Long-acting G-CSF primary prophylaxis for chemotherapy-induced febrile neutropenia: a systematic review of randomized trials and meta-analysis of first-cycle outcomes

The estimated benefit of long-acting granulocyte colony-stimulating factor as primary prophylaxis against chemotherapy-induced febrile neutropenia may vary with the outcome period and the treatments received by control participants. We assessed its effect during the first chemotherapy cycle while preserving the randomized comparisons of the included trials. We systematically reviewed randomized trials comparing long-acting granulocyte colony-stimulating factor with placebo or no planned primary prophylaxis. The primary random-effects synthesis required exact arm-level Cycle 1 febrile neutropenia counts, a maintained randomized contrast, and a protocol-defined fever trigger of at least 38.0 °C. This restriction was introduced after registration and before the arm-level event counts were inspected. Results from longer observation periods were summarized separately. Sparse-event sensitivity analyses varied the continuity correction and used Mantel–Haenszel pooling. Eleven trials met the review eligibility criteria, and four reported extractable Cycle 1 febrile neutropenia data. Three comparisons involving 595 participants met the additional fever-trigger criterion. Febrile neutropenia occurred in 10 of 381 participants receiving long-acting granulocyte colony-stimulating factor and 21 of 214 controls; the pooled risk ratio was 0.270 (95% confidence interval 0.129–0.566; I² = 0%). Adding Kosaka et al. yielded a risk ratio of 0.113 (95% confidence interval 0.023–0.569; four comparisons; I² = 77.0%). The certainty of the three-comparison Cycle 1 evidence was low. Low-certainty evidence suggests that long-acting granulocyte colony-stimulating factor as primary prophylaxis may reduce febrile neutropenia during the first chemotherapy cycle. Differences in febrile neutropenia definitions, chemotherapy regimens, and subsequent granulocyte colony-stimulating factor use limit the applicability of this estimate to other outcome periods or individual risk thresholds. PROSPERO, CRD420261371111.