Cancer Recurrence Following Immune-suppressive Therapies in Patients With Immune-mediated Diseases: a Systematic Review and Meta-analysis

Abstract

Background & Aims

Physicians frequently encounter patients with immune-mediated diseases and a history of malignancy. There are limited data on the safety of immunosuppressive therapy for these patients. Published studies have been small with few events, precluding robust estimates of risk.

Methods

We searched Medline, EMBASE, and conference proceedings for terms related to immune mediated disease, immune-suppressive therapy, and cancer recurrence from inception to April 2015. We included 16 studies (9 of patients with rheumatoid arthritis, 8 of patients with inflammatory bowel disease, and of patients with psoriasis) and stratified studies by type of immune-suppressive therapy (monoclonal antibodies to tumor necrosis factor [anti-TNF], conventional immune-modulatory agents, or no immune suppression). A random effects meta-analysis was performed to calculate pooled incidence rates as well as risk differences between the various treatments.

Results

Our analysis included 11,702 persons contributing 31,258 person-years (p-y) of follow up after a prior diagnosis of cancer. Rates of cancer recurrence were similar among individuals receiving anti-TNF therapy (33.8/1000 p-y), immune-modulator therapy (36.2/1000 p-y), or no immunosuppression (37.5/1000 p-y), but were numerically higher among patients receiving combination immune suppression (54.5/1000 p-y) (P>.1 for all). Subgroup analysis of new and recurrent cancers separately, type of immune-modulator therapy, or immune-mediated disease revealed similar results, with no increase in risk. We found similar pooled incidence values for new or primary cancers when immunosuppression was initiated within 6 years (33.6/1000 p-y for immune-modulatory agents and 43.7/1000 p-y for anti-TNF agents) vs more 6 years after the index cancer (32.9/1000 p-y for immune-modulatory agents; P=.86 and 21.0/1000 p-y for anti-TNF agents; P=.43)

Conclusion

In a meta-analysis of 16 studies, we observed similar rates of cancer recurrence among individuals with prior cancer who received no immunosuppression, anti-TNF therapy, immune-modulator therapy, or combination treatments. Prospective studies are needed to ascertain optimal intervals for re-initiation of immune suppressive therapies for individuals with specific cancers.

INTRODUCTION

Immunosuppression is the cornerstone of management of chronic inflammatory diseases including inflammatory bowel disease (IBD; Crohn’s disease (CD), ulcerative colitis (UC)), rheumatoid arthritis (RA), and psoriasis (Ps) which together affect more than 10 million individuals in the United States and several million more worldwide^1–3. These diseases share common underlying genetic and immunologic mechanisms⁴ and are progressive and disabling, exerting a considerable toll of physical, psychological, and socioeconomic impairment^1–3. Consequently, treatment paradigms have evolved, moving from amelioration of symptoms to early aggressive immunosuppressive treatment to achieve remission and prevent irreversible complications^{5, 6}. This has resulted in wider use of conventional immunosuppression (IMM) with thiopurines (azathioprine, 6-mercaptopurine) and methotrexate and targeted biologic therapy predominantly with monoclonal antibodies to tumor necrosis factor α (anti-TNF), often given in combination with conventional agents^7–13.

The improved efficacy of treatment regimens with conventional immunosupressants and anti-TNF agents must be balanced against an associated increased risk of cancer due to impaired immune surveillance¹⁴, facilitation of oncogenic viruses^15–17, and perhaps alteration in DNA¹⁸. For example, thiopurines are associated with an increased risk of lymphoma (primarily non-Hodgkin’s lymphoma) while this association for anti-TNF therapies is less clear^19–21. In addition, anti-TNF therapy, thiopurines, and methotrexate are all associated with an increased risk of skin cancers (melanoma and non-melanoma)^22–24. In contrast, other than cervical cancer,²⁵ the risk of solid organ cancer development appears unaltered. However, despite the modest increase in risk of malignancy, the absolute risk of incident cancers in individuals with no prior cancer remains very low, providing reassurance regarding the wide spread use of such treatment strategies in the population^{26, 27}.

Given improvements in cancer survival over the past few decades, physicians are increasingly treating patients with chronic inflammatory diseases and a prior cancer diagnosis. Many treatment guidelines and expert opinions recommend restricted use of immunosuppression for at least 5 years following a diagnosis of cancer^28–33. However, there are limited clinical data to support this recommendation or to establish a safe interval for restarting therapy. A few observational cohorts and registry-based studies suggested no increase in risk of cancer recurrence; however, the event rates were low, limiting statistical power to identify a true effect^34–38. In addition, few prior studies have been able to compare the rate of recurrence of prior cancer or development of a new primary cancer between the different types of immunosuppressive therapy^{34–36, 39–43}. Given the growing availability of different therapeutic targets, comparative safety studies are critical to appropriately position these therapies.

We performed this systematic review and meta-analysis to (1) identify the rates of recurrence of prior cancers and development of new primary cancers in individuals with chronic immunologic diseases treated with immunosuppressive therapy after a prior malignancy; and (2) perform pair-wise comparisons of such rates between the treatment groups in those on no immunosuppression, conventional immunosuppression, and anti-TNF biologic therapies.

METHODS

Literature Search

We conducted separate MEDLINE (inception to April 2015) and EMBASE (inception to April 2015) searches of all relevant English language articles and manually searched reference lists from potentially relevant studies. Additionally, abstracts of scientific meetings from American Gastroenterological Association, the American College of Gastroenterology, the United European Gastroenterology Week, the European Crohn’s and Colitis Organization, the Inflammatory skin disease summit, the American College of Rheumatology, the European League Against Rheumatism and the British Society for Rheumatology were searched until April 2015. Studies were identified by combining four search themes using the Boolean operator “AND”. The first theme, inflammatory mediated diseases, combined the terms Crohn’s disease, ulcerative colitis, inflammatory bowel disease, rheumatoid arthritis, psoriasis, systemic lupus erythematosus, or ankylosing spondylitis. The second theme, immunosuppressive therapies, combined the terms azathioprine, 6-mercaptopurine, thiopurine, methotrexate, immunomodulator, anti-tnf agent, tumor necrosis factor inhibitor, infliximab, adalimumab, golimumab, certolizumab, or etanercept. The third theme, recurrence, combined the terms recurrent, recurrence, prior, preexisting, previous, or survivor. The fourth theme, cancer, combined the terms cancer, malignancy or neoplasm. We did not include studies of cancer recurrence following organ transplantation as the immunosuppressive regimens used in that clinical setting are distinct from those used for inflammatory diseases.

Inclusion and Exclusion criteria

We included all cohort or case-control studies assessing the impact of thiopurines, methotrexate, or anti-TNF agents on the risk of recurrent or new primary cancer in patients with immune mediated diseases and a history of prior malignancy. Some studies in RA did not differentiate between immunosuppressive and non-immunosuppressive DMARDs (sulfasalazine, hydroxychloroquine). However, as the majority of patients in such cohorts were on immunosuppression, they were included in the conventional immunosuppressant category. Studies were categorized as population-based or clinic-based. Studies with or without a control population were eligible for inclusion. However, studies reporting incidence of primary cancers in an unselected population (not restricted to those with prior cancers) were not included.

Data Collection

The decision for inclusion of each study was made independently by 2 of the authors (E.S and A.A). Data was extracted including year of publication, study design and location, number of included patients, immune-mediated disease type, type of prior cancer (non-melanoma skin cancer, melanoma, lymphoma, breast, solid organ and mixed), treatment exposure, and number of new and/or recurrent cancers. Information was also obtained on the proportion of high-risk cancers (bladder, sarcoma, malignant melanoma, renal, non-melanoma skin cancer, and myeloma), stratifying by risk of cancer recurrence as previously described by Penn et al.⁴⁴. For each study, type of immunosuppression, person-years (p-y) of follow-up, and number of new primary and recurrent cancers were extracted from published findings and incidence rate per 1,000 p-y and confidence intervals (CI) were calculated assuming a Poisson distribution. Incident cancers were defined as new or recurrent, with new cancers being those that developed in a different organ from the organ associated with the previous cancer or in the same organ with a different histological type. In a majority of the studies, person-time of exposure was calculated from the date of diagnosis of the initial cancer. For patients re-initiating immunosuppression, this was done so at varying intervals after the index cancer. Where the complete data was not available in the abstract or full publication, personal communication was made with the corresponding authors to obtain missing information.

Study Quality

Due to the lack of applicability of existing study assessment quality scales, we developed a tool to assess quality of included studies by consensus among the study investigators. This score was based on six criteria: (1) inclusion of an unselected or consecutive cohort of patients with prior cancer exposed to immunosuppression; (2) availability of a control population within the same geographic area and disease type; (3) use of a systematic protocol to identify new or recurrent cancers; (4) distinction between recurrent cancers and new primary malignancy; (5) provision of time interval between index cancer and re-introduction of immunosuppression; and (6) sufficient follow-up (≥ 2 years) after exposure to develop new cancer or recurrent cancers. Studies meeting four or more criteria were considered high quality studies.

Statistical Analysis

Pooled incidence rates with their 95% CI for new primaries or recurrent cancers among patients with prior cancer were calculated per 1000 p-y of exposure, stratifying by exposure to anti-TNF biologic therapy, IMM therapy, or no immunosuppression (no IS). We calculated pooled incidence rates across exposures rather than meta-analyzing incident rate differences from each study as many of the studies included only one exposure group. In a secondary analysis, where data was available for more than one exposure group within the same study, we calculated a pooled incident rate difference for new primary or recurrent cancers between the different exposures. Owing to the heterogeneity between studies, meta-analysis was performed using a random effects model, using the DerSimonian-Laird weights⁴⁵. Heterogeneity between studies was assessed using the Cochran’s Q and i² statistic with values ranging from 0% (no heterogeneity) to 100%; values > 50% indicated significant heterogeneity. Between group comparisons in the pooled incidence rates of cancer were performed using meta-regression with p < 0.05 indicating a statistically significant treatment effect. Meta-regression was also performed to identify sources of heterogeneity including study design, location, age, proportion of high-risk cancers, prior cancer subtypes, number of participants and publication year. Where significant effects were identified, stratified analysis was performed. Additionally, pre-specified subgroup analyses were performed (i) by type of immune-mediated disease; (ii) new primary malignancies and recurrent cancers separately; (iii) proportion of high-risk cancer ≥ 50% of the cohort; (iv) by type of index cancer (hematologic, skin, solid organ); and (v) type of immunomodulator therapy (thiopurines, methotrexate). In this analysis, where separate data was not available for thiopurine and methotrexate users, studies were assigned the category of the most frequent agent. Publication bias was examined utilizing funnel plots by the Begg’s and Egger’s methods. All data were recorded in a Microsoft Excel spreadsheet (Microsoft Corp, Redmond, WA) and analyzed using Stata 13.2 (StataCorp, College Station, TX).

RESULTS

Literature Search

Our search identified 570 citations in MEDLINE and 1,526 citations in EMBASE (Figure 1). After reviewing the title and abstract and if necessary, full publications, 5 relevant studies (5 full papers) from MEDLINE and 24 relevant studies from EMBASE (10 full papers, 14 conference abstracts) were retrieved for full review. Following direct communication, the corresponding author provided 1 unpublished study currently in submission corresponding to a prior abstract⁴⁶. Eleven articles represented duplicate data^47–52 and we were unable to obtain sufficient information for analysis from 5 studies^53–57 resulting in a final cohort of 16 unique studies^{34–43, 46, 58–62} (Table 1). There was an equal number of studies in IBD^{35, 39–42, 46, 58} and RA^{34, 36, 37, 43, 60–62}, with 1 study combining both populations⁵⁹ and 1 study in psoriasis⁶². Ten studies were population-based^{34–38, 43, 46, 59, 60, 62} and 12 studies adopted a cohort design^{34–38, 40, 41, 43, 46, 59, 60, 62}. Three were from North America^{40, 58, 59}, twelve from Europe^{34–39, 41–43, 46, 60, 62}, and one from Asia⁶¹. The included studies investigated a spectrum of prior cancers (Table 1). Most commonly, studies grouped together all prior cancers (including non-melanoma skin cancers) (n=12). One study each examined only non-melanoma skin cancer⁵⁹, lymphoma⁵⁸ or breast cancer³⁷. The proportion of patients with a high risk index cancer subtype according to the Penn classification ranged from 0% to 100%. Thirteen studies reported on cancer recurrence after anti-TNF exposure^{34–38, 40–43, 46, 59–62}, 12 on IMM^{34–43, 58, 59}, 3 on combination therapy^{39, 40, 59} and 6 on patients with no IS^{35, 39, 40, 42, 58, 59} (Table 2). The interval between index cancer and initiation of immunosuppression was available in 9 studies (median interval 6.2 years, range 0.0 – 10.3 years).

Figure 1.

Flowchart of the literature search

Table 1.

Characteristics of included studies on immunosuppression and risk of recurrent cancer or new malignancy in patients with prior cancer

Author	Year	Study Region	Setting	Study Design	Disease	Number of patients	Mean age (in years)	Type of prior cancer	% High Risk Cancer	Medication groups	Median / Mean follow- up time (months)††
Bae	2014	Asia	Clinic	Case series	RA	22	63	Mixed	5	Anti-TNF	102
Fagerli	2014	Europe	Population-based	Cohort	Psoriasis	11		Mixed		Anti-TNF	-
Scott	2015	North America	Population-based	Cohort	IBD, RA	9,282	75	NMSC	100	Anti-TNF, Thiopurine, MTX	21–31
Raaschou	2014	Europe	Population-based	Cohort	RA	240	67	Breast	0	Anti-TNF	55–59
Aaltonen	2015	Europe	Population-based	Cohort	RA	177	63	Mixed		Anti-TNF, DMARD	26–28
Silvia-Fernandez	2014	Europe	Population-based	Cohort	RA	402	64	Mixed (excl NMSC)		Anti-TNF, DMARD	79–83
Poullenot	2014	Europe	Population-based	Cohort	IBD	79	47	Mixed	19	Anti-TNF	21
Onali	2014	Europe	Clinic	Cohort	IBD	15	41	Mixed	13	Anti-TNF, thiopurine	120
Beaugerie	2014	Europe	Population-based	Cohort	IBD	405	56	Mixed	13	Anti-TNF, Thiopurine, MTX	40
Strangfeld	2013	Europe	Population-based	Cohort	RA	254	64	Mixed		Anti-TNF, DMARD	19–42
Carmona	2011	Europe	Population-based	Cohort	RA	24		Mixed	17	Anti-TNF	19
El Mourabet	2011	North America	Clinic	Case series	IBD	9	43	Lymphoma	0	Thiopurine, MTX	36–73
Dixon	2010	Europe	Population-based	Cohort	RA	294	64	Mixed		Anti-TNF, DMARD	23–36
Axelrad	2015	North America	Clinic	Cohort	IBD	333	54	Mixed	29	Anti-TNF, Thiopurine, MTX	59–97
Rajca	2014	Europe	Clinic	Case-control	IBD	107		Mixed		Anti-TNF, Thiopurine, MTX	23–72
Algaba	2015	Europe	Clinic	Case series	IBD	46	43	Mixed	42	Anti-TNF, Thiopurine, MTX	34–38

^†- ranges represent the different median/median follow-up times for the various treatment arms

Table 2.

Rates of incident cancer in patients with prior cancer exposed to immunosuppression

Anti TNF
Author	Number	Person- years follow up	incidence per 1000pys	Number (new or recurrent cancers)
Bae	22		0.0	0
Fagerli	11		0.0	0
Scott (IBD)	202	416	67.3	28
Scott (RA)	582	1,528	53.0	81
Raaschou	120	592.0	15.2	9
Aaltonen	100	232.5	12.9	3
Silvia-Fernandez	243	1591.0	23.9	38
Poullenot	79	177.5	84.5	15
Onali	3	19.8	0.0	0
Strangfeld	124	419.4	50.1	21
Carmona	24	37.9	26.4	1
Dixon	177	515.0	21.4	11
Axelrad	55	284.6	24.6	7
Rajca	11	28.0	36.0	1
Immunomodulator
Author	Number	Person- years follow up	incidence per 1000pys	Number (new or recurrent cancers)
Scott (IBD)	390	738	85.4	63
Scott (RA)	3,157	6,640	67.8	450
Raaschou	120	550.0	16.4	9
Aaltonen	77	168.9	6.3	1
Silvia-Fernandez	159	855.0	46.8	40
Onali	12	106.0	0.0	0
Beaugerie	93	259.3	27.0	7
Strangfeld	130	415.8	33.7	14
El Mourabet	4	24.3	41.1	1
Dixon	117	235.0	38.3	9
Axelrad	78	383.0	57.5	22
Rajca	30	85.0	58.8	8
Algaba	15	47.6	0.0	0
No immunosuppression
Author	Number	Person- years follow up	incidence per 1000pys	Number (new or recurrent cancers)
Beaugerie	312	833.3	19.2	16
El Mourabet	5	15.0	0.0	0
Axelrad	149	852.0	54.0	46
Rajca	66	448.0	62.5	28
Algaba	29	83.4	24.0	2
Scott (IBD)	2,051	5,013.0	54.5	273
Scott (RA)	1,610	5,160.0	47.3	244
Combination
Author	Number	Person- years follow up	incidence per 1000pys	Number (new or recurrent cancers)
Axelrad	51	414.0	36.2	15
Algaba	4	12.6	0.0	0
Scott (IBD)	67	91.0	76.6	7
Scott (RA)	1,223	1,986.0	67.5	134

From the 16 included studies, all included patients were from an unselected cohort with 10 having a control population, 13 having a systematic method for identifying incident cancer, 10 differentiating new and recurrent cancer and 9 defining the interval between cancer and commencement of immunosuppressive therapy (Supplemental Table 1). The median person-years of follow-up were unable to be calculated for two studies. Fourteen studies met our pre-defined criteria of being of high quality.

Risk of cancer recurrence according to immunosuppression type

The pooled analysis included 11,702 persons contributing 31,258 p-y of follow-up after prior cancer diagnosis. There were 1,698 instances of a new primary or recurrent cancer (1,366 new cancers, 154 recurrent cancers, and 84 where this distinction could not be determined). In the no IS group (n = 4,222), there were 609 new or recurrent cancers over 12,404 person-years of follow-up, yielding a pooled incidence rate of 37.5 per 1000 p-y (95% CI 20.2 – 54.7) (Figure 2). However, there was significant heterogeneity between the studies (i²=94.3%, p < 0.001) with incidence rates ranging from 0⁵⁸ to 62.5 per 1000 p-y⁴². A total of 1,753 subjects contributing 5,842 person-years of follow-up were exposed to anti-TNF therapy after a prior cancer. There were 215 cases of new or recurrent cancer, yielding a pooled incidence rate of 33.8 per 1000 p-y (95% CI 22.3 – 45.2), similar to that observed in the no IS group (range 0⁴¹ to 84.5 per 1,000 p-y⁴⁶) (p=0.48). The pooled incidence rate in subjects placed on conventional immunosuppressant therapy after prior cancer was similar to the no IS (p=0.59) and anti-TNF groups with an incidence of 36.2 per 1000 p-y (95% CI 17.7 – 54.7) (p=0.84). This group included 718 incident cancers (new primary or recurrent) among 4,382 subjects contributing 10,508 person-years of follow-up. Three studies provided rates on cancer recurrence with combination immunosuppression therapy^{39, 40, 59} with a pooled incidence rate of 54.5 per 1,000 p-y (95% CI 29.7 – 79.3). While numerically greater, this was not statistically different from that with each of the individual therapies (p=0.23 compared to anti-TNF, p=0.27 compared to IMM) or with no immunosuppression (p=0.47).

Figure 2.

Forest Plot of Risk of recurrence of cancer or development of new primary by type of immunosuppression in individuals with a prior history of cancer

IBD – inflammatory bowel disease; RA – rheumatoid arthritis; NMSC – non-melanoma skin cancer

Subgroup analyses

We repeated the analysis separately for new cancers and recurrent cancers. The pooled incidence rates of new cancers were not statistically different between the no IS (24.9 / 1000 p-y; 95% CI 5.6 – 44.3), anti-TNF (28.8 / 1000 p-y; 95% CI 7.6 – 50.0, p=0.53) and IMM (38.4 / 1000 p-y; 95% CI 12.2 – 64.6, p=0.33) groups respectively though the latter had numerically the highest rates (Figure 3). Similarly, analysis of recurrent cancers also revealed no statistically significant difference in rates between the three groups (Figure 4). Heterogeneity between the study groups persisted in both the above analyses.

Figure 3.

Forest Plot of Risk of new primary cancer by type of immunosuppression in individuals with a prior history of cancer

Figure 4.

Forest Plot of Risk of recurrence of prior cancer by type of immunosuppression in individuals with a prior history of cancer

Analyses including only patients with IBD (Supplemental Figure 1) or RA (Supplemental Figure 2) yielded similar absolute and relative risk differences between the various treatment groups. Among 3,706 IBD patients contributing 10,332 person-years of follow-up, there were 539 cases of new or recurrent cancer. The pooled incidence rate among subjects on no IS, IMM, or anti-TNF was 35.7 per 1000 p-y, 37.9 per 1000 p-y and 48.5 per 1000 p-y respectively, with no statistical difference between the groups (p > 0.30 for all). Among 7,985 RA patients contributing 20,926 person-years of follow-up and 1,159 new or recurrent cancers, the pooled incidence rate among subjects on no IS, IMM, or anti-TNF was 47.3, 35.1, and 28.8 per 1000 p-y respectively (p > 0.2). We compared rates of recurrence of prior cancer or new primary between users of thiopurine therapy compared to those on methotrexate. This yielded similar rates of new or recurrent cancer with thiopurines (37.9 per 1,000 p-y, 95% CI 5.9 – 69.9) compared to methotrexate (38.9, 95% CI 14.7 – 63.0) (p=0.78).

Given heterogeneity of incident cancer subtypes, we repeated the analysis stratifying by type of index cancer. Among the two studies that included index skin cancer alone, the risk of new or recurrent cancers was statistically significantly greater with IMM (71.6 per 1000 p-y, 95% CI 58.9 – 84.2, p=0.035) when compared to no IS (50.8 per 1,000 p-y, 95% CI 43.7 – 57.8) and numerically (but not statistically) higher than anti-TNF therapy (55.5 per 1,000 p-y, 95% CI 44.7 – 66.3, p=0.22). In contrast, excluding studies with skin cancer as the index malignancy yielded similar rates of new or recurrent cancer with IMM (26.3 per 1,000 p-y, 95% CI 13.3 – 39.3), anti-TNF (24.3 per 1,000 p-y, 95% CI 16.0 – 32.5) or no IS (31.2 per 1,000 p-y, 95% CI 7.9 – 54.5) (p > 0.30 for all).

Meta-Regression analyses

Four variables met statistical significance on meta-regression – proportion of high risk cancers (p < 0.001), mean age (p=0.03), number of included patients (p=0.02) and inclusion of individuals with prior skin cancer (p < 0.001) (Supplemental Table 2). Type of inflammatory disease, study design, or interval between index cancer and initiation of immunosuppression were not predictive of increased rates of recurrence (p=0.25). Our estimates remained robust on sequential exclusion of each study (data not shown).

Meta-analysis of incidence rate differences

We repeated the random effects meta-analysis among studies that included a control population for comparison with an index group. In the random effects model, there was a lower rate of new or recurrent cancer with anti-TNF biologic therapy compared to IMM (8 studies, 5,867 patients, incidence rate difference= −9.8 per 1000 p-y, 95% CI = −19.5 to −0.1) (Figure 5). While there was no statistically difference for IMM therapy compared to no IS (4 studies, 7,884 patients; incidence rate difference = 11.9 per 1,000 p-y, 95% CI = −2.4 to 26.3), the number of included studies was small and consequently our statistical power was lower. There was also no difference in incident rates for anti-TNF therapy compared to no IS (−4.5 per 1,000 p-y, 95% CI −25.6 to 16.5) (3 studies, 4,726 patients).

Figure 5.

Incident rate differences between no-immunosuppression, conventional immunomodulator and anti-TNF therapy

(a) Conventional IMM vs. anti-TNF therapy

(b) Conventional IMM vs. no IS

(c) Anti-TNF vs. no IS

Sensitivity Analyses

We performed a number of sensitivity analyses to examine the robustness of our inclusion criteria and quality of included studies. Excluding studies that were published only in abstract form yielded similar pooled incidence ratios for all groups (IMM 38.7, Anti-TNF 34.9, Combination therapy 54.1, and no IS 38.7/1,000 p-y respectively). To avoid bias introduced by small studies or low frequency outcomes, we repeated the analysis including only studies with least 50 participants in each arm which also showed no statistically significant difference between the groups (IMM 41.9, Anti-TNF 34.4, Combination therapy 56.1, no IS 46.2/1,000 p-y) as did analyses which required at least 10 outcomes in each group (58.2, 44.6, 52.9, and 46.2/1,000 p-y respectively for IMM, anti-TNF, combination therapy, and no IS). We repeated the analysis stratifying studies by whether immunosuppression was introduced within or after our median interval of 6 years. When immunosuppression was initiated within a median of 6 years, the pooled incidence of new or recurrent cancers on IMM or anti-TNF therapy was 33.6 and 43.7/1000 p-y respectively which was similar in magnitude to when immunosuppression was introduced after 6 years (IMM 32.9/1000 p-y, p=0.86; anti-TNF 21.0/1000 p-y, p=0.43 when compared to within 6 years). All estimates were statistically similar to risk with no IS (p > 0.50). Stratifying by whether median follow-up was less than or more than the median for our analysis, we found numerical but no statistically significant difference between the various treatment arms. For studies with median follow up less than 36 months, the pooled incidence rates with no IS, IMM, anti-TNF, and combination therapy were 32.7, 41.4, 42.4, and 67.7/1,000 p-y respectively. For those with a median follow-up of over 36 months, the pooled incidence rates with no IS, IMM, anti-TNF and combination therapy were 56.6, 28.6, 24.2, and 35.6 per 1,000 p-y respectively. Excluding the two studies rated as not being high quality did not alter our estimates.

Publication bias

Both the Egger’s and Begg’s tests revealed no publication bias (p=0.42 and 0.78 respectively) (Supplemental Figure 3).

DISCUSSION

Physicians have historically been reluctant to recommence immunosuppressive therapy in patients with a history of cancer. For example, in a study from Saint-Antoine Hospital, IBD patients with prior cancer had comparable disease activity, but had lower use of IMM and higher rates of surgery than those without prior cancer⁴². Data on risk of new cancer in those with prior malignancy exposed to immunosuppressive therapy are sparse and the published studies to date are limited by the small number of events making it difficult to draw definitive conclusions^{25–29, 32–42}. In this meta-analysis, no increased risk of cancer recurrence was found in patients with chronic immune mediated diseases initiating immunosuppression with either conventional immunosuppressant or anti-TNF biologic therapy after a prior diagnosis of cancer. To our knowledge, ours is the first meta-analysis to analyze this risk, including over 11,702 persons and 31,258 person years of follow-up, combining data from diverse populations. These results were robust both in analyses using pooled incidence rates as well as including only those studies with a control group for comparison.

Among the various immunosuppressive regimens, no overall difference in risk was observed between conventional immunosuppressant and anti-TNF biologics using pooled incidence rates, while studies that included both IMM and anti-TNF arms showed a weak effect favoring lower recurrence rates with anti-TNF therapy compared to thiopurines. This is consistent with the pooled experience with these agents in patients without a history of cancer, where the risk of primary cancers such as lymphoma appears to be similar quantitatively with either therapy⁶³, or favors lower rates with anti-TNF compared to conventional immunosuppressants⁶⁴. However, it is important to note that while the risks were broadly similar for all cancers, there may be cancer-specific associations with individual therapies. Anti-TNF biologics have been associated with greater risk of melanoma but not non-melanoma skin cancers while conversely thiopurines may increase likelihood of non-melanoma cancers and less so for melanoma^{22, 23}. This interpretation is supported by a higher pooled incidence rate with IMM when compared to no IS or anti-TNF therapy when including only those with index skin cancer alone in contrast to the comparable incidence rates for studies not restricted to index skin cancer. Thus, individualized decision making is essential particularly for cancers that have been associated with a specific therapy.

Despite differences in demographics (older age among RA participants) and medication use (more common use of methotrexate in RA), we identified no differences in overall rates of recurrent or new cancers between both diseases and the relative safety of IMM and anti-TNF compared to no IS was also similar across the two. We also found no statistical difference in rates of new or recurrent cancer between thiopurines and methotrexate therapy though thiopurine use was almost entirely in those with IBD while methotrexate use was in those in RA. In the one study that contributed the most number of patients with each therapy, there was a numerically but not statistically higher rate of new NMSC with thiopurines (85.4 / 1,000 p-y) compared to methotrexate (67.8 / 1,000 p-y). We also, reassuringly, found a statistically similar pooled incidence rate of recurrence in those on combination immunosuppression though only a small subset of studies contributed to this analysis^{39, 40, 59}. This is in contrast to prior studies demonstrating an increased overall risk of index cancers with combination therapy compared with anti-TNF monotherapy in IBD⁶⁵.

The data presented have direct clinical relevance. Our study provides reassurance about restarting immunosuppressive therapy in patients with prior cancer, with the caveat that we were unable to ascertain the exact interval at which recommencement would be safe as this was at the discretion of the treating physician in each of the studies. There is heterogeneity in natural history of IBD with varying risks of recurrence of disease-related symptoms after therapy cessation. Comprehensive models that can predict this outcome may be helpful in identifying those at highest risk for IBD relapse, and consequently most likely to require early re-initiation of immunosuppressive therapy for IBD after the index cancer diagnosis^{29, 30, 66}. While many existing guidelines recommend avoiding immunosuppression for 5 years after index cancer^24–29, our data may provide reassurance that there is no increased risk. An important caveat exists that the median interval between index cancer and recommencement of immunosuppression in our cohort was 6 years and thus it cannot be robustly concluded that re-initiation of immunosuppression before the recommended 5 year interval is safe in all patients. However, a stratified analysis among studies where the median interval was shorter than 6 years revealed no difference in risk of new or recurrent cancers. As data becomes increasingly available, guidelines may ease the recommendation on restricting immunosuppression in this population also at risk for progressive damage from their inflammatory disease. In the meantime, one can continue to exercise caution in those with a very recent history of cancer, particularly those with high-risk index cancer. However, in a subset of patients, a shorter window of interval may be sufficient without increasing risk of adverse outcomes.

We readily acknowledge several limitations to this meta-analysis. There was significant heterogeneity between the studies and 3 of the included studies had fewer than 20 patients. Post-hoc power calculated revealed an 84% power to detect a statistically significant difference between the no IS and IMM group and 81% power to detect a difference between the anti-TNF and no IS groups. The process for detecting cancer recurrence was not uniform across the studies and may have resulted in missed cases. There was also significant variation among the cancers included as the index malignancy. Some studies were restricted to specific index cancers (for example, skin, breast), others included all prior cancer, while yet others included only prior solid organ or hematologic malignancies. It is also possible that patients on immunosuppression are under closer follow-up leading to a bias with a higher cancer rate than among controls. Person-time exposure was presented from the date of diagnosis of index cancer in most studies rather than from time to re-initiation of immunosuppressive therapy. However, the latter approach would also be susceptible to bias as time to re-initiation may be influenced by risk of recurrence over time. We were unable to calculate the effect of dose of immunosuppression or duration of exposure on the risk of recurrence and there were varying intervals between prior cancer and recommencement of immunosuppressive therapy. We also could not examine the impact of immunosuppression use prior to the development of the index cancer. As noted above, whether there are absolute or relative differences between each therapy and specific cancers could not be examined. It is also possible, and likely, that patients who were at high risk for recurrence were not recommenced on immunosuppression and consequently our findings may be more applicable to a population where the patient and physicians were comfortable with re-initiation of therapy. While pharmaco-epidemiologic studies of drug safety have traditionally relied on a new user design, such analyses were not possible in our study. However, for outcomes such as malignancy where a prolonged exposure to treatment is required for an elevation in risk, a new user design may not be feasible given the challenges in following large cohorts of patients for prolonged periods of time. The median follow-up in several studies was shorter than 36 months and though the risk of recurrence is highest soon after index cancer, the effect of immunosuppression on delayed recurrent or new primary cancer could not be examined.

In conclusion, treatment decisions following a cancer diagnosis are complex and must take into account the natural history of cancer, histological type and stage, time from diagnosis and course of underlying chronic inflammatory disease. Our findings suggest that anti-TNF therapy, conventional immunosuppressant therapy, or combination immunosuppression are not associated with increased risk of cancer recurrence in this population. However, there is a need for larger studies to prospectively monitor for cancer recurrence and new malignancies in this population to better inform our practice.

Abbreviations

CD

Crohn's disease

CESAME

Cancers Et Surrisque Associé aux Maladies inflammatoires intestinales En France

GETAID

Groupe d’Etude Thérapeutiques des Affections Inflammatoires du tube Digestif

IBD

Inflammatory bowel disease

NMSC

Non-melanoma skin cancer

UC

Ulcerative colitis