Food insecurity increases energetic efficiency, not food consumption: an exploratory study in European starlings

Food insecurity—defined as limited or unpredictable access to nutritionally adequate food—is associated with higher body mass in humans and birds. It is widely assumed that food insecurity-induced fattening is caused by increased food consumption, but there is little evidence supporting this in any species. We developed a novel technology for measuring foraging, food intake and body mass in small groups of aviary-housed European starlings (Sturnus vulgaris). Across four exploratory experiments, we demonstrate that birds responded to 1–2 weeks of food insecurity by increasing their body mass despite eating less. Food-insecure birds therefore increased their energetic efficiency, calculated as the body mass maintained per unit of food consumed. Mass gain was greater in birds that were lighter at baseline and in birds that faced greater competition for access to food. Whilst there was variation between experiments in mass gain and food consumption under food insecurity, energetic efficiency always increased. Bomb calorimetry of guano showed reduced energy density under food insecurity, suggesting that the energy assimilated from food increased. Behavioural observations of roosting showed inconsistent evidence for reduced physical activity under food insecurity. Increased energetic efficiency continued for 1–2 weeks after food security was reinstated, indicating an asymmetry in the speed of the response to food insecurity and the recovery from it. Future work to understand the mechanisms underlying food insecurity-induced mass gain should focus on the biological changes mediating increased energetic efficiency rather than increased energy consumption.

110 we programmed the variance in the inter-food access interval such that it was unpredictable to 111 the birds. Therefore, our operationalisation of food insecurity combined restricted access to food 112 with variability and unpredictability in the intervals between periods of access. This was done 113 intentionally to mimic the human experience of food insecurity and also the assumptions of the 114 optimality models cited above. It was not our intention in the current study to differentiate 115 between the effects of predictable and unpredictable variance in the timing of food, which have 116 been studied elsewhere (Bateson & Kacelnik, 1997;Cuthill et al., 2000). 117 The four experiments that we conducted used within-subjects designs, in which all 118 subjects experienced all treatments: birds were exposed to a baseline week of food security 119 (FS1), during which food was available immediately a bird sought access, followed by either one 120 or two weeks of food insecurity (FI), during which food was both limited and unpredictable. In 121 two of the four experiments (1 and 3), food security was reinstated for a final period (FS2) of 122 one or two weeks following the period of food insecurity in order to study how rapidly the effects 123 of food insecurity reversed. 124 In experiments 1-3, food insecurity was induced by withdrawing food access for a 125 randomly chosen 12 out of 20 contiguous 20-minute periods starting two hours after dawn each 126 day (designated 'total food removal' in Fig. 1). In experiment 1, this regimen resulted in a mean 127 of 4.42 intervals per day without food, ranging from 20-120 minutes (mean ± sd = 54.19 ± 128 36.22), and approximated the type of unpredictable access previously reported to induce mass 129 gain in starlings . Intervals of similar duration were generated in experiments 130 2 and 3. In experiment 4, food insecurity was induced via an operant schedule in place from 131 dawn until dusk (designated 'probabilistic schedule' in Fig. 1.), whereby the probability that a 132 single peck to an illuminated key was reinforced with 10 seconds of food access was reduced 133 from 1.0 under food security to 0.4 and later 0.2 under food insecurity (FI low and FI high 134 respectively). This probabilistic schedule would result in mean intervals of 20 ± 23.24 and 50 ± 135 53.66 seconds between periods of food access in FI low and FI high respectively, assuming the 136 birds pecked as soon as the key was illuminated. The schedule mimics an operant schedule 137 used to study diurnal patterns in responses to decreased food availability in coal tits (Periparus 138 ater) (Polo & Bautista, 2006). Thus, both methods for inducing food insecurity introduced 139 variable intervals between periods of free access to food, albeit on different time scales. Our 140 reasons for exploring the effects of the probabilistic schedule (experiment 4) in starlings for the 141 first time were as follows. First, because the schedule does not induce long periods without 142 food, it produces smoother and more consistent mass gain trajectories across the day that 143 should facilitate comparisons between treatments (see Cuthill et al. (2000) for a discussion of 144 the methodological problems inherent in total food removal methods). Second, using an operant 145 schedule makes it possible to measure individual foraging motivation via key pecking data. 146 Finally, the probabilistic schedule more closely embodies the assumptions of Anselme  Under both methods for inducing food insecurity, it was theoretically possible for the 151 birds to maintain or increase their total daily food consumption if they were motivated to do so. 152 They could do this, either by eating more during the periods of the day when food was freely 153 available (a total of 5 hours spread across the day in experiments 1-3 and during the 10-second 227 Procedure 228 All experiments began with birds being caught from their home aviary, manually weighed and 229 equipped with two plastic leg rings each of which had a unique RFID microchip attached (birds 230 wore two chips to guard against data loss in the event that one chip fell off, broke or was not 231 read due to poor alignment with the SFS aerial). Birds were then released into experimental 232 aviaries. Food was initially provided ad libitum by raising the hoppers of the SFS stations. Birds 233 were initially encouraged to visit the SFS by placing mealworms around the stations and in the 234 food hopper, but as soon as they were visiting the SFS readily these extra mealworms were 235 withdrawn. The first experimental treatment of experiments 1-3 began once the birds were 236 maintaining stable body masses on food obtained from the SFS. 237 Experiment 4 required additional training for the birds to learn that food could only be 238 accessed by pecking the illuminated key on the SFS. We used an auto-shaping procedure 239 whereby illumination of the pecking key signalled subsequent unconditional raising of the food 240 hopper. Birds first learned a Pavlovian association between the lit key and food and 241 spontaneously started performing appetitive pecks to the lit key. Pecks to the lit key were 242 reinforced by immediate hopper raising, thus creating an instrumental association between key 243 pecking and food. As soon as key pecking was established, hopper raising was made 244 conditional on key pecking and the birds were moved to continuous foraging, whereby they 245 earned all of their daily food by pecking at the lit key on the SFS. The first experimental 246 treatment of experiment 4 began once the birds were maintaining stable body masses on food 247 earned from the SFS. 248 The sequence and duration of the treatments experienced in each experiment is shown 249 in Fig. 1. In all experiments, birds were maintained in closed economy, obtaining the majority of 250 their daily food intake from the SFS. The diet was supplemented with four mealworms per bird 251 given during daily husbandry and supplied in two spatially separated bowls to prevent one bird 252 from monopolising the worms. The experiments ran seven days a week with no gap between 253 treatments. The birds' welfare was monitored daily via visual checks and using the body mass 254 data obtained from the SFS balances; criteria for consulting a vet and considering euthanasia 255 were either presence of puffed feathers and lethargy and/or a mass of <62 g. No bird became 256 unwell or dangerously thin at any point during the study and no birds died or were euthanized. 257 At the end of the final treatment of each experiment the birds were re-caught, manually weighed 258 and returned to their home aviaries where they were retained for further studies. Experiments 1-259 3 were conducted consecutively on the same group of 6 male birds with a 24-day break in the 260 home aviary between experiments 1 and 2 and an 18-day break between experiments 2 and 3. 261 Body mass 262 In all experiments, body masses were recorded by the SFS each day between lights-on and 263 lights-off; each mass was recorded with a bird identity corresponding to the microchip of the bird 264 on the perch. The balances measured masses at a frequency of 6 Hz. A stable mass was 265 recorded for a bird if the balance measured five consecutive masses of >50 g that were within a 266 range of 5 g. These criteria were chosen to eliminate masses from birds that were perching 267 incorrectly (e.g. by placing one foot on the food hopper), but to maximise the number of stable 268 masses recorded from moving birds. Once a stable mass had been recorded another stable 269 mass could not be recorded until the balance had measured a mass <10 g indicating that the 270 current bird had left the perch. Balances were checked with a 100-g test mass a minimum of 271 twice daily and calibrated if necessary. In order to control for build-up of guano on the perch 272 over the day, balances were automatically tared regularly throughout the day when no bird was 273 present on the perch. 274 The SFSs recorded a mean of 4.64 stable masses per bird per daylight hour over the 275 four experiments. The raw masses showed a clear trend of mass increase over the day as 276 expected ( Fig. S1), but there was a lot of random error due to the imprecision of the balances 277 and movement of the birds whilst on the perch, and masses were not always available at all 278 times of every day. To estimate comparable dawn and dusk masses for each bird on each day 279 we used the following procedure to model the available data. As long as a minimum of 10 280 masses were available, the mass data from each bird on each day were fitted with a cubic 281 polynomial (Fig. S1A). To remove biologically impossible outliers, any masses >3 g from the 282 fitted line were removed and a new cubic polynomial was refitted to the remaining data. This 283 latter fit was used to estimate dawn and dusk masses for that day. To avoid extrapolation 284 beyond the data, a dawn or dusk mass was only estimated if there was a data point within 1 285 hour of the estimate. The above procedure was devised based on a detailed exploration of the 286 data from experiment 4 and then applied unaltered in experiments 1-3. The one exception being 287 that due to differences in daylight hours between experiments, dawn mass was defined as the 288 fitted mass at 0900 for experiments 1-3 and 0615 for experiment 4; dusk mass was defined as 289 the fitted mass at 1800 for experiments 1-3 and 1815 for experiment 4 (Fig. S1B). Thus, the 290 dusk masses for experiments 1-3 and experiment 4 were estimated 9 and 12 hours after lights-291 on respectively. A later time was not chosen for the dusk masses in experiment 4 due to the fact 292 that birds often stopped foraging considerably before lights-off.
293 Operant foraging behaviour 294 In experiment 4, key pecks were recorded by the SFS between 0615 and 2045 each day. The 295 time of each key peck was recorded with a bird identity and whether or not the peck was 296 reinforced with food (probability of 1.0 under food security and either 0.4 or 0.2 under food 297 insecurity for FI low and FI high respectively).

Food consumption
299 Total food consumption in each aviary was estimated daily in all experiments by calculating the 300 difference in the mass of the SFS food hoppers at the beginning and end of the day and 301 subtracting any food collected in a spill tray located beneath each hopper. Hence, consumption 302 data were only available at the aviary level.
303 Energy density of guano 304 Energy density of guano was measured by bomb calorimetry in experiments 1 and 3. Guano 305 samples were collected daily from plastic trays positioned beneath perches in each aviary 306 avoiding any feathers or wood chippings. Hence, guano data were only available at the aviary 307 level. Samples were immediately frozen at -80 o C for storage. On completion of the experiment, 308 samples were dried in a flow oven at 55 o C for 48 hours until stable masses were obtained and 309 finely ground to homogenise. In experiment 1, samples from each aviary were pooled over three 310 consecutive days (yielding pooled samples centred on days 2 and 6 of FS1 and FS2 and days 311 2, 6, 9 and 13 of FI), whereas in experiment 3 all days were measured separately. For both 312 experiments, each pooled sample was divided into two technical replicates prior to analysis. 313 Bomb calorimetry was outsourced to Pemberton Analytical Services, Shropshire, UK and was 314 conducted blind to the treatment group and technical replicates using a PAR 1261 adiabatic 315 bomb calorimeter with a stated precision of ±0.1% on two determinations. 316 Physical inactivity 317 Data on physical inactivity were collected in experiments 1-3. As a metric of daytime physical 318 inactivity we measured roosting behaviour-defined as perching motionless on a high rope 319 perch. Video recordings of the aviaries were made for 15 minutes starting at 0900 and 1000 320 every day, before any food insecurity started, using a remotely operated wide-angle surveillance 321 camera mounted in the ceiling of each aviary. The videos were scored manually using BORIS 322 video coding software (Friard & Gamba, 2016). Scoring of roosting was blind to the 323 experimental treatment in place on the previous day (the relevant predictor variable). A scan 324 sampling method was used, whereby the instantaneous behaviour of each bird in the aviary was 325 scored every 30 seconds for each 15-minute video. Birds were not individually identifiable on 326 the videos, therefore the behavioural metric was the proportion of birds in the aviary observed 327 roosting on each scan. 328 Statistical analysis 329 Data were analysed using R version 3.5.1(R Core Team, 2018). Below we provide an overview 330 of our statistical approach. Further details of statistical models are given in the Tables S2-S9.  331 For analysis of the effects of food insecurity in individual experiments we used linear 332 mixed models (GLMMs) fitted using the R package 'lme4' (Bates et al., 2015). The models 333 contained random effects (intercepts) to account for sources of non-independence in the 334 datasets. In all models we analysed the effect of treatment: FS1 vs FI vs FS2 in experiments 1 335 and 3; FS1 vs FI in experiment 2; and FS1 vs FI low vs FI high in experiment 4. In cases where a 336 treatment continued for two weeks, the data were pooled from the first and second weeks of a 337 treatment for statistical analysis, but the graphs in Fig. 3 show the data broken down by week. 338 Since roosting data were proportions, they were arcsine square root transformed prior to 339 analysis. Following any necessary transformation, all models gave satisfactory distribution of 340 residuals, hence a Gaussian error structure was assumed throughout. We used restricted 341 maximum likelihood estimation (REML) and conducted overall significance tests of treatment in 342 the GLMMs using Satterthwaite's method. 343 For meta-analysis we used multi-level random effects meta-analysis models fitted using 344 the R package 'metafor' (Viechtbauer, 2010). The models contained random effects of aviary 345 nested in experiment. The effect sizes used in the meta-analyses were obtained from GLMMs 346 fitted to the data from each aviary, using the same models for the analysis of individual 347 experiments described above (with the exception that random effects of aviary were no longer 348 required). Effects of food-insecurity were the parameter estimates (β values) and associated 349 standard errors corresponding to: the difference between FS1 and FI for experiments 1-3 and 350 the difference between FS1 and FI high for experiment 4. Tests of modifiers were conducted using 351 meta-regression, whereby both baseline mass and competition (number of birds per SFS 352 station) were added to the meta-analytic models. Estimation was by REML in all meta-analytic 353 models. We used aviary-level data as the unit of analysis for the meta-analyses, because in the 354 experiments involving multiple aviaries (1 and 4) there was considerable heterogeneity in results 355 obtained from the different aviaries. Furthermore, increasing the number of replicates available 356 for the meta-analysis increased the power available for the meta-regressions, making it possible 357 to test for effects of baseline mass and competition.
358 Ethical statement 359 The study adhered to ASAB/ABS guidelines for the use of animals in research. Birds were taken 360 from the wild under Natural England permit 20121066 and the research was completed under 361 UK Home Office licence PPL 70/8089 with approval of the Animal Welfare and Ethical Review 362 Body at Newcastle University. Prior to each individual experiment, a detailed protocol was 363 approved by Newcastle University's Comparative Biology Centre, but these protocols were not 364 publicly pre-registered. Manuscript to be reviewed 410 because the mass estimates are based on multiple measurements and are therefore highly 411 precise. 412 To investigate the effect of reinstating food security in experiments 1 and 3, dawn and 413 dusk mass in FS2 were compared with those in FS1 and FI. Reinstating food security had no 414 effect on dawn mass in either experiment (Fig. 3A,C; Table S2), but dusk mass declined in both 415 experiments relative to FI back to a level not significantly different from that in FS1 (Fig. 3E,G; 416 Table S3).

Foraging behaviour
418 Key pecking data from the operant schedules in experiment 4 were used to investigate how 419 birds responded to the reduced probabilities of reinforcement in FI low and FI high . Birds increased 420 their frequency of key pecking during both periods of FI compared to FS1 (GLMM: F 2,118 = 87.16, 421 p < 0.001; Fig. 6A, Table S5). The birds also exploited each 10-s reinforcement more 422 effectively, consuming more food per second in FI (GLMM: F 2,55 = 23.07, p < 0.001; Fig. 6B, 423 Table S5). These responses were graded, with birds pecking and eating faster under FI high than 424 under FI low (Table S5). However, this increased foraging effort was insufficient to compensate 425 for the reduced probabilities of reinforcement in FI. The birds earned fewer reinforcements per 426 day (GLMM: F 2,118 = 32.95, p < 0.001; Table S5) and overall ate less food per day in FI (GLMM: 427 F 2,54 = 8.77, p < 0.001; Fig. 3L; Table S6).
428 Food consumption 429 The mean consumption per bird each day was computed (by dividing the measured 430 consumption in an aviary by the number of birds present) and compared on all days during FS1 431 with all days during FI for each experiment (Fig. 3I-L; Table S6). Overall, food consumption 432 dropped by 2.50 g.bird -1 .day -1 (95% CI: -3.13 to -1.87) when birds moved from FS1 to FI (RE 433 meta-analysis: z = -3.08, p = 0.002; Fig. 4C; Table S4). This corresponds to a ~13% drop in 434 daily consumption compared to the initial period of food security. There was heterogeneity 435 among the effects of FI on food consumption (tests for heterogeneity among associations: τ 2 = 436 3.20, Q7 = 61.76, p < 0.001; Table S3). Addition of baseline body mass and competition to the 437 meta-analytic model as moderators did not explain a significant percentage of this heterogeneity 438 (omnibus test of meta-regression: Q M 2 = 5.76, p = 0.056), although given the marginal nature of 439 this result there is potentially some evidence for an effect here. 440 Reinstating food security had no effect on consumption in experiment 1, but caused an 441 immediate increase in consumption in experiment 3 ( Fig. 3K; Table S6). However, consumption 442 during FS2 remained lower than during FS1 in both experiments (Fig. 3I,K; Table S6). 443 Energetic efficiency 444 We calculated energetic efficiency as the body mass maintained per mass of food consumed, 445 using dawn mass and the amount of food consumed on the previous day. We compared all 446 days during FS1 against all days during FI for all 4 experiments ( Fig. 3M-P, Table S7). Overall, 447 energetic efficiency increased by 0.85 (95% CI 0.65 to 1.06) when birds moved from FS1 to FI 448 (RE meta-analysis: z = 8.17, p < 0.001; Fig. 4D, Table S4). This corresponds to a ~18% 449 increase in energetic efficiency compared to FS1. The effect of FI on energetic efficiency (1.15 450 sd) was larger than that on dawn or dusk mass alone (0.17 and 0.42 sd respectively). 451 Furthermore, the effect of FI on energetic efficiency was less heterogeneous between 452 experiments and aviaries than the effects on body mass, although the heterogeneity was still 495 Discussion 496 Across four experiments, starlings responded to food insecurity by increasing their dusk body 497 mass by ~3%. Moreover, food insecurity caused increased body mass in experiment 4 where 498 there were no periods of food deprivation lasting more than a few minutes, indicating a 499 response to a relatively subtle increase in the short-term unpredictability of food availability 500 within a day. In none of the four experiments did the food-insecure birds consume more food in 501 total. In direct opposition to the predictions of recent mechanistic models (Anselme, Otto & 502 Güntürkün, 2017; Anselme & Güntürkün, 2019), food insecurity was associated with a ~13% 503 decrease in daily food consumption. Our data therefore lead to the novel conclusion that 504 starlings respond to food insecurity by increasing their energetic efficiency-the body mass 505 maintained per unit of food consumed per day-by ~18%. Although food-insecure birds gained 506 dusk mass on average, increasing energetic efficiency was the more robust response, 507 characterised by larger effect sizes and less variability in response. The increase in energetic 508 efficiency observed under food insecurity may be partially explained by the birds assimilating 509 more energy from their food, and in experiment 3, by reducing energy spent on physical activity. 510 The changes caused by food insecurity did not immediately reverse when food security was 511 reinstated: energetic efficiency, energy absorption and physical inactivity all remained higher 512 than they had been at baseline for 1-2 weeks following reinstatement of food security. 513 Although food insecurity caused an increase in body mass overall, there was variation 514 both within and between birds in the response: some birds maintained their baseline mass, 515 while others gained mass and the overall effect of food insecurity varied between experiments. 516 Our meta-regression results shed some light on the possible causes of this variation. In support 517 of previous results from birds, we showed that starlings were more likely to gain mass under 518 food insecurity if they were lighter at baseline (Pravosudov & Grubb, 1997; Witter & Swaddle, 519 1997) and if they faced greater competition for food in the aviary 520 Witter & Goldsmith, 1997). Both of these findings make adaptive sense. If increased mass 521 provides insurance against starvation, then thinner birds should obtain greater benefits from 522 mass gain than birds who already have sufficient fat reserves to survive periods without food. 523 Birds facing higher competition effectively face harsher food insecurity than birds facing lower 524 competition, increasing the risk of starvation and hence the fat reserves it is optimal to carry. 525 The finding that both baseline mass and competition moderate the effect of food insecurity on 526 body mass potentially helps to explain why published studies of the effects of unpredictable food 527 do not always report mass gain. 528 Increasing energetic efficiency implies either increasing the amount of energy absorbed 529 from food, or decreasing energy expenditure in some domain. We found suggestive evidence 530 for the former strategy and inconsistent evidence for the latter. The energy density of the birds' 531 guano decreased under food insecurity, suggesting that the birds were assimilating more 532 energy. Increased energy assimilation under reduced food intake has previously been reported 533 in starlings (Bautista et al., 1998). Moreover, rats and rhesus macaques subject to long-term 534 caloric restriction paradigms do not reduce their total daily energy expenditure by the amount 535 predicted by their reduced food intake, suggesting that these species too must increase their 536 energy assimilation when intake is restricted (Selman et al., 2005). Since energy assimilated is 537 likely to be a decelerating function of gut residence time, this result could simply be a passive 538 physical consequence of increased gut passage time resulting from slower food intake. Another 539 (not mutually exclusive) explanation, is that the birds responded to food insecurity by 540 strategically changing their gut anatomy or physiology to increase energy assimilation. Starlings 541 are able to adaptively alter their gut morphology in response to changes in diet (Al-Joborae, 542 1979; Geluso & Hayes, 1999), demonstrating that gut plasticity in response to food insecurity is 543 not implausible. Given that the change in energy density of guano that we observed was slow, 544 reducing over the 2-week period of food insecurity in experiment 1, and did not appear to 545 reverse immediately when food security was reinstated, our data favour a strategic adaptation. 546 Therefore, we hypothesise that starlings respond to food insecurity by altering their gut in some 547 way to increase energy assimilation. However, it seems unlikely that increased energy 548 assimilation could be the only explanation for increased energetic efficiency in the current 549 dataset, due to the relatively small size of the effect (~1% drop in energy density of guano) 550 compared with the large decrease in food consumption (~13% drop in daily food consumption). 551 In four of five aviaries where we measured physical inactivity, roosting behaviour 552 increased under food insecurity. An overall effect of food insecurity was rendered null by one 553 aviary where the effect went strongly in the opposite direction. Roosting occurs when the birds 554 are not engaged in other activities such as foraging, eating or bathing, and is likely to be 555 associated with the lowest levels of energy expenditure due to physical activity. We measured 556 roosting during the first two hours of the day when food was always available ad libitum, 557 meaning that any changes in roosting were not a direct response to the current unavailability of 558 food. The fact that the increase in roosting observed in experiment 3 was simultaneous with 559 body mass gain, suggests that reduced physical activity could have been causal in mass gain. 560 However, since food insecurity did not induce a significant increase in roosting in either 561 experiments 1 or 2, it seems unlikely that decreased physical activity is sufficient to explain the 562 increases in energetic efficiency observed in all experiments, even in conjunction with increased 563 energy assimilation. It is noteworthy that although zebra finches do not respond to unpredictable 564 food deprivation by increasing body mass (Dall & Witter, 1998;Marasco et al., 2015), they do 565 decrease physical activity (Dall & Witter, 1998), compatible with an increase in energetic 566 efficiency. 567 We did not measure metabolic rate in the current study, but reducing basal metabolic 568 rate could be a third mechanism by which food-insecure birds increased energetic efficiency 569 (Wiersma, 2005;Secor & Carey, 2016). Food insecurity might cause the birds to down-regulate 570 or turn off energetically expensive, but temporarily expendable, biological systems in order to 571 maximise probability of survival in the face of short-term energetic shortfalls. Candidate systems 572 include, somatic maintenance, the immune system and the reproductive system. The costs 573 associated with increased energetic efficiency might therefore be measured in terms of 574 accelerated biological ageing, increased risk of cancer and infectious disease, or reduced 575 reproductive success. No relevant data were collected in the current study, but previous studies 576 in passerine birds have provided evidence for the existence of such trade-offs (e.g.      and (M-P)energetic efficiency, calculated as the ratio of the mean dawn mass for a bird on a given day to the total mass of food eaten per bird on the previous day. Graphs are box plots with each box corresponding to 7 days of data, but the data from treatments lasting 14 days were pooled for statistical analysis. Mass data were available at the individual bird level, but consumption (and hence also efficiency data) were only available at the aviary level. For this figure, the three separate aviaries are combined for experiments 1 and 4. For display purposes only, data were within-subject centred and shown relative to the grand mean. All experiments involved 6 birds. Experiments 1-3 were run consecutively with the same 6 males (with breaks of 24 and 18 days between experiments), whereas experiment 4 was run with 6 females. Significance tests are presented in Tables S2, S3, S6 and S7: * p < 0.05, ** p < 0.01, *** p < 0.001.