The Value of Total Body Photography for the Early Detection of Melanoma: A Systematic Review

Early detection of melanoma is critical to reduce the mortality and morbidity rates of this tumor. Total body photography (TBP) may aid in the early detection of melanoma. To summarize the current evidence on TBP for the early detection of melanoma, we performed a systematic literature search in Medline, Embase, and the Cochrane Central Register of Controlled Trials (CENTRAL) for eligible records up to 6th August 2020. Outcomes of interest included melanoma incidence, incisional and excisional biopsy rates, as well as the Breslow’s index of detected tumors. Results from individual studies were described qualitatively. The risks of bias and applicability of the included studies was assessed using the QUADAS-2 checklist. In total, 14 studies published between 1997 and 2020 with an overall sample size of n = 12082 (range 100–4692) were included in the qualitative analysis. Individuals undergoing TBP showed a trend towards a lower Breslow’s thickness and a higher proportion of in situ melanomas compared to those without TBP. The number needed to excise one melanoma varied from 3:1 to 14.3:1 and was better for lesions that arose de novo than for tracked ones. The included studies were judged to be of unclear methodological concern with specific deficiencies in the domains “flow and timing” and “reference standard”. The use of TBP can improve the early detection of melanoma in high-risk populations. Future studies are warranted to reduce the heterogeneity of phenotypic risk factor definition and the technical implementation of TBP. Artificial intelligence-assisted analysis of images derived from 3-D TBP systems and digital dermoscopy may further improve the early detection of melanoma.


Introduction
The incidence of cutaneous melanoma continues to rise steadily in Western countries with Caucasian populations each year [1,2]. Surgical treatment is usually curative following early detection of the disease [3,4]. However, melanoma quickly becomes life-threatening once it metastasizes [3], although considerable progress has been made following novel therapeutic interventions including BRAF and MEK inhibitors and immunotherapies [5,6]. Thus, early detection of cutaneous lesions remains of paramount importance to significantly reduce the morbidity and mortality caused by melanoma, as thinner primary tumors are associated with an improved survival rate [7][8][9]. Hence, strategies to improve early detection are critical in patients at high risk for melanoma.
In Germany, the worldwide unique national skin cancer screening program was implemented in 2008. It involves a voluntary, standardized full-body examination by dermatologists or general practitioners specifically trained for this purpose. The costs are reimbursed by all German statutory health insurance companies every two years for members older than 35 years, while some health insurance companies also cover the costs for younger members [10]. Instead of mass screening of an entire population, other approaches aim to selectively identify and surveil individuals at the highest risk for melanoma [11,12]. One of the most promising approaches involves automated imaging of patients or total body photography (TBP), allowing objective documentation of all existing lesions and monitoring changes over time. TBP uses clinical photography of the patients' skin surface to provide a photographic record. This can be achieved by either 2-D or 3-D TBP. In 2-D TBP, common 2-D images of the skin are taken and composed to a body map while a 3-D representation of the patient's surface is created in 3-D TBP, partly even linked with dermoscopy images. Thus, 3-D TBP may particularly facilitate the localization of lesions for follow-up [13]. Both techniques may aid in the assistance of the early detection of melanoma as they improve the diagnostic accuracy of melanoma in high-risk patients [14][15][16][17]. TBP has the advantage of monitoring the patient's entire skin surface, rather than only a subset of individual lesions. Furthermore, TBP can be combined with sequential digital dermoscopy imaging (SDDI) as this "two-step approach" might provide more intensive surveillance in high-risk patients to assist with an early melanoma diagnosis.
However, until now, the value of TBP in the early detection of melanoma has not been systematically investigated. Here we performed a systematic review to summarize the current publication landscape of TBP in the prevention of melanoma. Our results will assist dermatologists as well as patients in choosing an appropriate and individual screening and surveillance strategy.

Materials and Methods
This review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [18], its extension for diagnostic test accuracy [19,20], and the Cochrane Handbook for Systematic Reviews [21].

Eligibility Criteria
We included studies investigating patients of any age with common melanocytic nevi (moles), congenital nevi, or dysplastic/atypical nevi undergoing 2-D TBP or 3-D TBP to detect melanoma early as part of screening for surveillance. Combination with other diagnostic measures, such as physical examination, self-examination, or SDDI was allowed but not obligatory for inclusion. In addition, studies investigating only TBP for keratinocyte cancer were excluded from the analysis.
Regarding the study design, we included randomized controlled trials (RCTs), clinically controlled trials (CCTs), non-controlled prospective trials, prospective observational studies, case-control studies as well as retrospective studies. Narrative reviews and case reports or series were excluded. No language restrictions were set. In RCTs and controlled observational studies, usual care served as control. If no control group was specifically included, data from external control groups or the general population served as control.

Types of Outcome Measures
The outcomes of interest were all related to the early detection of melanoma. Thus, the incidence of melanoma, the number of excisions and biopsies, the number of changed or new nevi, as well as the benign-to-malignant ratio of excised lesions were defined as outcomes of interest. Additionally, the mean or median vertical invasion depth measured in mm (Breslow's index) of detected melanomas was considered as an outcome. As we expected heterogeneously reported outcomes, we extracted the data as reported in the publication and did not perform any re-calculations.

Search Methods for Identification of Studies
We searched the electronic databases Medline, Embase (both via Ovid), and the Cochrane library CENTRAL until 6 August 2020 to identify all possibly relevant records (Table 1). For ongoing trials and completed trials without data publication, principal investigators or trial sponsors were contacted to obtain preliminary or unpublished data. Moreover, reference lists of included records and evidence-and consensus-based guidelines on the diagnosis of melanoma were screened.

Selection of Studies
Three authors (T.S., A.H., M.V.H.) independently screened titles and abstracts for eligibility that were identified in the electronic database searches. For records that were considered relevant according to title and abstract screening, full-text articles were obtained, and inclusion and exclusion criteria were applied by the same three authors. Whenever discrepancies arose, a resolution was achieved by discussion with another independent author (C.B., A.W., T.J.B., T.B., M.E.).

Data Extraction and Management
Information for each included study regarding study design, baseline characteristics of the included population, risk factors for melanoma, imaging technique, and intervals of TBP, follow-up strategy, main outcomes, and study limitations were collected and summarized independently by two authors (A.H., M.V.H.). Data were extracted to an internally piloted data extraction spreadsheet using Microsoft Excel 2010. If multiple reports of a primary study were identified, all available data were extracted. According to the principles of evidence-based medicine, we also decided to contact study authors for clarification in case of inconsistencies in reporting or overlapping study populations were identified. However, this was not the case for any of the identified records. The baseline characteristics and the outcomes of interest of each study were described qualitatively within the text, as a quantitative synthesis was not feasible due to the high heterogeneity of included studies.

Assessment of Methodological Quality
The risks of bias and applicability of the included studies were assessed by two reviewers (A.H., T.S.) independently using a modified version of the QUADAS-2 checklist [22]. QUADAS-2 consists of the four key domains "patient selection", "index test", "reference standard", and "flow and timing". Each domain was assessed by the two raters in terms of risk of bias. The first three domains were additionally rated in terms of concerns regarding applicability. Any disagreement was resolved by consensus with another review author (M.V.H.).

Study Identification
We identified 1573 references through database searching and cross-referencing in guidelines. After removal of duplicates and title-abstract screening, 48 records underwent full-text review. Thirty records were dismissed as the study design did not meet the predefined eligibility criteria or as the record did not refer to a study. In addition, two further duplicates were identified, and two references did not present any results. Hence, 14 published reports referring to 13 independent studies published between 1997 and 2020 with an overall sample size of n = 12082 (range 100-4692) were included in qualitative analysis ( Figure 1).

Prospective Studies
Banky et al. performed a cohort study of 309 patients at high risk for melanoma who underwent baseline TBP and were followed up at three-, six-, or twelve-month intervals [32]. Eighteen melanomas were detected. The overall incidence of melanoma was 19 per 1000 person-years. The benign-to-malignant ratio of lesion biopsies was almost 3:1.
In the prospective cohort study reported by Goodson et al., TBP was performed at baseline and patients were clinically followed-up at six-and twelve-month intervals ( Figure 2). Outcomes of interest were the biopsy rate, the efficiency of melanoma detection, and melanoma origin (de novo or nevus-derived). In total, 548 biopsies were performed in 1076 patients (273 during the initial visit and 275 on follow-up) [26]; 61% of the 275 biopsies at follow-up were performed due to photographic evaluation. At follow-up, 12 melanomas were detected, 5 were invasive. Five of them presented as changing and two as new lesions. Kelly et al. assessed 278 patients with dysplastic melanocytic nevi in a prospective cohort study [30]. Patients were followed up for a mean of 42 months, 210 biopsies were obtained. Overall, 20 new melanomas (n = 12 invasive, n = 8 in situ) were detected in 16 patients. The age-adjusted incidence was calculated to 1835/100,000 person-years. Notably, 11 melanomas were detected due to comparison with baseline photographs.
Moloney et al. performed a prospective observational study in Australia to examine the effect of regular full body examination in combination with dermoscopy and TBP [31]. SDDI was performed for ambiguous lesions. In 311 patients, 75 primary melanomas were detected, 14 of them at baseline visit; 38% were detected using TBP, and 39% with SDDI. Of the 770 excised lesions, 441 were melanocytic, including 82 melanomas, representing a 4.4:1 benign-to-malignant melanoma ratio. The overall rate of melanomas identified during follow-up per study year was 0.08. Nathansohn et al. established a pigmented lesions clinic based on a digital photography studio with TBP and dermoscopy. In the first 20 months of the prospective study, 895 patients were examined, 206 of them had follow-up visits [36]; 236 suspicious lesions were excised with 7 emerging as melanomas.
Taken together in all six prospective studies 7 to 98 melanomas were detected. Two of these studies indicated that up to 11 out of 20 melanomas were detected due to TBP. Three studies specified the number needed to excise, ranging from 3:1 to 11.9:1. Altogether the number of biopsies ranged from 210 to 1152 biopsies (Tables 3 and 4).

Retrospective Studies and Chart Reviews
In the retrospective study by Drugge et al., 4692 patients were analyzed regarding patient characteristics, self-reported melanoma risk factors, and the usage of TBP [23]; 2473 of the patients were scanned at least once with TBP, 268 biopsies were obtained in 218 patients. At least one malignant lesion was identified in 65 of the 218 patients (30%). The number needed to excise one melanoma for all 268 lesions was 3:1. The ratio of melanoma in situ to invasive melanoma was 1.56:1. Using data from the same cohort, the frequency of TBP and the invasion depth of identified melanoma were correlated in 218 patients. In total, 225 lesions were biopsied, 67 (29.8%) of which were malignant, including 44 in situ and 23 invasive melanomas [29]. The minimum interval between baseline and a follow-up TBP session to identify a new invasive melanoma was 1.29 years.
Feit et al. reported the results of a retrospective study including 576 patients [28]; 93 suspicious lesions were excised of which 77 were melanocytic. Photographic assistance identified 12 patients with 27 histologically confirmed melanomas (n = 21 in situ; n = 6 invasive); 74% of the melanomas (n = 27) were biopsied due to changes from baseline while 19% were biopsied because they newly emerged.
The retrospective registry review of two cohorts by Rademaker et al. investigated demographic and histological data from 100 melanomas diagnosed through self-referred TBP and SDDI service compared to those diagnosed through traditional methods [35]. Fifty-two invasive and 48 in situ melanomas were identified; 48 were diagnosed at the first visit and the remaining by serial digital imaging.
Risser et al. compared biopsy numbers in 946 patients with multiple atypical nevi in their first year of care who received either total body skin examination alone or in combination with TBP [24]. The mean number of performed biopsies was similar in both groups (skin examination only vs. skin examination plus TBP: 0.82 vs. 0.8). Three melanomas were diagnosed in 19 patients who did not receive TBP, whereas no melanoma was diagnosed in 16 patients receiving TBP.
Salerni et al. analyzed the use of TBP and dermoscopy in 618 high-risk melanoma patients. A total of 1152 lesions in 407 patients were excised during the surveillance period, corresponding to a mean of 2.83 lesions per patient during a 10-year follow-up period (global excision rate 1.86 per patient in 618 patients) [33,34]; 98 melanomas were detected in 78 patients (n = 53 in situ), comprising 8.5% of excised lesions. Sixty melanomas were tracked lesions (benign-to-malignant ratio 11.9:1) while 38 melanomas newly emerged (benign-to-malignant ratio 8.8:1).
Truong et al. reviewed the records of 926 patients who received TBP and had two or more follow-up visits over two years or longer [25]. Patients with TBP received 1.56 biopsies on average. In 1419 biopsies 93 primary melanomas were detected (49% melanoma in situ), corresponding to a combined benign-to-malignant ratio of 14.3:1.
Altogether in seven retrospective studies or chart reviews the number of invasive melanomas were reported in six studies and ranged from 0 to 70. The number needed to excise was specified in two studies and ranged from 3.1:1 to 14.3:1. The number of biopsies in patients with TBP ranged from 53 to 921 biopsies (Tables 3 and 4).

Breslow Index of Invasive Melanoma
The reported depth of detected melanoma varied across all studies (Tables 3 and 4). However, individuals undergoing TBP showed a trend towards a lower Breslow's thickness. Patients diagnosed by TBP and SDDI screening had thinner melanomas compared to the registry data in the study by Rademaker et al. (69% vs. 52% <0.75 mm; 1.9% vs. 10.8% >3 mm) [35]. Additionally, a lower index was associated with having one or more follow-up visits (median 0.83 vs. 0.33 mm) and photographic review (median 0.31 vs. 0.48 mm) in the study by Strunck et al. [27]. The median depth was 0.31 mm (range 0.11-1.5) if TBP could be used compared to 0.48 mm (0.10-3.3) without consulting of TBP. Additionally, depths of tumors detected with TBP ranged from 0.11 to 2.1 mm vs. 0.2 to 3.0 mm for those detected without TBP in the study by Truong et al. [25].

Methodological Quality of Included Studies
Overall, the included studies were judged to be of unclear methodological concern (Figure 3). At least half of the studies were at high or unclear risk of bias for "flow and timing" and all studies were at high risk for bias regarding the "reference standard", mostly since it was unclear whether the reference standard results had been interpreted without knowledge of the results of the index test. Approximately one-third of the studies showed an increased risk in the domain "patient selection" due to unclear sampling of participants. An additional third of the studies had an increased risk in the domain "index test" as it remained unclear if the index test results had been interpreted without knowledge of the results of the reference standard. However, the applicability of study findings was of low concern for the majority of studies in the domains "index test" and "patient selection", while the applicability was of unclear or high concern for the domain "reference standard" in less than 20% of the studies.

Discussion
In this study, we aimed to investigate the value of TBP in the early detection of melanoma. Overall, we have identified 14 studies with an overall sample size of n = 12,082 highlighting that individuals undergoing TBP showed a trend towards a lower Breslow's thickness and a higher proportion of in situ melanomas compared to those without TBP. In addition, our results show that the use of TBP can improve the early detection of melanoma in high-risk populations.
For the early detection and thus secondary prevention of melanoma, there are two distinct strategies to select the target population. Firstly, untargeted mass screening programs can be performed (population-based screening) regardless of individual risk profiles. However, this procedure is only carried out in a few countries such as Germany and is costly due to the high number of screened individuals [37]. In Australia, the number needed to screen to save one life from melanoma was calculated to 25,000 individuals [38]. Hence, the effect in terms of significantly reduced mortality is controversial and considerable resources are needed for a nationwide implementation of such a screening program. Secondly, screening efforts can be restricted to populations at increased risk of developing melanoma. This procedure is more targeted, and the expected effects are larger because the prevalence of the screened population is significantly higher compared with the normal population. The latter procedure was used in the studies reviewed here, yet the definition of risk factors was fairly heterogeneous in keeping with previous reviews on risk prediction models [39]. In most studies, the presence of >4 to 5 clinically atypical nevi or a large number (>100) of common nevi was required for investigation by TBP. A positive personal or family history of melanoma was also among the inclusion criteria in some studies while confirmed genetic susceptibility was reported only for two studies. From these data, we conclude that the screened populations had a high to very high risk for melanoma. However, four studies did not specify the risk profiles, underlining that there is currently no consensus on the best risk assessment [40,41]. For the conduct of RCTs, risk prediction tools should be harmonized and include a weighting of individual factors to enable cross-population comparisons of distinct regions and ethnicities.
The technical implementation and timing intervals of TBP were similarly heterogeneous as the definition of clinical risk factors. In most studies, 12-30 macroscopic images were acquired with conventional cameras to capture the entire skin surface in two dimensions (2-D TBP). As a limitation, multiple imaging systems were applied with distinct resolutions. Furthermore, the study protocols only referred to standard poses and often did not specify exactly how the images were acquired. Specific anatomic regions such as intertriginous areas, genitals, retroauricular area, or the flexor sides of the extremities may not always be adequately captured with 2-D TBP. Here, the use of 3-D TBP systems in defined postures can help to standardize the image acquisition as they generate a digital 3-D avatar of the imaged person that can be used to analyze and prospectively follow all nevi of interest [13,37,42]. Currently, an RCT is ongoing in Australia evaluating the efficacy of 3-D TBP combined with SDDI compared with regular skin examinations which may or may not use 2-D TBP with SDDI. The main outcomes to be prospectively assessed in this study are the number of incisional and excisional biopsies and the Breslow depth of detected melanomas [37]. The number of in situ tumors also represents a valid indicator for the early detection of melanoma.
The results of our study clearly show that 2-D TBP identified a higher proportion of in situ tumors and melanomas with a lower average Breslow thickness compared with the comparison groups without 2-D TBP. The number needed to excise one melanoma varied widely, ranging from 3:1 to 14.3:1. This ratio was better for lesions that arose de novo than for tracked ones. We conclude that the latter need to be assessed with additional methods such as dermoscopy and that the added value of TBP is highest for lesions that arose newly during the surveillance period.
Artificial intelligence (AI) was not used in any of the included studies to evaluate the captured images. The image analysis was performed by the respective physicians in all reports. Several landmark studies have recently shown that AI performed on par with dermatologists in the distinction of nevi from melanoma based on dermoscopic images [43][44][45]. Similarly, future studies are warranted to investigate whether the analysis of 3-D TBP images using AI is also feasible. AI ultimately should be integrated into 3-D TBP systems to support image analysis, pattern recognition, and detect changes of nevi in sequential examinations. Furthermore, the 3-D TBP images should also be equipped with interfaces to SDDI, which in turn could also be analyzed using AI to make screening efforts more accurate while maintaining cost efficiency.
Overall, our study shows that the use of 2-D TBP can improve the early detection of melanoma in high-risk populations. However, the definition of phenotypic risk factors, technical performance of TBP, and design of the included studies were highly heterogeneous. In the future, harmonized risk prediction tools should be developed to compare distinct populations. 3-D TBP with the generation of digital avatars is a game-changing technology and will enable standardized and gapless image acquisition.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.