Birthweight at birth is inherently retrospective and therefore primarily informs postnatal surveillance and counselling rather than antenatal triage. This study was not designed to redefine clinical thresholds, but to characterize how perinatal mortality varies across the continuous distributions of BW-for-GA and GA, supporting population-level interpretation and communication of graded mortality patterns. We found that perinatal mortality does not show discrete shifts at conventional SGA and LGA cutoffs, challenging the often implicit assumption that dichotomous thresholds reflect abrupt changes in outcome. While infants at the extremes of birthweight experience higher mortality, these patterns are continuous and GA–dependent. Continuous, GA–specific analyses reveal a smooth mortality gradient across the full birthweight distribution, with a consistent nadir around the 80th–83rd percentile. Although the results pertain specifically to perinatal mortality, they may generate hypotheses for other neonatal or long-term outcomes that warrant further study.

Our results are in line with previous studies. Vasak et al. found the lowest perinatal mortality between the 80th and 90th percentiles [9], and another study reported the lowest mortality around 1SD above the mean (≈84th percentile), despite increases in mean birthweight and decreases in absolute mortality rate over time [18]. In very preterm births, weight between the median and 85th percentile predicted the highest survival [19]. For risk stratification, the focus is generally on populations at increased risk. Consistent with other studies reporting relative risk, we observed the highest perinatal mortality rates in the smallest infants [10, 20, 21] Boulet et al. highlighted that defining SGA without reference to gestational-age-specific mortality is limited; they developed risk curves based on birthweights associated with 2-, 2.5-, and threefold increased neonatal death compared with infants in the 45th–55th percentile [22]. In our study, the 10th percentile did not represent a distinct inflection point in mortality. When infants with median birthweight were used as the reference population, the 10th percentile corresponded to a twofold increase in mortality only in term infants. When compared with the gestational-age-specific nadir of mortality (80th–83rd percentile), the 10th percentile was associated with an approximately twofold increase in perinatal mortality from 26 weeks’ gestation onward; however, similar elevations were observed across several adjacent percentiles, extending up to approximately the 25th percentile depending on GA (results not shown). These findings underscore that the perceived risk with any given percentile is strongly influenced by both the choice of reference point and the characteristics of the reference population, highlighting the need for careful interpretation of percentile-based thresholds in perinatal risk stratification.

Absolute and relative measures of mortality provide complementary perspectives on clinical relevance. While rate ratios were highest in (post-)term infants, the absolute burden of being SGA was much greater in preterm infants. When the baseline mortality rate is high, even a rate ratio close to 1 may reflect a clinically meaningful effect, whereas large rate ratios may not always translate into significant outcomes. Clinical relevance also depends on the feasibility of preventing the adverse outcome, determined not only by the potential effectiveness of an intervention, but also by available resources and clinical capacity. For example, mortality rates increased after 39–40 weeks, particularly among fetuses with abnormal growth; yet such growth abnormalities may be difficult to identify in late pregnancy [23]. In several countries, there is a growing trend toward offering elective induction of labor at 39 weeks to mitigate this risk [24]. Nevertheless, despite substantial increases in induction rates, population-level stillbirth and perinatal mortality rates have not shown clear improvement [25].

Our primary outcome, perinatal mortality, includes stillbirths, intrapartum fetal deaths, and early neonatal deaths, which is why we refer to rate rather than risk. Estimating perinatal mortality risk by GA and birthweight percentile is not feasible, as fetal weight is unknown until birth [26]. Size at birth serves as a retrospective proxy for fetal growth, with clinical utility mainly for population surveillance and counselling rather than antenatal triage, which relies on estimated fetal weight-based charts and multiple risk factors. Most deceased infants (69.9%) were antepartum stillbirths, for which birthweight and GA may be unreliable due to the timing of fetal demise and the degree of maceration. Fetal weight may decrease in utero over time [27], and GA at delivery is influenced by the timing of fetal death, potentially leading to misclassification of growth status. To assess the impact of these issues, we repeated analyses excluding antepartum stillbirths. The overall patterns remained consistent, confirming that our main conclusions are robust and not materially affected by these limitations.

We observed the highest mortality rates at both extremes of the birthweight percentile distribution, though mechanisms leading to adverse outcomes in LGA infants (e.g., shoulder dystocia, cesarean delivery) likely differ from those affecting SGA infants. While adverse outcomes in term LGA infants are often linked to macrosomia (> 4000 g) [28] and labor complications, we observed twofold increased rate ratios in LGA infants as early as 29 weeks’ gestation. At this gestation, the 99th percentile corresponds to 1842 g for boys and 1781 g for girls [14], suggesting that birthweight alone does not explain the elevated mortality. Similarly, Baer et al. reported increased neonatal mortality in preterm LGA infants (28–31 weeks), independent of maternal diabetes (adjusted RR 2.1; 95% CI 1.5–2.9) [29]. These observations suggest that preterm LGA infants may face distinct risks that warrant further investigation.

The principal strength of our study was the use of a very large, nationwide cohort of infants to investigate the distribution of perinatal mortality. Gestational age and birthweight percentile were analyzed as continuous variables, avoiding dichotomization or categorization into strata [7, 9, 10, 20]. To facilitate a nuanced interpretation and understanding of clinical relevance, we reported both absolute and relative measures of perinatal mortality. Additionally, a sensitivity analysis confirmed the robustness of our findings.

We acknowledge several limitations, most inherent to our retrospective design. Extreme outliers with implausible combinations of GA and birthweight were excluded. Their perinatal mortality rate was more than 20 times higher than the study population and was not limited to antepartum stillbirths, suggesting that some may reflect true pathology rather than random data errors [30]. Excluding these deaths may have led to underestimation of mortality at the extremes of the birthweight distribution. Birthweight at birth is an imperfect proxy for fetal growth, and residual confounding, management decisions around viability thresholds, and possible registry misclassification may have influenced gestational-age-specific patterns. Our database lacked sufficient clinical detail to determine causes of death, so we could not assign specific causes, potentially overestimating absolute perinatal mortality attributed to suboptimal growth. Nevertheless, there is little reason to suspect that mortality rate ratios were differentially affected, as all-cause mortality was included in both the numerator and denominator [22]. Finally, our data were collected between 2000 and 2015, during which the overall perinatal mortality rate decreased from 5.9 in 2000 to 3.1 in 2015. Temporal trends had minimal effect on rate ratios (results not shown) [18]. Importantly, our findings are descriptive and should not be interpreted as estimates of individual risk. Absolute mortality rates should be interpreted with caution in light of these considerations.

Worldwide, birthweight charts are updated periodically, yet, despite major differences in methodology, the thresholds to define abnormal birthweight are rarely questioned. Although infants at the extremes of the birthweight distribution do have higher mortality rates, this risk is neither constant across GA nor as clear-cut as our traditional SGA and LGA definitions suggest. Fixed birthweight cut-offs, such as the 10th and 90th percentiles, remain pragmatic; however, continuous GA-specific risk curves reveal graded, GA-dependent mortality that a single cut-off cannot capture. These fixed thresholds can exaggerate differences at the extremes and mask gradual changes in risk, underscoring the need for a more nuanced assessment. The aim of our study was neither to predict mortality nor to define an optimal cut-off for risk stratification. Its primary value lies in illustrating the continuous, GA-specific effects of birthweight percentile on perinatal mortality. A more nuanced understanding of these associations, and of the limitations of applying absolute thresholds, may help clinicians interpret risk more thoughtfully and guide decisions about the need for additional evaluation or monitoring. Future research should ideally be prospective and focused on other potentially preventable adverse perinatal outcomes, to determine whether clinically relevant cut-offs exist and ultimately improve patient outcomes.