Objective To test whether the assessment of growth in very low birth weight infants during the hospital stay using z-score differences (Zdiff) is confounded by gestational age (GA), birth weight percentiles (BW%ile), and length of the observation period (LOP). We hypothesize that Zdiff calculated from growth charts based on birth weight data introduces a systematic statistical error leading to falsely classified growth as restricted in infants growing similarly to the 50th percentile. Methods This observational study included 6,926 VLBW infants from the German Neonatal Network (2009 to 2015). Inclusion criterion was discharge between 37 and 41 weeks postmenstrual age. For each infant, Zdiff, weight gain velocity, and reference growth rate (50th percentile Fenton) from birth to discharge were calculated. To account for gestational age dependent growth rates, assessment of growth was standardized calculating the weight gain ratio (WGR) = weight gain velocity/reference growth rate. The primary outcome is the variation of the Zdiff-to-WGR relationship. Results Zdiff and WGR showed a weak agreement with a Zdiff of -0.74 (-1.03, -0.37) at the reference growth rate of the 50th percentile (WGR = 1). A significant proportion (n = 1,585; 23%) of infants with negative Zdiff had weight gain velocity above the 50th percentile's growth rate. Zdiff to WGR relation was significantly affected by the interaction of GA x BW%ile x LOP. Conclusion This study supports the hypothesis that Zdiff, which are calculated using birth weights, are confounded by skewed reference data and can lead to misinterpretation of growth rates. New concepts like individualized growth trajectories may have the potential to overcome this limitation.