia has a formal basis moving forward. Therefore, this study aims to integrate metric and visual methods of assessing marrow hyperplasia in the cranium using micro‐CT images. Initial quantitative data that can be used to establish measurement baselines are proposed, and recommendations on assigning a most‐likely diagnostic option for anemia in skeletal remains are presented. This study focuses on the cranium in individuals aged birth to 15 years, as skeletal manifestations of marrow hyperplasia are most widely studied in the cranium and are better understood in human skeletons undergoing growth and development.

2. Background

Anemia is a condition where the body cannot supply adequate oxygen to its tissues. Anemia can be caused by increased red blood cell (RBC) destruction, decreased RBC production, or excessive RBC loss (Walker et al. 2009) resulting from a variety of etiologies, including micronutrient deficiencies (e.g., iron, folic acid, vitamin B12, vitamin B9), parasitic infections, infectious disease, chronic illness, and genetic traits (e.g., sickle cell anemia or thalassemia) (McIlvaine ²⁰¹⁵; WHO ²⁰¹¹; WHO ²⁰²³). In response to a deficiency in RBCs, increased RBC production may occur, triggering the expansion of red blood cell‐producing bone marrow (Hoffbrand and Steensma 2020). This process, referred to as marrow hyperplasia, can potentially lead to skeletal changes as the bone adapts to accommodate the new volume of bone marrow (Brickley 2018). Increased trabecular separation to account for the new marrow is expected and thinned cortical bone and sparser/thinned trabeculae can develop as a result, leading to increased diploic/trabecular space width (Agarwal et al. 1970; Hoffbrand and Steensma ²⁰²⁰; Jaffe ¹⁹⁷²; Reynolds ^{1962, 1965}; Stuart‐Macadam ^1987a). In areas of the skeleton where cortical bone is already thin, such as the orbital lamina, the cortex may fully resorb in places, resulting in macroscopically visible porous lesions (Brickley 2018; Hengen ¹⁹⁷¹; Hoffbrand and Steensma ²⁰²⁰; Jaffe ¹⁹⁷²).

Not all forms of anemia lead to marrow hyperplasia, and the kinds of anemia associated with marrow hyperplasia are debated. Walker et al. (2009) argued that iron deficiency anemia cannot result in the excessive marrow proliferation that causes skeletal manifestations of marrow hyperplasia, a position contrasted by Oxenham and Cavill (2010). Additionally, McIlvaine (2015) and Schats (2023) pointed out the possibility for co‐occurrence of different forms of anemia, meaning that associating one form of anemia with skeletal changes is further complicated. In the current study, the term ‘anemia’ will be used as a general descriptor encompassing any type of anemia that can result in skeletal manifestations of marrow hyperplasia.

To recognize marrow hyperplasia when it does occur, identifying bone microarchitecture or trabecular width changes within the marrow space is necessary (Angel 1966; Brickley ²⁰²⁴; Wapler et al. ²⁰⁰⁴). The application of different visualization techniques, such as radiography (e.g., Stuart‐Macadam ^{1987a, 1987b}), CT imaging (e.g., Anderson et al. ²⁰²¹; Durdin ²⁰²⁰; Naveed et al. ²⁰¹²; O'Donnell et al. ²⁰²⁰; Panzer et al. ²⁰²³; Rivera and Mirazón Lahr ²⁰¹⁷; Saint‐Martin et al. ²⁰¹⁵; Zuckerman et al. ²⁰¹⁴) or micro‐CT imaging (e.g., Galea ²⁰¹³; Morgan ²⁰¹⁴; Morgan et al. ²⁰²⁴) allows for the marrow space to be evaluated non‐destructively, and for researchers to investigate the potential causes of cranial porosity.

As part of using imaging technology and visualization methods, quantitative methods of identifying skeletal changes associated with marrow hyperplasia have also been proposed and sometimes used (e.g., Durdin ²⁰²⁰; Galea ²⁰¹³; Morgan ²⁰¹⁴; Morgan et al. ²⁰²⁴; Panzer et al. ²⁰²³; Stuart‐Macadam ^1987a; Zuckerman et al. ²⁰¹⁴). Stuart‐Macadam (1987a) developed protocols for measuring the cranial vault and identifying pathological measurements, and various studies have used cranial vault thickness to investigate porous cranial lesions and anemia (e.g., Durdin ²⁰²⁰; Hengen ¹⁹⁷¹; Zuckerman et al. ²⁰¹⁴). Specific microstructure measurements, such as trabecular separation or trabecular number, are also likely to be affected by marrow hyperplasia and have been linked to the presence of porous orbital lesions in skeletal remains (e.g., Galea ²⁰¹³; Morgan ²⁰¹⁴).

Quantitative approaches often involve defining a range of non‐pathological ‘baseline’ measurements against which abnormal patterns can be identified, potentially signaling the presence of pathology. For example, in the clinical assessment of osteoporosis, a patient's bone densitometry results from certain skeletal sites are compared to those from a standard group of individuals without osteoporosis who have been measured on the same instrument using T‐scores to evaluate the significance of observed patterns (Binkley et al. 2014; Karaguzel and Holick ²⁰¹⁰). When using quantitative methods for assessing marrow hyperplasia related to anemia in skeletal remains, identifying a baseline group of individuals without anemia is challenging due to the lack of accompanying hematological data. However, the absence of any skeletal manifestations of marrow hyperplasia could be used as a conservative method to identify a baseline group.

Following the approaches to paleopathological diagnosis proposed by Mays (2018), quantitative analysis of archaeological bone can yield a “direct measurement of a diagnostic parameter” (Mays 2018, 13), which is fundamentally different from traditional lesion‐based methods used for anemia diagnosis (Brickley 2024). Although metric approaches require the application of skeletal imaging techniques which may be logistically challenging or prohibited in certain contexts due to cost and the accessibility of equipment, employing quantitative methods can help reduce discrepancies among observers and diminish overreliance on the visual assessment of gross porous lesions as the sole basis for anemia differential diagnosis (Brickley 2024).

3. Materials

Individuals from seven archaeological skeletal collections were evaluated: three samples were from 18th–19th century Quebec, Canada (N = 44), and four samples were from 18th–19th century Netherlands (N = 24) (Table 1). Consistent age‐related trends were observed at all sites. Consequently, all individuals were grouped into age categories (see Table 1) regardless of burial site so that metric analysis could be conducted with the largest sample size possible. For the purpose of differential diagnosis, all individuals underwent a full paleopathological macroscopic assessment with special focus on the identification of porotic orbital lesions. Age estimation was done by assessing dental development (Gustafson and Koch 1974), and all individuals under 15 years of age at the time of death with at least one intact orbit were included in the initial micro‐CT sample (N = 68, Table 1).

TABLE 1.

Summary of orbits that underwent micro‐CT imaging, by site and by age category.

Orbit micro‐CT sample‐site composition
Site	Context	Number of individuals
Saint‐Antoine	Quebec, CA, 1799–1855	19
Pointe‐aux‐Trembles	Quebec, CA, 1748–1878	17
Sainte‐Marie	Quebec, CA, 1709–1843	8
Alkmaar (GRK)	Netherlands, 1716–1830	8
Arnhem (ARJB)	Netherlands, 1650–1829	12
Eindhoven (EHV‐CK)	Netherlands, 1650–1850	3
Zwolle (ZW87)	Netherlands, 1675–1828	1

Orbit micro‐CT sample‐age composition
Age catergory	Number of individuals
0–1.9 years	38
2–5.9 years	19
6–15 years	11

Note: Table compiled from Arts and Altena ²⁰¹³; Baetsen ²⁰⁰¹; Baetsen and Zielman ²⁰²⁰; Bitter ²⁰⁰²; Clevis and Constandse‐Westermann ¹⁹⁹²; Ethnoscop ²⁰⁰⁶; Ethnoscop ^2016a; Ethnoscop ^2016b.

The Quebecois (Canadian) and Dutch collections were excavated prior to redevelopment projects. For Quebec, the collections represent settler communities, and no indigenous peoples are present. Permission for use of the collections has been provided by the City of Montréal, as a representative for the descendent communities. Dutch law does not prescribe any generic guidelines for the reburial of archaeological collections of any kind, so the decision to curate or rebury is made on a case‐by‐case basis. Following Dutch practice, community approval for the retention and research on past community members was gained for the sites studied. Permission for study was granted by the Faculty of Archaeology at Leiden University and Amsterdam UMC, University of Amsterdam.

4. Methods

4.1. Micro‐CT Imaging and Sample Preparation

Orbits were analyzed using the Nikon XTH‐225st micro‐CT scanner at the Ancient Images Lab of the Museum of Ontario Archaeology. Smaller single orbits were mounted inside Styrofoam cups and secured with low‐density green florist foam to minimize movement. For complete frontal bones that could not fit inside the Styrofoam cup, a block of green florist foam was carved to secure each bone. Samples were scanned using 95 kV, 165 μA, a molybdenum target, at 24.00 μm voxels. Each scan used 3141 frames, at a rate of 1 projection per second, and took 53 min. The 3D volume reconstruction was done in CT Pro 3D (v. 4.4.4), using a FDK filtered back projection reconstruction algorithm. Visualization, orbit microarchitecture measurements, and frontal bone ratio measurements were performed using Dragonfly (v. 2021.3) (https://www.theobjects.com/dragonfly/index.html).

4.2. Frontal Bone Ratio Measurements

Frontal bone ratios of trabecular space thickness to cortical thickness were calculated following Panzer et al. (2023). Measurements were taken on a mid‐sagittal view of the frontal bone, just above the nasal bones and the superior orbit (Figure 1). This area was only accessible in 16 of the scanned samples. Although the frontal sinus may interfere with these measurements in the oldest age category considered, this was not encountered in the current study. To assess the replicability of the ratio measurements, six frontal bones (~40% of the sample) were assessed by two observers (BM and ML), and the final ratio outcomes for each observer were compared. To evaluate if there was a relationship between ratio and age, a two‐tailed Pearson correlation test (α = 0.05) was performed using all 16 individuals who could be measured. The original research by Reynolds (1962, 1965) and Sebes and Diggs (1979) considered a ratio over 2.5 to be indicative of diploic expansion based on measurements from radiographs.

FIGURE 1.

Method of taking measurements for the frontal bone ratio calculation. (a) Arrow indicates location of frontal bone ratio measurement on 3D model. (b) Trabecular space measurement, which is a 2D linear measurement. Arrows indicate the length of the measurement. (c) Ectocranial cortical measurement, indicated by arrows. (d) Endocranial cortical measurement, indicated by arrows. Final ratio is calculated by dividing the measurement from (b) by the sum of the measurements from (c, d).

4.3. Visual Microstructure Assessment

Visual microstructure assessment of each orbit reconstruction was conducted with the criteria used in Table 2. Examples are provided in Figures 2, 3, 4. Cortical thinning of the orbital roof, increased trabecular separation, and trabecular thinning were all evaluated on a scale of 0 to 3, with 3 representing the most significant expression of each trait. Based on initial observations, changes were most evident in the inferior third of the orbit, just superior to the orbital lamina. Accordingly, features were scored in this location throughout the 3D orbit (see Figures 3 and 4).

TABLE 2.

Description of microarchitecture features for visual scoring.

Cortical thinning of orbital lamina
Score across entire orbital lamina
0‐Absent	No cortical thinning is present throughout the orbital lamina.
1‐Mild	Some patches of thinned cortex are present, but occur on less than 50% of the orbital lamina.
2‐Moderate	Mild cortical thinning is consistent across more than 50% of the orbital lamina, OR marked cortical thinning (as seen in scores of 3) is present in less than 50% of the orbital lamina.
3‐Marked	Marked cortical thinning is consistently present across more than 50% of the orbital lamina.

Increased trabecular separation
Score in lower third of marrow space above orbital lamina
0‐Absent	No abnormally large trabecular spacing is present throughout the marrow space.
1‐Mild	Abnormally large trabecular spacing is seen throughout less than 50% of the marrow space.
2‐Moderate	Mildly abnormally large trabecular spacing is seen throughout more than 50% of the trabeculae in the marrow space OR marked abnormally large trabecular spaces (score of 3) are present throughout less than 50% of the marrow space.
3‐Marked	Marked trabecular space enlargement is seen consistently throughout more than 50% of the marrow space.

Trabecular thinning
Score in lower third of marrow space above orbital lamina
0‐Absent	No trabecular thinning is present throughout the marrow space.
1‐Mild	Trabecular thinning is seen throughout less than 50% of the trabeculae in the marrow space.
2‐Moderate	Mild trabecular thinning is seen throughout more than 50% of the trabeculae in the marrow space OR marked trabecular thinning (score of 3) is present throughout less than 50% of the marrow space.
3‐Marked	Marked trabecular thinning is seen consistently throughout more than 50% trabeculae in the marrow space.

FIGURE 2.

Examples and description of cortical thinning for each scoring category. (a) Cortical thinning score of 0. (b) Cortical thinning score of 1. Arrow indicates a patch of mild thinning, which was present throughout this orbit. (c) Cortical thinning score of 2, where mild thinning is present over more than 50% of the visible cortex. (d) Cortical thinning score of 2, where marked thinning on is present on less than 50% of the visible cortex. (e) Cortical thinning score of 3, where marked thinning on is present on more than 50% of the visible cortex. In this case, the cortex is so thin that it has worn away in patches.

FIGURE 3.

Examples of increased trabecular separation for each scoring category, and location of scoring. (a) Location where orbit should be scored (lower third) is highlighted by the box. (b) Increased trabecular separation score of 0. (c) Increased trabecular separation score of 1. (d) Increased trabecular separation score of 2, where mildly abnormally large trabecular separation is seen throughout more than 50% of the marrow space. (e) Increased trabecular separation score of 2, where marked abnormally large trabecular separation (score of 3) is present throughout less than 50% of the marrow space. (f) Increased trabecular separation score of 3.

FIGURE 4.

Examples of trabecular thinning for each scoring category. (a) Location where orbit should be scored (lower third) is highlighted by the box. (b) Thinned trabeculae score of 0. (c) Trabecular thinning score of 1. The cortical bone is thin in general, and the trabeculae do not look markedly thin in comparison. (d) Trabecular thinning score of 2 where mild trabecular thinning is seen throughout more than 50% of the trabeculae in the marrow space. (e) Trabecular thinning score of 2 where marked trabecular thinning (score of 3) is present throughout less than 50% of the marrow space. (f) Trabecular thinning score of 3.

In addition to the three microarchitecture features, the possible causes of observable porous lesions were evaluated. Lesions could be assigned to any number of etiologies listed in Table 3 (examples provided in Figure S1). All features were evaluated for a final visual assessment of whether there was evidence of skeletal manifestations of marrow hyperplasia (Table 4). Individuals (n = 7) with internal evidence of post‐mortem damage (e.g., sediment within intertrabecular spaces) were excluded from visual analysis and further measurements, as taphonomic damage is almost certain to have affected measurements and scoring.

TABLE 3.

Description of porous lesion etiologies, and what they look like on micro‐CT images.

Porous lesion possible source
1‐Marrow hyperplasia	Porosity will originate from inside the orbital roof. The cortex may appear thin, and porosity may look as though the trabecular structure has penetrated through the cortex. Trabeculae may also extend through the cortex. Porosity may be irregular in shape and size.
2‐Vascular response	Signs of an inflammatory response are present. Individual pores, or channels, will be circular and less than 1 mm in diameter. Channels will generally originate on the outer surface of the cortex, although some may be more developed and will perforate through to the trabecular space, making this difficult to assess.
3‐Impaired mineralization	Impaired mineralization causes bone to look spiculated and porous. Pores may be relatively fine, but irregular in shape.
4‐Porous subperiosteal new bone formation	Layers of new bone may appear as highly porous. Observing new bone formation is key for this score.
5‐Post‐mortem damage	Porosity due to post‐mortem damage may be the result of corrosion, root damage, cracks, or weathering. Pores may have sharp, discolored edges with obvious soil inside.
6‐Normal growth and development	Porosity will be uniform and fine. Will be more common in those undergoing growth and development. In the orbit, this may be accompanied by lamination of bone, as the bone develops through mesenchymal ossification (Tawfik and Dutton 2018).

Note: Orbits can be scored with more than one lesion source.

TABLE 4.

Orbit visual assessment categories and features.

	Expected features
Evidence of Skeletal Manifestations of Marrow Hyperplasia	Porotic lesions linked to marrow hyperplasia (i.e., source score of 1) are present. Increased trabecular separation is present to at least some extent, and is accompanied by some level of cortical/trabecular thinning (scores ≥ 1).
No evidence of Skeletal Manifestations of Marrow Hyperplasia	Multiple key features expected in marrow hyperplasia (cortical/trabecular thinning, increased trabecular separation, porous lesions caused by cortical thinning) are all absent or mild. Increased trabecular separation is mild or absent (scores of 0–1). Cortical thinning or trabecular thinning may be present, but are more mild (scores of 1–2) and not accompanied by increased trabecular separation.

The visual assessment serves as a conservative method of identifying baseline individuals, as those with no evidence of skeletal manifestations of marrow hyperplasia were assigned to the ‘no evidence’ category. Visual assessment using this rubric is meant to be a simple, binary assessment that can contribute to a final assessment, but should be combined with other methods when assigning a most‐likely diagnostic option. If evaluating more‐variable single sections/slices of bone, the degree of certainty assigned must be viewed as a minimum, as there is potential that different conclusions could be made if additional slices are viewed.

To test for interobserver reliability, and to evaluate the scoring rubric's ease of use, seven (~10%) of the orbit reconstructions were evaluated by three additional observers with experience in evaluating porous lesions and/or micro‐CT images, and four who had some experience in evaluating skeletal remains, but not necessarily in metabolic bone disease or porous lesions. Cohen's Kappa coefficient was used to evaluate average interobserver agreement for the final visual assessments.

4.4. Orbit Microarchitecture Measurements

During macroscopic observation of orbits, the location of porous lesions was recorded on orbit diagrams. A 10 × 10 × 15 mm sample placed on the lateral two‐thirds of the orbital roof was used for microarchitecture measurements, as this was the most common lesion location from the diagrams, and because active marrow is present in only this section of the orbit for individuals over 8 years (O'Donnell et al. ²⁰²³). Instructions on placing this sample box are provided in Figure 5. Orbit measurements were taken using Dragonfly's Bone Analysis Wizard tool using Buie segmentation (Buie et al. 2007). These measurements are done using 3D volume reconstruction, so measurements are an average from across the entire bone, which differs from the 2D point measurement used for ratio assessments (e.g., Figure 1). Measurements taken included average cortical thickness of the orbital lamina (CtThOL, mm), average trabecular separation (TbSp, mm), and average trabecular thickness (TbTh, mm). To test intraobserver reliability in this method, the entire process of obtaining measurements was repeated on a random selection of seven (~10% of the sample) orbit reconstructions 1 week later, and differences between the two sets of measurements were calculated. Two‐tailed Pearson correlation tests at a 0.05 significance level were used to assess the relationship between an individual's midpoint age estimate and orbit microarchitecture measurements for the entire sample.

FIGURE 5.

Placement for the measurement sample box, where microarchitecture measurements were taken. The sample box measures 10 × 10 × 15 mm. The longest edge of the rectangular box was aligned parallel to the orbital lamina, and the lateral edge of the rectangle was aligned with the zygomatic process of the frontal bone, just before where the supraorbital margin begins to curve inferiorly (a, inferior view of right orbital lamina). On the medial‐lateral 2D view, the centre of the box was aligned with the orbital lamina, and oriented such that the most inferior edge was parallel to the centre of the orbital roof (b).

The method used to evaluate the relationship between microstructure measurements and marrow hyperplasia related to anemia used in this study was derived from the T‐score method used in osteoporosis diagnosis. T‐scores are calculated as: T‐score = (patient bone mineral density –normal mean bone mineral density)/standard deviation of the normal population. Scores of −1 to −2.5 standard deviations (SD) at particular sites are considered osteopenic, while scores of < −2.5 SD are considered osteoporotic (Binkley et al. 2014; Kanis et al. ¹⁹⁹⁴). To carry out the same procedure in this study, the average and standard deviation of all measurements for each age category were calculated using only the visually identified baseline group. T‐scores for all individuals not included in that group were calculated, and scores of ≤ −1 for average cortical thickness of the orbital lamina (CtThOL) and average trabecular thickness (TbTh) were considered evidence of cortical/trabecular thinning, and scores of ≥ 1 for average trabecular separation (TbSP) were considered evidence of increased trabecular spacing. In osteoporosis assessment, different T‐score thresholds are used for different skeletal sites (Binkley et al. 2014), and more metric data will help to refine these thresholds for the orbit in the future.

4.5. Assessment of Possible Anemia Status

Different degrees of diagnostic certainty exist when evaluating pathological lesions present on skeletal remains, and the framework by Appleby et al. (2015) sets out terminology that can capture these differences in certainty. To apply this framework to anemia assessment, we outlined important considerations in Figure 6, and defined terminology for assigning a most‐likely diagnostic option for anemia in Table 5. The basic steps in Figure 6 and the standardized terminology set out by Appleby et al. (2015) were used to construct a decision tree (Figure 7) to facilitate consistent evaluation and assignment of diagnostic options for the current study, where frontal bone ratios, visual assessments, and microstructure T‐scores were evaluated. Genetic anemias were not thought to be a factor for any of the individuals in this study, but in contexts where genetic anemias are considered, skeletal manifestations of marrow hyperplasia are frequently far more severe, and there may be additional bone abnormalities (Hershkovitz et al. 1997; Lewis ²⁰¹²). Assessments on the assigned most‐likely diagnostic option were made following the decision tree in Figure 7, and three observers repeated assessments on 20% of the sample to evaluate agreement. Although Figure 7 is specific to the data used in this study, it could be used as a model and adapted for future use depending on the data available.

FIGURE 6.

Flow diagram showing the various considerations for assigning diagnostic certainty during anemia assessment. D = diagnostic of, HC = highly consistent with, N = not consistent with.

TABLE 5.

Categories of most‐likely diagnostic options for anemia.

Category of certainty	Description
Diagnostic of	Features and metric changes could only have been caused by anemia. Visual features of the bone (e.g., hair‐on‐end appearance of the cranium) should be highly indicative of marrow hyperplasia. Metric data should strongly reflect skeletal manifestations of marrow hyperplasia, and be supported by robust clinical data.
Highly consistent with	Features and metric changes could have been caused by anemia, but a few other possible causes exist. Internal visual changes expected for skeletal manifestations of marrow hyperplasia (e.g., for the orbit, this includes cortical thickness, trabecular thickness, and trabecular separation) must be observable, and metric data supports skeletal manifestations of marrow hyperplasia.
Consistent with	Features and metric changes may be caused by anemia, but there are many other possible causes. Internal visual changes (e.g., for the orbit, this includes cortical thickness, trabecular thickness, and trabecular separation) are present but may be mild (scores of < 2). Metric data does indicate the potential for skeletal manifestations of marrow hyperplasia, but many other possible causes cannot be excluded as contributing to the observed features and measurements.
Not consistent with	Features and metric changes expected for skeletal manifestations of marrow hyperplasia caused by anemia are absent, or could not have been caused by anemia.

FIGURE 7.

Decision tree demonstrating how anemia assessments were done using the available data from this study.

Other pathological conditions (e.g., scurvy, rickets) or anatomical variations that could have resulted in microarchitectural changes were excluded through the process of differential diagnosis, based on a paleopathological analysis of the full skeleton (Klaus 2017; Mays ²⁰²⁰). When other conditions could not be excluded as contributing to the observed changes, this was reflected in the assigned diagnostic option. Although the protocol used in this study prioritizes micro‐CT analysis of the orbit, it is important that the entire skeleton is assessed macroscopically, so that other possible causes of bone microarchitecture changes could be evaluated.

4.5.1. ‘Diagnostic of’

At the ‘diagnostic of’ degree of certainty, only anemia could have caused the features and metric skeletal changes observed in an individual. The term ‘diagnostic of’ should be used in cases where observed features and measurements can be directly linked to those found in known cases of anemia. Visual changes that meet this criterion include hair‐on‐end trabeculae, and metric data must be supported by robust clinical data (e.g., ratio measurements derived from clinically‐diagnosed individuals). To use the ‘diagnostic of’ certainty category, other pathological conditions (e.g., osteoporosis, metabolic conditions, neoplastic conditions) or growth‐related features must be excluded as contributing to the observed metric and visual changes.

4.5.2. ‘Highly Consistent With’

In cases where observed bone metrics and features could have been caused by anemia, and few other possible causes for changes to exist, the term ‘highly consistent with’ can be used. The term ‘highly consistent with’ may be appropriate when observed visual features and metric changes are indicative of skeletal manifestations of marrow hyperplasia, but a limited number of other possible causes cannot be excluded with available evidence. Strong visual indicators and high‐level metric data that reflects skeletal manifestations of marrow hyperplasia should be present.

4.5.3. ‘Consistent With’

The ‘consistent with’ category can be used when metric and visual changes could have been caused by anemia, but several other possible causes cannot be excluded. Changes may be non‐specific, and there are many other possible causes of the observed features. When metric parameters are contradictory (e.g., the TbSp T‐score is over 1, but so is the CtThOL T‐score), ‘diagnostic of’ and ‘highly consistent with’ should not be used. Confounding factors, such as age, origin of reference data (i.e., imaging source), and other pathology, which could be causing the contradictions, should be considered in order to determine if ‘consistent with’ is still appropriate.

4.5.4. ‘Not Consistent With’

When observed features or metric data could not have been caused by anemia, the term ‘not consistent with’ should be used. This term may be appropriate in cases where skeletal manifestations of marrow hyperplasia can be evaluated, but there is no observable visual evidence, or no metric deviation from baseline data, such as the microarchitecture measurements in the current study, or ratio measurements derived from clinically‐assessed individuals. It may also be appropriate in cases where anemia can be excluded as a contributing factor to observed skeletal changes based on all available evidence.

4.5.5. Co‐Morbidity and Occurrence With Other Pathological Processes

Co‐occurrence of anemia and other conditions can contribute to variation in the expression of skeletal changes and may affect the most‐likely diagnostic option that can be assigned. If skeletal indicators of other conditions are present in addition to metric changes indicative of marrow hyperplasia related to anemia, then co‐occurrence should be considered. Conditions should be individually evaluated following established guidelines for assigning diagnostic options (e.g., Appleby et al. ²⁰¹⁵; Brickley and Morgan ²⁰²³). Expression of metric and visual changes when conditions co‐occur will depend on a variety of factors, such as order of disease development, severity, and age. Whether co‐occurrence may have affected metric data, or whether skeletal manifestations of marrow hyperplasia could have affected the expression of skeletal changes for other conditions should be considered. The final most‐likely diagnostic options for anemia and any co‐occurring conditions should be reported.

Skeletal manifestations of marrow hyperplasia may exacerbate the lesions caused by other diseases (Wapler et al. 2004). For example, cortical thinning of the orbital roof could enhance porosity caused by other conditions. Similarly, conditions that also affect bone microarchitecture may exacerbate any metric microarchitecture or ratio differences. Assessment of diagnostic options must therefore consider the idea that other diseases can no longer be excluded as causing the observed metric changes.

5. Results

5.1. Interobserver Error

For the orbit microstructure measurements, the average difference between the original and the repeated measurements was ≤ 0.005 mm for all three measurements, indicating a high level of repeatability in the sampling procedure and Dragonfly software. For the frontal bone ratio measurements, there was agreement on the ratio assessment (i.e., whether they were over or under 2.5) in five of six cases. A third set of measurements was taken for the individual who did not show agreement, and the average was used for the final frontal bone ratio.

The Cohen's Kappa coefficient across all observers in the visual rubric assessment was 0.34 (fair agreement). Between the experienced observers only, the average Cohen's Kappa was calculated to be 0.76 (substantial agreement) (McHugh ²⁰¹²), demonstrating that experience in evaluating porous lesions and familiarity with micro‐CT analysis did help in ensuring consistency when using the visual scoring rubric. Orbits with more extreme changes (e.g., Figure 8) showed the highest level of agreement, while those with more ambiguous changes showed less agreement. When using the decision tree in Figure 7 to assign a most‐likely diagnostic option, observers unanimously agreed on all assessments during the blind testing, indicating high replicability.

FIGURE 8.

Micro‐CT reconstruction of orbit (sagittal view, anterior towards the right) from individual 7A11‐S57 (approximately 1 year old), who displays significant trabecular and cortical thinning, and increased trabecular separation. All observers agreed that skeletal manifestations of marrow hyperplasia were present in this individual.

5.2. Age and Bone Microarchitecture Measurements

The three microarchitecture measurements are significantly associated with age (CtThOL: Pearson R = −0.73, p value ≤ 0.001; TbSp: Pearson R = −0.56, p value ≤ 0.001; TbTh: Pearson R = −0.49, p value ≤ 0.001). For all three measurements, a positive relationship exists (Figure 9), meaning that as age increases, so does average cortical thickness, trabecular thickness, and trabecular separation. For TbTh, the overall range of measurements was narrow (0.14–0.26 mm), while the range of variation is much wider for both CtThOL (0.11–0.52 mm) and TbSp (0.22–0.71 mm). In contrast, the frontal bone ratios are not significantly associated with age (Pearson R = −0.10, p value = 0.711).

FIGURE 9.

Bone microarchitecture measurements plotted against age estimate midpoints. Trendlines demonstrate the positive relationship between age and increased measurements.

5.3. Anemia Assessment Results

In total, 27 individuals were assessed as showing no visual evidence of skeletal manifestations of marrow hyperplasia and had frontal bone ratio measurements under 2.5. These individuals were therefore included in the baseline sample to calculate T‐scores.

Ten individuals were assessed as having evidence of skeletal manifestations of marrow hyperplasia using the metric and visual data in conjunction with the decision tree in Figure 7. Table 6 compiles all metric and visual analyses and presents the most‐likely diagnostic option assigned for each case. Measurements for the baseline group are presented in Table S1. All raw measurements and visual scores are presented in Table S2.

TABLE 6.

Compiled results of all evaluations, plus assessments for individuals with visual evidence of skeletal manifestations of marrow hyperplasia, which manifests as cortical/trabecular thinning and widened intertrabecular separation.

Skeleton ID	Site	Age (years) a	Cortical thinning score	Trabecular separation score	Trabecular thinning score	Overall visual	Frontal bone ratio b	CtThOL T‐score c	TbTh T‐score c	TbSp T‐score d	Anemia assessment
7A2‐S6	Pointe‐aux‐Trembles, CA	0.38	2	1	1	EMH	4.6	−0.8	−1.2	0.0	N
7A9‐S12	Pointe‐aux‐Trembles, CA	0.38	3	1	2	EMH	—	−0.2	−0.7	−0.9	N
20F‐S39	St‐Antoine, CA	0.50	2	1	1	EMH	2.0	−0.5	−0.7	0.5	N
7A2‐S10	Pointe‐aux‐Trembles, CA	0.50	3	1	2	EMH	—	−0.3	−0.3	−0.1	N
7A11‐S59	Pointe‐aux‐Trembles, CA	0.70	2	2	1	EMH	—	−0.3	0.2	0.9	N
ARJB‐V1625	Arnhem, NL	0.75	1	1	1	EMH	4.0 e	−0.2	−0.7	0.8	N
GRK‐V807	Alkmaar, NL	1.00	3	1	2	EMH	—	−1.4	−1.6	0.9	N
7A11‐S57	Pointe‐aux‐Trembles, CA	1.00	3	3	3	EMH	—	−0.1	−0.7	1.7	H
21E‐S8	St‐Antoine, CA	1.40	1	3	1	EMH	0.9	0.8	−1.2	1.6	C
GRK‐V326	Alkmaar, NL	1.50	2	2	2	EMH	—	−0.5	−1.6	0.5	N
GRK‐V532	Alkmaar, NL	1.50	3	2	1	EMH	—	−1.4	−1.6	−0.4	C
ARJB‐V1777	Arnhem, NL	1.50	3	1	2	EMH	4.5	−1.0	−1.2	−0.4	N
7A2‐S22	Pointe‐aux‐Trembles, CA	1.50	2	2	1	EMH	2.8	−0.1	1.1	1.6	C
7A2‐S24	Pointe‐aux‐Trembles, CA	1.50	2	2	2	EMH	—	−0.3	−0.3	1.2	H
17L‐S4	St‐Antoine, CA	1.00	2	2	3	EMH	—	0.2	−0.3	2.0	C
2E3 f	Ste‐Marie, CA	1.50	1	2	1	EMH	—	1.9	1.6	2.1	N
1G2	Ste‐Marie, CA	1.50	2	1	2	EMH	1.4	2.0	0.6	1.4	N
20A‐S3	St‐Antoine, CA	1.50	2	2	2	EMH	—	1.3	0.2	2.9	C
7A2‐S32	Pointe‐aux‐Trembles, CA	1.58	2	2	1	EMH	4.9	−0.2	0.2	1.4	H
ARJB‐V478	Arnhem, NL	2.00	2	1	1	EMH	—	−1.7	−0.5	0.4	N
ARJB‐V1812	Arnhem, NL	2.50	2	1	1	EMH	3.4	−1.1	−0.05	0.9	N
2G3	Ste‐Marie, CA	2.50	3	2	2	EMH	6.9	−1.9	−0.3	0.9	C
ARJB‐V667	Arnhem, NL	3.00	0	2	3	EMH	—	0.8	−0.6	0.9	N
21E‐S4	St‐Antoine, CA	3.00	0	2	1	EMH	—	1.5	−0.5	1.0	C
21F‐S5	St‐Antoine, CA	3.00	1	2	2	EMH	—	−1.2	−0.3	1.3	H
ARJB‐V1319	Arnhem, NL	4.00	1	2	2	EMH	—	0.6	−0.5	1.7	H
1G6	Ste‐Marie, CA	4.00	3	2	3	EMH	—	−0.3	−0.4	1.2	H
20F‐S2	St‐Antoine, CA	7.00	2	2	2	EMH	4.0	−1.4	−0.1	0.4	C
21E‐S14	St‐Antoine, CA	7.00	0	2	1	EMH	—	1.1	−0.4	0.8	N
21R‐S5	St‐Antoine, CA	7.60	2	3	3	EMH	0.6	−1.1	−1.6	0.8	C
30R‐S7	St‐Antoine, CA	7.6	2	3	3	EMH	3.5	0.8	−1.3	3.5	H
ARJB‐V1258	Arnhem, NL	9.00	1	2	0	EMH	—	−0.1	−2.2	1.2	H
ARJB‐V1735	Arnhem, NL	9.50	2	2	1	EMH	—	−1.6	−1.3	2.6	H
21U‐S6	St‐Antoine, CA	10.25	2	3	2	EMH	—	−3.0	−0.1	1.8	H

Abbreviations: C = Consistent with, CtThOL = Average cortical thickness of the orbital lamina, D = Diagnostic of, EMH = Evidence of skeletal manifestations of marrow hyperplasia, H = Highly consistent with, N = Not consistent with, TbSp = Average trabecular separation, TbTh = Average trabecular thickness.

^a Midpoint age estimate.

^b Bolded values are over 2.5, which were considered pathological by Panzer et al. (2023) and Sebes and Diggs (1979), based on radiographic ratios from Reynolds (1962, 1965). See Section 5 for further discussion on how frontal bone ratios were used in this study.

^c Bolded values ≤ −1, and are considered evidence of cortical or trabecular thinning.

^d Bolded values are ≥ 1, and are considered evidence of increased trabecular separation.

^e The frontal bone ratio was not considered for diagnosis, due to poor preservation of the cortical bone.

^f As this individual had significantly positive values for all T‐scores, their positive TbSp score is not likely to be indicative of anemia and may reflect that their bone microarchitecture was generally more developed for their age group.

6. Discussion

Anemia assessment in the current study used a variety of metric data and followed a consistent diagnostic framework that considered the biological mechanisms of bone microarchitecture changes, relative bone ratios, and lesion development. The quantitative approach taken here is rooted in the biological approach (Mays 2018) and emphasizes consideration of the underlying process (i.e., marrow hyperplasia) that contributes to metric changes during anemia. In contrast, scoring and evaluating gross porous lesions is not always useful for assessing lesion etiology (Anderson 2023; Brickley ²⁰²⁴), and recent studies have demonstrated the high amount of interobserver error that exists when recording porotic lesions (Anderson 2023; Santos et al. In press). Additionally, not all porous lesions are caused by skeletal manifestations of marrow hyperplasia (Wapler et al. 2004); many individuals in the current study, such as individual 7A9‐S21, showed no internal visual evidence of microarchitectural changes related to anemia, but did have porous orbital lesions consisting of new bone formation and cortical porosity. These issues demonstrate the importance of understanding skeletal manifestations of marrow hyperplasia and analyzing internal changes as part of anemia diagnosis in paleopathology (Grauer 2019; Brickley ²⁰²⁴).

Using a standardized framework and terminology for paleopathological diagnosis helps ensure consistency and rigor in the process (Buikstra et al. 2017; Klaus and Lynnerup ²⁰¹⁹). For anemia, the proposed framework can help ensure that the pathogenesis of lesions and skeletal changes (i.e., skeletal manifestations of marrow hyperplasia) is considered more consistently and allows for improved comparisons of anemia across contexts. Metric methods help to improve reproducibility and reduce subjectivity, further improving consistency. However, metric measurements should not confer a ‘diagnostic of’ assessment without considering the other factors that may have contributed to calculated metric differences.

In the current study, for example, frontal bone ratios over 2.5 did not necessarily confer a high degree of certainty, even though this threshold value is derived from individuals clinically diagnosed with anemia (Reynolds 1962, 1965; Sebes and Diggs ¹⁹⁷⁹), and is therefore a parameter that could be used to assign the ‘diagnostic of’ degree of certainty (Table 5). However, this value is derived from measurements taken on radiographs, which have been found to differ from those performed during micro‐CT analysis (Belgın et al. 2019). The greater resolution of micro‐CT analysis allows for more accurate placement of the cortical bone borders, which can result in smaller cortical thickness measurements, and therefore larger ratios. For this reason, frontal bone ratios had to be interpreted conservatively and were not as highly prioritized in the decision tree (Figure 7) created for this study, but could be for future analyses where imaging modalities are more comparable. Improved consistency and diagnostic rigor are achievable using metric data and standardized frameworks, but researchers must still consider confounding technical and biological factors as part of diagnosis.

As more metric data become available for comparison, it is expected that some of the assessments presented in the current study may change, and additional cases could be assigned as ‘diagnostic of.’ Further adjustments could result from improved frontal bone ratio comparisons, and adding more data to baseline measurements may change T‐score calculations. Metric evaluation of anemia in skeletal remains is still in the early stages, and may be difficult to compare across studies due to variation in instruments or measurement protocols. In the future, new visual or metric criteria from a variety of skeletal areas may also be sufficient to propose most‐likely diagnostic options. Accumulating measurement data is a key step towards differentiating between normal variation and possible pathology, and will be a necessary part of adopting these techniques in the future. Furthermore, while the decision tree in Figure 7 is specific to the observations produced by the current study, it could be adapted for future research depending on the metric data that is available and the relationship of that data to the recommended terminology in Table 5 and Section 4.5, and the considerations in Figure 6. For example, in our decision tree, trabecular separation T‐scores were prioritized, as cortical thinning was thought to be most sensitive to taphonomic factors, and trabecular thickness was less variable than the other two scores, but studies in other contexts may produce data that support the adoption of different decision‐tree configurations.

6.1. Interobserver Error

The results from interobserver error testing of the visual assessment rubric show high agreement among individuals with experience in evaluating metabolic conditions/porotic lesions, and less agreement for those with limited experience of evaluating gross porous lesions. Microarchitecture measurements had strong agreement and were highly replicable when using the Dragonfly program. Frontal bone ratios showed good agreement but did introduce additional subjectivity, as the users had to determine where to take the measurements and assess the borders of the cortical bone. For the one individual in which there was no agreement on the outcome of the ratio, there was considerable weathering of the cortical bone, resulting in a high degree of variation in cortical thickness. Diploë measurements were consistent between observers, but differences in cortical thickness measurements affected the final ratios. The frontal bone ratio for this individual was not considered for the final diagnosis, as it may have been skewed by preservation. It is recommended that researchers repeat frontal bone ratio measurements a minimum of two times and use an average to reduce error.

6.2. Age and Bone Microarchitecture Changes

Age was found to be a major factor when considering bone microarchitecture measurements, which is expected as part of normal bone growth and development. The orbit undergoes 50% of its expected growth by the age of 2 years (Berger and Kahn 2012), meaning that individuals under this age may show significantly different microarchitecture and increased variation compared to those who are past this developmental threshold. The significant positive relationship between age and microstructure measurements means that age must be considered when selecting a metric baseline for determining pathological ratios and/or T‐scores. It is expected that possible cases of anemia in individuals under 2 years of age may be less likely to be classified as ‘diagnostic of’, since they are less likely to have straightforward metric results.

Frontal bone ratios were not significantly associated with age since ratio calculations help to standardize values. Non‐ratio measurements of cranial vault thickness have been found to increase with age in those undergoing growth and development (De Boer et al. 2016), but Panzer et al. (2023) found that frontal bone ratio values did not appear to be associated with age.

6.3. Future Directions

Establishing baseline measurements and corresponding ratios of normal and abnormal orbital characteristics is crucial for evaluating marrow hyperplasia and anemia‐related changes in the cranium with a greater degree of confidence and rigor. Additional measurements across various age categories are especially needed to further explore how skeletal manifestations of marrow hyperplasia are affected by growth and development. In this study, individuals older than 2 years were underrepresented, so more data on older age groups will be essential, particularly for calculating more robust T‐scores. Evaluating microstructure measurements in individuals clinically diagnosed with anemia would be ideal and would contribute greatly to establishing baseline measurements.

Specific investigation of measurement differences between micro‐CT, CT, and conventional radiography will be essential for improving the use of frontal bone ratios. Micro‐CT and CT‐derived measurements tend to be more equivalent (Lillie et al. 2015), but data on the amount of error that exists when comparing ratios derived via conventional radiographs versus micro‐CT analysis will help to clarify when certain metric data can be used, and what limitations researchers should be aware of when making such comparisons. Previous studies have measured frontal bone thickness at different locations (e.g., Stuart‐Macadam ^1987a; De Boer et al. ²⁰¹⁶; Rivera and Mirazón Lahr ²⁰¹⁷), and it will also be important to establish if differences exist between specific anatomical sites.

7. Conclusion

This study demonstrates that metric observations can provide another line of evidence for anemia diagnosis in skeletal remains, particularly when assessing crania of growing individuals. The results presented here reiterate the importance of specifically evaluating changes related to marrow hyperplasia for anemia assessment and highlight the utility of methods that allow for analysis of the internal marrow space. Developing a standardized framework for the differential diagnosis and reporting of anemia allows for continued progress by ensuring consistency and repeatability across studies. Evaluation of porous cranial lesions can play a role in this framework, but is not the primary feature for assessing anemia. By incorporating metric data into the diagnostic process, researchers can better understand the relationship between marrow hyperplasia and anemia, leading to improved diagnostic accuracy and rigor in skeletal remains.

Author Contributions

Brianne Morgan: conceptualization (lead), data curation (lead), formal analysis (lead), funding acquisition (supporting), investigation (lead), methodology (lead), software (lead), validation (supporting), writing – original draft (lead), writing – review and editing (equal). Meghan Langlois: formal analysis (supporting), methodology (supporting), validation (equal), writing – review and editing (equal). Rachel Schats: project administration (equal), resources (equal), validation (equal), writing – review and editing (equal). Alie E. van der Merwe: project administration (equal), resources (equal), writing – review and editing (equal). Isabelle Ribot: project administration (equal), resources (equal), writing – review and editing (equal). Andrea Waters‐Rist: funding acquisition (supporting), project administration (equal), resources (equal), writing – review and editing (equal). Megan B. Brickley: conceptualization (supporting), funding acquisition (lead), project administration (equal), resources (equal), supervision (lead), writing – review and editing (equal).

Ethics Statement

This research has received ethics approval from the Human Tissue Committee of the Hamilton Integrated Research Ethics Board (HiREB) (project ID 16436). Research outcomes will be shared with the descendant communities in Quebec and the Netherlands at the close of the SSHRC‐funded project.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Examples of different porous lesion etiologies seen on micro‐CT reconstructions. (a) Cortical thinning and subsequent porosity due to marrow expansion. (b) Perforations due to vascular response. (c) Gaps in the bone due to impaired mineralization. (d) Periosteal new bone formation on the surface of the orbit, indicated by the arrow. (e) Dirt and minerals inside the trabecular space mean that taphonomy cannot be excluded as a possible source of lesions.

Table S1: Visual assessment scores for individuals used as part of the baseline group for measurement comparisons.

Table S2: Raw microstructure measurement data.

Morgan, B. , Langlois M., Schats R., et al. 2025. “A Framework for Anemia Differential Diagnosis in Paleopathology Incorporating Metric Methods.” American Journal of Biological Anthropology 188, no. 1: e70125. 10.1002/ajpa.70125.

Data Availability Statement

Data are available in the Supporting Information of this article.