Association of assisted reproductive technology with offspring growth and adiposity from infancy to early adulthood

Importance: People conceived using assisted reproductive technology (ART) make up an increasing proportion of the worlds population, and their numbers are expected to continue rising. Objective: Investigate association of ART conception with growth and adiposity outcomes from infancy to early adulthood in offspring from a large multinational multi-cohort study. Design: 26 population-based cohort studies. Setting: Europe, Asia-Pacific, and North America Participants: Infants, children, adolescents, and young adults born from 1984 to 2018, with mean ages at assessment of growth/adiposity outcomes ranging from 0.6 month to 27.4 years. Exposures: Conception by ART (conventional in vitro fertilisation and intracytoplasmic sperm injection) versus natural conception (NC). Main Outcomes and Measures: Length/height, weight, and body mass index (BMI). Each cohort was analysed separately with adjustment for maternal BMI, age, smoking, education, parity, ethnicity, and offspring sex and age. Cohort results were combined in random effects meta-analysis for thirteen age groups. Results: Up to 158,066 offspring (4,329 conceived by ART) were included in each age-group meta-analysis; 47.6% to 60.6% were female. Compared with NC, ART-conceived offspring were slightly shorter, lighter, and thinner from infancy to early adolescence. The differences in growth/adiposity outcomes were largest at the youngest ages and attenuated with older child age, e.g., adjusted standardised mean differences (95% confidence intervals) in offspring weight at age <3 months, 17 to 23 months, 6 to 9 years, and 14 to 17 years were -0.27 standard deviation (SD) units (-0.39 to -0.16), -0.16SD (-0.22 to -0.09), -0.07SD (-0.10 to -0.04), and -0.02SD (-0.15 to 0.12), respectively. There was no evidence that results were driven by parental subfertility or of difference between conventional in vitro fertilisation and intracytoplasmic sperm injection however, smaller offspring size appeared to be limited to offspring conceived by fresh but not frozen embryo transfer, compared with NC. More marked but less precise differences were observed for body fat measurements. There was imprecise evidence that offspring conceived by ART may develop greater adiposity by early adulthood. Conclusions and Relevance: People conceiving or conceived by ART can be reassured that differences in early growth and adiposity are small and no longer evident by late adolescence.

or overadjustment for possible mediators (15-17). A recent study which examined trajectories 162 of change in height, weight, and body mass index (BMI) from birth to age 7 years (n=81,461 163 offspring with 1,721 conceived by ART) in the Norwegian Mother, Father and Child Cohort 164 Study (MoBa) found that ART-conceived offspring started smaller and grew faster than NC 165 offspring (18). Another more recent but considerably smaller Singaporean birth cohort study 166 (n=1,180 offspring with 85 conceived by ART) discovered smaller height and lower skinfold 167 thickness at age 6 years in ART-conceived than NC offspring (19). 168 Our primary aim was to conduct a multi-cohort study to provide evidence on the associations 169 of ART conception (compared with NC) with offspring growth and adiposity from infancy to 170 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Eligible cohorts were identified from the European Union Child Cohort Network (21-23) and 181 by searching cohort profile papers. We targeted population-based cohorts without selection or 182 oversampling of ART-conceived offspring to reduce potential for selection bias and to ensure 183 identical growth and adiposity assessments for ART-conceived and NC offspring. All cohorts 184 from any geographical region with birth years from older to more contemporary cohorts were 185 eligible for inclusion provided they had data on whether offspring were conceived by ART or 186 NC, and one or more offspring growth or adiposity outcome measures assessed from age one 187 month (including repeated measurements). A total of thirty cohorts were invited to participate 188 and 26 were included in this study (eTable 1). Detailed description of the 26 included cohorts 189 is provided in the eMethods. 190 All cohorts had ethical approval from the relevant local or national ethics committees and all 191 offspring gave informed consent/assent to participate in the respective cohorts and secondary 192 data analyses. Details on ethics approvals and consent are in eMethods. 193 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) Offspring growth/adiposity outcomes 206 (kg)), BMI (kg/m 2 ), and secondary outcomes were waist circumference (cm), total body fat 208 %, and fat mass index (FMI; kg/m 2 ). Length/height and weight were obtained from research 209 clinics, child health records, and maternal-/self-reports, with BMI calculated from these as 210 weight in kilograms divided by height in metres squared. Waist circumference was mostly 211 measured during research clinics. Body fat % was calculated from bioelectrical impedance 212 analysis done during research clinics, and FMI was derived as fat mass in kilograms from 213 dual-energy x-ray absorptiometry scans divided by height in metres squared. Details on 214 outcome measurements and ages in each cohort are in the eMethods. Descriptive data on 215 outcomes and ages at outcome assessments are in eTable 2. 216 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) Outcome age groups for this study were determined by available data from each cohort i.e., 217 ages at outcomes assessment. Cohorts were allocated to meta-analysis age groups by mean 218 age at outcome assessment with the aim of maximising cohort numbers in each age group 219 meta-analysis. The primary outcomes were allocated to thirteen age groups, and secondary 220 outcomes (available in 17/26 cohorts) were allocated to four age groups. If a cohort had >1 221 outcome assessment in an age group, we selected the one with the biggest sample size. 222 223

Confounders 224
We used a Directed Acyclic Graph (eFigure 1) to identify and adjust for confounders i.e., 225 anything that could plausibly cause ART use and influence offspring growth/adiposity (25, 226 26). This process identified the following maternal factors as potential confounders: age at 227 pregnancy/birth, socioeconomic position (education), pre/early-pregnancy BMI, pre/early-228 pregnancy smoking, parity, and ethnicity. Nineteen cohorts were able to adjust for all these 229 confounders, four did not adjust for ethnicity but were ethnically homogeneous, one did not 230 adjust for parity because it only included nulliparous women, and two were unable to adjust 231 for BMI and smoking. Details on the available confounders in each cohort is in eMethods. 232 233

Statistical analysis 234
Analyses were performed separately in each cohort applying standard statistical code, with 235 results combined using meta-analysis. In cohort-specific analyses, we estimated associations 236 of ART conception (versus NC) with offspring outcomes using linear regression adjusted for 237 confounders (plus offspring age and sex). Analysis was done in offspring with data on mode 238 of conception, ≥ 1 growth/adiposity outcome, and confounders. To facilitate comparison of 239 results for different outcomes and ages, outcomes were analysed in age-, sex-cohort-specific 240 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. by random-effects meta-analyses in sub-groups defined by mean age at outcome assessment. 242 Variability in the pooled estimates that is due to between-cohort heterogeneity was quantified 243 by the I 2 statistic (27). Influential cohorts (i.e., whose exclusion led to considerable change in 244 the meta-analysis model) were identified by repeating each meta-analyses with each cohort 245 left out in turn. 246 To separate effects of ART from any effects of parental subfertility, we repeated analyses 247 comparing ART-conceived with NC offspring of sub-fertile parents and separately for NC 248 offspring from fertile parents. Differences by sex and ART treatment types were explored by

256
A total of 26 cohorts with participants from Europe (n=20 cohorts), Australia (n=2 cohorts), 257 and one each from New Zealand, China, Singapore, and Canada was included in this study 258 (eTable 1). Most (n=23 cohorts) were population-based cohorts, two were twins register 259 cohorts, and one was a clinical cohort of ART-conceived young adults and NC controls from 260 the same source population. Birth years were from 1984-2018, with most (n=19 cohorts) born 261 >2002. Mean age at outcome ranged from 0.6 month to 27.4 years. Fifteen cohorts included 262 singletons and multiple births (proportion of multiple births across these ranged from 0.9% to 263 12.9%), nine included singletons only, and two included twins only. Between 3 to 16 cohorts 264 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 21, 2022. ; https://doi.org/10.1101/2022.03.20.22272579 doi: medRxiv preprint were included in each meta-analysis with numbers of participants in each meta-analysis 265 ranging from 158,066 (4,329 ART) for weight at age 3 to 5 months to 3,111 (151 ART) for 266 FMI at age >17 years. 267 Mean length/height was on average smaller in ART-conceived than NC offspring across all 268 age groups, although for some ages, point estimates were close to the null value which was 269 included in the CIs (Figure 1). The largest differences in length/height were at the youngest 270 ages and these differences attenuated with older child age. ART-conceived offspring were 271 more similar in height to NC older adolescents and young adults although estimates were 272 imprecise (Figure 1). 273 Mean weight was lower in ART-conceived than NC offspring from age <3 months up to age 274 10 to 13 years though CIs included the null at ages 6 to 8 months, 9 to 11 months, and 10 to 275 13 years (Figure 1). As for length/height, differences were greatest at the youngest ages and 276 smaller at older offspring ages. The difference in mean weight was close to the null in older 277 adolescents and mean weight in young adults was slightly higher in ART-conceived than NC 278 offspring but with wide CIs (Figure 1). 279 Differences in mean BMI followed a similar pattern to that of weight, with mean BMI lower 280 in ART-conceived than NC offspring up to age 10 to 13 years, with differences being greatest 281 at youngest ages but with wide CIs that included the null value for some results (Figure 1). 282 As for weight, difference in mean BMI was closest to the null in older adolescents and mean 283 BMI was somewhat slightly higher for ART-conceived than NC young adults although this 284 was imprecisely estimated (Figure 1). 285 Results for waist circumference, total body fat %, and FMI were like those observed for 286 weight and BMI, with lower mean adiposity measurements in ART-conceived than NC 287 offspring during childhood and adolescence, though with larger differences that were 288 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted March 21, 2022. ; https://doi.org/10.1101/2022.03.20.22272579 doi: medRxiv preprint imprecisely estimated for several time points (Figure 2). As for weight and BMI, adiposity 289 measures were higher in ART-conceived than NC adults, but with larger mean differences 290 and wider CI's that included the null (Figure 2). 291 Between-cohort heterogeneity was low to moderate for all outcomes at all ages, with a few 292 exceptions. There was substantial between-cohort heterogeneity in results for length/height, 293 weight, and BMI at ages <3 months and 3-5 months (Figure 1). Sensitivity analysis showed 294 that results for outcomes at both ages were robust to influential cohorts, although they were 295 slightly attenuated when the MoBa cohort was omitted (eFigure 8). 296 For some additional analyses there were too few ART conceptions to include all older age 297 groups. Results were similar when ART was compared with sub-fertile NC and fertile NC 298 (Figure 3), when ICSI and IVF were compared with NC (Figure 4), and in females and 299 males (eFigure 9). Mean length/height, weight, and BMI were on average lower in those 300 conceived by fresh ET compared with NC offspring across all available age groups i.e., from 301 age <3 months to age 6 to 9 years ( Figure 5). Conversely, differences were closer to the null 302 for FET compared with NC, though results were imprecise (Figure 5). The differences in all 303 growth and adiposity outcomes were only partially attenuated when analyses were restricted 304 to singletons (eFigure 10), whereas differences between ART-conceived and NC offspring 305 (eFigure 11), and between fresh ET and NC offspring (eFigure 12) were considerably 306 attenuated after further adjustment for birth weight and gestational age, particularly at 307 younger ages. 308 309 DISCUSSION 310 We used data from 26 multinational cohort studies to investigate the association between 311 ART conception and offspring growth and adiposity. The large number of offspring, and 312 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. length of follow-up, allowed us to explore findings in subgroups by age from infancy to early 313 adulthood. We found that offspring conceived by ART were on average shorter, lighter, and 314 thinner from infancy to early adolescence than NC offspring. Differences were largest earlier 315 in life but were small in magnitude across all ages. There was little evidence that differences 316 were driven by parental subfertility given similar results when we compared ART with NC 317 where parents conceived after 12 months and where conception occurred within a shorter 318 period since start of trying. Those conceived from fresh embryos were smaller than NC 319 offspring whereas frozen-thawed embryos were comparable to NC. Results appeared 320 independent of multiple births and were at least partly mediated by birth weight and 321 gestational age, particularly at younger ages. 322 Our findings are in line with previous studies and reviews of outcomes at birth and in young 323 children (6, 15-17). Although not directly comparable with our study, our finding of smaller 324 associations among older children is consistent with results from a recent study that found 325 more rapid growth from birth to 3 years in ART-conceived than NC offspring (18). That 326 study also examined outcomes at age 17 years in individuals screened for conscription in 327 Norway and found no difference between ART-conceived and NC offspring at that age, 328 which is consistent with our finding of no difference in growth in older adolescence. 329 Our results for fresh ET and FET are consistent with previous studies showing smaller birth 330 weight in fresh ET compared with NC, and higher birth weight and large-for-gestational-age 331 in FET compared with fresh ET (12, 13). Our study also agrees with results from a UK record 332 linkage study that assessed birth size and body size at 6-8 weeks and 5 years in offspring born 333 between 1997-2009, showing that compared with NC, offspring born by fresh embryos were 334 lighter, and those born by FET were heavier at birth and 6-8 weeks, and that all groups had 335 similar weight at 5 years (16). 336 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The reasons for lower birth weight and higher risk of small-for-gestational-age shown in 337 previous studies (8, 15) and the (modestly) smaller infant/child size in our study in ART-338 conceived offspring are not fully understood. The gametic and embryonic manipulations 339 associated with ART may impact embryonic/fetal development in a manner that is reflected 340 in different growth patterns relative to NC individuals. Growth differences could also reflect 341 physiological responses to ART-induced lower birth size (and gestational age), and unlikely 342 to be sex-specific or differ by ART type given our finding of broadly similar results in males 343 and females and conventional IVF and ICSI. This is supported by our finding that differences 344 attenuate by adjustment for birthweight and gestational age, though, this should be interpreted 345 with caution since assumptions for such analyses and potential for collider bias makes them showing slightly increased obesity risk in young ART-conceived adults (32). One possible 358 reason for this result is that the rapid infant growth we observed in ART-conceived offspring 359 continues (at a decelerating rate) for extended time. This is consistent with prior evidence of 360 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is important to note that our pooled effect sizes were small across all age groups, including 363 at the youngest ages where they were largest in magnitude. For example, when expressed in 364 its natural units, the largest differences in weight, observed at age <3 months, was 183 grams 365 (95%CI: 105 to 261) lower in those conceived by ART. Therefore, it is unlikely that these 366 differences will result in any clinically meaningful differences at any age. It is also worth 367 acknowledging that our pooled results represent average differences in outcomes across all 368 populations from all included cohorts, and there was some evidence of heterogeneity for 369 some outcomes. However, sensitivity analyses indicated results were robust to influential 370 cohorts and heterogeneity was due to differences in directionally consistent effect sizes. Limitations include low precision/power at older ages, which highlights the importance of 385 measuring outcomes in adult life. Those with outcomes at older ages were exposed to ART 386 some decades ago using treatments and embryo culture techniques that are less relevant to 387 contemporary practices, thus making it difficult to know the extent to which findings would 388 generalise to more recently born cohorts. Therefore, there is a need to promote collection of 389 data on mode of conception from birth cohorts and to ensure those conceived by ART are 390 included so that future analyses can continually add new cohorts to examine changes in 391 associations by birth years and age. Our analysis was restricted to those with complete data 392 on mode of conception, outcomes, and confounders which may have reduced precision of 393 estimates and introduced bias due to missing data. Residual confounding by unmeasured 394 factors (e.g., paternal health) is possible and might influence our findings. 395 In conclusion, we found that ART-conceived offspring were on average slightly smaller and 396 had modestly lower adiposity than NC offspring during early life, with associations reduced 397 with older child age, with some imprecise evidence for higher adiposity by early adulthood 398 with ART conception. Overall, our findings are reassuring since differences in early growth 399 were small though, there is a need for additional follow-up and studies with larger numbers 400 into older ages to is investigate the possibility of greater adiposity in adulthood.   I² represents the percentage of total variability that is due to between cohort heterogeneity. 525 Cohort-specific results are provided in eFigure 2-4. 526 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  conceived and NC offspring (ART minus NC). Cohort-specific results were adjusted for 531 maternal age, parity, BMI, smoking, education, ethnicity/country of birth, plus offspring sex 532 and age. St. is the number of cohort studies; NC is the number of NC ofspring; ART is the 533 number of ART-conceived offspring; I² represents the percentage of total variability that is 534 due to between cohort heterogeneity. Cohort-specific results are provided in eFigure 5-7. 535

Figure 3. Mean difference in growth and adiposity outcomes between ART-conceived 536
and NC offspring, separately for NC offspring of sub-fertile and fertile parents. Figure  537 shows the pooled adjusted mean differences in SD units and 95% confidence intervals in 538 growth and adiposity outcomes at each age group between ART-conceived and NC offspring 539 (ART minus NC) of fertile parents (≤12 months to pregnancy and sub-fertile parents (>12 540 months to pregnancy). Cohort-specific results were adjusted for maternal age, parity, BMI, 541 smoking, education, ethnicity/country of birth, plus offspring sex and age. The number of 542 offspring at each age for the primary outcomes (length/height, weight and BMI) varied from 543 2,955 ART, 93,877 fertile NC, and 11,153 sub-fertile NC for weight at age 3 to 5 months to 544 51 ART, 3,350 fertile NC, and 494 sub-fertile NC for BMI at >17 years. 545 in growth and adiposity outcomes at each age group between ART-conceived and NC 549 offspring (ART minus NC), separately for offspring conceived by conventional IVF and 550 ICSI. Cohort-specific results were adjusted for maternal age, parity, BMI, smoking, 551 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)  and 95% confidence intervals in length/height, weight, and body mass index at each age 560 group between ART-conceived and NC offspring (ART minus NC), separately for offspring 561 conceived using fresh embryo transfer and frozen-thawed embryo transfer. Cohort-specific 562 results were adjusted for maternal age, parity, BMI, smoking, education, ethnicity/country of 563 birth, plus offspring sex and age The number of offspring at each age varied from 1,904 fresh 564 embryo transfer, 303 frozen-thawed embryo transfer, and 78,128 NC for weight at age 3 to 5 565 months to 433 conventional fresh embryo transfer, 84 frozen-thawed embryo transfer, and 566 15,490 NC for BMI at age 17 to 23 months. 567 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)