Evaluation of the Number-Needed-to-Biopsy Metric for the Diagnosis of Cutaneous Melanoma

Key Points

Question

Does the number-needed-to-biopsy for cutaneous melanoma differ among clinicians according to the medical literature?

Findings

In this systematic review of 46 articles on the number-needed-to-biopsy metric for melanoma diagnosis, the number ranged from 2.2 to 287.

Meaning

The number needed to biopsy appeared to vary significantly across geographic regions and according to prevalence of disease and clinician characteristics, including level of training, age, and use of dermoscopy; standardization of NNB and its reporting may be warranted.

Abstract

Importance

To date, no concerted effort has been made to date to evaluate the literature on number-needed-to-biopsy (NNB) metrics, particularly to account for the differences in clinician type and melanoma prevalence in certain geographic locations.

Objective

To review and synthesize worldwide data for NNB for the diagnosis of cutaneous melanoma.

Data Source

MEDLINE, Embase, and PubMed databases were searched for English-language articles published worldwide from January 1, 2000, to November 28, 2018.

Study Selection

A total of 46 studies were included that addressed NNB for at least 3681 clinicians worldwide and included 455 496 biopsied tumors and 29 257 melanomas; primary care practitioner (PCP) data were only available from Australia.

Data Extraction and Synthesis

Articles were screened for eligibility, and possible overlapping data sets were resolved. Data extracted included clinician specialization, use of dermoscopy, geographic region and location-specific health care system, study design, number of benign tumors, number of melanomas, and NNB. The review followed the PRISMA guidelines.

Main Outcome and Measures

The NNB for the diagnosis of cutaneous melanoma.

Results

A total of 46 studies were included that addressed NNB for at least 3681 clinicians worldwide and included 455 496 biopsied tumors and 29 257 melanomas; primary care practitioner (PCP) data were only available from Australia. The reported NNB ranged from 2.2 to 287, and the weighted mean NNB for all included publications was 15.6. The exclusion of publications structured as all biopsied tumors, owing to variable data characterization, resulted in reported NNB ranging from 2.2 to 30.5, with a global weighted mean NNB of 14.8 for all clinicians, 7.5 for all dermatologists, 14.6 for Australian PCPs, and 13.2 for all US-based dermatological practitioners, including dermatologists and advanced practice professionals. The summary effect size (ES) demonstrates that a mean 4% of biopsies demonstrated melanoma for study stratum A (all biopsied skin tumors, ES, 0.04; 95% CI, 0.03-0.05), and a mean 12% of biopsies demonstrated melanoma for study strata B (melanocytic tumors on pathology review, ES, 0.12; 95% CI, 0.10-0.14) and C (clinical concern for melanoma, ES; 0.12; 95% CI, 0.09-0.14).

Conclusions and Relevance

The existing NNB for cutaneous melanoma appeared to vary widely worldwide, lacking standardization in the metric and its reporting, and according to clinician characteristics as well; the NNB of US-based clinicians may warrant further exploration.

This systematic review and meta-analysis evaluates the worldwide variations in the number-needed-to-biopsy (NNB) metric, lack of related data on US-based dermatologists, and nonstandardized processes for reporting NNB for diagnosing cutaneous melanoma.

Introduction

The number needed to biopsy (NNB) quantifies biopsies that result in benign diagnoses in the course of identifying cutaneous melanoma. Biopsy of benign lesions not only has fiscal implications in an era of increasing insurance deductibles and co-insurance rates, but it also contributes to patient morbidity. Patients may experience anxiety while waiting for the laboratory report as well as morbidity from the biopsy procedure and resulting scar.

In assessments of the accuracy and cost-effectiveness of cutaneous melanoma diagnostic tools, 2 metrics are frequently identified in the literature: diagnostic accuracy and NNB. Diagnostic accuracy refers to the ability of a test to distinguish a specific condition (eg, melanoma) from normalcy (eg, benign nevus). The NNB, a metric similar in purpose to the number needed to treat, is an indicator of diagnostic sensitivity (or willingness to assume risk of missing a subtle melanoma) and melanoma prevalence in a patient population (eg, a higher melanoma prevalence patient population results in a lower NNB). Although an imperfect metric, NNB allows for a clinically understandable measure of the number of benign (ie, false-positive) biopsies performed.

To our knowledge, no concerted effort has been made to date to evaluate the literature on NNB metrics, particularly to account for the differences in clinician type and melanoma prevalence in certain geographic locations. We conducted a systematic review of articles published worldwide between January 1, 2000, and November 28, 2018, including information on NNB variance based on clinician type (when specified), to provide a perspective on the strengths and weaknesses of data categories. We also elaborated on the design for future studies of NNB in the setting of cutaneous melanoma.

Methods

This systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline. The following bibliographic databases were searched from January 1, 2000, through November 28, 2018: MEDLINE, Embase, and PubMed. The results were limited to publications in the English language. All search strategies were developed and tested by a medical librarian, with input from our lead author (K.C.N.). The search strategies contained a combination of controlled vocabulary (eg, MeSH or Emtree) and keyword terms to identify articles concerning a diagnosis of melanoma, nevi, seborrheic keratosis, basal and squamous cell carcinoma, and skin neoplasms in general. The sensitivity of preliminary search strategies was tested by their ability to retrieve a pool of known, relevant citations. The precision of the strategies was strengthened through author feedback on preliminary results (eMethods in the Supplement). An EndNote library was created for managing the retrieved records and for de-duplication.

The inclusion criteria were publication date between January 1, 2000, and November 28, 2018, to account for variability in pathologic interpretation, patterns of dermoscopy use, and changes in melanoma prevalence over time; clearly cited and internally consistent tumor counts for benign neoplasms and melanoma, or clearly cited author-calculated NNB for melanoma; availability of the full text article in the English language; and inclusion of at least 1 melanoma in the data set. Clinicians of all medical specialties were included, as were available data from advanced practice professionals (ie, nurse practitioners and physician assistants), although specific data pertaining to primary care practitioners (PCPs) were only identified for Australian PCPs. In addition, the data cohorts were reviewed for each publication, and subsequent publications that used overlapping data over the same period were excluded.

We identified 770 records through database searches and 5 records through other sources. After removal of 125 duplicate records, one of us (K.C.N.) analyzed the remaining 650 records and summarily excluded 551 of these records for not meeting the inclusion criteria. Ninety-nine full-text articles were assessed for eligibility, and of these 99 articles, 47 were excluded after a single-reviewer evaluation (K.C.N.). The remaining 52 publications received a dual-reviewer assessment (K.C.N. and C.C.L.), resulting in the exclusion of an additional 6 articles. Ultimately, we selected 46 publications for inclusion (eFigure 1 in the Supplement).

Selected studies were evaluated by multiple reviewers (K.C.N., C.C.L., and S.C.C.) to find the total numbers of tumors and melanomas included and to compare them with author-reported NNB, if provided. If the author-reported NNB did not align with the provided tumor counts (mainly owing to the number of tumors that may be “concerning for melanoma” not being clearly specified), we included total tumor counts in Table 1.^{1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47}

Table 1. Studies Selected for Inclusion in Systematic Review, 2000-2018.

Study	Diagnosing Specialists	Dermoscopy	Years Included	Location	Design Strata	No. of Clinicians	Tumors, Total No.	Melanomas, Total No.	NNBa
Duff et al,1 2001	Plastic surgeons	Yes	1993-1998	Frenchay Hospital, Bristol, United Kingdom	All biopsied tumors	2	2372	586	4.0
Westbrook et al,2 2006	Dermatologists	Not referenced	2003	Royal Hallamshire Hospital, Sheffield, United Kingdom	All biopsied tumors	Not referenced	145	21	6.9
Rosendahl et al,3 2011	PCPs	Yes	Not referenced	University of Queensland, Brisbane, Australia	All biopsied tumors	1	463	29	16.0
Wilkinson et al,4 2006	PCPs	Yes	2005	Skin Alert Clinics, Northern Territory, Australia	All biopsied tumors	20	3317	116	28.6
Har-shai et al,5 2001	Plastic surgeons	Not referenced	1997-1998	Linn Clinic of Plastic Surgery, Israel	All biopsied tumors	4	835	28	29.8
Scrace et al,6 2009	PCPs	Yes	2001-2008	Royal Flying Doctor Service, Australia	All biopsied tumors	6	269	7	38.4
Byrnes et al,7 2007	PCPs	Yes	2005-2006	Queensland Innovative Practices, Queensland, Australia	All biopsied tumors	23	1220	31	39.4
Matteucci et al,8 2011	Plastic surgeons	Not referenced	2006	Castle Hill Hospital, Cottingham, United Kingdom	All biopsied tumors	Not referenced	1187	24	49.5
Heal et al,9 2006	PCPs	Not referenced	2004-2005	Queensland (North), Australia	All biopsied tumors	16	1205	20	60.3
Youl et al,10 2007	PCPs	Not referenced	Not referenced	Queensland, Australia	All biopsied tumors	16	2544	42	60.6
Green et al,11 2004	Dermatologists	Mixed	2002	University of Miami, United States	All biopsied tumors	22	11 072	156	71.0
Youl et al,12 2007	PCPs	Mixed	2005	Queensland (Southeast), Australia	All biopsied tumors	154	11 116	152	73.1
Rolfe,13 2012	Dermatologists	Not referenced	2009-2010	Queensland public hospital, Australia	All biopsied tumors	Not referenced	5039	55	91.6
Moffatt et al,14 2006	PCPs	Not referenced	2003	Queensland skin cancer clinic, Australia	All biopsied tumors	Not referenced	287	1	287.0
Ahnlide et al,15 2014	Dermatologists	Yes	2009-2012	Helsingborg Hospital, Lund, Sweden	All melanocytic tumors + SKs on path review	10	1717	252	6.8
English et al,16 2004	PCPs	Not referenced	1998-2000	Perth, Australia	All melanocytic tumors + SKs on pathologic review	468	4699	160	29.4
Hansen et al,17 2009	PCPs	Not referenced	2005-2007	Skin Cancer Clinics, Australia	All melanocytic tumors + SKs on pathologic review	57	10612	348	30.5
Esdaile et al,18 2014	Dermatologists	Yes	2006, 2011	Churchill Hospital, Oxford, United Kingdom	All melanocytic tumors on pathologic review	Not referenced	1380	454	3.0
Moloney et al,19 2014	Dermatologists	Yes	2006-2009	Sydney Melanoma Diagnostic Center, Australia	All melanocytic tumors on pathologic review	4	441	82	5.4
Argenziano et al,20 2014	Dermatologists	Yes	2011	Italian Pigmented Lesion Clinics, Italy	All melanocytic tumors on pathologic review	22	5402	866	6.2
Sidhu et al,21 2012	Dermatologists	Yes	2005-2009	Welsh Institute of Dermatology; Swansea University, United Kingdom	All melanocytic tumors on pathologic review	9	4691	750	6.3
Tschandl et al,22 2013	Dermatologists	Mixed	1998-2008	Medical University of Vienna, Vienna	All melanocytic tumors on pathologic review	Not referenced	9100	1365	6.7
Carli et al,23 2004	Dermatologists	Mixed	1997-2001	Department of Dermatology, Florence, Italy	All melanocytic tumors on pathologic review	6	3053	319	9.6
Soares et al,24 2009	Dermatologists	Mixed	2005	Mayo Clinic, Scottsdale, Arizona, United States	All melanocytic tumors on pathologic review	Not referenced	1547	147	10.5
Lott et al,25 2018	Not referenced	Not referenced	2007-2012	Kaiser Permanente, Washington, United States	All melanocytic tumors on pathologic review	Not referenced	18 715	1609	11.6
Argenziano et al,26 2012	Multidisciplinary	Mixed	1998-2007	International	All melanocytic tumors on pathologic review	21	300 215	17 172	17.5
Goodson et al,27 2010	Dermatologists	Yes	1999-2004	University of Utah, Salt Lake City, United States	All melanocytic tumors on pathologic review	Not referenced	548	28	19.6
Koelink et al,28 2014	PCPs	Mixed	2010-2011	Northeastern Netherlands	Clinical concern for melanoma	55	22	10	2.2
Rauniyar et al,29 2003	Not referenced	Not referenced	2000	Kathmandu Valley, Nepal	Clinical concern for melanoma	Not referenced	21	9	2.3
Chia et al,30 2008	Dermatologists	Not referenced	2005-2006	Sydney, Australia	Clinical concern for melanoma	35	686	195	3.5
Brown et al,31 2016	Multidisciplinary	Not referenced	2015	Royal Devon and Exeter NHS (National Health Service) Foundation Trust, United Kingdom	Clinical concern for melanoma	Not referenced	775	186	4.2
Carli et al,32 2003	Dermatologists	Yes	1998	University of Florence, Italy	Clinical concern for melanoma	2	13	3	4.3
Verykiou et al,33 2012	Dermatologists	Not referenced	2009, 2011	Royal Victoria Infirmary, Newcastle upon Tyne, United Kingdom	Clinical concern for melanoma	Not referenced	1313	288	4.6
Aitken et al,34 2006	PCPs	Not referenced	1998-2001	Queensland, Australia	Clinical concern for melanoma	Not referenced	161	33	4.9
Carli et al,35 2003	Dermatologists	Mixed	1997-1999	Italian Pigmented Lesion Clinics, Italy	Clinical concern for melanoma	6	114	16	7.1
Rosendahl et al,36 2012	PCPs	Mixed	2008-2012	Skin Cancer Audit Database, Australia	Clinical concern for melanoma	193	21 900	2367	9.3
Terushkin et al,37 2010	Dermatologists	Mixed	2004-2005	NYU (New York University) Langone, New York, United States	Clinical concern for melanoma	2	438	42	10.4
Lynch et al,38 2013	Dermatologists	Yes	Not referenced	St John's Institute of Dermatology, London, United Kingdom	Clinical concern for melanoma	Not referenced	121	10	12.1
Wilson et al,39 2012	Dermatologists (6 MDs, 2 APPs)	Mixed	2008-2009	Wake Forest University Medical Center, Winston-Salem, North Carolina, United States	Clinical concern for melanoma	8	410	28	14.6
van der Rhee et al,40 2010	Dermatologists	Yes	Not referenced	Leiden University Medical Center, Leiden, Netherlands	Clinical concern for melanoma	17	209	14	14.9
Soltani et al,41 2015	Dermatologists	Not referenced	2013	University of Utah, Salt Lake City, United States	Clinical concern for melanoma	43	2643	165	16.0
Waldman et al,42 2012	Dermatologists and PCPs	Not referenced	2003-2004	Schleswig-Holstein region, Germany	Clinical concern for melanoma	1789	7408	452	16.4
Baade et al,43 2008	PCPs	Mixed	2005	Queensland (Southeast), Australia	Clinical concern for melanoma	154	2977	152	19.6
Nault et al,44 2015	Dermatologists (13 MDs; 5 APPs)	Mixed	2010	University of Wisconsin, Madison, United States	Clinical concern for melanoma	18	492	23	21.4
Anderson et al,45 2018	Dermatologists (15 MDs; 15 APPs)	Not referenced	2011-2015	University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania, United States	Clinical concern for melanoma	30	4039	149	27.1
English et al,46 2003	PCPs	Not referenced	1998-2000	Queensland, Australia	Clinical concern for melanoma	468	8563	295	29.0
Totals	NA	NA	NA	NA	NA	3681	455 496	29 257	15.6
Totals excluding studies in design stratum A, all biopsied tumors	NA	NA	NA	NA	NA	3417	414 425	27 989	14.8

Abbreviations: APP, advanced practice professional; NA, not applicable; NNB, number needed to biopsy; PCP, primary care practitioner; SK, seborrheic keratoses.

^aThe NNB, a metric similar in purpose to the number needed to treat, is an indicator of diagnostic sensitivity (or willingness to assume risk of missing a subtle melanoma) and melanoma prevalence in a patient population (eg, a higher melanoma prevalence patient population results in a lower NNB).

All extracted data of the selected studies were imported into Stata, version 15 (StataCorp LLC). The proportion of melanoma diagnoses was calculated by dividing the number of melanomas by the number of biopsies performed (1/NNB). For the meta-analysis, the 1/NNB data were stratified by study design and presented in forest plot format (Figure 1).^{1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27} Heterogeneity was quantified using I², and effect sizes were expressed with 95% CI.

Figure 1. Proportions of Melanoma Diagnoses in All Biopsied Tumors (Stratum A Studies) and Melanocytic Tumors (Stratum B Studies).

Data represent the number of melanoma diagnoses as a proportion of the number needed to biopsy (1/NNB). Design strata A and B demonstrate statistically significant heterogeneity (I²>98.37%); thus, combining studies within individual study design stratum should be approached with caution. The summary effect size (ES) for each stratum demonstrates that a mean 4% of biopsies demonstrated melanoma for study stratum A (ES, 0.04; 95% CI, 0.03-0.05), and a mean 12% of biopsies demonstrated melanoma for study stratum B (ES, 0.12; 95% CI, 0.10-0.14). The horizontal lines adjacent to each square indicate CIs. The diamonds represent the meta-analyzed effect size for the individual study strata.

Results

The 46 publications selected for inclusion addressed the NNB of at least 3681 clinicians and 455 496 biopsied tumors. Most of the included articles were retrospective medical record reviews, with most categorized as level 3 Quality of Evidence on SORT (Strength of Recommendation Taxonomy).⁴⁷

The data in Table 1^{1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46} and Figure 1^{1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27} and Figure 2^{28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46} represent 3 study design strata: all biopsied tumors (stratum A), all melanocytic tumors on pathologic review (stratum B), and clinical concern for melanoma (stratum C). Inclusion or exclusion of seborrheic keratosis varied across all 3 design strata or types, as did the specificity of histopathologic diagnostic parameters.

Figure 2. Proportion of Clinical Concern for Melanoma in Stratum C Studies.

Data represent the number of melanoma diagnoses as a proportion of the number needed to biopsy (1/NNB). Design stratum C demonstrates statistically significant heterogeneity (I²>98.37%); thus, combining studies within individual study design stratum should be approached with caution. The summary effect size (ES) demonstrates that a mean 12% of biopsies demonstrated melanoma for study stratum C (ES, 0.12; 95% CI, 0.09-0.14). The horizontal lines adjacent to each square indicate CIs. The diamond represents the meta-analyzed effect size for the individual study stratum.

To facilitate estimations of variance (Figure 1 and Figure 2), the data were structured as proportions, or number of melanomas per biopsies performed (1/NNB). All strata demonstrated statistically significant heterogeneity (I² = 98.37% to 99.53%; P < .001); thus combining studies within individual study strata should be approached with caution. The summary effect size (ES) for each stratum demonstrates that a mean 12% of biopsies demonstrated melanoma for study stratum A (ES, 0.04; 95% CI, 0.03-0.05), and a mean 12% of biopsies demonstrated melanoma for study strata B (ES, 0.12; 95% CI, 0.10-0.14) and C (ES, 0.12; 95% CI, 0.09-0.14).

Analysis of all 3 strata demonstrated NNB ranging from 2.2 to 287 (Table 1; eFigure 2 in the Supplement), with data representing at least 3681 clinicians (as not all studies explicitly referenced the number of clinicians) and with 29 257 cutaneous melanomas and 455 496 biopsied tumors. Analysis of all selected worldwide publications and clinicians included a weighted mean NNB of 15.6 (455 496 tumors divided by 29 257 melanomas) (Table 2). Exclusion of publications structured as stratum A (all biopsied tumors), owing to variable data characterization, resulted in author-reported NNB that ranged from 2.2 to 30.5, with a weighted mean NNB of 14.8 for all worldwide clinicians, 7.5 for all worldwide dermatologists, 14.6 for Australian PCPs, and 13.2 for all US-based clinicians. Source cutaneous melanoma and biopsied tumor data were used to calculate these reported weighted means. When the Australian article by Rosendahl et al³⁶ was excluded, because most of the included Australian PCPs reported a high level of dermoscopic proficiency, the weighted mean NNB of the Australian PCPs increased to 27.3.

Table 2. Summary Data of Systematic Review, 2000-2018.

Data Categorization	Location	NNB, Weighted Mean
All publications	Worldwide	15.6
All publications, excluding all biopsied tumor data structure
Dermatologists	Worldwide	7.5
Primary care practitioners	Australia	14.6
Dermatologists and APPs	United States	13.2

Abbreviations: APP, advanced practice professional; NNB, number needed to biopsy.

Discussion

Identified publications addressing NNB for cutaneous melanoma appeared to document statistically significant heterogeneity and widely variable NNB. A direct comparison, particularly between US and Australian publications, was complicated by varying study design, clinician specialization, use of dermoscopy, and location-specific melanoma prevalence. Even single-institution publications demonstrated clinician-specific NNB variance.⁴⁵ This variance, assuming equivalent cutaneous melanoma prevalence within the institution’s general dermatological patient population, highlights a potential association between clinician-specific factors and NNB, and thus clinician-directed educational opportunities may improve NNB.

In the field of dermatology, NNB varies from physician to physician, with an even greater variance reported when nondermatologist clinicians (physicians and advanced practice professionals) are considered.⁴⁵ Additional clinician factors, such as older age,¹¹ completion of dermatology residency training,^36,44,45 and skillful use of dermoscopy^3,32 and total body photography,²⁷ have been associated with lower NNB. Patient factors, such as melanoma prevalence per geographic location and in the clinician’s patient cohort,⁴⁸ melanoma subtype, and patient anxiety,⁴³ may also be associated with changing NNB, but patient factors are seldom included as companion data to country-specific articles on NNB.

Additional database challenges were present in this NNB metric review. The degree of clinician concern for melanoma was not always precisely represented in the laboratory report (with potential concern for either biasing the pathologist or seeking to avoid denial of insurance coverage for removal of a clinically benign skin tumor), such as categorizing tumors that did not truly worry the clinician as concerning for melanoma. For borderline melanocytic neoplasms (eg, nevus with severe atypia or dysplasia vs melanoma in situ), interobserver variability in pathologic interpretation⁴⁹ may result in dissimilar NNB. What may be most challenging to overcome was the lack of representation of false-negatives (ie, missed melanomas) in most of the included articles; these data would be of value in cost and cost-effectiveness analyses.

The published NNB studies were structured according to 1 of 3 distinct design strata: (A) all biopsied tumors, (B) all melanocytic tumors on pathologic review, and (C) clinical concern for melanoma. Stratum A had the advantage of representing a real-world picture of clinical practice and presented greater diversity in clinician specialization, with the inclusion of more PCPs. However, all PCP data we identified in this systematic review originated from Australia, which had higher melanoma prevalence and a well-established culture of PCP-based melanoma screening³⁴ with dermoscopic proficiency³⁶; thus, stratum A may not align with the PCP-based NNB in the United States, which was not represented in the data set. The lack of information on the degree of clinician concern for melanoma was also a substantial challenge with this data set, as tumors that were not concerning for melanoma were likely included in the total tumor counts, thus potentially falsely elevating the clinicians’ NNB. Two other concerns with stratum A included the interobserver pathologic interpretation variance and discordant diagnostic groupings.

Stratum B design included all tumors that were interpreted as melanocytic on pathologic review, with variable inclusion and exclusion of seborrheic keratosis, the most common lesion mimicking cutaneous melanoma. This strata excluded most nonmelanocytic tumors from evaluation, but to the detriment of excluding nonmelanocytic tumors that may have also simulated melanoma, such as pigmented basal cell carcinomas.

Stratum C design included all lesions for which the clinician indicated a clinical concern for melanoma; however, the number of articles with this structure was relatively low, representing only 5 US-based publications with data for 101 clinicians. Thus, the true range of NNB for US-based dermatologists when evaluating a lesion concerning for melanoma was likely not precisely represented in this data grouping. In addition, the practice of routinely listing rule out melanoma as the clinical impression on a pathologic order, in an effort to avoid insurance denial for the removal of clinically and dermoscopically innocuous melanocytic neoplasms, may disproportionately skew the data in a US-based clinician cohort.

The absence of identified US-based PCP data was notable and represented an undefined portion of the melanoma screening biopsies performed. Primary care practitioners address dermatological concerns with variable expertise and, often out of necessity, in geographic regions with low-level access to dermatological services. Clinics that publicly profess their dermatological expertise but are staffed by physicians who lack dermatology residency training and board certification are increasing in number. The NNB for cutaneous melanoma of these clinicians is entirely absent in the literature we reviewed.

The wide range of NNB, paucity of US-based clinician data, and varying study designs reported in the included publications highlight the potential advantage of developing consensus-led, structured clinical impression and pathologic diagnostic data to standardize NNB reporting across electronic health records and health care systems. Performance metrics, when paired with opportunities for clinical practice improvement, have the potential to optimize patient comfort and satisfaction, improve clinicians’ self-efficacy regarding noninvasive diagnosis, improve busy clinicians’ allocations of time and effort (for both dermatologists and pathologists), and reduce health care costs.

The number of studies that had to be excluded highlights the need for minimal metrics for meaningful NNB comparison across publications, including explicit reference to the number of participating clinicians, clinician specialization and training, reported use of dermoscopy by clinicians, specific criteria for data inclusion or exclusion, methods of data extraction and validation, and reporting of total tumor counts to support NNB metric calculations. Removal of benign skin lesions is entirely expected, given that perfect diagnostic accuracy in cutaneous melanoma has not been demonstrated even by pigmented lesion experts,⁵⁰ and such removal may be appropriate to avoid missing a clinically subtle melanoma. However, these data indicate an opportunity for education-driven and technology-supported melanoma screening programs that may mitigate the concern for patient harms by optimizing NNB when indicated.⁵¹

Limitations

This systematic review, despite the careful structuring and multireviewer evaluations of identified articles, carries the inherent limitations of retrospective analyses: specifically, publications appropriate for study inclusion were not identified through the search process. The paucity of data on US-based clinicians and varying study designs also limited our consolidated data analysis and highlighted the need to develop consensus-led, structured clinical impression and pathologic diagnostic data to standardize NNB reporting across electronic health records and health care systems. Given the statistically significant heterogeneity of the data we reviewed, representing patient populations with variable melanoma prevalence and clinicians with diverse training and experience, the interpretation of any summary metrics should be approached with extreme caution.

Conclusions

The reported NNB for cutaneous melanoma varies widely across clinicians, geography, and patient populations. Variations in study designs compromise the ability to compare publications across design types. Standardization of NNB reporting offers an opportunity to improve the quality of published reports on this important but imperfect metric, which quantifies the volume of biopsies that result in benign diagnoses in the course of identifying cutaneous melanoma.

Supplement.

eMethods. MEDLINE Search Strategy

eFigure 1. PRISMA 2009 Flow Diagram

eFigure 2. Systematic Review of Publications Addressing Number Needed to Biopsy for Melanoma