Infections After Radiotherapy May Signal Poorer Survival in Glioblastoma Patients

Glioblastoma remains one of the most feared diagnoses in modern oncology. Even with maximal surgical resection, radiotherapy and temozolomide-based chemotherapy, median survival for most patients is measured in months rather than years, and the therapeutic arsenal has expanded only incrementally over the past two decades. Against this grim backdrop, clinicians have long noticed that patients with aggressive brain tumors frequently develop serious infections during or after their treatment, yet the microbiological character of these infections and their possible relationship to survival have never been systematically documented. A new retrospective cohort study from Akdeniz University Faculty of Medicine in Antalya, Turkey, published in BMC Cancer, now offers one of the most detailed pictures to date of what actually grows in the cultures of these patients, and how a documented infection relates to the time they have left.

The research team, led by radiation oncologist Rahmi Atıl Aksoy together with colleagues from radiation oncology and medical microbiology, followed 152 adults with glioblastoma who had completed a course of radiotherapy at their institution. Rather than relying on billing codes or broad clinical impressions, the investigators anchored their analysis to microbiological evidence: every infection counted in the descriptive part of the study was culture-positive, meaning that a pathogen had actually been isolated and identified in the laboratory. Patients were categorized over the twelve months following radiotherapy completion into three groups: those with a documented culture-positive infection, those who underwent a culture evaluation that came back negative, and those from whom no culture was ever obtained during the observation window.

The descriptive findings are the study’s most robust contribution and paint a strikingly consistent picture. Of the 152 patients, 48, or 31.6 percent, experienced at least one culture-positive infection within a year of finishing radiotherapy. Another 22 patients, 14.5 percent, had samples sent for culture that yielded no pathogen, while the majority, 82 patients or 53.9 percent, never had a culture obtained at all. Among the infections that were microbiologically confirmed, respiratory tract infections were the most common, accounting for 16 cases, closely followed by urinary tract infections with 15. This distribution is telling: rather than the brain abscesses or wound infections that might intuitively be associated with neurosurgery and irradiation of the head, the dominant infectious burden in this population arose in the lungs and the urinary tract, sites classically linked to hospitalization, catheterization, aspiration and declining performance status.

The microbiological profile reinforced that impression of a healthcare-associated flora. Of the 60 isolates recovered across the cohort, 41 were Gram-negative bacteria, a family of organisms that includes Escherichia coli, Klebsiella species and Pseudomonas aeruginosa and that is strongly associated with hospital environments, indwelling devices and broad-spectrum antibiotic pressure. Gram-negative organisms are also notorious for their resistance mechanisms, and the study’s focus on antimicrobial resistance patterns reflects a practical concern for clinicians: when a glioblastoma patient deteriorates with fever, the empirical antibiotics chosen at the bedside must cover organisms that are often far hardier than those circulating in the community. The predominance of respiratory and urinary sources, together with the Gram-negative skew, suggests that these infections are less a direct consequence of the tumor or its treatment of the brain itself and more a marker of the general frailty and healthcare exposure that accumulate as the disease progresses.

Beyond describing which microbes were present, the investigators asked a more provocative question: does a documented infection carry prognostic information? To address this, they moved beyond simple comparisons of groups defined at baseline, which are notoriously vulnerable to bias, and instead modeled infection status as a time-dependent covariate in a Cox proportional hazards framework. In practical terms, each patient’s status could change over the follow-up period: a person might begin in the no-culture-yet state, transition to a culture-positive or culture-negative state on the date a sample was collected, and contribute survival time in each state accordingly. This time-updated design prevents the classic immortal-time bias that arises when patients are classified by events that occur after the start of follow-up, and it allows the hazard of death to be estimated in relation to the current microbiological state rather than a fixed label assigned at the outset.

The results of that secondary analysis were dramatic. In the multivariable time-dependent model, adjusted for baseline variables that were significantly associated with overall survival in univariable screening, being in the culture-positive state was associated with a substantially higher hazard of subsequent death compared with the no-culture-yet state: an adjusted hazard ratio of 6.60, with a 95 percent confidence interval of 4.38 to 9.93 and a p value below 0.001. In other words, once a patient had a documented culture-positive infection, the rate of dying in the following period was more than six times higher than before such an infection was documented, after accounting for other measured prognostic factors. By contrast, a culture-negative evaluation, which indicates that a clinician suspected infection strongly enough to order cultures but no pathogen was recovered, was not significantly associated with mortality, with an adjusted hazard ratio of 1.65 and a confidence interval spanning 0.96 to 2.83.

The contrast between the two states is perhaps the most intellectually interesting aspect of the analysis. If the mere act of becoming sick enough to be sampled were driving the mortality signal, one might expect the culture-negative group to show a similar elevation, since those patients also presented with clinical concern sufficient to prompt microbiological workup. Instead, mortality in the culture-positive state was 4.01 times higher than in the culture-negative state, with a confidence interval of 2.22 to 7.22. This graded pattern, in which confirmed infection carries a far stronger association than suspected-but-unconfirmed infection, is consistent with the idea that the isolation of a true pathogen identifies a qualitatively more severe physiological insult, although the authors are careful not to overinterpret this hierarchy.

Recognizing that some deaths may have occurred within days of culture collection, potentially reflecting patients who were already moribund when sampled, the team performed a sensitivity analysis in which all state transitions were delayed by 30 days. This lagged design asks whether the association survives when deaths immediately following the sampling event are effectively excluded from the culture-positive exposure window. It did: the adjusted hazard ratio for the culture-positive state attenuated from 6.60 to 3.36, with a confidence interval of 2.09 to 5.39, but remained highly significant. The persistence of the signal after lagging argues against the crudest form of reverse causation, in which the infection is merely a terminal event rather than a contributor to decline, though it cannot eliminate subtler forms of the same problem.

The authors themselves are admirably explicit about the limits of their prognostic finding. Because cultures were obtained on clinical indication rather than on a fixed schedule, the sampling process itself is entangled with patient condition: sicker patients are more likely to be tested, and the decision to test is made by clinicians responding to deterioration. Residual confounding by performance status, steroid use, disease progression and other unmeasured factors remains possible, and reverse causation, in which advancing tumor drives both infection susceptibility and death, cannot be excluded in a retrospective design. For these reasons, the study’s conclusion frames the survival association as hypothesis-generating rather than causal. A documented culture-positive infection after radiotherapy should be understood as a marker of increased subsequent mortality, not as a proven independent killer.

Even with those caveats, the study has clear practical implications. For radiation oncologists and neuro-oncologists, the finding suggests that a confirmed infection in the months after chemoradiation is a meaningful clinical event that should prompt renewed attention to supportive care, medication review and candid conversations about goals of treatment, rather than being dismissed as an incidental complication. For microbiologists, the predominance of Gram-negative, respiratory and urinary, frequently healthcare-associated organisms provides an evidence base for empirical antibiotic choices in this population. And for researchers, the work establishes a methodological template: time-dependent modeling with lagged sensitivity analyses offers a rigorous way to interrogate the prognostic weight of complications that occur on a rolling timeline. Whether infection is purely a mirror of tumor progression or an active participant in it remains an open question, but this study ensures that the question will now be asked with far better data behind it.

Subject of Research: Microbiological profile of post-radiotherapy infections and their prognostic significance in glioblastoma

Article Title: Microbiological profile and prognostic significance of infections following radiotherapy in glioblastoma

Article References: Aksoy, R. A., Şahiner, M. B., Kahya, U. K., Koca, Ö., Koca, T., & Korcum, A. F. (2026). Microbiological profile and prognostic significance of infections following radiotherapy in glioblastoma. BMC Cancer. https://doi.org/10.1186/s12885-026-17033-8

Image Credits: AI Generated

DOI: 10.1186/s12885-026-17033-8

Keywords: glioblastoma, radiotherapy, infection, microbiology, Gram-negative bacteria, antimicrobial resistance, overall survival, time-dependent Cox regression, prognostic markers, respiratory infections, urinary tract infections, healthcare-associated infections