Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2020 Dec 1.
Published in final edited form as: J Genet Couns. 2019 Jul 17;28(6):1208–1213. doi: 10.1002/jgc4.1155

Insurance Coverage Does Not Predict Outcomes of Genetic Testing: The Search for Meaning in Payer Decisions for Germline Cancer Tests

Laura M Amendola 1, M Ragan Hart 1, Robin L Bennett 1, Martha Horike-Pyne 1, Michael Dorschner 2, Brian Shirts 3, Gail P Jarvik 1,4
PMCID: PMC6901727  NIHMSID: NIHMS1040447  PMID: 31317629

Abstract

In this work, we explore the results of germline cancer genetic tests in individuals whose insurance would not cover this testing. We enrolled thirty-one patients with a personal history of cancer whose health insurer denied coverage for a clinical germline cancer panel genetic test recommended by a medical genetics provider into a study providing exome sequencing and return of cancer related results. Five participants (16%) had a pathogenic variant identified related to increased cancer risk. Three participants (10%) had a variant of uncertain significance (VUS) in a gene related to their cancer history. These rates are not significantly different than the 12% rate of pathogenic or likely pathogenic (P/LP) variants and VUS in 1,462 patients approved by insurance to have a similar clinical germline cancer test (p=0.59 for P/LP variants; p=0.87 for VUS; Shirts et al 2016). Health insurance guidelines may not meaningfully differentiate between patients with cancer who are likely to benefit from germline cancer genetic testing and those who will not. Failure to identify pathogenic variants in this research cohort would have led to suboptimal care. Strategic evaluation of current germline cancer genetic testing coverage policies is needed to appropriately deliver precision medicine.

Keywords: Health insurer coverage, insurer guidelines, hereditary cancer, germline genetic testing, Genetics, Genetic Counseling, Clinical Genetics, Cancer genetics

Introduction

The equitable integration of genomic sequencing into routine clinical care requires patient access to this technology. Health insurer coverage policies do not exist for all types of genetic testing, and those that do may not be consistent across payers(Graf, Needham, Teed, & Brown, 2013; Kutscher, Joshi, Patel, Hafeez, & Grinspan, 2017; Wang, Beattie, Ponce, & Phillips, 2011) or conflict with professional society guidelines (Spoonamore & Johnson, 2016). A lack of transparency and standardization in health insurer coverage policies for genetic testing may be a barrier for patients who are recommended to have genetic testing.

Payers have proposed that clinical utility, exhibited by improved outcomes and well-informed decision making, is necessary for an intervention to be a covered benefit(Pezalla, 2016). There is evidence to support the diagnostic utility of certain next generation sequencing genetic tests, for example germline cancer tests in patients being evaluated for hereditary cancer(Desmond et al., 2015; Ricker et al., 2016). The results of hereditary cancer genetic testing inform cancer screening recommendations and surgical considerations, and multi-gene germline cancer tests can identify hereditary cancer syndromes that would not be found on a targeted single gene test. However, payer identified barriers to multi-gene hereditary cancer testing coverage remain, including insufficient evidence, lack of rigor in test design, and the departure from diagnostic testing to genetic screening (Trosman et al., 2017).

In this work we explore whether having insurance coverage for cancer genetic testing predicts the outcomes for patients being evaluated for hereditary cancer. Our objective is to highlight a potential lack of evidence supporting the implementation of current cancer genetic testing coverage policies and encourage further evaluation of their utility, consistency and evidence gaps.

Methods

Participants

All participants were patients seen at the University of Washington, Genetic Medicine Clinic between August 2013 and August 2017. Thirty three patients were identified as eligible for enrollment. Eligibility criteria included 1) a personal history of cancer, 2) a referral for a hereditary cancer evaluation, 3) a clinical germline cancer panel test recommended by a certified genetic counselor after in-person genetic counseling, and 4) a request for insurance pre authorization for this testing that was denied by the patient’s health insurer. The specific germline cancer panel test recommended was either the University of Washington Coloseq™ or BROCA gene panel dependent on patient personal and family history of cancer. Patients were consented for research participation via telephone by their clinical genetic counselor and returned signed consent documents via mail in summer 2017. This research was approved by the University of Washington, Institutional Review Board as part of a larger project investigating the incorporation of genomic sequencing into clinical evaluation for hereditary cancer(Gallego et al., 2014).

Procedures

Exome sequencing was performed using an in-house optimized, xGEN Exome Research Panel v1.0 (Integrated DNA Technologies, Coralville, IA) sequence capture system to enrich participant-specific libraries for the coding portions of the genome. Enriched libraries were subjected to sequencing (paired-end, 2 × 101bp) on a HiSeq 4000 (Illumina, San Diego, CA), using standard chemistry. Resulting sequence data was processed with a standard pipeline, including: 1) Burrows Wheeler aligner (BWA) for alignment of reads to the reference genome, 2) Genome Analysis Tool Kit (GATK) for variant calling and 3) variant annotation with an in-house tool based on SNPeff. All participants had variants in 65 genes associated with hereditary cancer annotated and interpreted for return. Pathogenic and likely pathogenic variants (P/LP) were reported in any cancer gene. Variants of uncertain significance (VUS) were only reported in genes related to the participant’s cancer history. Participants were also given the option to receive P variants in incidental finding genes, which included adult onset medically actionable conditions, select carrier status results and select pharmacogenomics variants. The development and content of the incidental finding gene list has been described elsewhere(Amendola et al., 2015; Berg et al., 2013; Dorschner et al., 2013). Results were returned to participants via telephone by their clinical genetic counselor and a report was mailed to them and entered into their electronic medical record.

Data Analysis

We compared P/LP variant rates and, separately, VUS rates, in this research cohort to the rates in a cohort of patients who had a similar clinical germline cancer test that was approved by insurance(Shirts et al., 2016) using a two-tailed Fisher’s exact test. The majority of these patients who underwent clinical cancer genetic testing had a personal history of cancer (1204/1462, 82%). Forty-four percent had a complex family history of cancer, which was defined as the presence of multiple cancer types in the proband and their first-degree relatives that was not clearly consistent with a single P variant in an established cancer gene passing through the family.

Results

Participants

Thirty-one of the thirty-three participants eligible for enrollment consented for participation. The majority of participants had European Ancestry (27/31, 87%), were female (94%) and had a personal history of breast cancer (71%). Ten of the twenty-two participants with a history of breast cancer had previously undergone normal BRCA1/2 gene sequencing. The majority of participants had commercial health insurance (87%) at the time of the insurance pre authorization request. A reason for denial of coverage was documented for twenty-five (81%) of the participants (Figure 1). Reasons for denial included; testing was considered experimental or investigational (13/25, 52%), the patient did not meet medicacriteria (24%), testing was not medically necessary (16%) and testing was out of network (8%).

Figure 1.

Figure 1.

Open in a new tab

Reason for Denial of Coverage (N = 25)

Return of results

Five of the thirty-one participants (16%) had a P variant related to increased cancer risk identified, none had a LP variant. An additional three participants (10%) had a VUS in a gene consistent with their cancer history (Table 1). Changes to medical management and cascade testing were discussed with the participants identified to have a P variant in the BRCA2 and PALB2 genes. The participant with a P variant in BRCA2 was referred to a local high risk breast and ovarian cancer clinic to discuss breast MRI, bilateral salpingo-oophorectomy and pancreatic cancer screening due to a family history of pancreatic cancer in her father. Her two sisters and her mother, as well as multiple second and third degree relatives, were eligible for cascade testing. The participant with a P PALB2 variant had a personal history of breast cancer so was already receiving high risk breast cancer screening. She was referred to a local high risk cancer clinic to discuss the risks and benefits of pancreatic cancer screening. Her adult daughter and her sister were eligible for cascade testing. Cascade testing of adult first degree relatives was also discussed with the additional three participants with a P variant that did not impact their medical management: the participant with a P variant in the CHEK2 gene had two daughters and two siblings eligible for testing, the participant with a P variant in the HOXB13 had a brother and father eligible for testing, and the participant with a BRIP1 pathogenic variant had a daughter and a sister eligible for testing.

Table 1.

Cancer Related Variants Returned to Participants

Finding Participant Information
Gene Variant Interpretation Cancer Diagnosis Sex Previous Genetic Testing Reason for denial
BRCA2 p.Glu1285fs Pathogenic Gastric cancer diagnosed age 36 Female None E/I
BRIP1 p.Arg798* Pathogenic Ovarian cancer age 63 Female None E/I
HOXB13 p.Gly84Glu Pathogenic Two primary breast cancers diagnosed age 34 Female BRCA1/2; sequencing and del/dup NMC
PALB2 p.Thr799fs Pathogenic Breast cancer age 54 Female None E/I
CHEK2 p.Thr410fs Pathogenic, low penetrance Breast cancer diagnosed age 49 Female None NMC
ATM p.Cys977Tyr VUS Breast cancer age 54 Female None E/I
BARD1 p.Arg751Trp VUS Breast cancer age 51 Female BRCA1/2; sequencing and del/dup NMC
BMPR1A p.Glu468Lys VUS Colon cancer diagnosed age 48 Female None Unknown
Open in a new tab

VUS = variant of uncertain significance; E/I = experimental/investigational; NMC = patient did not meet medical criteria

Two participants with a history of breast cancer and no known personal or family history of colorectal cancer had a single P variant in the MUTYH gene which is associated with autosomal recessive MUTYH associated polyposis. Colorectal cancer risk in monoallelic MUTYH P variant carriers depends on family history(Jones et al., 2009; Win et al., 2014), and increased colorectal cancer screening is only recommended for carriers with a first degree relative with a diagnosis of colorectal cancer based on current by National Comprehensive Cancer Network guidelines. To be conservative, we did not count these two participants in our total number of participants with P variants related to increased cancer risk identified.

Comparison to covered tests

Shirts et al.,(Shirts et al., 2016) reported a 12% rate of P/LP variants (179) in 1,462 patients who had a similar clinical germline cancer genetic test, which was approved by insurance; the VUS rate was 10.5% (157/1,462). This reported rate of P/LP variants in patients with testing covered by insurance was not significantly different than the rate in our cohort of participants whose insurer denied coverage for testing (p=0.59). The reported VUS rate in patients who received this clinical test was also not significantly different than the VUS rate in this research cohort (p = 0.87).

Discussion

Patients denied coverage for germline cancer genetic tests had the same rate of P/LP variants and VUS as a published series of patients who had similar testing that was covered by their health insurer. This finding suggests that health insurance guidelines may not meaningfully differentiate between cancer patients who are likely to benefit from germline cancer genetic testing and those who will not. In this research cohort, changes to medical management were indicated for two of the five participants with P variants identified. A discussion of cascade testing of first degree relatives was also indicated for all five of these participants. Thus, failure to identify these variants would have led to suboptimal care.

Other authors have documented a similar rate (7.4 – 9.6%) of P/LP variants on four multi-gene germline cancer genetic tests in over 2000 patients referred for testing at a commercial laboratory(LaDuca et al., 2014). Interestingly, of those patients found to have a P/LP variant, 30% had clinical histories that did not meet National Comprehensive Cancer Network diagnostic/testing criteria(LaDuca et al., 2014). While these patients did receive a test, and thus are likely to have had insurance coverage, not meeting diagnostic/testing criteria could have resulted in failure to test. These results and ours raise concerns for similar patients and their families with undiagnosed cancer predisposition syndromes and suggest the need for broader germline cancer panel testing criteria.

Heterogeneity in payer decisions across patients is caused in part by different health insurer evidence thresholds for coverage of germline cancer genetic tests(Phillips, Deverka, Trosman, et al., 2017). Threshold variability can arise when payers use a range of health technology assessment (HTA) methods in reimbursement coverage decisions for genetic testing. Notably, an analysis of multiple HTA organizations and their associated review processes identified shortcomings in HTA methods that may be informative to payers developing genetic testing policies(Trosman, Van Bebber, & Phillips, 2011). These shortcomings include technology evidence reviews that do not adequately incorporate nonclinical factors, disparate recommendations for clinical guidelines, and inconsistent technology comparisons relevant to genomic testing. Threshold variability can also arise because payers cover and administer health insurance plans based on employer required inclusion of specific health services and for patient populations with different needs. Health plans apply different evidence preferences regarding the time horizon to assess the value of a genetic test and need to cover and account for different budgets to manage the risk pool. Finally, payers must account for a “number-needed-to test” that requires demonstrating clinical and economic utility within a given population.

Health insurer coverage decisions also vary because of differences in payer value frameworks that are intended to aid coverage determination assessments for health care services(Pearson, 2018). Some of this variation is due to stakeholder healthcare decision makers prioritizing different attributes of interventions, which can result in inconsistent value determinations. In addition, the budget impact of covering germline cancer genetic testing is specific to the health insurance decision makers, and is tied to the differences across payers described above. The opportunity cost and cost offsets related to re-allocating resources (away from health spending for other non-genetic testing health services) for genetic testing coverage also contribute to the variability. Thus, identifying HTA methods that can be standardized and developing best practice value frameworks when consensus is previously established for clinical utility could improve the transparency and efficiency of the coverage decision making process, and decrease inconsistency in patient access to appropriate germline cancer gene testing(Faulkner et al., 2012).

Limitations

Given the small number of participants and limited follow up time in this research cohort, we could not meaningfully assess how the diagnostic information returned from cancer genetic testing informed subsequent clinical care management and cascade testing of at-risk relatives. Outcomes evidence is a critical component of assessing the clinical utility of an intervention(Parkinson et al., 2014; Phillips, Deverka, Sox, et al., 2017). Future assessments of patients having cancer genetic testing should include tracking patient outcomes following results disclosure to contribute clinical utility evidence. Of note, the current phase of the National Institutes of Health (NIH) funded Clinical Sequencing Evidence-Generating Research (CSER) consortium, which includes a clinical project providing hereditary cancer panel genetic testing, is focused in part on the clinical utility of genomic sequencing, and exploring medical follow up and cascade testing of relatives(Amendola et al., 2018).

The small size of our research cohort also limited us from exploring insurer or patient phenotype specific differences in reason for denial and diagnostic rate. Investigating this question in a larger cohort of patients is necessary to identify factors that may put some patients without coverage at a higher chance of remaining with an undiagnosed hereditary cancer syndrome. The participants in this research cohort were all referred for medical genetics evaluation due to suspicion for hereditary cancer. Thus, the diagnostic rate found here may not be generalizable to all patients having hereditary cancer genetic tests. Finally, the majority of our participants had European ancestry. The rate of P/LP variants and VUS may be different in patients with an ancestral background where genetic variation has been less studied.

Current insurer policies for germline cancer genetic testing do not consistently provide coverage to patients who are likely to have hereditary cancer. This could be due in part to variability in how the evidence that supports these policies is assessed, and variation in how payers assign value to the different attributes of genetic testing. This work highlights the need to strategically evaluate the utility, consistency and evidence gaps of current germline cancer genetic testing coverage policies, in order to appropriately deliver on the promise of precision medicine.

Implications for Genetic Counseling Practice

This work highlights the importance of clinical judgement when considering genetic testing for patients referred for hereditary cancer evaluation. Strict adherence to health insurer and/or other national testing eligibility criteria is likely to lead to missed hereditary cancer diagnoses. Genetic counselors should explore research opportunities and self-pay options with patients whenever possible, and encourage patients who cannot access hereditary cancer genetic testing to continually check with their provider or the genetics clinic regarding updates to testing guidance and to reevaluate their coverage should they change health insurers.

Conclusions

Lack of access to genetic testing due to limitations of insurance coverage is a barrier to implementing clinical genomic medicine. Patients likely to benefit from germline cancer genetic testing may be denied testing coverage by health insurance guidelines, which can lead to suboptimal care. Health insurers’ genetic testing policy guidelines require an evidence base that supports their implementation.

Acknowledgements

This work was funded by the NHGRI & NCI CSER Program (U01HG007307 U01HG006507).

Footnotes

Conflict of Interest

Laura M Amendola, M. Ragan Hart, Robin L Bennett, Martha Horike-Pyne, Michael Dorschner, Brian Shirts and Gail P Jarvik declare that they have no conflict of interest.

Human Studies and Informed Consent

All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1975, as revised in 2000 (5). Informed consent was obtained from all patients for being included in the study.

References

  1. Amendola LM, Berg JS, Horowitz CR, Angelo F, Bensen JT, Biesecker BB, . . . Jarvik GP (2018). The Clinical Sequencing Evidence-Generating Research Consortium: Integrating Genomic Sequencing in Diverse and Medically Underserved Populations. American Journal of Human Genetics, 103(3), 319–327. doi: 10.1016/j.ajhg.2018.08.007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Amendola LM, Dorschner MO, Robertson PD, Salama JS, Hart R, Shirts BH, . . . Jarvik GP (2015). Actionable exomic incidental findings in 6503 participants: challenges of variant classification. Genome Research, 25(3), 305–315. doi: 10.1101/gr.183483.114 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Berg JS, Amendola LM, Eng C, Van Allen E, Gray SW, Wagle N, . . . Jarvik GP (2013). Processes and preliminary outputs for identification of actionable genes as incidental findings in genomic sequence data in the Clinical Sequencing Exploratory Research Consortium. Genetics in Medicine, 15(11), 860–867. doi: 10.1038/gim.2013.133 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Desmond A, Kurian AW, Gabree M, Mills MA, Anderson MJ, Kobayashi Y, . . . Ellisen LW (2015). Clinical Actionability of Multigene Panel Testing for Hereditary Breast and Ovarian Cancer Risk Assessment. JAMA Oncology, 1(7), 943–951. doi: 10.1001/jamaoncol.2015.2690 [DOI] [PubMed] [Google Scholar]
  5. Dorschner MO, Amendola LM, Turner EH, Robertson PD, Shirts BH, Gallego CJ, . . . Jarvik GP (2013). Actionable, pathogenic incidental findings in 1,000 participants’ exomes. American Journal of Human Genetics, 93(4), 631–640. doi: 10.1016/j.ajhg.2013.08.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Faulkner E, Annemans L, Garrison L, Helfand M, Holtorf AP, Hornberger J, . . . Watkins J (2012). Challenges in the development and reimbursement of personalized medicine-payer and manufacturer perspectives and implications for health economics and outcomes research: a report of the ISPOR personalized medicine special interest group. Value in Health, 15(8), 1162–1171. doi: 10.1016/j.jval.2012.05.006 [DOI] [PubMed] [Google Scholar]
  7. Gallego CJ, Bennette CS, Heagerty P, Comstock B, Horike-Pyne M, Hisama F, . . . Veenstra DL (2014). Comparative effectiveness of next generation genomic sequencing for disease diagnosis: design of a randomized controlled trial in patients with colorectal cancer/polyposis syndromes. Contemporary Clinical Trials, 39(1), 1–8. doi: 10.1016/j.cct.2014.06.016 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Graf MD, Needham DF, Teed N, & Brown T (2013). Genetic testing insurance coverage trends: a review of publicly available policies from the largest US payers. Personalized Medicine, 10(3), 235–243. doi: 10.2217/pme.13.9 [DOI] [PubMed] [Google Scholar]
  9. Jones N, Vogt S, Nielsen M, Christian D, Wark PA, Eccles D, . . . Sampson JR (2009). Increased colorectal cancer incidence in obligate carriers of heterozygous mutations in MUTYH. Gastroenterology, 137(2), 489–494, 494.e481; quiz 725–486. doi: 10.1053/j.gastro.2009.04.047 [DOI] [PubMed] [Google Scholar]
  10. Kutscher EJ, Joshi SM, Patel AD, Hafeez B, & Grinspan ZM (2017). Barriers to Genetic Testing for Pediatric Medicaid Beneficiaries With Epilepsy. Pediatric Neurology, 73, 28–35. doi: 10.1016/j.pediatrneurol.2017.04.014 [DOI] [PubMed] [Google Scholar]
  11. LaDuca H, Stuenkel AJ, Dolinsky JS, Keiles S, Tandy S, Pesaran T, . . . Chao E (2014). Utilization of multigene panels in hereditary cancer predisposition testing: analysis of more than 2,000 patients. Genetics in Medicine, 16(11), 830–837. doi: 10.1038/gim.2014.40 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Parkinson DR, McCormack RT, Keating SM, Gutman SI, Hamilton SR, Mansfield EA, . . . Kelloff GJ (2014). Evidence of clinical utility: an unmet need in molecular diagnostics for patients with cancer. Clinical Cancer Research, 20(6), 1428–1444. doi: 10.1158/1078-0432.Ccr-13-2961 [DOI] [PubMed] [Google Scholar]
  13. Pearson SD (2018). The ICER Value Framework: Integrating Cost Effectiveness and Affordability in the Assessment of Health Care Value. Value in Health, 21(3), 258–265. doi: 10.1016/j.jval.2017.12.017 [DOI] [PubMed] [Google Scholar]
  14. Pezalla EJ (2016). Payer view of personalized medicine. American Journal of Health System Pharmacy, 73(23), 2007–2012. doi: 10.2146/ajhp160038 [DOI] [PubMed] [Google Scholar]
  15. Phillips KA, Deverka PA, Sox HC, Khoury MJ, Sandy LG, Ginsburg GS, . . . Douglas MP (2017). Making genomic medicine evidence-based and patient-centered: a structured review and landscape analysis of comparative effectiveness research. Genetics in Medicine, 19(10), 1081–1091. doi: 10.1038/gim.2017.21 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Phillips KA, Deverka PA, Trosman JR, Douglas MP, Chambers JD, Weldon CB, & Dervan AP (2017). Payer coverage policies for multigene tests. Nature Biotechnology, 35(7), 614–617. doi: 10.1038/nbt.3912 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Ricker C, Culver JO, Lowstuter K, Sturgeon D, Sturgeon JD, Chanock CR, . . . Gruber SB (2016). Increased yield of actionable mutations using multi-gene panels to assess hereditary cancer susceptibility in an ethnically diverse clinical cohort. Cancer Genetics, 209(4), 130–137. doi: 10.1016/j.cancergen.2015.12.013 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Shirts BH, Casadei S, Jacobson AL, Lee MK, Gulsuner S, Bennett RL, . . . Pritchard CC (2016). Improving performance of multigene panels for genomic analysis of cancer predisposition. Genetics in Medicine, 18(10), 974–981. doi: 10.1038/gim.2015.212 [DOI] [PubMed] [Google Scholar]
  19. Spoonamore KG, & Johnson NM (2016). Who Pays? Coverage Challenges for Cardiovascular Genetic Testing in U.S. Patients. Frontiers in Cardiovascular Medicine, 3, 14. doi: 10.3389/fcvm.2016.00014 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Trosman JR, Van Bebber SL, & Phillips KA (2011). Health technology assessment and private payers’ coverage of personalized medicine. Journal of Oncology Practice, 7(3 Suppl), 18s–24s. doi: 10.1200/jop.2011.000300 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Trosman JR, Weldon CB, Douglas MP, Kurian AW, Kelley RK, Deverka PA, & Phillips KA (2017). Payer Coverage for Hereditary Cancer Panels: Barriers, Opportunities, and Implications for the Precision Medicine Initiative. Journal of the National Comprehensive Cancer Network, 15(2), 219–228. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Wang G, Beattie MS, Ponce NA, & Phillips KA (2011). Eligibility criteria in private and public coverage policies for BRCA genetic testing and genetic counseling. Genetics in Medicine, 13(12), 1045–1050. doi: 10.1097/GIM.0b013e31822a8113 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Win AK, Dowty JG, Cleary SP, Kim H, Buchanan DD, Young JP, . . . Jenkins MA (2014). Risk of colorectal cancer for carriers of mutations in MUTYH, with and without a family history of cancer. Gastroenterology, 146(5), 1208–1211.e1201–1205. doi: 10.1053/j.gastro.2014.01.022 [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES