Research Signal – Source date 2022-04-13
Forty-two studies compared cancer patients with COVID-19 who had previously received immune checkpoint inhibitors with patients who had not. Their pooled results did not show a statistically clear difference in death, severe illness, or hospitalisation. But the most informative comparisons were sparse, and some estimates were wide enough to leave substantially different effects plausible.
Immune checkpoint inhibitors are cancer drugs that lift restraints on immune cells, helping the body mount an anti-tumour response. That immune shift raised competing questions during COVID-19: might it help antiviral defence, or could it add to inflammatory injury? The review assessed patient outcomes across published studies. Its results do not show that either proposed mechanism occurred, and they do not establish that the drugs had no effect.
The pooled numbers leave room for more than one answer
For death, eight studies contributed adjusted estimates: comparisons that attempted to account for differences between patients. The pooled odds ratio was 1.31, with a 95% confidence interval from 0.58 to 2.95. An odds ratio of 1 would mean equal odds in the two groups; this estimate was not statistically clear, and its interval includes both lower and considerably higher odds among patients who had received checkpoint inhibitors.
Thirty studies contributed unadjusted death results. Their pooled odds ratio was 1.06, with an interval from 0.85 to 1.32. Combining adjusted and unadjusted results also did not produce a statistically clear association. The adjusted estimate is especially useful for understanding the uncertainty: “not statistically clear” does not mean the groups were shown to have the same risk.
The comparison matters because patients who receive cancer treatment can differ in ways that also affect COVID-19 outcomes. Age and other illnesses are relevant, as are cancer type and stage, other cancer therapies, and how ill a person was with COVID-19. A simple count of deaths in each group cannot separate the drug’s possible contribution from those differences.
A fair comparison was available in only a few studies
The researchers sought studies that compared patients within the same study: those previously treated with a checkpoint inhibitor and those who had not received one. They used adjusted results when available. Seven studies reported adjusted outcomes for at least one measure, and two more supplied enough individual patient data for the investigators to calculate adjusted results. Only eight studies contributed adjusted estimates for death.
Adjustment helps make groups more comparable, but it depends on which factors were measured and included. In eight of the nine studies with adjusted results for at least one outcome, models accounted for seven or fewer variables. Age, sex, high blood pressure, diabetes, heart disease, and chronic lung disease were among the factors commonly considered. The studies did not consistently capture the broader set of cancer and infection details needed to untangle the comparison.
Most of the evidence also came from observational research: 38 of the 42 studies included were retrospective, meaning investigators examined records of care that had already occurred. This allowed many reports to be assembled, but it did not make treatment assignment comparable to a controlled experiment. The review therefore describes associations in the available patient groups, not a causal effect of the drugs.
Severe illness and hospitalisation were not the same measure
“Severe events” combined different outcomes across studies. Depending on the report, the measure could include respiratory failure, other organ failure, sepsis, or intensive care admission. Five studies supplied adjusted results for these events. Their pooled odds ratio was 1.20, but the 95% interval ran from 0.30 to 4.74. That range is too broad to settle whether prior treatment raised or lowered the odds of severe illness.
Hospitalisation had an even thinner adjusted evidence base: only two studies reported adjusted results, and no patient-level data were available for a new adjusted analysis. The review combined those with unadjusted results from 13 other studies. Across all 15, it found no statistically clear association with hospitalisation. This pooled result is informative as a summary of reports, but it cannot substitute for a well-adjusted comparison.
These outcomes answer different questions. Death is a final outcome; the severe-event category gathered a variety of serious complications; hospitalisation reflects whether a patient was admitted. A finding for one measure would not automatically answer the others, and the evidence was limited in different ways for each.
Small treated groups blur possible differences
The number of patients who had received checkpoint inhibitors was strikingly small in many reports. Twenty-six studies included fewer than ten such patients, while only one included more than 100. A handful of outcomes in a group that small can leave an estimate highly uncertain, especially when researchers also need to account for several other factors.
Other gaps made the results harder to interpret. Follow-up duration or completeness was unclear in 18 studies, and the studies did not provide enough detail to assess specific checkpoint inhibitor regimens. Only four studies focused specifically on prior checkpoint inhibitor treatment; in 13 studies, the reviewers had to clarify how many patients had received checkpoint inhibitors and what happened to them. The evidence also raised concerns about bias, inconsistency, indirectness, and imprecision.
Some subgroup analyses raised possible exceptions. Death odds were higher in studies that included patients whose last treatment was at least 60 days before COVID-19 diagnosis, compared with studies limited to more recent treatment or with a shorter average interval. For severe events and hospitalisation, differences also appeared in some analyses involving treatment combinations or possible overlap in patient enrolment. Those patterns did not line up consistently across outcomes, and sensitivity analyses were inconclusive. They cannot establish a timing rule or show that a particular regimen changes COVID-19 risk.
For a person making decisions about cancer care, these findings are not a basis to stop, continue, or change treatment. They show why a broad comparison across small, varied patient groups cannot answer an individual treatment question.
What this could make possible
Imagine a person receiving a checkpoint inhibitor who develops COVID-19. If future studies could estimate risk specifically for this treatment while accounting for the person’s cancer and infection, a care team could discuss those estimates alongside the patient’s cancer needs and the severity of the infection. The missing piece is a dependable estimate for checkpoint inhibitor treatment itself, separate from other cancer therapies.
The review identifies what stronger comparisons would need to capture. A nearer milestone would be a substantially larger analysis that clearly separates checkpoint inhibitors from other cancer treatments, accounts for cancer type and stage, other therapies, and COVID-19 severity, and reports treatment-specific outcomes with more precise estimates. That would give clinicians and patients a firmer basis for discussing risk, while still leaving decisions grounded in the full clinical picture.
Key Sources
- Primary source: Comprehensive adjusted outcome data are needed to assess the impact of immune checkpoint inhibitors in cancer patients with COVID-19: Results of a systematic review and meta-analysis.
DOI or Journal Page · Review or Meta-Analysis · 2022-04-13
Keep exploring