Birth size and cancer prognosis: a systematic review and meta-analysis

Abstract There is an established link between birth parameters and risk of adult-onset cancers. The Developmental Origins of Health and Disease concept provides potential underlying mechanisms for such associations, including intrauterine exposure to endogenous hormones (androgens and estrogens), insulin-like growth factors, etc. However, there is conflicting evidence on the association between birth parameters and the cancer mortality risk. Therefore, we aimed to review and analyse the available data on the association linking birth weight and birth length with cancer mortality. Eleven studies were identified, published until April 2019. A significant association between birth weight and the prognosis of cancer (overall) was found (relative risk, RR 1.06, 95% confidence interval, CI: 1.01, 1.11), with low heterogeneity (I2 = 27.7%). In addition, higher birth weight was associated with poorer prognosis of prostate cancer (RR 1.21, 95% CI: 1.02, 1.44). However, the association of birth weight with breast cancer mortality risk in women was not significant (RR 1.16, 95% CI: 0.93, 1.44), which might be due to high statistical heterogeneity (I2 = 67.9%). Birth length was not associated with cancer mortality risk (RR 1.0, 95% CI: 0.90–1.11). It might be inferred that birth parameters are not associated with cancer mortality as strongly as with the risk of developing cancer. Also, the association between birth parameters and cancer mortality risk is not uniform and varies according to its subtypes, and study characteristics/design. This highlights the need for further prospective studies.


Introduction
Substantial evidence implicates the crucial role of early life factors in the occurrence of adult-onset diseases, including cancers. 1 The results of many studies suggest that birth parameters, a proxy for cumulative intrauterine exposures, are associated with cancer risk in adulthood. 2 These findings are plausibly explained by the Developmental Origins of Health and Disease hypothesis, which suggested that fetal adaptive strategies to the adverse intrauterine environment generate long-term consequences for poor health. 3 Birth weight in relation to gestational age is one of the most intensively studied birth parameters in predicting adult cancers' risk. 2,4 Of all the widely prevalent cancers, there are convincing results linking higher birth weight with increased risk of incident breast-, prostate-and testicular cancer. 5 However, less attention has been given to explore the effects of birth parameters in predicting prognosis (survival or mortality) among adult cancer patients. A meta-analysis, determining associations between birth weight and various causes of adult mortality, concluded that there is a strong positive association of birth weight with cancer mortality among men, but not in women. This impressive literature is limited by the bias of incomplete identification of studies, the number of included studies (only five) and the study of birth weight as the only birth parameter. 6 Given that many high-quality articles investigating the potential effect of birth parameters on cancer prognosis have not been included in this review, and that there is conflicting evidence on such associations from the existing literature, we conducted the systematic review and meta-analysis. Thus, the aim of this systematic review and meta-analysis was to determine whether birth size is associated with prognosis of cancer in adults.

Research question
Are markers of birth size at delivery (birth weight, birth length and head circumference) related to the prognosis of adult-onset cancers?

Search strategy
We systematically searched MEDLINE (Pubmed), EMBASE and Web of Science from inception until the end of April 2019 to identify observational studies that have reported on associations between birth parameters (birth weight, birth length and head circumference) and prognosis of any adult-onset cancer, using overall or cancer-specific mortality as defined by the authors. Besides, reference lists and citation indices of the included studies were also searched. Inclusion was limited to studies, which reported cancer mortality in patients aged more than 20 years. All the studies that met the inclusion criteria were included irrespective of the study site, hospital or community-based settings, or sample size. For meta-analysis, the studies that have reported risk estimates and the confidence intervals (CIs) of the association between birth weight or birth length and adult-onset cancer mortality (overall or specific type) were included. We did not include conference abstracts or unpublished or grey literature, anecdotal reports, or case series, or manuscripts published in any language other than English.
Searches were performed by two reviewers (SS and CK) independently. Keywords used for search engine were ('Birth weight' OR 'Birth size' OR 'Birth parameters' OR 'Birth length' OR 'Gestational age' OR 'Head Circumference') AND (cancer OR neoplasm OR carcinoma OR 'Neoplasms' OR tumor OR tumour) AND ('prognosis' OR survival OR mortality). We included studies of all designs (case-control, cohort, nested case-control, and longitudinal cohort studies). We contacted corresponding authors of two studies, which did not report risk estimates for the associations between birth weight and cancer prognosis.

Data extraction
Two reviewers (SS and CK) independently assessed titles and abstracts to select articles based on a-priori established in-and exclusion criteria. In case of discordance between the two reviewers, the consensus was made with the help of other coauthors. All the search results were imported into Endnote (X9.0.1) and duplicates, if any, were removed. In case of overlap of the study population from two or more papers, the most recent one was selected. When multiple studies were published with a substantial geotemporal overlap in the cohort population, the most recent publication was selected.

Quality assessment
One of the reviewers (SS) performed the quality assessment of the papers using the Newcastle-Ottawa Scale for cohort and casecontrol studies. 7 The scale uses a star system to evaluate the cohort and case-control studies and a maximum of eight stars could be achieved.

Statistical analysis
For studies reporting results with categorical birth weight or birth length, we used the most fully adjusted risk estimates and 95% CI for the highest versus lowest reported (reference) category. Hazard ratios and other reported risk ratios were considered interchangeably. For studies reporting results with continuous birth weight or birth length, we used the reported most fully adjusted risk estimates and their CI per unit of increase, respectively, 0.5-1 kg increase in birth weight, or per cm increase in birth length and their standard errors for all the studies. The pooled and weighted relative risks were calculated using a random effects meta-analytic model, and statistical heterogeneity was assessed by means of the I 2 statistic, with values <50% considered as low statistical heterogeneity. Subgroup analyses were done by cancer type (overall, breast or prostate), exposure (categorical or per unit of increase) and for both sexes separately if feasible.
Publication bias was assessed by generating a funnel plot and assessing its symmetry. Because of the relatively low number of studies, further statistical testing (e.g., Egger's or Begg's test) was considered unreliable. The statistical analyses were performed using Stata version 14 (Stata Corp., College Station, TX, USA).

Results
The Preferred Reporting Items for Systematic Reviews and Metaanalyses (PRISMA) flow chart for the systematic review is shown in Fig. 1, resulting in 11 eligible articles, of which 8 reported the association between birth weight and cancer mortality only, 3 on both birth weight and birth length, and none on head circumference. The main characteristics of these studies are presented in Table 1. All the studies scored six or more stars on the Newcastle-Ottawa Scale. Among 11 studies, there were 7 cohort studies and 4 case-control studies. [8][9][10][11][12][13][14][15][16][17][18] Seven out of 11 studies were cohort-based by design [9][10][11]13,14,16,17 and 4 were case-control studies. 8,12,15,18 Ten of these 11 included studies used data on birth weight from hospital archives, medical records or registers, 8,[9][10][11][13][14][15][16][17][18] and 9 studies were from Nordic countries. 8,9,11,[13][14][15][16][17][18] However, only three studies were identified for associations between birth length and cancer prognosis. 8,11,15 Two studies were excluded from the meta-analysis because one did not report the risk estimates of the association between birth weight and death due to prostate cancer, and the other did not specify the risk estimate for the mortality due to neoplasms in adulthood. 17,18 The authors of the studies were contacted, however, they did not respond.
The average age at death due to cancers among studies was low (51.5, 58.8 years). 11,14 The median years of follow-up of the cancer cases post-diagnosis ranged between 8 and 11 years in the studies. 13,15,16 Birth weight and prognosis of breast cancer All the studies except one 10 reported poorer breast cancer prognosis in individuals with a larger birth weight. 11,12,15,16 However, the associations did not reach statistical significance in all but the study by Sovio et al. 16 Two of these studies from Norway and USA were nested case-control studies 12,15 and the remaining studies from Sweden, Finland and England had a cohort study design. 11,16 Birthweight was adjusted for gestational age in two of the five studies. 12,15 Birth weight and prognosis of prostate cancer Of the five studies reviewed, 8,10,11,13,18 all suggested a positive association between larger birth weight and mortality due to prostate cancer, only one of which reached statistical significance. 11 The study by Gerdtsson et al. did not show the risk estimates, but revealed that the birth weight was not associated with the mortality due to prostate cancer. 18 Three of the five studies were from different counties in Sweden (Malmö, Gothenburg, and Uppsala) 8,13,18 and the other two were from Finland and England. 10,11 The studies from Sweden together cover a wide range of time period, including individuals born before 1900. Ekbom et al. adjusted for a large
number of confounders including socio-economic status, history of pre-eclampsia in mother, etc. 8 Eriksson et al. reported that the association between birth weight and mortality due to prostate cancer is not age-dependent. 13 Birth weight and prognosis of cancer (overall) All four included studies of birth weight and cancer mortality among men and women reported that higher birth weight is associated with increased cancer mortality. [9][10][11]14 Such associations were statistically significant for men in two of the four studies. 9,10 All the included studies were cohort by design and from different countries (England, Denmark, Sweden, and Finland).

Meta-analysis of birth weight and cancer mortality
Risk estimates were pooled from nine studies for the risk of mortality from cancers per kg increase in birth weight using the random effects model (Fig. 2). We found that birth weight was associated with a small increased risk of overall mortality (relative risk, RR 1.06, 95% CI: 1.01, 1.11), and the heterogeneity among studies was low (I 2 = 27.7%). Further, subgroup analyses of the cancer types (breast and prostate cancers) showed different associations with varying amounts of heterogeneities. Higher birth weight was associated with a 21% increased risk of prostate cancer mortality (RR 1.21, 95% CI: 1.02, 1.44) and the heterogeneity was low (I 2 = 15%). On the contrary, the association of birth weight with breast cancer was not significant (RR 1.16, 95% CI: 0.93, 1.44), with a high amount of heterogeneity (I 2 = 67.9%).

Birth length and prognosis of cancers
The relationship between birth length and cancer mortality was not uniform across studies. 8,11,15 Moreover, the association was statistically significant only in one of the three studies. 15 Maehle et al.
in his study suggested that higher birth length was associated with poorer prognosis of breast cancer. 15 The three studies were from different Nordic countries (Finland, Sweden and Norway).

Meta-analysis of birth length and cancer mortality
We separately conducted a meta-analysis of the three studies for mortality per cm increase in birth length (Fig. 3). We found that birth length was not associated with cancer mortality (RR 1.0, 95% CI: 0.90, 1.11), which might be due to the fact that the unit of increase may be too small to have sufficient power to detect a difference.

Discussion
This systematic review and meta-analysis included data on nearly 9601 adult deaths among some 311,970 individuals. Our study suggests that higher birth weight is associated with an increased cancer overall mortality. The association of birth weight with cancer mortality risk was stronger for prostate compared to breast cancer. Birth length has not been found to be associated with cancer mortality.
The results are compatible with the findings from another meta-analysis of five studies relating birth size with cancer mortality. 6 The study suggested 13% increased risk of mortality due to cancers (Hazard Ratio, HR 1.13, 95% CI: 1.07,1.19) for   3. Birth length (per cm) in relation to risk of cancer mortality (overall (n = 1), prostate (n = 1) and breast (n = 1)) using a random effects model.
men and 4% increased risk (HR 1.04, 95% CI: 0.98, 1.10) among women (P interaction = 0.03). 6 Contrary to our findings, Belbasis et al. observed no convincing evidence supporting association between high birth weight and later health outcomes, including cancers. Instead, they suggested the associations of low birth weight with increased risk for all-cause mortality. 2 Our findings and the results from a meta-analysis by Zhou et al. on prostate cancer risks (2015) together suggest the hypothesis that higher birth weight is a risk marker in the entire trajectory of prostate cancer, from incidence to metastasis to mortality. 19 The study provided evidence that heavier birth weight may be associated with modest increased risks of total and aggressive/lethal prostate cancer. 19 The differential associations of birth weight with the risk of cancer mortality (overall) and the mortality of its subtypes highlight that the relationship is not uniform across cancers. Prostate and breast cancers are among the few cancers, which have been studied extensively for their relations with birth parameters. 5 It has been established that higher birth weight and birth length are associated with an increased risk of prostate and breast cancer incidence. However, the associations of birth parameters with cancer prognosis are not as strong and significant compared to their associations with cancer incidence. 19,20 It is well known that birth weight and gestational age are highly correlated. 2 Despite that, a few studies have adjusted birth weight for gestational age while assessing associations with cancer mortality risk. 11,16,17 In addition, only a few studies have adjusted for socio-economic factors, 16 which are potential confounders linked with both exposure (birth weight) and the outcome (cancer mortality). 21 Some of the studies have a very long period of follow-up of the cancer cases post-diagnosis, which minimises the bias of missing potential cases because cancer-related mortalities occur late in life. This provides an advantage over studies with a shorter period of follow-ups. 22 The sex-segregated analysis reveals stronger and significant association of birth weight with cancer mortality risk among men than women. This is in agreement with the findings from a study by Risnes et al., which reported stronger positive associations for men than women. In addition, significant association of birth weight with prostate cancer risk compared to breast cancer risk in our results supports this statement. 6 Although multiple studies have shown associations between head circumference and the risk of brain, breast and colorectal cancer incidence, 23,24 we did not find any study assessing the association with cancer survival.
The strength of our study includes thorough systematic search of the studies and addition of a novel research to the literature. Our meta-analysis has several limitations. Firstly, there was a moderate amount of heterogeneity among studies as a result of which a definitive association could not be concluded. This could be due to different study designs, and different outcomes of the study (overall cancer mortality or specific cancer mortality). Secondly, we used the published, aggregated risk estimates and our meta-analysis was not based on individual participant data (IPD). IPD allow more powerful and uniformly consistent analyses as well as better characterisation of subgroups and outcomes, compared to those which are based on aggregate data extracted from published risk estimates. 25 All the studies in our analysis were from high-income countries, which limit the generalisability of the results. Despite clinical and methodological heterogeneity, statistical heterogeneity was limited, Fig. 4. Birth weight in relation to risk of cancer mortality segregated by sex using a random effects model. and associations were found between birth weight and cancer prognosis. However, residual confounding cannot be ruled out, and the findings may not be generalised to less highly developed countries. Lastly, we could not exclude the possibility of publication bias, given that only nine studies published results for the associations between birth size and risk of cancer mortality.
In conclusion, the systematic review and meta-analysis suggest that higher birth weight is associated with increased overall cancer mortality and prostate cancer mortality. The effect of birth weight on the risk of mortality due to breast cancer is not convincing. This suggests the need for prospective studies and elaborative research on birth cohorts for conclusive results on such associations.