But Did You See the Gorilla?

Twenty years ago, two cognitive psychologists implemented an experiment that asked participants to watch a video and count the number of times players on opposite teams, one in black shirts and one in white shirts, pass a ball between them. Midway through the video, a gorilla enters the game, stands in the middle, pounds his chest and exits. Then, study participants were asked, ‘…did you see the gorilla?’ More than half the time, subjects missed the gorilla entirely, and even after being told about it, they were certain they could not have missed it. According to the authors of The Invisible Gorilla [1], our intuition is that we will notice something visible that is distinctive, but that intuition is consistently wrong. While this story is frequently used to describe the perils of inattentional blindness, it also speaks to our human tendency to miss seeing the obvious when our attention is focused elsewhere. When it comes to preterm birth (PTB; defined as delivery before 37 weeks' completed gestation), the perinatal community is constantly watching for some exciting new gorilla—can we see the thing that will unveil the secret to reducing the relatively constant proportion of births delivered preterm within a given racial or ethnic group. Its causes have been the subject of research since the late 19th century, with an exploration into preterm delivery's physiologic mechanisms following shortly thereafter. Despite decades devoted to understanding and ameliorating PTB, it remains stubbornly resistant to interventions. Further, its individual-level risk markers are stable worldwide, and many appear non-manipulable. Within the wealthier Organisation for Economic Cooperation and Development (OECD) countries, non-white race or ethnicity, both young and advanced maternal age, and low educational attainment remain robust predictors for the risk of PTB [2]. The inability to meaningfully reduce the prevalence of PTB has encouraged investigators to consider other individual-level features that might help account for its persistence. Topics, including substance use, in vitro fertilisation, stress and access to prenatal care, have all been used to help explain why some pregnancies experience PTB while others do not. Researchers exploring more multidimensional social indicators, including social deprivation, have mostly concluded that material and social disadvantage measures, although variably constructed, are consistently associated with PTB. Into this long-standing conversation about the social predictors of PTB, Gottardi and colleagues [3] offer their findings from the French PreCARE study. The authors analysed the association between PTB, its phenotypes and their social deprivation index (which included indictors of social isolation, insecure housing, no income from work and absence of standard health insurance). They reported generally null findings for the relationship between social deprivation and preterm birth or its phenotypes (spontaneous labour, preterm prelabour rupture of membranes and placental vascular pathologies). The concept of social deprivation has a rich history in France, and various indicators have been used to estimate its presence and magnitude [4]. In this commentary, we take as a given that social deprivation is an important structural determinant of health, including pregnancy-related health and will note a few relevant topics that this paper enables us to consider. When we think about potential causes or correlates of adverse outcomes, we must interrogate one's substantive theory of the problem. The authors' Directed Acyclic Graph (DAG, eFigure 1) helpfully clarifies Gottardi et al.'s thinking about the role of social deprivation in the risk of PTB [3]. Their proposed causal structure is not atypical. Still, by indicating that maternal age, parity and education level contribute to one's social deprivation, it ignores the body of work suggesting that young maternal age [5], parity [6] and education [7] are all caused by social deprivation. Individuals from socially deprived families often begin bearing children at younger ages, consequently having more children, which further impedes educational attainment. In addition, one's social deprivation at birth is also predictive of their risk of social isolation, unemployment and insecure housing [8]. The intersection of low education, high parity, unemployment and insecure housing reduces the likelihood of upward socioeconomic mobility, essentially ensuring that socially deprived individuals remain so. Studies on socioeconomic mobility show a strong income persistence, or class inertia, in France compared to many other OECD studies. Specifically, about 10% of children from the lowest income quintile in France make it to the top 20% of wage earners [9]. Gottardi et al.'s DAG helps us think more carefully about the role of social deprivation on PTB, the correlations among the individual indicators of social deprivation, and whether their analytic approach is likely to capture the relationship the authors set out to explain. Given that maternal age, parity and education are mediators on the pathway from social deprivation to preterm birth, their approach may have led them astray. Investigators of the PreCARE study reported that the maternity units within the ‘North Paris University Hospital Group’ were located in a region with a high prevalence of social disadvantage [3]. This assertion, however, is not supported by their data, which indicated that two-thirds (66.3%) of the parents reported no indicators of social deprivation. Potentially, the intersection of high-area deprivation with a relatively advantaged sample may obscure the author's ability to identify any effects of social deprivation. The sample may not have met the author's expectations for their sample’s social standing from an area enriched for disadvantage, with 84% of the sample reporting experiencing zero (66.3%) or one (17.8%) of the deprivation criteria. These participants may be regional exceptions, which would likely make them different in other important ways related to their relationship with their deprived area of residence, their decision to participate in the study, or pregnancy outcomes. The limited variability of the social deprivation score among these participants would make it challenging to observe any effects of social deprivation on a relatively rare outcome like PTB. Alternatively, given the high proportion of participants reporting 0 or 1 indicators of social deprivation and their residence in a relatively disadvantaged area, their sample may represent a sizable number of deprived parents with health insurance but non-zero income. The 4-item measure may fail to capture other important dimensions of social deprivation. Therefore, many of those with a score of 0 or 1 may be struggling. Still, their exposure metric will have driven the effect estimate toward the null by failing to capture other relevant aspects of deprivation. One must also wonder about the validity of the constructed social deprivation variable applied in this study. The four measures used in the index were combined into a single score by giving them equal weight, with no justification offered for why these indicators should all contribute equally to PTB risk and no exploration in sensitivity analyses of different weighting schemes. Likely, there is effect measure modification occurring in this study as other work indicates that deprivation may interact with nativity or ethnicity in ways that enable deprivation to exert a more considerable influence on some pregnancy experiences than others. The power for evaluating this type of intersectionality is relatively limited with this cohort size, although new methods allow for exploring these interactions and could be pursued [10]. Among the most socially deprived (a score of 3), there were 44 preterm deliveries, and 61 PTBs occurred to participants with a score of 2 on the index. Given these numbers, the study was underpowered to observe all but implausibly large effects of social disadvantage. While the authors report the association between deprivation and PTB as null, their adjusted estimates are imprecise and do not rule out moderately sized deprivation effects. The upper limits of the adjusted CIs in Table 2 are all around 1.3, consistent with the range of modest associations seen in other studies. In Table 3, with stratification by PTB subtype, the upper CI limits are mostly 1.8–2.0 and are so imprecise as to preclude the identification of any realistic magnitude of association. Power for the sensitivity analyses with PTB defined before 35 and 32 weeks is even more hopelessly inadequate. We agree with Gottardi and colleagues that social deprivation is an important exposure warranting deeper investigation in the European context. Also, it likely has synergistic effects with area-level deprivation and other social determinants experienced during pregnancy [3]. We do not doubt that the social deprivation gorilla entered into this aetiologic scenario and beat its chest and that the authors, attending diligently to their adjustments and confidence intervals, failed to observe its effect on the pregnancies in this cohort. Their findings suggest renewed attention to exposure definition, cohort selection, sample size and statistical modelling. It does not exonerate social inequality in the aetiology of preterm birth because the absence of evidence is not evidence of absence. What this result does instead is to inspire us to refine our measures and designs in future work. Lynne C. Messer conceptualized and drafted the commentary. Jay S. Kaufman contributed text, editing, and intellectual collaboration. The authors declare no conflicts of interest. The authors have nothing to report.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.