Skin Malignancy in the Military: A Number Needed to Biopsy Analysis

ABSTRACT

Introduction

Skin malignancy has increased in prevalence over the last 15 years and effective diagnosis is required for adequate treatment. Retrospective data analysis of skin biopsy data has shown correlation between various independent variables, but no studies have been shown to directly assess skin malignancy risks for military personnel. Assessing correlation could lead to more effective, targeted screening programs that could lead to decreased mortality from skin malignancies. We present a 1-year analysis of the number needed to biopsy (NNB) to detect skin cancer and analysis of military-specific risk factors in a military dermatology training program. The present study aims to (1) compare skin biopsy yields to civilian institutions and patient populations and (2) determine significance of exposure variables including age, gender, military beneficiary status, branch of service, and military rank.

Materials and Methods

We performed a retrospective observational study over 1 year by identifying all skin biopsies performed in the Walter Reed National Military Medical Center dermatology clinic from August 2015 to July 2016. Utilizing the pathology reports, we manually excluded biopsies performed for the purpose of ruling out inflammatory/immunologic conditions or cosmeses and focused only on encounters performed to rule out basal cell carcinoma, squamous cell carcinoma, or melanoma. We decided to exclude malignant diagnoses that were exceedingly rare or could mimic inflammatory conditions, such as cutaneous T-cell lymphoma. For uncertain diagnoses with vague context per pathology report, previous office clinic notes and pre-biopsy differential were referenced and included only if melanoma or non-melanoma skin cancer (NMSC) diagnosis was the intended indication.

Results

A total of 3,098 biopsies were included in the study, diagnostic for 1,084 total skin malignancy and 54 melanoma diagnoses. Melanoma comprised 4.98% of all skin malignancy diagnosed. The NNB for all skin malignancy was 2.86 (95% CI 2.76-2.96) and NNB for melanoma and NMSC was 20.93 (95% CI 19.70-22.15) and 1.91 (95% CI 1.83-2.00), respectively. Patient age, gender, and military rank significantly impacted NNB values (P < .001).

Conclusions

The proportion of melanoma skin cancers is notably increased in our population compared to published population statistics with comparable total biopsy yields. Skin biopsy for purpose of screening for malignancy should be performed in the military population and consideration should be made for gender, age, and rank. Our findings can further expand on military risk factors for skin cancer and aid in further multivariant modeling.

INTRODUCTION

Skin malignancy has increased in prevalence over the last 15 years and effective diagnosis is required for adequate treatment. The US Preventive Services Task Force has published general guidelines for screening and diagnostic follow-up but remains inconclusive on whether increased screening has led to decrease in all-cause mortality from melanoma.¹ They state that future research on skin cancer screening should focus on evaluating the effectiveness of targeted screening in persons considered to be in high risk for skin cancer. Retrospective data analysis of skin biopsy data has shown correlation between independent variables, but no study has been shown to directly assess skin malignancy risks for military personnel. Assessing correlation between military-specific demographics and skin biopsy yields could lead to more effective, targeted screening programs and ideally decrease mortality from skin malignancies.

In a large ecological study of Medicare beneficiaries over 15 years, the number of skin biopsies performed is increasing at a greater rate (146%) than the number of skin cancer treatments performed (56%).² Furthermore, the same study demonstrated, from 2000 to 2015, the rate of skin biopsies that resulted in a subsequent procedure fell from 76% to 49%. Presumably, the decline in percentage of biopsies requiring subsequent treatment is a result of dermatologists having a lower threshold to perform a skin biopsy than existed in the past, leading to more biopsies being performed. Therefore, it remains critical to assess individual patient risks before proceeding with skin biopsy as a means of raising this threshold and approaching a more consistent biopsy rate. As such, determining the accuracy of clinical suspicion is a powerful tool when assessing outcome measures in a healthcare organization.

Various studies have demonstrated either number needed to screen or number needed to biopsy (NNB) as tools to explore the yield of confirmed skin malignancy.^3–5 Often, these studies stratify risk factors and biopsy yields of various types of malignancy as the dependent variable. Well-described risk factors for malignancy such as male gender and increasing age have demonstrated increasing yield for malignancy. In a unique population such as the military, unexpected risk factors may present due to occupational exposure or socioeconomic status. Several studies have claimed increased risk for those with military service, active duty, and veterans alike.^6–8 In this study, we attempt to assess if risk factors specific to military beneficiaries could improve quality of care by increasing the accuracy of skin malignancy detection rates relative to biopsies performed. By studying specific military risk factors, we hope to provide more consistent and efficient dermatologic care for all those serving in the military.

METHODS

We identified all skin biopsies performed in the Walter Reed National Military Medical Center (WRNMMC) dermatology clinic from August 2015 to July 2016. Utilizing the pathology reports, we manually excluded biopsies performed for the purpose of ruling out inflammatory/immunologic conditions or cosmeses and focused only on encounters performed to rule out basal cell carcinoma (BCC), squamous cell carcinoma (SCC), or melanoma. We decided to exclude malignant diagnoses that were exceedingly rare or could mimic inflammatory conditions, such as cutaneous T-cell lymphoma. For uncertain diagnoses with vague context per pathology report, previous office clinic notes and pre-biopsy differentials were referenced and included only if melanoma or non-melanoma skin cancer (NMSC) diagnosis was the intended indication.

Data Collection

For each biopsy, determination of inclusion was performed based on a pathologic diagnosis for which malignancy could be considered and a pre-biopsy differential which included skin malignancy. After inclusion, patient’s age, gender, military beneficiary status (sponsor versus dependent), branch of service, and military rank category were obtained from the electronic medical record (EMR).

Validation by ICD Diagnosis

To ensure accurate inclusion of data, we performed a validation study based on the International Classification of Diseases (ICD) 9-238.2 code “Neoplasm of uncertain behavior.” Of the 200 randomly chosen patients included in the study, 96.5% had this diagnosis in the EMR. Alternate ICD codes included other benign neoplasm of skin (ICD-9-216.9) and skin neoplasm, unspecified (ICD-9-239.2).

Statistical Analysis

Number needed to biopsy was calculated as the inverse of absolute risk per biopsy performed for malignancy rule out. Absolute risk was calculated as the number of positive results divided by the total biopsies performed and clinically indicates malignancies detected per biopsy performed. Chi-squared tests were performed for all EMR demographic data, including military beneficiary status, gender, age, and rank category. Type 1 error was controlled at 5% for all tests and reported 95% confidence intervals (CIs). Additionally, we calculated relative risk (RR) within each demographic variable. Relative risk is the comparison of an outcome, in our case diagnosis of skin malignancy, occurring between mutually exclusive groups. Relative risk was calculated by selecting a control group within each demographic variable and dividing the absolute risk of each other group by the control and reported with a 95% CI. A statistically significant result within RR was established if the CI excluded “1.” All calculations were performed via computer software (Excel). For purposes of this study, a multivariate model for risk stratification was not attempted.

Ethical Considerations

The study was given a waiver by the Institutional Review Board of WRNMMC as part of quality assurance and performance improvement policies in place. To preserve integrity of personal identifiable information, age and rank categories were used to avoid extrapolation of patient identity.

RESULTS

Population

A total of 5,391 biopsies were performed from August 2015 to July 2016. A total of 3,098 biopsies met inclusion criteria, of which 1,084 were positive for BCC, SCC, melanoma, and atypical nevus/squamous proliferation recommended for re-excision. Fifty four of the positive results were for melanoma. Among the total malignancies diagnosed, the percentage of melanoma was 4.98%. Further population statistics can be found in Table I.

TABLE I.

Patient Demographics and Calculated number needed to biopsy (NNB)

	Patients, n (%)	NNB	95% Confidence Interval	P-value
N, total	3,098	2.86	(2.76-2.96)
Pigmented lesions	1,100	20.93	(19.71-22.15)
Nonpigmented lesions	1,998	1.91	(1.83-2.00)
Sex				<.001
Male	1,996 (64.43)	2.68	(2.56-2.79)
Female	1,102 (35.57)	3.26	(3.07-3.45)
Age, mean (SD)	54.65 (±18.24)			<.0001
<35	493 (15.91)	17.61	(16.05-19.16)
35-49	710 (22.92)	4.93	(4.57-5.29)
50-65	945 (30.50)	2.45	(2.29-2.60)
65<	950 (30.66)	1.81	(1.69-1.92)
Military status				.10
Sponsor	2,163 (70)	2.78	(2.67-2.90)
Dependent	935 (30)	3.05	(2.85-3.24)
Branch of service (% of sponsors)				.07
U.S. Army	688 (31.81)	2.74	(2.54-2.95)
U.S. Navy	593 (27.42)	2.64	(2.42-2.85)
U.S. Air Force	494 (22.84)	2.86	(2.60-3.11)
Marine corps	237 (10.96)	3.59	(3.13-4.05)
Otherb	134 (6.20)	2.53	(2.10-2.96)
Enlisted service members	618 (28.57)	3.68	(3.39-3.97)	<.0001a
E1-E4	50 (2.31)	6.25	(4.52-7.98)
E5-E7	345 (15.95)	3.63	(3.25-4.01)
E8-E10	223 (10.31)	3.43	(2.98-3.88)
Officer service members	1,451 (67.08)	2.53	(2.40-2.66)
O1-O3	135 (9.80)	5.19	(4.32-6.07)
O4-O6	1,104 (67.08)	2.57	(2.42-2.72)
O7-O10	212 (28.57)	1.81	(1.57-2.06)

a

Value <.0001 when comparing rank subcategories of officer and enlisted.

^b

Includes public health service and coast guard.

Number Needed to Biopsy

Our NNB for all skin malignancy was 2.86 (95% CI 2.76-2.96). A total of 1,100 biopsies were performed on pigmented lesions and 1,998 were performed on nonpigmented lesions. Of those biopsied, we diagnosed 626 BCC, 363 SCC, 54 melanoma, and 30 cases of atypical nevi that were recommended to undergo further excision. When further broken down into melanoma and NMSC, our NNB was 20.93 (95% CI 19.70-22.15) and 1.91 (95% CI 1.83-2.00), respectively.

Sex

1,996 males and 1,102 females were biopsied over the year. The NNB was 2.68 (95% CI 2.56-2.79) for males and 3.26 (95% CI 3.07-3.45) for females with a P-value of <.001. Relative risk within the sex category demonstrated statistical significance at 1.22 (95% CI 1.10-1.35) for males compared to females.

Age

For age, we categorized patients by 15-year increments, with lower limit <35 years old and upper limit >65 years old. Age was determined at the time of biopsy rather than at the time of diagnosis or data collection. Within each category, we report an NNB of 17.61 (95% CI 16.05-19.16) for ages <35, 4.93 (95% CI 4.57-5.29) for ages 35-49, 2.45 (95% CI 2.29-2.60) for ages 50-65, and 1.81 (95% CI 1.69-1.92) for age >65. With advancing age, we demonstrated incremental decrease in the NNB with a P-value of <.001. Relative risk within the age category demonstrated statistical significance at 0.10 (95% CI 0.07-0.15) for ages <35, 0.37 (95% CI 0.31-0.43) for ages 35-49, and 0.74 (95% CI 0.67-0.81) for ages 50-65 when compared to the >65 age category.

Sponsors versus Dependents

In the military health system, patients are identified using a “20” digital identifier if they are the primary DoD beneficiary or sponsor and alternate numbers if they are dependents. A total of 2,163 sponsors were biopsied over the year, with approximately 935 dependents biopsied. We report an NNB of 2.78 (95% CI 2.67-2.90) for sponsors and an NNB of 3.05 (95% CI 2.85-3.24) for dependents. Relative risk within the military beneficiary status did not demonstrate statistical significance at 0.91 (95% CI 0.82-1.02) for dependents when compared to sponsors.

Branch of Service

Branch of service was separated by three major branches of the military (army, navy, and air force), the marine corps, and the remaining sponsors, which include the coast guard, public health service, and DoD contractors. Although not statistically significant, there was a slight increase in the NNB for the marine corps at 3.59 (95% CI 3.13-4.05) that excluded 95% CI for the remaining branches of service (see Table I). Relative risk within the branch of service category demonstrated statistical significance for the marine corps, 0.76 (95% CI 0.61-0.96), when compared to the army. Remaining RRs were not statistically significant (see Table II).

TABLE II.

Relative Risk Based on Demographics

	Relative risk	95% Confidence Interval
Sex
Male	1.22	(1.10-1.35)
Female	1	–
Age, mean (SD)
<35	0.10	(0.07-0.15)
35-49	0.37	(0.31-0.43)
50-65	0.74	(0.67-0.81)
65<	1	–
Military status
Sponsor	1	–
Dependent	0.91	(0.82-1.02)
Branch of service (% of sponsors)
U.S. Army	1	–
U.S. Navy	1.04	(0.90-1.20)
U.S. Air Force	0.96	(0.82-1.12)
Marine corps	0.76	(0.61-0.96)
Othera	1.08	(0.86-1.37)
Enlisted service members	0.69	(0.60-0.79)
E1-E4	0.29	(0.15-0.55)
E5-E7	0.50	(0.40-0.62)
E8-E10	0.53	(0.42-0.67)
Officer service members	1	–
O1-O3	0.35	(0.24-0.50)
O4-O6	0.71	(0.61-0.81)
O7-O10	1	–

a

Includes public health service and coast guard.

Rank Category

Rank category is arranged by ranges that are approximated by milestones in the rank structure and are listed as pay grades. “E” designation signifies enlisted, while “O” designation signifies commissioned officers. Out of the 2,163 sponsors, 1,451 were officers and 618 were enlisted (94 patients were DoD contractors or were unable to be found in the EMR). Analysis was run separately by rank category as well as officers versus enlisted with a statistical significance <0.0001 for both. The NNB and respective 95% CI can be found in Table I. Relative risk within the rank category demonstrated statistical significance for enlisted members at 0.69 (95% CI 0.60-0.79) when compared to officers. All individual rank categories demonstrated statistical significance when compared to the O7-O10 group with results in Table II.

DISCUSSION

Comparison to Other Studies

Our NNB for all skin malignancy (2.86), melanoma (20.93), and NMSC (1.91) was comparable to other similar reports.^3–5,9–13 Most civilian institutions reported NNB for two major subsections: NMSC and melanoma. These values ranged from 1.6-1.86 for NMSC to roughly 9.2-27.8 for melanoma. Despite similarities between NNB, several assumptions must be made to interpret the data, namely provider-to-provider variation in clinical practice. Additionally, data collection was variable as many studies drew data directly from the EMR. Since NNB is the inverse of absolute risk, its utility to describe disease prevalence can be skewed by varying diagnostic accuracies or sensitivity thresholds. Many of the studies focused on both primary clinic and tertiary medical center populations, which impacts biopsy yields as the patient populations would likely differ. Furthermore, differing referral thresholds for primary care providers would affect overall number of biopsies performed and thus affect NNB at the primary care level. Overall, it is difficult to extrapolate NNB data to draw conclusions on population risk or as a metric for skin cancer care in the absence of measurement standardization.¹⁴

We were able to reproduce increased RR for well-known risk factors for skin malignancy such as age and gender. Given we studied a single population at a tertiary care center, statistical significance would suggest that our reported values have some utility in assessing prevalence rather than simply diagnostic variability. Within demographic variables with greater than two categories, we demonstrated RR to a control group within our population. By comparing independent groups within each variable, we were able to quantify risk in relation to our selected control. Relative risk, in this case, allows for more useful clinical application to approximate relative prevalence within each category. Therefore, trends in our data would support diagnostic consideration for military branch and rank potentially increasing diagnostic yields, leading to more efficient clinical screening.

Military Incidence

In population statistics, melanoma accounts for about 1% of all skin malignancies.¹⁵ In our population, melanoma accounted for nearly 5% of diagnosed skin malignancy. Although increased incidence of melanoma could partially be explained by the data collection from a tertiary center, our data are congruent with proposed increased risk in the military population found in the study by Lea et al. As such, increased education, screening, and access to care would provide greater yield for an increased risk population. A large retrospective review of melanoma incidence from 2005 to 2014 showed marine corps incidence of 7.7 per 100,000 person-years, which was lower than air force 16.0, navy 12.3, and army 9.3. Additionally, there was an increased incidence of 31.1 per 100,000 person-years in officers when compared to 7.0 in enlisted.¹⁶ The differences in incidence would seem to correlate with our findings of decreased RR for both marine corps (RR 0.78, 95% CI 0.68-0.87) and enlisted members (RR 0.76, 95% CI 0.70-0.82) in our studies. We hypothesize that certain career fields limited to officers such as pilots could contribute to increased occupational exposure, resulting in increased RR for skin malignancy.

Data Collection

The purpose of this project was to compare civilian and military rates of malignancy and to assess military-specific risk factors. By focusing inclusion on provider and clinical variables, we were able to gain insight on both accurate clinical assessment as well as disease prevalence within a population. We predict our data provide insight to a provider’s clinical assessment as we utilized pre-biopsy differential in our inclusion criteria. Therefore, the NNB can be used to predict diagnostic accuracy rather than simply population risks since similar pathologic diagnoses could be either included or excluded. Although the method of biopsy (excisional versus incisional) may alter diagnostic accuracy, our intent was to study biopsy yields within the confines of current dermatological standard of care and physician evaluation and thus was not included as a study variable. Furthermore, the study design and limitations to the EMR do not allow for accurate assessment of environmental and occupational exposures such as number deployments, job duties, etc. As such, the presented data corresponding to military branch would benefit from further subcategory analysis based on the military-specific exposure. We suggest a future cohort or case–control studies to assess these exposures.

Given the retrospective nature of the study, we were unable to include several patients in our study due to incomplete information within the EMR. With regard to rank category, 36 with malignancy and 58 in the inclusion population were either DoD contractors or unlisted in the EMR. Furthermore, nine patients with malignancy and eight in the inclusion population did not have their branch of service listed. We suspect that, given the proportionally low number of these patients, this discrepancy did not affect our statistical significance and thus our conclusions.

Beneficiary Status

There is data to suggest that certain military occupations increase the risk of skin malignancy—particularly in aviation or with exposure due to occupational duties.^6–8 Although our results did not demonstrate statistical significance in the NNB between individuals who served on active duty versus their dependents, our data would suggest a slight trend toward increased malignancy risk (NNB of 2.78 [95% CI 2.67-2.90] for sponsors and NNB of 3.05 [95% CI 2.85-3.24] for dependents). This trend toward increased risk is supported by numerous variables, such as deployments in equatorial latitudes (Iraq and Afghanistan), occupational exposures, and increased active lifestyles, which could not be distinguished with our study parameters and limitations to the EMR.⁶ Additionally, military families are subjected to more frequent relocations and are possibly more active outdoors, which may present confounding variables in distinguishing statistical significance. Sub-analysis of specific occupational hazards would be beneficial but are not available solely via EMR.

Veterans/socioeconomic Gap

Perhaps one of the most obvious selection biases within our sample population is the exclusion of former military members, with significant occupational exposures, that have either exited the military before retirement or obtained dermatologic care elsewhere. The overwhelming majority of these service members would have left before retirement to pursue further careers at the end of their enlistment or service requirement. Data on these individuals are limited, with the exception of those who pursue medical care in the Veterans Administration (VA) healthcare system. Studies of the VA patient population have shown an increased risk for all skin malignancy yet this patient population does not fully reflect all who have served.⁸ We describe three clinically distinct populations of patients with military exposures: active duty/retired service members, separated/retired members receiving care from the VA, and separated/retired members receiving care in the civilian sector. Further assessment of these distinct populations would be best assessed in prospective studies with long-term follow-up. Therefore, it is difficult to assess absolute correlation between risk for skin malignancy and military service as a distinct population of service members cannot be measured in this study.

Use of the Number Needed to Biopsy

Utilization of the NNB as a metric for dermatologic care should be carefully considered. Although it is a commonly utilized metric for measuring diagnostic accuracy, the value itself leads to many statistical ambiguities and avenues for bias. Reported NNB has ranged between 2.2 and 287 per large systematic review demonstrating variability, which can be highly influenced by data manipulation.⁹ Until a standard method of measuring NNB is adopted, it would be more valuable to compare populations or diagnostic accuracy with single data sets to avoid differences in inclusion criteria. Furthermore, meta-analyses would provide limited insight unless standardized physician populations (specialist versus primary care) and data collection (prospective versus retrospective) were utilized.

CONCLUSION

The proportion of melanoma skin cancers is notably increased in our population compared to published population statistics with comparable total biopsy yields. Skin biopsy for the purpose of screening for malignancy should be performed in the military population and consideration should be made for gender, age, and rank. Our findings can further expand on military risk factors for skin cancer and aid in further multivariant modeling. We recommend further exploration into military-specific risk and increased screening of the military population.

ACKNOWLEDGMENTS

There are no further acknowledgments.

FUNDING

None declared.

CONFLICT OF INTEREST STATEMENT

Authors report no conflict of interest.

Author notes

Presented at the Uniformed Services University Research Day, May 2019.

The views and opinions of authors expressed herein are those of the authors and do not reflect the official policy or position of the Department of the Air Force, Department of the Army, DoD, or US government.