The use of digital photographs for the diagnosis of hand osteoarthritis: the AGES-Reykjavik study

Hand osteoarthritis (HOA) is an important cause of pain and disability in the middle aged and elderly [1–3]. In addition to symptoms directly related to the hand itself, it is also related to osteoarthritis at other sites and there is evidence that the presence of HOA increases the propensity for the development and progression of both knee and hip osteoarthritis (OA) [4]. This relationship, usually referred to as "generalized osteoarthritis" (GOA) has proved hard to define, despite some serious efforts [5, 6]. The reasons are complex and among them are gender and age related differences and problems with measures of the disease activity and progress. Imaging is particularly problematic, radiography has been considered the gold standard for diagnosis and monitoring of HOA, but the method is basically a delayed reflection of damage and repair caused by OA, showing only moderate association with symptoms and giving little information about prognosis [7, 8]. Other imaging modalities such as scintigraphy, ultrasound and magnetic resonance imaging (MRI) have the advantage of being more dynamic and thus give a better indication of disease activity in the various tissues of the joints. All of these methods may be useful, but for various reasons including cost, availability and interpretation, none of them is likely to replace radiography in larger studies. The use of hand photographs as a screening method for HOA has been investigated in a few studies but found to be less sensitive than radiography [9–11].

The pathogenesis of osteoarthritis is still obscure and theories are evolving. Previously, OA was considered a degenerative disease, and simply an inevitable part of ageing. Now, however, OA is increasingly viewed as a dynamic process, one that is metabolically active, with the process of the disease involving both destruction and repair that may be triggered by a variety of biochemical as well as mechanical insults [12].

Members of our group have considerable experience in using different diagnostic methods for HOA including clinical examination [13], scintigraphy and MRI, both cross-sectional and longitudinal [14–16] and radiography with both the Kellgren-Lawrence and the Verbruggen-Veys scoring systems [17, 18]. Different subsets have also been studied including hypermobile subjects and individuals with known genetic mutations [19, 20]. Based on this experience, we think it may be time to consider HOA as trait or disease burden which is not well measured by any of the current methods and that larger studies involving investigations of other organ systems may be necessary for further understanding of the systemic factors involved in osteoarthritis. There is evidence that HOA is a systemic disorder interacting with other organ systems and sharing pathophysiological pathways with other conditions such as atherosclerosis [21, 22]. This association was discovered only by linking photographic HOA information to the extensive database of information gathered in the AGES-Reykjavik Study [21].

Hand photography is a simple and inexpensive imaging method involving no ionizing radiation or discomfort. This project of using photographs to diagnose and grade HOA is based on the belief that by standardizing the taking and the reading of the photographs, we would acquire a method that would be very suitable for use in large studies, allowing a wide variety of association studies. This article describes the process of establishing this method in four main steps: (1) To develop a standardized reproducible grading system for the diagnosis of hand osteoarthritis from high quality hand photographs. (2) To compare the photographic scoring system with the two main diagnostic methods currently used; radiography and clinical examination. (3) To analyze the relationship between the three methods and hand pain, the main symptom of hand osteoarthritis (4). To construct a shortened version of the photographic method as a diagnostic tool suitable for very large samples, where a simple global assessment of HOA may be preferable.

The Age, Gene/Environment Susceptibility-Reykjavik Study (AGES-Reykjavik) study is a population based study of approximately 5700 elderly individuals from the 40 year long Reykjavik study. They were aged between 66 and 96 and randomly recruited between 2002 and 2005. Details of the investigations are described in the study's baseline article [23]. The participants had extensive laboratory and imaging investigations including high quality hand photographs.

Photographic reading procedure

All photographs were taken with a Fuji Finepix 6800 zoom camera with images taken at 2800 × 2200 pixels. The camera was mounted on a tripod with a fixed distance to a black velvet board with markers for thumb positioning. The quality of the digital images and correct thumb positioning is important in order for the readers to be able to visually assess the degree of enlargement and deformity.

A photographic scoring system was developed. Initially, the readers (HJ, GPH) examined a few photographs at a time recording a number of variables that were suspected to be related to hand osteoarthritis in each joint. Each individual hand joint was graded separately. Subsequently the observers results were compared with each other and with hand radiographs. After a number of sessions, the variables most likely to be associated with clinical and radiographic hand OA were determined. Several factors were found to be of importance, such as hard tissue enlargement, visible soft tissue swelling, position and deformity.

The distal interphalangeal (DIP) and the proximal interphalangeal (PIP) joints were scored on a 0-3 scale as follows: 0 = no evidence of OA, 1 = suspected but not definite OA, 2 = definite moderate OA, 3 = severe OA.

For the DIP joints, the deformity of a joint without hard tissue enlargement did not justify the diagnosis of hand OA on its own but when deformity was severe (> 30°), the recorded score was raised by one (1) unit (to the maximum score of 3).

Reference photographs for the grading of DIP and PIP joints are shown in Figures 1 and 2. For uniformity of presentation the right second DIP and third PIP joints are shown.

For assessment of OA of the first carpometacarpal (CMC1) joints, a slightly different approach was needed. Two different findings, enlargement of the joint and abnormal positioning, were related to OA in that joint. Abnormal positioning reflects palmar migration of the base of the first metacarpal bone and is reflected on photography by a number of factors, including disappearance of the normal configuration of the CMC1 joint, medial rotation of the thumb showing increased folding of the skin over the first metacarpal joint (MCP1) and sometimes hyperextension of that joint.

Both enlargement and position were scored on a 0-3 scale, (0 = no evidence of OA, 1 = suspected but not definite OA, 2 = definite moderate OA, 3 = severe OA.) and subsequently added, giving a score of 0-6 which was translated into a 0-3 score as follows: (0 = Normal joint, 1 = Doubtful OA, 2-3 = Definite OA and 4 + = Severe OA). Reference photos for the CMC1 joints are shown in Figure 3.

The reference photographs were subsequently used as assistive tools in the reading of all photographs. Interreader agreement measured by ICC was good (average 0.83) and intrareader agreement (50 photographs re-read at 4 weeks intervals, GPH) was excellent (ICC 0.89) (Table 1). Finally the readers re-examined all discordant readings and decided upon a consensus score.

	Joint	Interobserver	Interobserver	Intraobserver
		kappa	ICC	ICC
Left	DIP5	0.83	0.85	0.90
	DIP4	0.87	0.83	0.91
	DIP3	0.85	0.84	0.93
	DIP2	0.80	0.84	0.95
	PIP5	0.92	0.79	0.81
	PIP4	0.94	0.78	0.90
	PIP3	0.86	0.86	0.92
	PIP2	0.84	0.81	0.81
	CMC1	0.87	0.88	0.91
Right	DIP5	0.82	0.84	0.87
	DIP4	0.88	0.80	0.94
	DIP3	0.88	0.85	0.93
	DIP2	0.79	0.78	0.95
	PIP5	0.95	0.83	0.88
	PIP4	0.97	0.80	0.89
	PIP3	0.89	0.87	0.89
	PIP2	0.84	0.81	0.84
	CMC1	0.89	0.89	0.95
Average for all 18 joints		0.87	0.83	0.89

This paper describes the use of high quality digital hand photographs for the diagnosis and severity grading of hand osteoarthritis. The study population was a random sample of 381 elderly (mean age 76) participants in the AGES-Reykjavik Study who also had standard hand radiographs and expert clinical hand examination in addition to the digital photographs.

In the first step of the study, we developed a set of reference photographs to facilitate the reading and grading of the hand photographs. These were chosen after repeated assessments. By the help of these reference photographs, we managed to achieve an interobserver agreement measured by ICC which is comparable to that reported in radiological studies [24–26].

In the second phase of the study, we compared hand photographs with the results of radiography and clinical examination in the same group of individuals with regard to the diagnosis of HOA, and the grading of HOA severity. Unfortunately, the definition of HOA is very problematic because of lack of an absolute clinical, radiological, or pathological standard that the epidemiology of hand OA can be compared with. The ACR criteria for clinical diagnosis of HOA [27] are useful for identifying HOA patients with persistent symptoms but the prevalence of HOA by those criteria is low. A study of an elderly population in Iceland based on the ACR criteria [13] found that the prevalence of symptomatic hand OA was 3% in men and 7% in women. The symptoms criterion, however, showed considerable variation with time and thus the symptomatic OA group was not stable.

Radiological changes are most commonly used to grade hand OA. At present, several different radiographic classification systems are used but the Kellgren-Lawrence (K-L) scale for grading of radiological changes has been most widely used in the past [17]. In a review by Marshall and colleagues in 2008 it was reported that in 1996-2005 thirty epidemiological studies, all using the K-L scale, used 13 different cut-off points for diagnosis of systemic HOA [28]. The prevalence of symptomatic HOA is also low using the radiographic criteria, in the Framingham Study it had a point prevalence of 1.8-5.5% [29].

In the present study, we found several differences between the three methods. Radiography was more sensitive than either of the other methods when the cut-off was set at K-L ≥ 2. Radiography also had a higher relative prevalence in the PIP joints than the other methods. Aggregate scores for all methods showed highly significant correlations and with few exceptions, they tended to identify the same individuals as having severe HOA. Not unexpectedly, photographs and clinical examination results were more closely correlated to each other than to radiography. There were a few individuals with high radiography scores and low photographic and clinical scores (non-nodal hand osteoarthritis), which may constitute a relevant subset of HOA and will be the subject of further studies. Agreement between the three methods was considerably better in females than males. Weight and BMI were negatively associated with photographic HOA scores but despite reports indicating a positive association between radiographic HOA and weight we found no such association [30]. The photographic finding is not entirely unexpected and probably related to increased finger soft tissues hindering the detection of nodes and deformities.

Pain is the central symptom of OA and in the next phase of the study, we investigated the relationship between pain and HOA detected by the three methods.

Somewhat surprisingly, we found no association between HOA and pain in males except in those with CMC1 OA who admitted to pain in that joint. This could be related to the high age of our participants but some previous studies have reported a weaker association between radiographic HOA and pain in males than females [31, 32]. In females, we found only a modest similar association between "pain in hands lasting at least one month" and aggregate severity scores for all three methods. The prevalence of this symptom and the weak association are in analogy with that found in previous studies [7]. Intermittent pain (pain sometimes) was more strongly related to HOA findings, both aggregate HOA scores and pain in individual joints. All three methods performed similarly in the case of the DIP joints, but radiography showed a somewhat stronger association with pain than photography for PIP and CMC1 joints. For the PIP joints, radiography also appeared superior to clinical examination.

In the final step of the study plan, we developed a shortened version of the photographic system. If information about HOA severity could be collected from some of the large detailed studies like the AGES-Reykjavik Study, it would open up a number of possibilities to examine the relationship of HOA to lifestyle and all kinds of conditions. If we consider HOA severity as a continous trait, the exact prevalences are less important than information the relative burden of HOA in each individual.

Using photographs for diagnosing HOA has a number of advantages. The method is simple, inexpensive, and involves little discomfort and no radiation. This study shows that the taking and the reading of the photographs can be standardized in a reproducible fashion with adequate inter- and intraobserver variation, at least in this age group. The photographic method in this age group is also in many ways comparable to the other methods, identifying mainly the same patients and showing comparable or only slightly inferior association with symptoms. Compared with clinical examination it also has the advantage of having an image for later analysis. The shortened version (HOASCORE) has practical advantages, speeding up the reading and appears to be particularly suitable for analysis of very large studies such as the AGES-Reykjavik Study where information about HOA status can be analysed in relation to the extensive health-related information available on each participant. By applying the HOASCORE to the AGES-Reykjavik Study population, we have discovered a number of potentially important new systemic associations, undetectable except in large studies [4, 21, 33].

The disadvantages of photography is that it is less sensitive on joint for joint analysis than either radiography and clinical examination particularly on PIP and CMC1 analysis. Also, compared with radiography it reflects anatomy less well. Bone damage or repair cannot be evaluated and the method cannot be used to diagnose erosive OA. The photographic scores also negatively associated with individual weight suggesting lower sensitivity in heavy subjects. Finally, a recent study suggests that photographic scoring of HOA is relatively insensitive to change, at least in this age group [34].

Of course, our conclusions regarding the use of photographs to diagnose HOA are limited to the current age group. In many ways this is a suitable age group to examine since it reflects cumulative disease burden and organ damage aquired over a long time. There is no reason to believe that photography performs differently in other age groups. In the future, it is even possible that photographs will prove to be more sensitive than radiographs in younger subjects with early disease who have nodal HOA but have not had time to develop radiological changes.

High quality hand photographs can be used to diagnose and grade hand osteoarthritis in the elderly. The method has the advantage of being inexpensive and easy to perform. By using a slightly simplified method of reading, it appears to be highly suitable for use in large studies.