Anticipation in multiple-case lymphoid cancer families after controlling for ascertainment biases

Anticipation is an inheritance pattern where disease severity increases and/or age of onset decreases in successive generations. It has been documented in neurodegenerative disorders [1,2], autoimmune diseases [2], and several cancers, including familial leukemias and lymphomas [2–10]. In neurodegenerative disorders such as Huntington's disease, genetic anticipation is due to the expansion of trinucleotide repeat sequences [1]; in cancer, the molecular mechanism remains unknown. Although there is evidence for anticipation in lymphoid cancer families, this phenomenon is difficult to interpret due to the potential effects of biases.

Ascertainment bias is a systematic difference between participants in a study compared to the general population. We examined age of onset patterns in 200 intergenerational, multiple-case lymphoid cancer families while adjusting for several types of ascertainment bias and controlling for known factors that affect lymphoid cancer risk.

This study was approved by the University of British Columbia/BC Cancer Research Ethics Board. All participants provided informed consent.

Eligible families had two or more members with lymphoid cancer. Families were recruited by physician referral or other means between 2006 and 2018. Living participants provided cancer diagnosis information for themselves or close relatives. Reported diagnoses were validated by review of original medical, clinical and/or laboratory records, by referring physician report, or through examination of histopathology slides by an expert oncology pathologist. Lymphoid cancers were classified according to the InterLymph hierarchical classification of lymphoid neoplasms for epidemiologic research [11]. Participation was not geographically limited, though most families were identified through a member residing in British Columbia (BC), Canada.

A sign test was used to assess if the observed ages of onset were less than the population median age of onset. An F-statistic was used to assess the association between age of onset and generation for each lymphoid cancer subtype. For each cancer type, the age of onset was randomly permuted 10,000 times within each family, to generate a reference distribution of F-statistics on which to calculate a p-value.

Possible ascertainment biases were corrected using 3 methods: 1) removing cases diagnosed before reproductive age (≤25 years old), as these individuals may have limited reproductive capabilities, resulting in oversampling of parent generations; 2) removing probands, as they are more likely to be self-selected because of an earlier age at diagnosis; and 3) removing cases with short duration of follow-up, for which insufficient time has elapsed for normal or late disease development among family members in similar birth cohorts. A short duration was classified as a current age below the population median age of onset. Affected members of families that no longer had 2 or more cases after controlling for ascertainment biases were also removed. Subtypes with 10 or fewer cases were not examined, nor were generations with 2 or fewer cases.

BC Cancer Registry data [12] were used to control for age incidence patterns among families with heterogenous lymphoid cancer subtypes. Registry data was used to create a cumulative incidence distribution for each combination of sex, subtype and age of onset (1-year interval) observed. The cumulative incidence was converted to a percentile. All statistical tests were repeated using percentiles in replacement of age of onset.

We included 527 cases in 200 intergenerational, multiple-case lymphoid cancer families (Table S1). Cases were predominantly male (53.3%), with the exception of follicular lymphoma (FL), marginal zone lymphoma (MZL) and nodular sclerosis (NS) subtypes. The median age of onset was 57 years. Medical records and/or histopathology slides were available for 252 (47.8%) cases, all of which supported the self-reported diagnosis.

The sign test (Table 1) indicates that the familial ages of onset were significantly earlier than the population median age of onset for all lymphoid cancers together, and for non-Hodgkin lymphoma (NHL), Hodgkin lymphoma (HL) and several subtypes including diffuse large B-cell lymphoma (DLBCL), FL, MZL, chronic lymphocytic leukemia (CLL), T-cell NHL, and classic HL (CHL). After controlling for age of onset, sex and subtype, the familial percentile of onset were earlier than the population median for NHL and most NHL subtypes. Familial lymphoplasmacytic lymphoma (LPL)/Waldenstrom macroglöbulinemia (WM), mantle cell lymphoma (MCL) and multiple myeloma (MM) cases were also diagnosed earlier than population cases (age and percentile), however, only MCL and MM percentiles were statistically significant.

Table 1. Age of onset (AoO) and age of onset percentiles in multiple-case lymphoid cancer families after controlling for ascertainment bias.

								Controlling for ascertainment bias
	Population		All cases					1. Reproductive (dx < 25 years) cases removed					2. Probands removed					3. Short duration of follow-up removed
Subtype	AoO	per	n	Median AoO	p	Median per	p	n	Median AoO	p	Median per	p	n	Median AoO	p	Median per	p	n	Median AoO	p	Median per	p
NHL^a	66	50%	254	57	**	29%	**	234	57	**	29%	**	106	60	**	34%	**	110	65		44%
B-cell NHL^b	66	50%	304	58	**	28%	**	287	59	**	28%	**	107	60	*	29%	**	129	67		43%	*
DLBCL	67	50%	34	53.5	**	27%	**	27	56	**	27%	**	8	57.5	***	27%	***	9	60		29%
FL	62	50%	57	56	**	32%	**	54	55.5	**	32%	**	18	65		59%		19	60		49%
LPL/WM	67	50%	18	60		20%		18	60		20%	*	9	61		21%		7	67		38%
MCL	69	50%	9	65		36%	*	8	64.5		36%	***	4	54		16%		5	68		49%
MZL	66	50%	12	53	**	19%	**	11	53	**	19%	**	5	52	***	17%	***	7	53	**	19%	**
CLL	71	50%	150	62	**	29%	**	149	62	**	29%	**	57	64	**	31%	**	72	68.5	*	48%	**
T-cell NHL	65	50%	9	45	*	22%	*	9	45	*	22%	*	5	45		22%		2	48.5		26%
HL	39	50%	79	31	**	49%		43	38		65%	**	43	31	*	44%		46	36		56%
CHL^c	40	50%	30	32.5	**	55%		18	35	*	65%	**	10	35	*	56%		14	35	***	65%
NS	29	50%	27	31		56%		17	35	**	65%	**	9	35		57%		12	35.5	***	65%	*
Myeloma	69	50%	34	64		28%	*	33	64		28%	*	17	64		28%		22	68		44%
All cases		50%	517	57	**	30%	**	459	59	**	32%	**	223	58	**	34%	**	250	64		48%	**

Significance: **p < 0.01; *p < 0.05; ***p < 0.10.

Abbreviations: AoO: age of onset; per: percentile; n, sample size; p: p-value.

Subtype abbreviations: NHL: non-Hodgkin lymphoma; DLBCL: diffuse large B-cell lymphoma; FL: follicular lymphoma; LPL/WM: lymphoplasmacytic lymphoma/Waldenstrom macroglobulinemia; MCL: mantle cell lymphoma; MZL: marginal zone lymphoma; CLL: chronic lymphocytic leukemia; HL: Hodgkin lymphoma; CHL: classic Hodgkin lymphoma; NS: nodular sclerosis; MM: multiple myeloma.

Notes: ^aNHL excluding CLL cases. ^bB-cell NHL including CLL cases. ^cCHL includes nodular sclerosis, mixed cellularity, lymphocyte depleted, lymphocyte rich, but not CHL not otherwise specified (NOS) cases.

We performed three tests to control for possible types of ascertainment bias, on the age of lymphoid cancer onset. First, cases diagnosed before the age of 25 years were excluded. Familial NHL cases (and B-cell NHL, DLBCL, FL, LPL/WM, MZL, CLL and T-cell subtypes) were diagnosed earlier than sporadic lymphoma cases (age and percentile). Second, probands were removed. Familial lymphoid cancers were considered as a group and NHL, B-cell and CLL were diagnosed significantly earlier than comparable population data (age and percentile); while familial DLBCL and MZL (age and percentile) cases were earlier and marginally significant. Familial HL and CHL age of onset was significantly earlier than population cases (age but not percentile). The third correction for ascertainment bias was attempted by excluding cases whose current age was below the population median age of onset. Most familial lymphoid entities were diagnosed earlier than population cases; however, the relationship was significant for all lymphoid cancers together (percentile), B-cell (percentile), MZL (age and percentile) and CLL (age and percentile).

The median age of onset was significantly different between generations, and the ages of onset were younger for later generations for all familial lymphoid cancers together (p < 0.0001), and for NHL, B-cell NHL, DLBCL, LPL/WM, CLL, and HL subtypes (Table 2). After controlling for age of onset, sex, and subtype using percentiles, anticipation is still observed. MCL, MZL and T-cell NHL lacked a sufficient sample size for generational analyses (data not shown).

Table 2. Anticipation effects for familial lymphoid cancers and subtypes after controlling for ascertainment biases.

				Controlling for ascertainment bias
	All cases			1. Reproductive (dx < 25 yrs) cases removed			2. Probands removed			3. Short-duration of follow-up
Subtype	n	Median AoO	Median percentile	n	Median AoO	Median percentile	n	Median AoO	Median percentile	n	Median AoO	Median percentile
NHL
Gen 1	49	67	51.0	45	67	51,2	16	68.5	50.1	19	74	64.8
Gen 2	132	56	27.5	120	57	29.4	51	61	45.4	65	65.5	48.9
Gen 3	70	52	21.0	61	52	22.3	37	49	17.0	29	60	34.1
Gen 4	10	44.5	12.0	8	45.5	12.0	8	42.5	9.1	4	46.5	19.3
p-value		<0.0001	<0.0001		<0.0001	<0.0001		<0.0001	<0.0001		0.0022	0.0068
B-cell
Gen 1	61	68	48.7	57	68	51.0	21	70	48.7	23	70	52.6
Gen 2	163	58	27.1	153	58	27.5	51	60	37.0	74	67	43.4
Gen 3	66	55	22.0	60	55.5	22.5	31	58	17.8	31	60	27.1
Gen 4	18	52	11.1	17	52	10.2	8	44.5	6.0	5	52	21.3
p-value		<0.0001	<0.0001		<0.0001	<0.0001		0.0029	0.0021		0.0048	0.0083
DLBCL
Gen 1	5	64	43.8	4	71	58.0	–	–	–	–	–	–
Gen 2	22	53.5	23.4	18	55.5	26.5	5	60	27.5	6	58.5	28.2
Gen 3	7	48	11.5	5	48	18.5	3	27	1.7	2	–	–
p-value		0.0006	0.0004		0.0031	0.0021
FL
Gen 1	12	68.5	64.5	10	68.5	64.5	4	77	88.7	3	79	92.8
Gen 2	29	56	31.2	28	56	33.0	9	58	44.2	11	58	44.2
Gen 3	13	51	22.7	13	51	22.7	3	46	18.0	3	49	22.7
Gen 4	3	46	12.1	3	46	12.1	2	–	–	2	–	–
p-value		0.12	0.25		0.50	0.25		0.51	1.0		1.0	1.0
LPL/WM
Gen 1	7	67	38.6	7	67	38.6	3	61	21.5	1	–	–
Gen 2	7	57	13.7	7	57	13.7	5	60	19.5	5	60	19.5
Gen 3	6	48.5	4.2	4	48.5	4.2	3	68	40.2	3	68	40.2
p-value		0.035	0.033		0.033	0.034		1.0	1.0
CLL
Gen 1	28	70.5	51.5	28	70.5	51.5	10	71	54.7	14	70.5	53.1
Gen 2	81	62	28.6	78	62	29.3	24	65	34.1	41	68	42.7
Gen 3	21	58	17.8	21	58	17.8	12	56.5	16.0	10	68	45.7
Gen 4	10	52	9.2	10	52	9.2	4	42	3.2	–	–	–
p-value		<0.0001	<0.0001		<0.0001	<0.0001		0.059	0.045		0.17	0.20
HL
Gen 1	11	50	68.5	8	50.5	69.1	3	29	43.0	7	50	68.5
Gen 2	31	36	56.3	22	39	60.6	13	40	56.6	19	40	64.6
Gen 3	24	25	24.3	8	36	53.2	18	25	26.5	18	25	24.3
Gen 4	6	20	17.6	–	–	–	4	20.5	19.0	–	–	–
p-value		0.0001	0.0006		0.36	0.39		0.049	0.039		0.082	0.049
NS
Gen 1	3	60	89.7	2	–	–	–	–	–	2	–	–
Gen 2	14	34.5	59.8	10	35.5	65.7	3	35	64.6	8	35.5	65.7
Gen 3	4	22	25.1	1	–	–	3	27	39.1	1	–	–
p-value		0.081	0.085
MM
Gen 1	8	76	62.7	7	76	62.7	4	72	53.2	5	72	53.2
Gen 2	16	62.5	26.3	16	62.5	26.3	6	62.5	26.3	11	68	43.6
Gen 3	10	63.5	27.6	9	63	26.8	7	62	26.8	6	71	51.4
p-value		0.25	0.25		0.25	0.25		0.33	0.51		0.49	0.38
All types
Gen 1	99	67.5	51.5	91	68	51.5	36	69	48.7	48	70	54.8
Gen 2	264	58	31.1	240	58	31.9	97	61	48.2	138	65	49.4
Gen 3	131	53	22.3	104	55	23.9	79	50	21.0	66	58.5	30.2
Gen 4	31	45	11.9	22	51	11.9	17	35	9.4	7	52	28.3
p-value		<0.0001	<0.0001		<0.0001	<0.0001		<0.0001	<0.0001		0.0004	0.0020

Abbreviations: AoO: age of onset; dx: diagnosed; yrs: years; per: percentile; n: sample size; p: p-value.

Subtype abbreviations: NHL: non-Hodgkin lymphoma; DLBCL: diffuse large B-cell lymphoma; FL: follicular lymphoma; LPL/WM: lymphoplasmacytic lymphoma/Waldenstrom macroglobulinemia; MCL: mantle cell lymphoma; MZL: marginal zone lymphoma; CLL: chronic lymphocytic leukemia; HL: Hodgkin lymphoma; NS: nodular sclerosis; MM: multiple myeloma.

After controlling for reproductive ascertainment bias, all lymphoid cancers considered as a group and NHL, B-cell, DLBCL, LPL/WM and CLL cases were diagnosed earlier in later generations (Table 2), and the F-test shows the ages of onset were significantly different between generation (ages and percentiles). After controlling for ascertainment bias by excluding the probands, the intergenerational age of onset was significantly different between generations for NHL, B-cell NHL, CLL, HL and all familial lymphoid cases, with the exception of familial CLL percentiles which showed a marginal trend toward significance (p = 0.059). The observed ages of onset were younger across generations for these types of familial lymphoid cancers, as well as DLBCL and NS subtypes, which had an insufficient sample size for analysis. Anticipation effects were observed for NHL, B-cell, FL, HL and all lymphoid cancers grouped together after removing cases with a short duration of follow-up. The ages of onset were significantly different between generations for these subtypes with the exception of familial FL cases. Age of onset by generation was limited by sample size for MCL, MZL and T-cell subtypes, as well as DLBCL, LPL/WM, and NS after controlling for some potential ascertainment biases.

Most types of familial lymphoid malignancies occurred, on average, at a substantially younger age than sporadic cases, even after controlling for age incidence patterns, sex, and subtype. The younger onset of familial cases is consistent with other studies [4,5,8–10]. We also observed anticipation across 3 or more generations of lymphoid cancer families for NHL, DLBCL, LPL/WM, CLL and HL cases. Anticipation has previously been described for NHL, HL and CLL [5,6,10,13,14].

Apparent anticipation may be caused by ascertainment biases, and three approaches to mitigate this was used. Our observations seem unlikely to be caused by ascertainment or other bias, which suggests anticipation exists in familial lymphoid cancers. A possible non-genetic explanation of anticipation is the simultaneous exposure of family members to a causative environmental agent; however, this seems unlikely as it would be limited to families that live together and fails to address a larger percentage of families with multigenerational cases, or cases where the parents (or later) generations were diagnosed decades after their children. Our observation of anticipation in NHL and HL cases implies that changes in surveillance for CLL are not the sole explanation.

There are several possible explanations for observing anticipation in lymphoid cancer families. The accumulation of germline variants, mutations initiated by a defective DNA repair gene, and inherited telomere abnormalities or shortening [1,8] could contribute to earlier disease onset in later generations. The expansion of unstable repetitive sequences has been described in both familial and sporadic CLL, leukemia and other cancers [8]; however, results are inconsistent [8,15].

Limitations of this study include other types of ascertainment bias and recall bias. Our data contains detailed age of onset and histological subtype information, cross-checked through interviews with multiple family members and verified by medical records and/or an expert oncology pathologist, when possible. There was no inconsistency in lymphoid cancer diagnosis among 252 cases that had medical records and/or pathology slides for examination. The numbers of familial MCL, MZL and T-cell cases were limited, providing an inadequate sample size for investigation.

The substantially earlier age of onset may reflect underlying susceptibility factors, some of which may predispose to more than one lymphoid cancer type. Earlier age of lymphoid cancer onset in this collection of multiple-case families supports the application of genomic methods to identify genes and genetic variants that underlie familial lymphoid cancers. Awareness of familial lymphoid cancer patterns and the identification of susceptibility genes has the potential to inform future screening methods for affected families.

Acknowledgments

We thank the families for their participation. This study was approved by the BC Cancer – University of British Columbia Clinical Research Ethics Board.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work is supported by the Canadian Institutes for Health Research under Grant [MOP-130311].