Objective
This study was undertaken to explore the association between maternal coffee consumption during pregnancy and childhood acute leukemia (AL).
Study Design
The PubMed database was used to search studies up to May 5, 2013, and the lists of references of retrieved articles were also screened to identify additional relevant studies. Studies were included if they reported the odds ratio and corresponding 95% confidence interval (CI) of childhood AL, including childhood acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML), with respect to maternal coffee consumption during pregnancy.
Results
Compared with non/lowest drinkers, the combined odds ratio regarding the relationship of maternal coffee consumption during pregnancy and childhood AL was 1.22 (95% CI, 1.04–1.43) for ever drinkers, 1.16 (95% CI, 1.00–1.34) for low to moderate-level drinkers, and 1.72 (95% CI, 1.37–2.16) for high-level drinkers. When analysis was conducted by subtypes of childhood AL, maternal coffee consumption (high-level drinkers vs non/lowest drinkers) was statistically significantly associated with childhood ALL (1.65; 95% CI, 1.28–2.12) and childhood AML (1.58; 95% CI, 1.20–2.08). We observed the linear dose-response relationship of coffee consumption and childhood AL ( P for nonlinearity = .68), including childhood ALL and childhood AML; with increased coffee consumption, the risk of childhood AL increased.
Conclusion
The findings of the metaanalysis suggest that maternal coffee consumption during pregnancy may increase the risk of childhood AL. Because of limited studies, further prospective studies are urgently needed to explore the adverse effect of coffee consumption on childhood AL.
Childhood cancer has been a major global health issue. Each year, nearly 100,000 children age <15 years die of cancer, >90% of them in resource-limited countries. Acute leukemia (AL) is one of the most common childhood cancers, and morbidity has been increasing in Europe and the United States in past decades. Among AL, acute lymphoblastic leukemia (ALL) was the most common subtype, followed by acute myeloid leukemia (AML), also called acute nonlymphoblastic leukemia.
Despite decades of research, the cause of childhood AL remained unclear with the exception of radiation exposure and a few chromosomal and genetic abnormalities. The attention of recent research on finding risk factors of childhood AL has been focused on parental environmental exposures during pregnancy. Factors such as maternal alcohol consumption and smoking during pregnancy have been extensively explored. According to 2 recent metaanalyses, positive association was found between childhood leukemia and maternal alcohol during pregnancy, but not maternal smoking.
Among environmental exposures, the impact of maternal coffee consumption during pregnancy on childhood AL was another research hotspot. Coffee, one of the most widely consumed beverages in the world, contains caffeine, which may result in childhood AL. For instance, caffeine may act as a topoisomerase (topo) II inhibitor, a DNA repair inhibitor or a carcinogen metabolism inhibitor. It induces chromosomal aberrations and translocations, such as abnormalities of chromosome 11q23, which was taken as a cause for the pathogenesis of infant leukemia. However, the results based on epidemiological studies on the association between coffee consumption during pregnancy with childhood AL were inconsistent. Although the risk estimates concerning the association of maternal coffee consumption during pregnancy and childhood ALL from previous published studies were pooled by Milne et al in their original paper, but only 2 published studies were included and the impact of coffee consumption during pregnancy on childhood AL and childhood AML remains unknown. Therefore, we aimed to conduct a comprehensive metaanalysis of identified studies to evaluate the relationship of maternal coffee consumption during pregnancy and childhood AL and its subtypes (ALL, AML).
Materials and Methods
Search strategy
The PubMed database was searched for studies up to May 5, 2013, with the following key words: “coffee,” “caffeine,” “beverage,” “leukemia,” and “AL.” All subterms were included and no language restrictions were imposed. In addition, the lists of references in previous studies (including reviews) were also screened to identify additional relevant studies.
Study selection
We included the identified studies that meet the following criteria: (1) case-control or cohort study design; (2) main exposure of interest was coffee consumption during pregnancy; (3) outcome of interest was childhood AL, ALL, or AML (or acute nonlymphoblastic leukemia); (4) targeted population was children <18 years; and (5) they provided the odds ratio (OR) and corresponding 95% confidence interval (CI) (or data allowed to recalculate them).
Data extraction
Data from identified studies were extracted independently by 2 investigators using a standardized data collection form, and then compared. For each included study, the data extracted were the following: first author’s last name, year of publication, location, study period, children’s age range, the number of cases and controls, type of leukemia, level of coffee consumption during pregnancy, effect estimates (OR and 95% CI), adjustment for potential confounding factors, and assessment methods of coffee consumption (interview or self-administrated).
Statistical analysis
We extracted study-specific risk estimates and calculated the summary estimates by combining log risk estimates weighted by the inverse of their variances. First, we compared the risk of childhood AL in ever drinkers with never/lowest drinkers. Several studies did not provide the risk of childhood AL for ever drinkers. For those studies, a summary OR was calculated for ever drinkers using the reported risk estimates for each coffee consumption level. Then the summary OR was used in final metaanalysis when pooling the overall risk of childhood AL for ever drinkers.
Second, to estimate the summary OR for various levels of coffee consumption during pregnancy, we also calculated the study-specific estimates for low to moderate coffee consumption (defined as >0 to ≤3 cups/d for Menegaux et al ; >0 to ≤1 cup/d and >1 to ≤2 cups/d for Bonaventure et al ; >0 to <3 cups/d for Clavel et al ; >0 to <2 cups/d for Milne et al ; >0 to <3 times/wk for Ross et al ) and for high coffee consumption (defined as >4 to ≤8 cups/d and ≥8 cups/d for Menegaux et al ; >3 cups/d for Menegaux et al ; >2 cups/d for Bonaventure et al ; ≥3 for Menegaux et al ; ≥2 cups/d for Milne et al ; ≥4 times/wk for Ross et al ). Statistical heterogeneity between studies was evaluated with Cochran Q and I 2 statistics with corresponding P value and P < .1 was considered statistically significant. Fixed effect model was used to calculate the pooled OR if no heterogeneity was observed, otherwise, random effect model was used.
Since different levels of coffee consumption were analyzed in different studies, we also conducted a dose-response metaanalysis to explore the pooled dose-response relationship between coffee consumption during pregnancy and childhood AL. Studies included in final dose-response had to meet the following criteria: (1) at least 3 levels of coffee consumption, including the reference level were presented; and (2) number of cases and controls, and OR (for case-control study) with corresponding 95% CI for each level of coffee consumption were reported. The median or mean coffee consumption for each level was assigned to each corresponding OR for each study. The midpoint of the upper and lower boundaries in each level was assigned as the average intake if the median or mean coffee consumption for each level was not provided. If the upper boundary of the highest level was not provided, we assumed that the boundary had the same amplitude as the adjacent category.
To examine the nonlinear relationship between coffee consumption during pregnancy and childhood AL, a 2-stage random effects dose-response metaanalysis was performed, taking into account the heterogeneity among studies recently proposed by Orsini et al. Briefly, for each identified study contributing to this metaanalysis, the log OR was calculated for various levels of coffee consumption. Then a restricted cubic spline model, with 4 knots at percentiles (5%; 35%; 65%; 95%) of levels of coffee consumption, was estimated using generalized least square regression, taking into account the correlation within each set of published ORs. The study-specific estimates were combined using fixed effects restricted cubic spline models with 4 knots and using the method of Greenland and Longnecker to estimate the covariances of OR. If P < .05, the nonlinearity dose-response relationship was considered to exist.
We conducted stratified analysis by the subtype of childhood AL and investigated the following potential sources of heterogeneity: assessment methods of coffee consumption and study population. Meanwhile, publication bias was also assessed by constructing funnel plots and by Egger regression asymmetry test. P < .05 was considered representative of statistically significant publication bias. All statistical analyses were conducted using software (Stata, version 11; StataCorp, College Station, TX).
Results
A total of 7 published studies ( Figure 1 ) regarding the relationship between coffee consumption during pregnancy and childhood AL were included in our metaanalysis. All the studies were case-control design ( Table 1 ), of which 4 were conducted in France, 1 in Australia, 1 in Greece, and 1 in North America. Among these studies, 5 used self-administrated questionnaires for coffee consumption assessment, and the others used interview technology for coffee consumption assessment. Five evaluated the association between coffee consumption during pregnancy and childhood AL, 5 assessed childhood ALL, and 4 evaluated childhood AML.
| Reference | Country and study; period | Age range | Cases/controls | Type of leukemia (no.) | Coffee consumption | Effect estimation (95% CI) | Adjustments | ||
|---|---|---|---|---|---|---|---|---|---|
| AL | ALL | AML (ANLL) | |||||||
| Ross et al, 1996 a | United States and Canada; NA | <12.5 mo | 84/97 | AL (84) ALL (54) AML (30) | Never ≤3 times/wk ≥4 times/wk | Reference 1.5 (0.7–3.3) 2.5 (1.0–6.2) | Reference 1.1 (0.4–3.0) 2.3 (0.7–8.2) | Reference 2.4 (0.6–9.2) 2.6 (0.7–10.0) | Socioeconomic status, maternal education |
| Petridou et al, 1997 a | Greece; 1993 through 1994 | 0–14 y | 153/300 | AL (153) | Never Ever | Reference 0.89 (0.55–1.46) | Gender, age, place of residence | ||
| Clavel et al, 2004 a | France; 1995 through 1999 | <15 y | 219/105 | AL (219) | Never Ever ≤3 cups/d >3 cups/d | Reference 1.8 (1.0–3.3) 1.6 (0.9–2.9) 4.1 (1.6–10.1) | Gender, age, ethnic origin, hospital, socioeconomic status, maternal education | ||
| Menegaux et al, 2005 a | France; 1995 through 1999 | <15 y | 280/288 | AL (280) ALL (240) ANLL (40) | Never ≤3 cups/d 4-8 cups/d >8 cups/d | Reference 1.1 (0.7–1.8) 2.4 (1.3–4.7) 3.1 (1.0–9.5) | Reference 1.6 (0.6–4.3) 2.8 (0.7–10.4) 3.0 (0.3–35.1) | Gender, age, ethnic origin, socioeconomic status | |
| Menegaux et al, 2007 b | France; 1995 through 1998 | <15 y | 472/567 | AL (472) ALL (407) AML (62) | Never Ever ≤3 cups/d >3 cups/d | Reference 1.1 (0.9–1.5) 1.1 (0.8–1.4) 1.5 (0.9–2.4) | Reference 1.1 (0.8–1.4) 1.1 (0.8–1.4) 1.5 (0.9–2.4) | Reference 1.6 (0.8–2.9) 1.6 (0.8–3.0) 1.4 (0.5–4.4) | Gender, age, region, socioprofessional category, and birth order |
| Milne et al, 2011 b | Australia; 2003 through 2006 | ≤14 y | 337/697 | ALL (337) | Never Ever <2 cups/d ≥2 cups/d | Reference 0.89 (0.61–1.30) 0.77 (0.51–1.16) 1.12 (0.72–1.74) | Age, sex, state of residence, and maternal education | ||
| Bonaventure et al, 2013 a | France; 2003 through 2004 | ≤14 y | 764/1681 | AL (764) ALL (648) AML (101) | Never/occasional <1 cup/d 1-2 cups/d >2 cups/d | Reference 1.0 (0.8–1.3) 1.3 (1.0–1.7) 1.6 (1.2–2.1) | Reference 1.3 (0.7–2.1) 1.8 (1.0–3.3) 2.4 (1.3–4.3) | Reference 1.0 (0.8–1.3) 1.3 (1.0–1.7) 1.5 (1.1–2.0) | Gender, age, maternal education, and socioeconomic status |
a Studies used interview for coffee assessment.
b Studies used self-administered questionnaire for coffee assessment.
Maternal coffee consumption during pregnancy on childhood AL
Risk estimates for childhood AL are shown in Figure 2 . Compared with never/lowest drinkers, adverse effect of maternal coffee consumption during pregnancy on childhood AL was observed in ever drinkers (1.22; 95% CI, 1.04–1.43), low to moderate-level drinkers (1.16; 95% CI, 1.00–1.34), and high-level drinkers (1.72; 95% CI, 1.37–2.16). There was no statistically significant heterogeneity among studies (I 2 = 32.3%, P = .206; I 2 = 0.0%, P = .615; I 2 = 35.0%, P = .202). Regarding the assessment of publication bias, neither Begg bias ( Figure 3 ) nor Egger bias was observed ( P = .406).
Maternal coffee consumption during pregnancy on childhood ALL
Figure 4 presents the summary ORs for childhood ALL. Compared with the never/lowest drinkers, the summary OR was 1.26 (95% CI, 1.05–1.50) for ever drinkers, 1.09 (95% CI, 0.91–1.31) for low to moderate-level drinkers, and 1.65 (95% CI, 1.28–2.12) for high-level drinkers. Significant between-study heterogeneity was not noted in last 2 (I 2 = 24.1%, P = .261; I 2 = 47.0%, P = .110). When the results on ever drinkers were pooled in random effect model, the adverse effect of ever drinkers was not observed: 1.30 (95% CI, 0.97–1.73). With respect to publication bias, no bias was observed (Egger test, P = .596).
Maternal coffee consumption during pregnancy on childhood AML
The summary ORs for childhood AML are presented in Figure 5 . The significant association between childhood AML and coffee consumption during pregnancy was observed for ever drinkers (1.35; 95% CI, 1.10–1.66) and high-level drinkers (1.58; 95% CI, 1.20–2.08), but not for low to moderate-level drinkers (1.18; 95% CI, 1.00–1.40). No heterogeneity was observed (I 2 = 17.0%, P = .306; I 2 = 0.0%, P = .643; I 2 = 0.0%, P = .486). There was publication bias among studies included in this metaanalysis (Egger test, P = .018).
Dose-response relationship
Six studies were included in final dose-response metaanalysis with one removed due to no available data presented for cases and controls. Among these studies, 3, 4, and 3 were used to explore the dose-response relationship between coffee consumption during pregnancy and childhood AL, childhood ALL, and childhood AML, respectively.
For childhood AL, the nonlinearly relationship was not observed regarding maternal coffee consumption with childhood AL ( P for nonlinearity = .68) ( Figure 6 , A). Compared with never/lowest drinkers, the pooled ORs were 1.10 (95% CI, 0.94–1.29) for 1 cup/d, 1.24 (95% CI, 1.06–1.46) for 1-2 cups/d, 1.53 (95% CI, 1.21–1.93) for 2-3 cups/d, and 1.96 (95% CI, 1.32–2.92) for 4-5 cups/d.
