Systematic review of observational studies reporting antiphospholipid antibodies in patients with solid tumors

Antiphospholipid syndrome (APS) is an acquired autoimmune prothrombotic disease characterized by persistent elevation of antiphospholipid (aPL) antibodies, lupus anticoagulant (LA), immunoglobulin G (IgG) and/or IgM isotype of anticardiolipin (aCL), or anti-β2-glycoprotein I (anti-β2-GPI) antibodies, leading to recurrent thromboembolic and pregnancy adverse events (abortion, fetal death, or premature birth).^1-3  Multiple organ failure known as catastrophic APS occurs in ∼1% of patients with APS.^4-6 

The prevalence of elevated aPL antibodies is 1% to 5% among healthy young individuals increasing to 50% among elderly patients with chronic diseases. However, it is unclear how many healthy people with positive antibodies eventually develop APS.^7-11  A systematic review of observational studies excluding patients with autoimmune diseases reported a pooled prevalence rate of aPL antibodies in up to 23.3% of patients with stroke, 23% with myocardial infarction, 15.8% with deep vein thrombosis, and 13% of women with pregnancy adverse events.¹²  aPL antibodies develop in genetically susceptible individuals following an infection or in the setting of autoimmune diseases as a “first hit”; evidence suggests a “second hit” is also required for thrombosis to occur.¹³  Potential putative candidates for this second hit are infection, cancer, other procoagulant conditions, and certain drugs (eg, cytotoxic chemotherapy).¹⁴ 

Patients with cancer are at increased risk of thrombosis (four- to 60-fold higher) compared with the general population.^15,16  Elevation of aPL antibodies has been reported in various solid and hematological malignancies, suggesting a possible pathogenic role of aPL in triggering thrombosis in cancer patients.^17-19  We conducted a systematic review and meta-analysis of observational studies evaluating the development of aPL antibodies and related thromboembolic events in patients with solid tumors.

Data sources included Medline, EMBASE, Web of Science, PubMed ePubs, and the Cochrane Central Register of Controlled Trials through August 2019 with no restrictions. The Medline search strategy is detailed in supplemental Appendix 1.

Publications were screened by 6 independent reviewers (in pairs) using a 2-step approach. First, titles and abstracts were reviewed to identify relevant citations. Then, the full text of these citations was reviewed. Discrepancies were resolved by adjudication by a third reviewer.

We included observational studies reporting patients with solid tumors who were evaluated for aPL antibodies, measured at least once after the diagnosis of cancer. When multiple citations of the same study were published, we considered the most recent full text.

We excluded studies that reported patients with prior infection, autoimmune diseases, or surgery (eg, laparoscopy) and those receiving certain drugs (eg, interferon α) that could explain positive aPL antibodies. Studies were also excluded if they reported patients with prior history of thromboembolic/pregnancy events or elevated aPL antibodies diagnosed before the diagnosis of cancer, measured rare aPL subtypes not included in the diagnostic criteria of APS,^1,2  or did not provide sufficient information about participants. Studies not published in English were excluded.

Data were extracted and quality was assessed by one reviewer (N.A.-W.) and cross-checked by another (F.F.). Disagreements were resolved by consensus. We extracted data on study characteristics (design, number of participants, type of cancer, subtype of aPL antibodies and isotype distribution, persistence of aPL antibodies, aPL-related thromboembolic or adverse pregnancy events, timing of aPL testing in relation to the clinical events, and related death). Because of the laboratory differences and variation in the cutoffs for aPL positivity, elevated titers were considered according to the study authors’ definition. When patients had >1 elevated aPL isotype, the overall prevalence was considered if the data were provided. Otherwise, the highest prevalent isotype was considered. Clinical features associated with APS (eg, thrombocytopenia) but not in the diagnostic criteria were not considered.^1,2 

Risk of bias in individual studies was assessed using the Newcastle-Ottawa Scale in 3 key quality domains (selection, comparability of cohorts, and outcome); total maximum score is 9, and a higher score indicates a lower risk of bias.²⁰ 

We performed a meta-analysis and estimated the risk ratios (RRs) with 95% confidence intervals (CIs) of developing aPL antibodies in patients with solid tumors compared with healthy controls via a random-effects model using RevMan version 5.3.²¹  We also estimated the RRs with 95% CIs of developing thromboembolic and/or pregnancy adverse events and related death in patients with solid tumors and positive aPL antibodies compared with patients with negative antibodies when data were available. Then, we pooled prevalence rates of aPL antibodies and related clinical events for each tumor type using STATA statistical software version 15 (STATA Statistical Software 2017: Release 15; STATA Corp., College Station, TX). Our numerators were the number of patients with solid tumors and positive aPL antibodies and/or aPL-related clinical events in each study. For the denominator, we considered all patients with solid tumors in each study. Statistical significance was set at 2-sided P < .05. Study heterogeneity was assessed by using the I² statistic.

A funnel plot and a regression asymmetry test was performed to assess publication bias and small-study effects in the meta-analyses for having elevated aPL antibodies (aCL, LA, or anti-β2-GPI antibodies) in patients with solid tumors compared with healthy individuals.²² 

Our search yielded 11 639 unique citations. Of these, 33 observational studies (40 publications) met eligibility criteria. The reasons for exclusion in each step are shown in supplemental Figure 1. References for selected studies are included in supplemental Table 1.

The publication bias graph is shown in supplemental Figure 2; Egger’s test showed no evidence for publication bias, even when outliers were removed.

Of the 33 studies included, 31 reported the prevalence of aPL antibodies among patients with solid tumors, 11 also reported the prevalence among healthy controls as a comparison group, and 10 studies reported on the prevalence of thromboembolic and/or pregnancy events. Eighteen studies (54.5%) were prospective, 11 (33.3%) cross-sectional, and 4 (12.1%) retrospective cohort studies. Gastrointestinal (GI) tumors were the most frequently evaluated in 11 studies (33.3%), followed by genitourinary (GU) in 9 studies (27.3%). Twelve studies reported the stage of malignancy and/or the presence of metastasis; 1 reported that 80% of patients diagnosed with thromboembolism at presentation had distant metastasis at the same time.²³  A total of 2773 patients with solid tumors were included (871 with GU, 428 with GI, 414 with breast, 353 with lung, 136 with melanoma, 19 with bone/soft tissue, 16 with head and neck, 11 with central nervous system [CNS], and 525 with various types of solid tumors); 877 healthy controls were evaluated. The most frequent aPL antibodies examined were aCL (19 studies; 57.6%), followed by anti-β2-GPI antibodies in 8 studies (24.2%) and LA in 7 (21.2%). Most studies used the 5th percentile as the cutoff for aCL and the 99th percentile as the cutoff for anti-β2-GPI; 11 studies did not report cutoffs used (supplemental Tables 2 and 3). Ten studies evaluated aPL antibodies but did not specify the antibody subtype. At least 2 aPL subtypes were examined in 8 studies (24.2%), and all 3 antibodies were examined in 2 studies.^24,25 

A total of 235 patients with GI cancer were evaluated in 5 studies; patients were 5.1 times more likely than controls to develop aCL antibodies (95% CI, 2.6-9.95) (Table 1).^26-30  Two studies included patients with GU cancer (n = 49) with increased pooled risk of developing aCL antibodies compared with healthy controls (RR, 7.3; 95% CI, 3.3-16.2).^30,31  Pooled risk estimates from 2 studies evaluating 62 patients with breast cancer were not statistically significant (RR, 1.9; 95% CI, 0.19-19.9).^30,32  One study evaluated patients with CNS, lung, and melanoma³⁰ ; a significant higher risk was observed in patients with lung cancer compared with controls (RR, 5.2; 95% CI, 1.3-20.6), whereas in CNS and melanoma, the increased risk was not statistically significant. Two additional studies evaluated 165 patients with various solid tumors, reporting a significantly increased risk compared with controls (RR, 6.5; 95% CI, 2.1-20.3)^25,30  (supplemental Figure 3).

Two studies measured these antibodies in patients with GI cancer,^26,33  and 1 study measured them in patients with various solid tumors (Table 1).²⁵  No statistically significant risk was observed when compared with healthy controls (supplemental Figure 4).

Font et al measured LA antibodies in patients with various solid tumors.²⁵  Only 1 patient had LA, and the risk was not statistically significant when compared with controls.

Three additional studies evaluated the presence of aPL antibodies in different cancer types without specifying the antibodies subtype (Table 1).^25,34-39  A statistically significant increased risk was observed in patients with breast, GI, head/neck, and lung cancers.

For this analysis, we included the prevalence of antibodies for all studies, with or without controls (Table 2). The pooled prevalence rates of the various types of cancer ranged between 7% and 28% for aCL antibodies,^{19,24-32,40-50}  between 5% and 53% for anti-β2-GPI antibodies,^{24-26,33,40,42,44,45,51,52}  and from 3% to 80% for LA.^{23-25,32,45,49,50,53,54}  Two studies examined all 3 aPL antibodies, but none of them reported the percentage of patients who were positive for all 3; therefore, the pooled prevalence rate of triple-positive aPL antibodies could not be determined.^24,25 

Ten additional studies evaluated the presence of aPL antibodies in different cancer types but did not specify the antibody subtype.^{25,34,35,41,51,52,55-59}  The highest pooled prevalence rate was 31% (22% to 41%) in patients with GU cancer.

Ten studies reported the isotype distribution of aCL antibodies in patients with solid tumors, and 3 included healthy controls.^{19,24-26,29,41-43,48,50}  IgM was the most frequent isotype (87/601; 14.5%) compared with IgA (9/106; 8.5%) and IgG (41/532; 7.7%). Patients with breast and GI cancers had a significant increased risk of IgM compared with healthy controls (RR, 4.5; 95% CI, 1.1-18.9), while the increased risks of IgG (RR, 3.0; 95% CI, 0.44-20.8) and IgA (RR, 4.5; 95% CI, 0.52-39.7) were not statistically significant.

Four studies reported the isotype distribution of anti-β2-GPI antibodies; 2 of them included healthy controls.^24-26,42  IgA was the most frequent isotype (33/130; 25.4%) compared with IgM (29/307; 9.4%) and IgG (8/309; 2.6%). A numerical nonsignificant increase in risk was observed in patients with breast and GI cancers compared with controls for all isotypes, but it did not reach statistical significance.

Four studies (12.1%) examined aPL antibodies in 2 consecutive occasions at least 12 weeks apart and the frequency of persistent positive antibodies was variable.^24,25,30,41  One study evaluated patients with lung cancer every 6 months for up to 2 years, and there was no change in aPL positivity upon retesting, yet the number of patients who were tested more than once was not specified.²⁴  Another study evaluated patients with breast cancer; all patients with positive IgM aPL (n = 20) and IgM aCL (n = 24) at baseline became negative after chemotherapy, while 4 out of 8 patients (50%) with positive IgG aPL and 8 out of 12 patients (33.3%) with positive IgG aCL at baseline remained positive.⁴¹  An additional study re-evaluated 3 patients (2 with colorectal and 1 with lung cancer) and reported that all 3 lost their positivity for aCL and remained thromboembolic-free for 12 months of follow-up.³⁰  Only one study reported follow-up in 1 patient and 2 healthy controls who tested positive at baseline, but all 3 turned to be negative after 12 weeks.²⁵ 

Ten studies reported thromboembolic events. Of these, only 4 reported the events for patients with positive and with negative aPL antibodies (Table 3). Patients with lung cancer and positive aPL antibodies had a statistically significant higher risk of developing thromboembolic events compared with patients without aPL (RR, 3.8; 95% CI, 1.2-12.2).^24,45,53  Two studies evaluated women with GU cancer for thromboembolic and/or pregnancy events and compared patients with positive and with negative aPL antibodies, but the increased risk was not statistically significant^43,56  (supplemental Figure 5). One study reported deaths secondary to thromboembolic events in patients with lung cancer comparing those with positive and with negative aPL antibodies.⁵³  Deaths were more frequent among those with aPL antibodies (RR, 2.1; 95% CI, 1.4-3.3).

The pooled prevalence rates of thromboembolic events for the various cancers ranged between 16% and 23% (Table 4). Both arterial and venous thrombosis was reported (stroke, transient ischemic attack, myocardial infarction, deep venous thrombosis, sinus venous thrombosis, arterial embolism, catheter thrombosis), as well as recurrent abortion. Few patients had >1 thromboembolic event. The pooled prevalence rates of thromboembolic events in patients with solid tumors and positive aPL antibodies was 35% (95% CI, 27% to 43%) for patients with lung cancer^24,45,53  and 21% (95% CI, 8% to 38%) for GU cancer.^43,56  One study examined all 3 aPL antibodies in patients who developed clinical events,²⁴  another examined only 2 antibodies,⁵⁰  and the remaining 8 examined only 1 subtype.^{23,43,51,53,55,56,60,61}  LA was the most frequent antibody reported in 48% of patients who developed clinical events, followed by aCL (26.3%) and anti-β2-GPI (19.5%). Isotype distribution of aCL antibodies was reported in only 3 studies, and IgM was the most frequent (26.3%) compared with IgG (14.0%).^24,43,50  As for anti-β2-GPI isotype, IgM (28.0%) was also more frequent than IgG (4.8%).²⁴  Three additional studies evaluated the presence of aPL antibodies in patients with GU cancer but did not specify the antibody subtype; IgG was more frequent (36.5%) compared with IgM (19.0%) and IgA (48.3%).^51,55,56 

Of note, one study evaluated all 3 aPL antibodies as well as other immunologic and clotting factors (fibrinogen, factor VIII, factor IX, protein C, D dimmer, interleukin-6, and tumor necrosis factor levels) in 66 patients with lung cancer and identified only LA as a significant risk factor for increased thrombosis in a multivariable regression model.²⁴  The same study evaluated aPL persistent positivity in lung cancer patients who developed thrombotic events and reported no change in aPL positivity in a 6-month interval. None of the studies reported the interval between positive aPL test result and occurrence of clinical events.

Overall Newcastle-Ottawa Scale scores ranged from 3 to 6 (supplemental Table 4). All studies reported adequately representative cohorts, but only 11 (33.3%) included healthy individuals from the same community and matched exposed and nonexposed participants. For the outcome domain, all studies assessed ≥1 aPL antibody, 22 (67%) specified the reference range used to define antibody elevation, and only 4 (12.1%) reported follow-up for aPL antibodies, and the duration of follow-up was enough to evaluate if the antibodies were persistent or transient.^24,25,30,41  As for adequacy of follow-up, only 1 study reported follow-up for all participants who initially tested positive for aPL antibodies⁴¹ ; the follow-up rate was <80% in the other 3 studies, without description of those lost.^24,25,30 

Patients with cancer have an increased risk of thromboembolic events resulting in significant morbidity and mortality.^16,62-65  Thromboembolic events are the second most common cause of death in cancer patients, occurring in up to 20%.^65-67  The most widely used clinical tool for identification of risk of cancer venous thromboembolism (VTE) is the Khorana score.⁶⁸  However, the clinical value of this score for thromboprophylaxis in cancer patients is uncertain.⁶⁹  Moreover a recent meta-analysis of studies evaluating the Khorana score concluded that most of VTE events occurred outside the high-risk group.⁷⁰  Therefore, there is an unmet need for newer tools incorporating other biomarkers to better identify cancer patients at high risk of VTE.^71,72  There is no prediction model tool for cancer VTE that incorporates the risk associated with the presence of aPL in cancer patients.

To our knowledge, this is the first systematic review and meta-analysis to synthesize the evidence from observational studies evaluating the prevalence of aPL antibodies and associated thromboembolic events in patients with solid tumors. We pooled results from 33 studies, including 11 in which patients were compared with healthy controls. Pooled results showed an increased risk of developing aPL antibodies for almost all solid tumors examined when compared with controls, but for some cancers, the differences did not reach statistical significance, as samples were small. The highest risks were observed for GI, GU, and lung cancers. While many studies measured aCL antibodies, few examined anti-β2-GPI antibodies or LA, limiting potential inferences on causation, as these antibodies are important in the development of clinical events. No statistically significant association was observed between elevated LA or anti-β2-GPI antibodies and solid tumors, but the available evidence may not be sufficient to draw a definitive conclusion. Nevertheless, our data showed a high prevalence of all 3 aPL antibodies in various types of solid tumors. For instance, the pooled prevalence rates from all studies showed elevated aCL antibodies in more than one-fourth of patients with GU cancer, elevated anti-β2-GPI antibodies in more than half of patients with breast cancer, and positive LA antibodies in approximately half of patients with lung cancer.

The clinical significance of aPL antibodies in cancer patients and their potential impact on patients’ survival remain uncertain. To quantify the role of these antibodies in pathogenesis of thromboembolic events in patients with solid tumors, we evaluated the risk of thrombosis in cancer patients with aPL vs those without aPL. Few studies were identified showing a significant increased risk in patients with lung cancer, whereas the increased risk in GU cancer did not reach statistical significance, as the sample was small. Notably, LA seemed to be a significant risk factor for thrombosis in cancer, being positive in approximately half of the patients who developed clinical events. Only De Meis et al’s study was identified reporting persistent positive aPL in patients with lung cancer who developed thrombosis.²⁴  However, most studies reported only 1 aPL titer, and none reported the interval between laboratory diagnosis of positive aPL and occurrence of thromboembolic events, so we cannot reach definitive conclusion regarding the attribution of clinical events to aPL antibodies. In an attempt to quantify the influence of elevated aPL antibodies on survival, we also evaluated thromboembolic event–related mortality in cancer patients with aPL vs those without aPL. Only Fei et al’s study was identified to show a higher mortality in patients with lung cancer and positive aPL antibodies.⁵³ 

Gomez-Puerta et al previously described the clinical and immunological profiles of 120 patients with cancer and positive aPL antibodies identified from the literature but had a more limited search (only one database) and included 45 cases with concomitant autoimmune diseases.¹⁷  Similarly, Islam documented 31 case reports where APS coexisted with cancer and reviewed observational studies reporting aPL antibodies in cancer.¹⁸  Both reviews identified the presence of aPL antibodies in different types of cancer and reported that aPL-related thromboembolic events could be the first manifestation of underlying cancer and that the presence of aPL could influence cancer treatment and prognosis. Nevertheless, it is important to note that in both reviews, cases diagnosed with cancer after the thrombotic manifestations of APS occurred were included. Moreover, the reviews did not follow the specific steps required for systematic reviews and did not pool the prevalence of aPL antibodies and related thromboembolic events or compare the results between cancer patients and controls.

aPL antibodies are well known as an acquired immune-mediated risk factor for thromboembolic events^1,2,73 ; however, the exact mechanisms mediating the development of aPL antibodies in cancer are poorly understood. Conceivably. aPL antibodies could be produced as a humoral response against tumoral antigens. Also, some patients with monoclonal gammopathies may secrete immunoglobulins with aCL activity.^30,46,74-77  Finally, the potential effect of cancer therapy is not well understood, but in the new era of immunotherapy, development of autoimmune antibodies is considered part of the spectrum of immune-related adverse events. Studies have shown that polymorphism mucin-like molecules such as P-selectin may contribute to the thrombotic risk of aPL.⁷⁸  Several GI malignancies are mucin secreting, a known risk factor for thrombosis in cancer.⁷⁹  However, specific tumor histologies were not consistently reported in the studies included in our review, and therefore, the effect of mucin-secreting tumors (adenocarcinomas) vs other types of tumors of the GI tract (ie, neuroendocrine) could not be assessed. Also, mucin-like substances are not routinely measured in the clinical setting for the assessment of thrombotic risk; hence, the effect of those biomarkers could not be incorporated in our review.

Other factors contribute to hypercoagulability in cancer, including activation of platelets and endothelial cells and enhanced expression of tissue factor, ultimately disrupting the coagulation cascade and fibrinolysis.^80-82  In primary APS, in the presence of aPL antibodies, these mechanisms actively participate in triggering thrombosis.^13,83,84  Yet, it is unclear how aPL antibodies may add to the preexisting risk of thrombosis in cancer patients. It is postulated that genetic factors play a role in susceptibility to thrombosis in cancer as well as primary APS.^85,86  It should also be noted that the presence of aPL antibodies in cancer may carry additional risk as these antibodies could accelerate tumor angiogenesis and progression through a tissue factor–mediated effect.^16,87  Moreover, whether the presence of aPL antibodies could increase the risk of developing immune-related adverse events in cancer patients treated with immunotherapies is still undetermined. Two patients with aPL antibodies induced by the use of checkpoint inhibitors have been reported, 1 with APS and the other without clinical manifestations^88,89  In addition, aPL antibodies were observed in 18 of 30 patients with melanoma treated with interferon α and/or interleukin-2.⁴⁶ 

Our review included an extensive search of 5 databases and specific criteria for inclusion and quality appraisal. We excluded studies in which patients had thromboembolic events or elevated aPL antibodies before the diagnosis of cancer and studies reporting patients with concurrent autoimmune diseases or infection to avoid overestimating the risk of developing aPL antibodies and related thrombosis. Nevertheless, our review was limited by the data available in original studies. For instance, most studies did not assess all recommended aPL antibodies or used different cutoffs to define aPL antibody elevation, which might affect the interpretation of RR across subgroups. Moreover, most studies measured aPL antibodies on 1 occasion, did not report isotype distribution or measured only one isotype, and did not report the 3 aPL antibodies or isotype distribution in cancer patients who had thrombosis. Most studies did not specify the stage of cancer and whether cancer progression or specific therapies were associated with aPL and thrombosis, and therefore, we were not able to discern local vs locally advanced vs distant metastatic tumors, and further stratification on tumor histology and stage was not possible. No studies reported specifically whether any patients developed catastrophic APS. Data were pooled from different types of observational studies; few studies on specific malignancies had small sample sizes, and quality was a concern.

In conclusion, evidence in the literature supports an increased risk of developing aPL antibodies, primarily in patients with GI, GU, and lung cancers. Although it would be difficult to attribute clinical events to aPL antibodies drawn on 1 occasion, clinicians should be aware that these antibodies might contribute to developing thrombosis in cancer. While aPL positivity could help in predicting risk of thrombosis, the available evidence is not enough to recommend routine aPL testing for patients with solid tumors. Yet, aPL testing should be considered for patients presented with clots and should be repeated for any patients with positive aPL to determine if antibodies were clinically significant. Preclinical studies are warranted to help understanding the proposed mechanisms for aPL development and pathogenicity in cancer. Further well-designed longitudinal studies are crucial to evaluate the clinical outcomes of cancer patients developing these antibodies.

This study was supported in part by National Institutes of Health, National Cancer Institute grant P30 CA016672.

Contribution: M.E.S.-A. had full access to all of the data in the study and takes responsibility for the integrity and the accuracy of the data analysis; M.E.S.-A., M.A.L.-O., and N.A.-W. contributed to study concept and design; M.E.S.-A., M.A.L.-O., J.H.T., and N.A.-W. acquired data; N.A.-W., M.A.L.-O., J.H.T., F.F., G.S., T.S., A.Y., and S.A.-H. analyzed and interpreted data; N.A.-W., F.F., and G.S. performed quality appraisals; N.A.-W. and M.E.S.-A. drafted the manuscript; M.E.S.-A., J.H.T., M.A.L.-O., C.M.R.-H., N.A.-W., F.F., G.S., T.S., A.Y., and S.A.-H. critically revised the manuscript for important intellectual content; N.A.-W. and M.A.L.-O. performed statistical analyses; and M.E.S.-A. provided administrative, technical, or material support and supervised the study.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Maria E. Suarez-Almazor, Department of Health Services Research, Unit 1444, The University of Texas MD Anderson Cancer Center, 1400 Pressler St, Houston, TX 77030; e-mail: msalmazor@mdanderson.org.