In this issue of Blood, Ammann et al used the Medicare Surveillance, Epidemiology, and End Results (SEER-Medicare) linked databases, which include comprehensive clinical information for a representative cohort of cancer patients in the United States, to identify an increased risk of thromboembolic events (TEEs) in patients with secondary hypogammaglobulinemia who receive a first dose of IV immune globulin (IVIg) in comparison with a propensity-matched comparison group of 8035 nonusers of IVIg.1
IVIg is often administered to persons with hypogammaglobulinemia and chronic lymphocytic leukemia (CLL) or multiple myeloma (MM) in order to avoid infections. Based on potential biological mechanisms, the authors hypothesized that prothrombotic effects of IVIgs would most likely be acute, leading to high rates of TEEs shortly after IVIg infusion. The study hypothesis was supported by a 3.4-fold increased risk of arterial TEEs identified immediately after IVIg infusion.
The safety concern is important. Reports of TEEs led to voluntary withdrawals of 1 IVIg product in the United States in 2010 and a second product in Israel in 2011. In 2013, the US Food and Drug Administration required IVIg manufacturers to include a black box warning about increased risks of TEEs.2
The study by Ammann et al was based on data obtained from the SEER cancer registries linked to Medicare claims and enrollment data for the years 1992 to 2010: “big data” to the rescue.3 The major difference between observational studies such as this and randomized clinical trials is patient selection. The authors avoided pitfalls of selection bias by using a comparison group of non-IVIg users that was formed on an individual basis in order to adjust for baseline risks of clinically important thrombosis risk factors that could bias the analysis. This effort facilitated identification of statistically significant and clinically relevant rates of TEEs among CLL patients who received first doses of IVIg by adjusting for confounding factors. Thus, an improved approach to detection of signals for rare serious adverse drug reactions has been described.
It should be noted that drug-associated TEEs have been identified in other cancer settings. Probably the most unexpected discovery was the 1.6-fold increased TEE risk identified with erythropoiesis-stimulating agents administered to cancer patients.4 Previously, fivefold increased TEE risks were identified with thalidomide treatment of persons with MM.5 The underlying pathophysiology for these serious adverse drug reactions remains unknown, but the risk-benefit calculus for each drug has been altered. For erythropoiesis-stimulating agents, TEE risks (as well as cancer progression risks) are so strong that clinicians rarely administer these agents to cancer patients receiving chemotherapy for curative intent. For thalidomide, most patients receive some form of TEE prophylaxis: aspirin, warfarin, or low-dose heparin. For IVIg, the implications of TEE risks are yet to be seen.
It all boils down to risk-benefit ratios. The risks of drugs are not absolute but must be weighed against benefits. In oncology and hematology, most treatments have serious and even life-endangering side effects. Because most cancers lead to early death if left untreated, great risks are accepted when striving for cure. On the other hand, when treating with the intent of prolonging life instead of curing, the same risks cannot be accepted.
Both MM and CLL provide illustrative examples of the spectrum of difficult risk-benefit decisions in the setting of hematologic malignancies. Depending on the disease stage and risk factors, survival rates vary, but in both cases, many patients may live for many years with minimal symptoms. Patients with MM or CLL are usually treated with the intention of controlling the disease while allowing patients to maintain a good quality of life. However, the only potentially curative treatment of both MM and CLL remains high-dose chemotherapy followed by an allogeneic stem cell transplant. This treatment involves significant morbidity and mortality, and requires in-depth consideration of patient risk factors that could affect patient survival following transplantation.
Preventing and treating infections of MM and CLL patients is essential. Some patients with MM or CLL are more susceptible to infections than others, and controlling their infections could be crucial when carrying out intensive chemotherapy. Antimicrobial agents are needed but they also have side effects and, in addition, the risk of antibiotic resistance must be taken into account.
When using IVIgs, risk of TEEs is 1 factor to consider when treating certain CLL patients with serious risk or life-threatening infections. The risk-benefit calculus differs for patients with CLL or MM who do vs do not have high baseline risks of infection. The risk-benefit ratio becomes even more challenging when taking into account that in randomized controlled trials, IVIg reduces risk of major infection, but not overall mortality.6
But are we even in a position to make solid recommendations on use of IVIgs for persons with CLL or MM? On the risk side of the equation, do we know enough about TEEs and their relation to IVIgs? Big data are the cornerstone of many pharmacovigilance studies. And yet clinical inquiry, rather than big data, must play the more important role. In 1854, Louis Pasteur said that “in the field of observation chance favors only the prepared mind.”7 Hematologists must have prepared minds to link drugs with rare serious adverse events such as IVIg-associated TEEs and quinine-associated hemolytic uremic syndrome.8 Big data did not identify these associations. Big data opens the door, but scientific inquiry closes it. Here is a call for inquisitive hematologists to provide solid laboratory evidence that IVIgs actually cause TEEs. Absent these findings, hematologists will continue to wonder to what extent the statistical findings of increased risks of TEEs reflect clinically significant concerns.
Conflict-of-interest disclosure: The authors declare no competing financial interests.
