Association between family history and onset age of cancer in China.

Introduction
Although cancer is generally considered to be a disease predominantly affecting people ≥50 years of age, emerging evidence indicated that a concerning upward trend in the incidence rates of various malignancies—including breast, colorectal, esophageal, renal, hepatic, pancreatic, prostate, thyroid, and stomach cancers—among adults <50 years of age in many parts of the world.[1–9] Notably, early-onset cancers exhibit distinct clinical, pathological, and molecular characteristics compared to these later-onset cancers, with a higher proportion of advanced stage and poorer survival.[10–14] These alarming situations have promoted extensive research into primary prevention and cancer screening for early-onset cancers.
Many factors are associated with the early onset of cancer, and of these, family history (FH) is considered one of the most important factors.[13,15–17] In fact, FH not only serves as a potential marker of inherited genetic susceptibility but also reflects shared environmental risk factors within families.[18] Therefore, although FH is not directly modifiable, its identification enables targeted preventive interventions, including environmental risk factor modification and enhanced cancer screening. Moreover, individuals with a positive FH of cancer demonstrate higher compliance with both organized and opportunistic screening programs, leading to earlier detection and potentially contributing to the observed association between FH and younger age at onset to a certain extent.[13]
Despite most previous epidemiological studies have focused on familial risks of cancer, specific studies on the difference in onset age between familial and sporadic cancers are less common. Among these studies conducted in China, most of these have focused on single cancer types, such as esophageal or stomach cancer,[12,19–22] and reliable estimates among the main types of cancer in China were lacking. To address the above research gap, the primary aim of our study was to quantify the relationship between FH of concordant cancer and onset age among main types of cancer in China. Furthermore, we aimed to explore the magnitude of potential bias introduced by earlier detection among individuals with a positive FH, which may lower the observed age at onset. These results hold significant implications for clinical counseling and the design and implementation of cancer screening programs, especially in determining the appropriate starting age for screening.

Methods

Ethics approval
This study was approved by the Ethics Committee of National Cancer Center/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College (approval No. 18-016/1645). Written signed consent was obtained from patients.

Study design and participants
The study was based on a multicenter, hospital-based cancer registration program, which was launched by the National Cancer Center of China. This program has been conducted in 25 areas across 13 provinces, representing six geographical regions of China (north, northeast, northwest, east, central, and south) with different socioeconomic statuses and lifestyle habits in China. The study design has been described in detail previously.[23] In brief, we identified all patients diagnosed with first primary, invasive cancer between January 1, 2016 and December 31, 2017, using the electronic health records and cancer registration record. Cancer cases within any of the following categories were eligible: (1) diagnosed at these hospitals during the period and (2) diagnosed elsewhere but received all or part of the first course of therapy at these hospitals. After rigorous quality control, a total of 52,103 cases of breast, lung, esophageal, stomach, or colorectal cancer were included in this program.[23]
In the present study, we restricted our study population to patients diagnosed with lung, esophageal, stomach, or colorectal cancer and excluded 1633 participants without available information on FH. These four types of cancers collectively account for nearly 47.8% of the incident cancer cases and 56.0% of the cancer deaths in China in 2020.[24] The baseline characteristics of participants between the included and excluded populations were compared [Supplementary Table 1, http://links.lww.com/CM9/C443]. The excluded populations appear to be socioeconomically deprived populations. However, since the excluded population is only a small fraction of the total population (3.82%), we believe that the representativeness of the population will be not affected. The final sample consisted of 41,072 cancer cases from 23 hospitals across 12 provinces in six regions [Supplementary Figure 1, http://links.lww.com/CM9/C443]. In our study, data on breast and liver cancer were collected in the program but excluded in our analysis, as some crucial variables were not collected (hepatitis B virus/hepatitis C virus infection status for liver cancer) or missing for a large proportion of participants (age at menarche, oral contraceptive use, breastfeeding history, and age of menopause for breast cancer).

Data collection
Demographic and health-related data were obtained by trained physicians via face-to-face interviews using a standardized questionnaire. Demographic and health-related variables included sex, onset age, smoking status, FH of cancer (and if yes, the cancer type and relation to the participant), drinking status, health insurance status, marital status, etc. FH of cancer in our study was considered positive if at least one first-degree relative (parent, sibling, or offspring) had cancer at the same site as these patients. Cancer staging information was obtained from electronic medical records, pathological reports, imaging tests, and surgical reports by trained investigators at each hospital. We further classified the stage at diagnosis according to the Manual of The American Joint Committee on Cancer Staging (7th edition).[25] The early stage at diagnosis was defined as stages I–Ⅱ and late stage as stages III–IV.

Statistical analysis
Continuous variables were presented as medians (Q1–Q3), and categorical variables were reported as counts and percentages. Demographic and clinical characteristics were assessed and compared between participants with and without an FH of concordant cancer using the Wilcoxon test for continuous variables and the chi-squared test for categorical variables. To examine the hypothesis that having a positive FH of concordant cancer would promote the early detection of cancer, the distribution of cancer stage was compared by FH of concordant cancer with adjustment for potential confounders using multivariable logistic regression. The baseline characteristics of participants between participants with and without known stages was compared in Supplementary Table 2, http://links.lww.com/CM9/C443. In addition, quantile regressions with sequential modeling were fitted to estimate the associations between FH of concordant cancer and the onset age of cancer. The quantile regression is similar to multiple linear regression, that is, estimating the difference of median or other quantiles for the outcome variable. Since the age distribution was significantly highly skewed in our study, multiple linear regression is not applicable and the quantile regression can be fitted without assuming normality or homoscedasticity of the underlying distribution.[26–28] The statistical methodology has been described in full detail elsewhere.[29,30] The first model was adjusted for sex, area, and cancer type (for overall cancer only), and yielded the median difference 1 (MD1). Model 2 further included common cancer risk factors (smoking history, drinking history, health insurance status, and marital status), and yielded the median difference 2 (MD2). In addition to model 2, model 3 was further adjusted for cancer stage to assess potential overestimation of the association due to earlier detection among individuals with a positive FH, yielding median difference 3 (MD3). The changes in MDs were defined as 100×(MD2–MD3)/MD2. Moreover, the effects of FH on onset age at the 20th, 40th, 60th, and 80th quantiles were also accessed. We analyzed all cancers combined and separately for each cancer type. In sensitivity analysis, we redefined patients with a positive FH of cancer as having at least one first-degree relative with upper gastrointestinal cancer in the analyses of esophageal and stomach cancer. Moreover, we additionally adjusted for body mass index (BMI) as a confounder. We explored the impact of FH across various subgroups defined by sex, geographic area, smoking history, drinking history, or cancer stage. Furthermore, we investigated the relationship between the number of first-degree relatives and cancer stage as well as age at cancer onset. Additional details of our sensitivity analysis results are reported in Supplementary Tables 3–8, http://links.lww.com/CM9/C443. All statistical analyses were performed through the R software (version 4.2.2, R Foundation for Statistical Computing, Vienna, Austria). All tests were two-sided, and statistical significance was considered as P <0.05.

Results
Of the 41,072 participants under study, 3054 (7.44%) reported an FH of cancer in at least one first-degree relative. The demographic and clinical characteristics of study participants with and without an FH of concordant cancer are shown in Table 1. Of the eligible participants, 27,735 (67.5%) participants were male and 27,807 (67.7%) participants were living in urban areas. The median age of cancer onset was 63.06 (Q1–Q3: 56.21–69.30) years. Individuals with an FH of concordant cancer were more likely to be previous or current smokers and drinkers, younger at cancer onset, diagnosed at an early stage, and covered by urban insurance (all P <0.001) [Table 1].
Table 2 shows the estimated association between FH of concordant cancer and being late-stage (stages III–IV) at diagnosis. Of 31,693 participants with known cancer stage, 12,360 (39.0%) participants were diagnosed at stages I–II, while in colorectal and esophageal cancers, this proportion exceeded 45%. About 65% of individuals with stomach and lung cancer were diagnosed at a late stage (64.8% and 67.0%, respectively). The proportion of stages I–II was higher in individuals with an FH of concordant cancer than individuals without an FH across all cancer types combined and for lung, stomach, esophageal, and colorectal cancers individually. For most of the cancer types, these associations remained significant after adjusting for potential confounding. Multivariable-adjusted odds ratios (ORs) were 0.67 (95% confidence interval [CI]: 0.61 to 0.73) for all cancer combined, 0.62 (95% CI: 0.54 to 0.71) for lung, 0.81 (95% CI: 0.67 to 0.98) for stomach, and 0.62 (95% CI: 0.51 to 0.76) for esophagus cancer, respectively [Table 2]. However, no statistically significant association was observed for colorectal cancer (0.89, 95% CI: 0.72 to 1.11). Supplementary Table 2, http://links.lww.com/CM9/C443, showed the participants without known cancer stage (excluded in the analyses of Table 2) tended to be male, unmarried, divorced, or widowed, older at cancer onset, smokers and alcohol drinkers, rural residents, and covered by rural insurance.
The relationships between the FH of concordant cancer and the age of onset are summarized in Table 3. Among 41,072 participants, the median onset ages of lung, stomach, esophagus, and colorectum cancer were 62.47, 63.21, 64.80, and 62.74 years, respectively. The age of onset significantly differed between participants with and without an FH of concordant cancer, with earlier diagnoses observed in those with an FH (MD2: –1.21, 95% CI: –1.60 to –0.83). By specific cancer type, significantly younger ages at onset were observed for lung, stomach, and colorectal cancers among individuals with a positive FH compared to those without. However, no significant differences were found for esophagus cancer (MD2: −0.63, 95% CI: −1.56 to 0.30). Adjustment for cancer stage attenuated the magnitude of the median differences, with reductions ranging from 3.4% (stomach cancer) to 15.4% (colorectal cancer) relative to unadjusted estimates (MD2). The most pronounced difference in age at onset was observed for colorectal cancer (MD3: –2.25, 95% CI: –3.31 to –1.19), followed by lung cancer (MD3: –1.24, 95% CI: –1.84 to –0.64) and stomach cancer (MD3: –1.13, 95% CI: –2.03 to –0.23).
Table 4 presents the effect of FH of concordant cancer on cancer onset age across onset age quantile. For all cancers combined, no significant reduction in onset age was observed at the 20th percentile for the group with a positive FH. However, the magnitude of the reduction increased progressively across higher quantiles. Compared to the 20th percentile, the reduction in onset age was larger at its 80th percentile (−0.12 and −1.74, respectively). Similar trends were observed for individual cancer types, with the exception of colorectal cancer, where the largest effect was found at the 60th percentile (−2.69, 95% CI: −4.03 to −1.36).

Discussion
Although numerous studies have characterized the high risk of developing cancer in individuals with a positive FH, evidence regarding the difference in age at onset between familial and sporadic cancers remains limited. Based on a large multicenter, hospital-based observational study across six regions with different socioeconomic statuses, we examined the association between FH of concordant cancer and onset age across main cancer types in China. Our findings demonstrated that individuals with a positive FH of concordant cancer were diagnosed at earlier ages than their counterparts, with the most pronounced difference observed in colorectal cancer. In addition, we confirmed that patients with a positive FH of concordant cancer tended to be diagnosed at an earlier stage than those without an FH. Importantly, our results highlight that the relationship between FH and age of onset might be overestimated if the effect of FH on promoting cancer screening and early detection was ignored. These results are informative for the etiological understanding and design of cancer screening programs in China.
Owing to the early-onset cancer epidemic observed in recent years, the US National Cancer Institute and Cancer Research UK have listed this phenomenon as a research priority in one of its “Cancer Grand Challenges” for 2023.[31] While increased exposure to shared environmental and lifestyle risk factors during early life and young adulthood may contribute substantially to this trend, the role of FH in early-onset cancer development warrants careful consideration. Our data indicated that patients with parents, siblings, or offspring affected by cancer at the same site as these patients were diagnosed at earlier ages than patients without any affected first-degree relatives, with the most pronounced onset age difference observed in colorectal cancer (about 2.3 years). The effect of FH on cancer onset age remained stable across all subgroups examined [Supplementary Table 6, http://links.lww.com/CM9/C443]. Previous evidence have also confirmed that early onset of breast, colorectal, endometrial, head and neck, liver, prostate, stomach, and esophageal cancers were associated with FH of cancer.[12,13,32–35] Gausman et al[36] estimated that early-onset colorectal cancer patients (<50 years of age) were more than eight times as likely to have an FH of colorectal cancer compared to cancer-free controls and nearly three times more likely than late-onset colorectal cancer patients (≥50 years of age). Several studies conducted in China, focusing on esophageal and stomach cancers, found that familial cancer cases exhibited a significantly younger age of onset compared to sporadic cases, with estimates ranging from –1.66 to –1.10 years across studies.[12,19,21,22,37] A meta-analysis published in 2020 found that the familial risk of lung cancer increased for younger individuals, which might indicate a genetic component that would favor a tumor suppressor model with cellularly recessive effects commonly found in familial cancers.[38] Moreover, Lichtenstein et al[39] found that heritable factors accounted for a higher proportion of cancer susceptibility in colorectal cancer than in stomach and lung cancer, a finding that corroborates our observations.
Our study found that individuals with an FH of concordant cancer were associated with a decreased likelihood of a late-stage diagnosis. Other previous research also observed similar findings and supported our conclusions.[40,41] This could be due to increased awareness and high uptake of health screening in these participants. Studies examining the association between FH of cancer and cancer screening compliance have supported this speculation.[42,43] Moreover, this association might contribute to the difference in onset age between individuals with and without an FH, as this difference might be partly explained by the phenomenon that individuals with an FH were more likely to participate in cancer screening and be diagnosed with earlier stage and at a younger age. This bias results from the lead time of illness produced by the diagnosis of a condition during its latency period and is commonly called lead time bias in the evaluation of the efficacy of screening. Although this bias has been noticed in other studies, it is frequently ignored in the assessment of the relationship between onset age and FH of cancer, which may lead to overestimating the effect of FH on onset age. Several statistical approaches have been developed to correct lead time.[44] However, as the information on cancer screening uptake was not collected in our study, we could only roughly adjust for this bias by incorporating the cancer stage as an adjusted confounder. Our results showed that the differences in onset age between individuals with and without an FH were attenuated after adjusting for cancer stage (3.4% to 15.4% across cancer types), implying that the advanced time due to early detection in screen-detected patients should not be ignored. The magnitude of this bias was modest in the present study, as the uptake of cancer screening was low in our study area.[23] However, as the cancer screening participation rate gradually increases in the future, the impact of this bias may become non-negligible. Additional large studies that collect detailed information on cancer screening participation are therefore needed to accurately quantify this bias.
In addition, our study examined the associations between FH of concordant cancer and onset age across quantiles of onset age. Some studies reported stronger association in patients with relatives who developed cancer at earlier ages,[45–48] but few studies have examined the heterogeneity of effect of FH on onset age across patients’ onset age levels. A population-based case-control study included 2386 breast cancer cases and 2502 control subjects and found that a positive FH of breast cancer had a greater impact on breast cancer risk in women aged 40–49 years compared to women aged 25–39 years.[49] However, the age interaction did not reach statistically significance, which might be due to the limited sample size. Similarly, a population-based cohort study of 1149 men calculated the age-specific relative risk of prostate cancer associated with FH and found that the effect of FH increased with the age of participant.[50] Our study reported the evidence of heterogeneity of FH effect on onset age among lung, stomach, esophagus, and colorectal cancer in China. While the mechanism underlying this phenomenon was quite vague, some hypotheses might explain the phenomenon. Multiple environmental factors and genetic risk were found to be synergistic in many studies. Thus, the phenomenon that the effect of FH on cancer onset age was more pronounced in the older population may be a reflection of cumulative environmental exposure.[51] Further study is needed to clarify the reasons for the heterogeneity of the effect of FH across age levels.
Our study has some limitations. First, this is a hospital-based observational study on FH of concordant cancer and onset age, and some inherent biases, such as residual bias and admission rate bias, are unavoidable. Specifically, the FH data in this study are self-reported, which may introduce reporting biases, such as recall bias or underreporting. Reliable inference of causality cannot be made in our study. Second, although our study included cancer cases from multiple hospitals in six of seven regions of China with different socioeconomic status and lifestyle habits, the study participants are not perfectly representative of the entire Chinese population. Third, this study only allows for some exploration of bias due to lead time, and we cannot completely rule out overestimation due to this bias as the detailed cancer screening-related variables are lacking in our study. Fourth, our study fails to reach stable estimators in the association between number of affected first-degree relatives and cancer stage as well as age at cancer onset, as the number of first-degree relatives with positive cancer history rarely exceeded one in our participants [Supplementary Tables 7 and 8, http://links.lww.com/CM9/C443]. Fifth, although this study focused on four specific cancer types, the exclusion of other cancers, such as ovarian and breast cancer, which are also strongly associated with FH, is a limitation. Future studies should consider a broader range of cancers to provide a more comprehensive understanding of the effect of FH on cancer onset age. In summary, our findings provide valuable insights into the role of FH in cancer onset, which could inform the development of more tailored screening guidelines. Individuals with an FH of specific cancers may benefit from earlier or more frequent screenings. These results highlight the importance of integrating familial risk factors into personalized medical strategies, potentially improving early detection and patient management. Furthermore, the study underscores the need for future research to explore genetic and environmental interactions that contribute to familial cancer risk, advancing the field of precision medicine and guiding the design of targeted preventive and therapeutic interventions.

Funding
This work was supported by grants from the National Natural Science Foundation of China (No. 82273721), the Cooperation Project in Beijing, Tianjin, and Hebei of China (No. J200017), the Sanming Project of Medicine in Shenzhen (No. SZSM201911015), and National Key Research & Development Program of China (No. 2016YFC1302502).

Conflicts of interest
None.

Supplementary Material