Background
Cancer family history (CFH) has many important clinical applications.1 Traditionally, inadequate time and attention are spent on collecting CFH.2–4 Cancer genetic counselors (CGCs) are experts in the collection and assessment of CFH5; however, genetic counseling is labor-intensive and requires a highly skilled workforce.6
Ways to streamline CFH collection and assess validity have relied on outdated family history collection tools.7 Novel electronic tools are in development, including questionnaires eliciting personal and family history.8 Other novel tools include interactive chatbots to simulate natural conversational dialogue and clinical-decision support tools integrated with electronic medical records.4,9
CFH is increasingly important to contextualize unexpected findings identified on widespread tumor and germline sequencing of patients with cancer.10 Almost half of pathogenic germline variants identified in one study of 1,566 patients with advanced cancer were discordant with the individual’s cancer type.11
Patient-reported CFHs have a wide range of accuracy when compared with a reference standard, such as pathology reports, clinical notes, or death certificates.12–15 Scalable tools for collecting CFH may enable easier collection of accurate, standardized CFH information. We sought to compare pedigrees generated from an electronic CFH questionnaire (eCFHQ) alone with the same pedigrees augmented by CGCs for patients seeking cancer genetics services at an academic cancer institute. We hypothesized that an eCFHQ could be used as a standalone tool to determine whether NCCN genetic testing (GT) criteria are met.16,17
Methods
This study was approved by the Dana-Farber Cancer Institute (DFCI) Institutional Review Board. Patients are referred to the Division of Cancer Genetics and Prevention at DFCI for CGC related to a personal and/or family history of cancer. At the time of this study, patients with cancer treated by DFCI oncologists were seen by CGCs alone, whereas patients without cancer (unaffected patients) had appointments with a CGC followed immediately by an appointment with a physician (either a geneticist, oncologist, or gastroenterologist with expertise in cancer genetics) who provides cancer screening recommendations based on family history. Adults scheduled in the Division of Cancer Genetics and Prevention at DFCI were sent a secure link to the eCFHQ, which builds a preliminary pedigree (pre-GC; Figure 1) using the data directly input by patients. The commercially licensed product, the Family History Questionnaire (Progeny Genetics) has been used for standard clinical care since 2016. At the time of this study, the eCFHQ was not integrated with the electronic medical record and was only available in English. Patients whose family history was previously collected from another family member’s genetic counseling (GC) appointment were not sent the eCFHQ. Unique sequential pedigrees collected between June 1, 2018, and December 31, 2018, were included, and all GC appointments were in-person with board-certified CGCs.
Pedigree analysis workflow.
Abbreviation: GC, genetic counseling.
Citation: Journal of the National Comprehensive Cancer Network 20, 8; 10.6004/jnccn.2022.7022
Schedulers encouraged patients to complete the eCFHQ and sent email reminders once a week for up to 3 weeks followed by a telephone reminder (if still incomplete). Patients are able to curate their family history and edit the eCFHQ up until they submit it. Once the eCFHQ is submitted, it is locked and cannot be revised remotely by the patient. Patients met with a CGC within 3 months after being sent the eCFHQ. During the visit, patient-reported personal and CFH information was reviewed and clarified as per standard-of-care practice. This process often includes refining ages at cancer diagnosis and differentiating between reports of various abdominal, pelvic, bowel, and gynecologic cancers. The CGC updated the pedigree with additional information elicited from the patient during the GC appointment.
CGCs independently reviewed paired pre-GC and post-GC pedigrees for differences in CFH. No additional healthcare providers were involved in the review of pedigrees for the purposes of this study. The post-GC pedigrees (n=1,113) were reviewed first by 3 CGCs, and the pre-GC pedigrees (n=1,113) were reviewed afterward by 4 CGCs to determine whether any NCCN GT criteria were met. Pedigrees were sorted and distributed so that no counselor reviewed both the pre-GC and post-GC pedigree of the same patient. The specific NCCN criteria reviewed included those from the NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines) for Genetic/Familial High-Risk Assessment: Breast and Ovarian, Version 2.201916 and the NCCN Guidelines for Genetic/Familial High-Risk Assessment: Colorectal, Version 1.2018,17 in addition to other NCCN Guidelines (see supplemental eTable 1, available with this article at JNCCN.org).
Interrater agreement was assessed for pre-GC and post-GC pedigrees upon a second independent review of 10% of the pedigrees. Agreement was determined based on whether the pedigree indicated that the patient met NCCN GT criteria. Unaffected patients and affected male patients (ie, male patients with cancer) were selected for interrater agreement analysis because they were less likely to meet NCCN GT criteria, which have traditionally focused on female cancers. Before the interrater agreement analysis, the team decided that an acceptable amount of agreement between CGCs would be at least 70% for whether a pedigree did or did not satisfy any NCCN GT criteria.
The time between the eCFHQ being sent and submission by the patient was recorded in days (for submitted pedigrees only). Notably, patient-entered information in the eCFHQ is visible even if the patient does not submit the finalized questionnaire. Regardless of submission status, a nonblank pedigree was defined as one in which at least some family history information was provided. Pre-GC and post-GC pedigrees were further analyzed for changes in demographic information, clinical history, meeting NCCN GT criteria, and family structure. Post-GC pedigrees were used to gather family characteristics, including cancer burden (reported as cancer ratio).
Pedigrees that were left blank before the GC visit were removed (n=85). Only pedigrees with at least some family history information, whether submitted (n=952) or not (n=76), were included in the statistical analysis.
Statistical Analysis
Descriptive statistics were used to summarize patient and family characteristics. Cancer ratio was defined as the number of first-, second-, or third-degree relatives (FDR, SDR, TDR, respectively), excluding the patient with a personal cancer history, divided by the total reported number of FDRs, SDRs, and TDRs. Step-up logistic regression models for the following covariates were performed: sex, affected status, age at visit (split into 3 categories: <50 years, 50–65 years, >65 years), race/ethnicity (as mandated by the NIH), completion of the Family History Questionnaire (self, others, or unknown), prior patient GT (“Have you ever had GT for cancer?”), and prior family GT (“Has anyone in your family had cancer GT?”) were used for 2 different outcome variables. The first was all discrepancies, and the second was only discrepancies that went in the direction of not meeting the NCCN GT criteria based on the pre-GC pedigree to meeting the NCCN GT criteria based on the post-GC pedigree. Race was self-reported, and because of the low numbers of underrepresented patients (each subgroup had ≤50 patients), data from these patients were combined for the logistic regression models.
Results
Demographics
Of the 1,113 patients, 780 (70%) had cancer (affected), 869 (78%) were female, 964 (87%) were White, and the median age was 57 years (interquartile range [IQR], 45–66 years; range, 21–91 years). There were 333 (30%) patients without a cancer history (unaffected); of those, 58 (17%) had gastrointestinal polyps, 19 (5.7%) had nonmelanoma skin cancer, and 11 (3.3%) had other benign tumors (lobular carcinoma in situ, uterine polyps, atypical ductal hyperplasia/atypical lobular hyperplasia, endometrial intraepithelial neoplasm). Prior personal GT was self-reported in 132 (12%) pre-GC pedigrees, and 240 (21%) patients had a family history of previous GT (Table 1).
Patient Characteristics
Among the 1,113 pre-GC pedigrees, 85 (8%) did not have any information filled out by the patient (or person completing it on the patient’s behalf) and were therefore considered blank. The 85 blank pedigrees were removed from the logistic regression models and were analyzed separately (Table 1). The median time to submission of the eCFHQ was 4 days (IQR, 0–12 days; range, 0–448 days).
Agreement
Interrater agreement was assessed between pre-GC and post-GC pedigrees upon a second independent review of 10% of the pedigrees. Agreement was determined based on whether the pedigree indicated that the patient met the NCCN GT criteria. Unaffected patients (77 pre-GC and 77 post-GC) and affected male patients (35 pre-GC and 34 post-GC) were selected for interrater agreement analysis because they were less likely to meet NCCN GT criteria, which have traditionally focused on female cancers. Interrater agreement for pre-GC pedigrees was 95% and for post-GC pedigrees was 87%, surpassing the 70% a priori requirement. Despite this high agreement, NCCN-discrepant pedigrees were reviewed by all 4 CGCs and differences were adjudicated by consensus for confirmation. We defined NCCN-discrepant pedigrees as all pedigrees in which there was a difference in meeting the NCCN GT criteria between pre-GC and post-GC.
Family Characteristics
Based on post-GC pedigrees, the sample population had a median family size (FDR, SDR, and TDR) of 30 (IQR, 22–40), with a median of 6 FDRs (IQR, 4–7), 13 SDRs (IQR, 10–17), and 10 TDRs (IQR, 6–17) (Table 2). The median number of relatives with cancer was 5 (IQR, 3–7). The cancer ratio was slightly higher among unaffected patients (0.19; IQR, 0.13–0.28) than affected patients (0.16; IQR; 0.09–0.23). Overall, 147 (13%) patients reported having at least one half-sibling, 12 (1%) reported being a twin, and 12 (1%) reported that they were adopted.
Family Characteristics
GC Modifications
There were 32,368 individuals (including the patient) in the pre-GC pedigrees and 37,303 individuals in the post-GC pedigrees. Of those added, 917 (14.6%) were FDRs, 2,591 (41.3%) were SDRs, and 2,770 (44.1%) were TDRs.
Modifications to relatives’ ages occurred in 63% of pedigrees as follows: added when previously missing (60%; 95% CI, 56%–63%) and changed by >5 years (12%; 95% CI, 10%–15%) (Table 3). These changes were more frequent among NCCN-discrepant pedigrees (82%; 95% CI, 73%–89%) than NCCN-conforming pedigrees (61%; 95% CI, 57%–64%). At least 1 cancer diagnosis was added to 467 (45%; 95% CI, 42%–49%) nonblank pedigrees, and at least 1 cancer diagnosis was removed from 216 (21%; 95% CI, 19%–24%) nonblank pedigrees. The most frequently added cancer was breast cancer, whereas the most frequently deleted cancer was an unknown/missing cancer type.
Summary of Changes (Nonblank Pedigrees, by Pedigree)
NCCN Cancer GT Criteria
Of the 1,028 nonblank pedigrees, 95 (9%) were found to be discrepant in meeting any NCCN GT criteria. Of these, 17% went from “meeting” to “not meeting” any NCCN GT criteria and 83% went from “not meeting” to “meeting” any NCCN GT criteria.
In a step-up logistic regression model, patients with discrepant pre- and post-GC pedigrees were more likely to report non-White race, belonging to racially and/or ethnically underrepresented groups (odds ratio [OR], 1.91; 95% CI, 1.08–3.25; P=.02), or no prior genetic testing (OR, 4.54; 95% CI, 1.66–18.70; P=.01) or to not answer whether they had prior genetic testing (OR, 18.5; 95% CI, 5.04–88.73; P<.0001) (Table 1). The other covariates (sex, affected status, age at visit, and the individual who completed the eCFHQ) did not add significantly to the model. For patients who went from not meeting to meeting NCCN GT criteria, prior GT was the most significant single covariate (yes vs no: OR, 0.28; 95% CI, 0.07–0.76; P=.03; unknown vs no: OR, 5.11; 95% CI, 2.14–11.3; P<.001), and no other covariates were added significantly in a step-up model (see supplemental eTable 2).
Blank Pedigrees
Of the 1,113 patients, 85 (8%) had blank pre-GC pedigrees. Whether the pedigree was left blank before the GC visit was used as an outcome variable in post hoc logistic regression models. Patient characteristics were used as covariates (sex, age, race, and cancer status). Affected status (OR, 4.01; 95% CI, 2.09–8.71; P=.0001) and belonging to a racially and/or ethnically underrepresented group (OR, 2.60; 95% CI, 1.51–4.35; P=.0004) were the only covariates selected in the final model using step-up model selection (see supplemental eTable 3).
Discussion
Completion of the eCFHQ was high, with only 8% of pedigrees left blank. Our findings differ from those of a recent national pilot study using an enhanced text chat (Chatbot) to gather CFH, in which only 20% of participants completed the intervention.8 This difference may result from sample characteristics of the studied populations. Patients in our study were specifically referred for cancer genetics evaluation based on their personal or family history of cancer, whereas the Chatbot study assessed a population-based cohort.
For most of the patients in this study, a standalone eCFHQ was sufficient to identify candidates for GT based on NCCN GT criteria. Using an eCFHQ alone was often sufficient to determine whether NCCN GT criteria were met for 91% (n=933) of nonblank pedigrees. However, 9% (n=95) of pedigrees were discrepant in meeting any NCCN GT criteria when compared with the same pedigrees augmented by CGCs. Of those individuals with pedigree discrepancies, 83 patients (7% of the overall cohort) became eligible for testing by any NCCN GT criteria after GC.
We found that prior GT status and self-reported race were characteristics associated with discrepancies among pre- and post-GC pedigrees. As expected, individuals who have not had GT or do not know whether they have had GT have a higher likelihood of having a change in their NCCN GT criteria. These patients may be less likely to have previously thoroughly reviewed their CFH compared with those who previously had GT. During a GC appointment, CGCs ask probing questions to elicit additional details, such as clarifying cancer diagnoses, age at diagnosis, and other family history details that the patient may not have provided using the eCFHQ. Such details, including the age of death of unaffected family members and limited family structure, can make the difference in whether a patient meets NCCN GT criteria.
These data suggest that the eCFHQ tool may not be widely applicable across all patient populations and GT indications. Certain patients may benefit from meeting with a CGC to gather further clinical information. Given the diversity within racially and/or ethnically underrepresented populations, it is difficult to determine why this tool was not as effective for them. Previous research found limited knowledge of CFH among historically underserved patients, ascertained through a federally qualified health center.18 This finding could be explained by a lack of access to health information for patients whose relatives live abroad, cultural differences in health communication, or familial estrangement.19,20 Asian patients were the largest group (n=50) among our racially and/or ethnically underrepresented population. This grouping itself combines extremely diverse and heterogeneous patients. Although little has been published about the GC needs of Asian Americans, lower levels of communication about risks and results have been noted.21 A patient’s CFH is increasingly important to contextualize GT results. The presence or absence of affected family members may contribute to the reclassification of variants of uncertain significance. This factor makes an accurate family history especially important for those who may be more likely to have variants of uncertain significance identified, which is more frequent in patients who belong to racially and/or ethnically underrepresented groups.22
We found that CGC modifications of family history were most frequent to third-degree relatives (44%). This finding is consistent with prior studies. In a study of CFH validity in family registries of patients with breast, ovarian, and colorectal cancer, reporting of CFH by first-degree relatives was found to be highly reliable. However, reliability sharply dropped off in second- and third-degree relatives.12 Similar results were observed in an unselected primary care–based study of prostate, lung, breast, and colorectal cancers.23
The significance of meeting NCCN GT criteria aside, CGC augmentation of CFH is an important consideration for cancer screening recommendations. For example, breast cancer risk estimation models and national guidelines for pancreatic cancer screening depend on knowledge of family history of breast and pancreatic cancers, respectively. When a germline pathogenic variant is identified, the accuracy of CFH is integral in the development of tailored screening recommendations. Moreover, a well-delineated family structure can facilitate efforts to identify relatives for cascade testing.
The findings in this study must be interpreted in the context in which they were conducted. Important limitations include that the study population was drawn from a single institution with predominantly White patients. In addition, although race was used a covariate in our logistic regression models, it is a social construct with limited utility in understanding medical research, practice, and policy. Because of the relative lack of diversity in this sample, we combined disparate racial and ethnic groups to allow for analysis. We did not collect data on household income, education, literacy skills, or other measures of social determinants of health that might have provided a more nuanced evaluation of the tool. The eCFHQ was only in English. It required computer and internet access to complete and an email address. Patient-reported CFH was not validated with family members’ medical records and is the subject of future research by this team. The eCFHQ version we used did not allow for sex diversity and some family structures (eg, gamete donation, nonheteronormative partnerships), which may have been additional barriers to patients using the tool.
Our eCFHQ was not designed to systematically collect benign features of cancer predisposition syndromes or to integrate somatic data, which are both used for genetics evaluation. It is important to note that our analysis was limited to the NCCN Guidelines for Genetics/Familial High-Risk Assessment16,17 and therefore did not include indications for GT embedded within the treatment guidelines for specific cancers. These guidelines are updated annually, and our analysis was at a specific time point. We limited our analysis to patients’ changes in meeting NCCN GT criteria, which evolve over time.
Future Research
Although cultural and linguistic adaptations of a family history tool for Spanish speakers have been studied and found to be acceptable,24 further study of electronic tools in diverse populations is needed. Further research may clarify why the eCFHQ was less effective at capturing CFH for patients who belong to racially and/or ethnically underrepresented groups, and will hopefully help improve its performance for all.
Conclusions
When clinicians use an eCFHQ to triage patients for GT, special attention is needed to identify patients for whom standalone eCFHQ may not be as effective and may perpetuate healthcare disparities. Ultimately, this research supports the utility of a standalone eCFHQ for identifying patients who meet NCCN GT criteria.
References
- 1.↑
Office of Disease Prevention and Health Promotion. Genomics. Accessed May 27, 2022. Available at: https://www.healthypeople.gov/2020/topics-objectives/topic/genomics
- 2.↑
Powell KP, Christianson CA, Hahn SE, et al. Collection of family health history for assessment of chronic disease risk in primary care. N C Med J 2013;74:279–286.
- 3.↑
Rich EC, Burke W, Heaton CJ, et al. Reconsidering the family history in primary care. J Gen Intern Med 2004;19:273–280.
- 4.↑
Welch BM, Wiley K, Pflieger L, et al. Review and comparison of electronic patient-facing family health history tools. J Genet Couns 2018;27:381–391.
- 5.↑
Bennett RL, French KS, Resta RG, et al. Standardized human pedigree nomenclature: update and assessment of the recommendations of the National Society of Genetic Counselors. J Genet Couns 2008;17:424–433.
- 6.↑
Schuette JL, Bennett RL. The ultimate genetic tool: the family history. In: Uhlmann WR, Schuette JL, Yashar BM, eds. A Guide to Genetic Counseling, 2nd ed. Hoboken, NJ: Wiley-Blackwell;2009:37–70.
- 7.↑
Murff HJ, Byrne D, Syngal S. Cancer risk assessment: quality and impact of the family history interview. Am J Prev Med 2004;27:239–245.
- 8.↑
McCain LA, Milliron KJ, Cook AM, et al. Implementation of INHERET, an online family history and cancer risk interpretation program for primary care and specialty clinics. J Natl Compr Canc Netw 2022;20:63–70.
- 9.↑
Welch BM, Allen CG, Ritchie JB, et al. Using a chatbot to assess hereditary cancer risk. JCO Clin Cancer Inform 2020;4:787–793.
- 10.↑
Seifert BA, O’Daniel JM, Amin K, et al. Germline analysis from tumor-germline sequencing dyads to identify clinically actionable secondary findings. Clin Cancer Res 2016;22:4087–4094.
- 11.↑
Schrader KA, Cheng DT, Joseph V, et al. Germline variants in targeted tumor sequencing using matched normal DNA. JAMA Oncol 2016;2:104–111.
- 12.↑
Ziogas A, Anton-Culver H. Validation of family history data in cancer family registries. Am J Prev Med 2003;24:190–198.
- 13.↑
Soegaard M, Jensen A, Frederiksen K, et al. Accuracy of self-reported family history of cancer in a large case-control study of ovarian cancer. Cancer Causes Control 2008;19:469–479.
- 14.↑
Chang ET, Smedby KE, Hjalgrim H, et al. Reliability of self-reported family history of cancer in a large case-control study of lymphoma. J Natl Cancer Inst 2006;98:61–68.
- 15.↑
Murff HJ, Spigel DR, Syngal S. Does this patient have a family history of cancer? An evidence-based analysis of the accuracy of family cancer history. JAMA 2004;292:1480–1489.
- 18.↑
Rana HQ, Cochrane SR, Hiller E, et al. A comparison of cancer risk assessment and testing outcomes in patients from underserved vs. tertiary care settings. J Community Genet 2018;9:233–241.
- 19.↑
Thompson T, Seo J, Griffith J, et al. The context of collecting family health history: examining definitions of family and family communication about health among African American women. J Health Commun 2015;20:416–423.
- 20.↑
Campbell-Salome G, Rauscher EA, Freytag J. Patterns of communicating about family health history: exploring differences in family types, age, and sex. Health Educ Behav 2019;46:809–817.
- 21.↑
Young JL, Mak J, Stanley T, et al. Genetic counseling and testing for Asian Americans: a systematic review. Genet Med 2021;23:1424–1437.
- 22.↑
Bylstra Y, Lim WK, Kam S, et al. Family history assessment significantly enhances delivery of precision medicine in the genomics era [erratum in Genome Med 2021;13:109]. Genome Med 2021;13:3.
- 23.↑
Mai PL, Garceau AO, Graubard BI, et al. Confirmation of family cancer history reported in a population-based survey. J Natl Cancer Inst 2011;103:788–797.
- 24.↑
Cerda Diez M, Cortés DE, Trevino-Talbot M, et al. Designing and evaluating a digital family health history tool for Spanish speakers. Int J Environ Res Public Health 2019;16:E4979.
