Description of the condition
Necrotising enterocolitis (NEC), a syndrome of acute intestinal necrosis of unknown aetiology, affects about 5% of very preterm (< 32 weeks) or very low birth weight (VLBW) (< 1500 grams) infants (Holman 1997). The associated mortality rate is > 20%. Infants who develop NEC experience more nosocomial infections, have lower levels of nutrient intake, grow more slowly, and have longer durations of intensive care and hospital stay than gestation-comparable infants who do not develop NEC (Bisquera 2002; Guthrie 2003). NEC is also associated with a higher incidence of long-term neurological disability, which may be a consequence of infection and under-nutrition during a critical period of brain development (Stoll 2004; Soraisham 2006; Rees 2007; Pike 2012).
Description of the intervention
Short gestational age at birth is the major clinical risk factor for developing NEC (Beeby 1992; Luig 2005). The other putative major risk factor is intrauterine growth restriction, especially if associated with absent or reversed end-diastolic flow velocities in Doppler studies of the fetal aorta or umbilical artery (Bernstein 2000; Garite 2004; Dorling 2005). Most very preterm or VLBW infants who develop NEC have received enteral milk feeds. Evidence exists that feeding with artificial formula rather than human milk increases the risk of developing NEC (Quigley 2007). Other differences in enteral feeding regimens, such as the timing of introduction of feeds and the size of the daily volume increments, may also contribute to inter-unit variation in the incidence of NEC. Multicentre benchmarking studies have found that those neonatal centres where enteral feeding is introduced earlier and feeding volumes advanced more quickly tend to have higher incidences of NEC (Uauy 1991). Observational studies have suggested that delaying the introduction of enteral feeds beyond the first few days after birth, or increasing the volume of feeds by less than about 24 ml/kg body weight each day, may be associated with a lower risk of developing NEC in very preterm or VLBW infants (Brown 1978; McKeown 1992; Patole 2005; Henderson 2009).
Why it is important to do this review
There are potential disadvantages associated with slowing the advancement of enteral feed volumes, such as delaying the establishment of full enteral nutrition (Flidel-Rimon 2004). Prolonged use of parenteral nutrition is associated with infectious and metabolic risks that may have adverse consequences for survival, growth, and development (Stoll 2004). It has been argued that the risk of NEC should not be considered in isolation of these other potential clinical outcomes when determining feeding policies and practice for very preterm or VLBW infants (Flidel-Rimon 2006; Chauhan 2008; Hartel 2009).
Other Cochrane reviews address the questions of whether delaying the introduction of any enteral milk feeding or restricting feed volumes to trophic levels (minimal enteral nutrition) affect the risk of NEC in very preterm or VLBW infants (Bombell 2009; Morgan 2011). This review focused on the question of whether advancing feed volumes at slow rates compared to faster rates affects the risk of NEC, mortality and other morbidities.
To determine the effect of slow rates of enteral feed advancement on the incidence of NEC, mortality and other morbidities in very preterm or VLBW infants.
Criteria for considering studies for this review
Types of studies
Controlled trials utilising either random or quasi-random patient allocation.
Types of participants
Enterally-fed very preterm (< 32 weeks) or VLBW (< 1500 grams) newborn infants.
Types of interventions
Advancement of enteral feeds at no more than 24 ml/kg (birth weight or current body weight) per day versus faster rates of feeds advancement. Infants should have received the same type of milk and in both groups the advancement of feed volume should have commenced within five days of introduction of enteral feeds.
Types of outcome measures
1. NEC confirmed by at least two of the following features:
- abdominal radiograph showing pneumatosis intestinalis or gas in the portal venous system or free air in the abdomen;
- abdominal distension with abdominal radiograph with gaseous distension or frothy appearance of bowel lumen (or both);
- blood in stool;
- lethargy, hypotonia, or apnoea (or a combination of these);
or a diagnosis confirmed at surgery or autopsy (Walsh 1986).
2. All-cause mortality during the neonatal period and prior to hospital discharge.
(i) Time to regain birth weight and subsequent rates of weight gain, linear growth, head growth, or skinfold thickness growth up to six months (corrected for preterm birth).
(ii) Long-term growth: weight, height, or head circumference (or proportion of infants who remained below the 10th percentile for the index population's distribution) assessed at intervals from six months of age.
(i) Death or severe neurodevelopmental disability defined as any one or a combination of the following: non-ambulant cerebral palsy, developmental delay (developmental quotient less than 70), auditory and visual impairment. Each component was to be analysed individually as well as part of the composite outcome.
(ii) Neurodevelopmental scores in survivors aged greater than, or equal to, 12 months of age measured using validated assessment tools.
(iii) Cognitive and educational outcomes in survivors aged more than five years.
5. Time to establish full enteral feeding (independently of parenteral nutrition).
6. Time to establish oral feeding (independently of parenteral nutrition or enteral tube feeding, or both).
7. Feed intolerance (defined as a requirement to cease enteral feeds).
8. Incidence of invasive infection as determined by culture of bacteria or fungus from blood, cerebrospinal fluid, urine, or from a normally sterile body space.
9. Duration of hospital stay (days).
Search methods for identification of studies
We used the standard search strategy of the Cochrane Neonatal Review Group (http://neonatal.cochrane.org/).
We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library 2012, Issue 12), MEDLINE (1966 to December 2012), EMBASE (1980 to December 2012), and CINAHL (1982 to December 2012) using a combination of the following text words and MeSH terms: [Infant, Newborn OR Infant, Premature OR Infant, Low Birth Weight OR Infant, Very Low Birth Weight/ OR infan* OR neonat* OR preterm OR prem*] AND "Infant-Nutrition"/ all subheadings OR Infant Formula OR milk OR formula OR trophic feeding OR minimal enteral nutrition OR gut priming]. The search outputs were limited with the relevant search filters for clinical trials. We did not apply a language restriction.
Searching other resources
We examined the reference lists in all studies identified as potentially relevant.
We searched the abstracts from the annual meetings of the Pediatric Academic Societies (1993 to 2012), the European Society for Pediatric Research (1995 to 2012), the UK Royal College of Paediatrics and Child Health (2000 to 2012), and the Perinatal Society of Australia and New Zealand (2000 to 2012). Trials reported only as abstracts were eligible if sufficient information was available from the report, or from contact with the authors, to fulfil the inclusion criteria.
Data collection and analysis
We used the standard methods of the Cochrane Neonatal Review Group (http://neonatal.cochrane.org/).
Selection of studies
Two review authors screened the titles and abstracts of all studies identified by the above search strategy. We assessed the full texts of any potentially eligible reports and those studies that did not meet all of the inclusion criteria were excluded. We discussed any disagreements until consensus was achieved.
Data extraction and management
We used a data collection form to aid extraction of relevant information from each included study. Two review authors extracted the data separately. Any disagreements were discussed until consensus was achieved. We contacted the investigators for further information if data from the trial reports were insufficient.
Assessment of risk of bias in included studies
We used the criteria and standard methods of the Cochrane Neonatal Review Group to assess the methodological quality of any included trials. Additional information from the trial authors was requested to clarify methodology and results as necessary. We evaluated and reported the following issues in the 'Risk of bias' tables.
- Sequence generation. We categorised the method used to generate the allocation sequence as:
- low risk, any random process e.g. random number table; computer random number generator;
- high risk, any non-random process e.g. odd or even date of birth; patient case-record number;
- Allocation concealment. We categorised the method used to conceal the allocation sequence as:
- low risk, e.g. telephone or central randomisation; consecutively numbered sealed opaque envelopes;
- high risk, open random allocation; unsealed or non-opaque envelopes; alternation; date of birth;
- Blinding. We assessed blinding of participants, clinicians and caregivers, and outcome assessors separately for different outcomes and categorised the methods as:
- low risk;
- high risk;
- Incomplete outcome data. We described the completeness of data including attrition and exclusions from the analysis for each outcome and any reasons for attrition or exclusion where reported. We assessed whether missing data were balanced across groups or were related to outcomes. Where sufficient information was reported or supplied by the trial authors, we re-included missing data in the analyses. We categorised completeness as:
- low risk, < 20% missing data;
- high risk, > 20% missing data;
Measures of treatment effect
We calculated risk ratio (RR) and risk difference (RD) for dichotomous data and weighted mean difference (WMD) for continuous data, with respective 95% confidence intervals (CI). The number needed to treat for benefit (NNTB) or harm (NNTH) was determined for a statistically significant difference in the RD.
Unit of analysis issues
The unit of analysis was the participating infant in individually randomised trials and the neonatal unit (or subunit) for cluster randomised trials.
Assessment of heterogeneity
If more than one trial was included in a meta-analysis, we examined the treatment effects of individual trials and the heterogeneity between trial results by inspecting the forest plots. We calculated the I² statistic for each analysis to quantify inconsistency across studies and describe the percentage of variability in effect estimates that may be due to heterogeneity rather than sampling error. If substantial (I² > 50%) heterogeneity was detected, we explored the possible causes (for example differences in study design, participants, interventions, or completeness of outcome assessments) in sensitivity analyses.
We used the fixed-effect model in RevMan 5 (RevMan 2011) for meta-analysis.
Subgroup analysis and investigation of heterogeneity
We planned the following subgroup analyses:
- trials in which most infants were exclusively formula fed;
- trials in which most infants were at least partially fed with human milk (maternal or donor);
- trials in which most participants were of extremely low birth weight (ELBW) (< 1000 grams) or extremely preterm gestational age (< 28 weeks);
- trials in which participants were infants with intrauterine growth restriction, or infants with absent or reversed end-diastolic flow velocities detected on antenatal Doppler studies of the fetal aorta or umbilical artery.
Description of studies
A total of 588 infants participated in the five included trials. The trials were undertaken in neonatal care centres in North America (Rayyis 1999; Caple 2004), India (Salhotra 2004; Krishnamurthy 2010) and Turkey (Karagol 2012) within the past 10 to 15 years.
All of the trials specified participant birth weight eligibility criteria:
- Rayyis 1999 < 1500 grams;
- Caple 2004 1000 to 2000 grams;
- Salhotra 2004 < 1250 grams;
- Krishnamurthy 2010 1000 to 1500 grams;
- Karagol 2012 750 to 1250 grams.
Since most participants in Caple 2004 were of birth weight < 1500 grams or gestational age < 32 weeks, a consensus decision to include the trial was made. Infants born 'small for gestational age' (birth weight < 10th percentile of the index population's distribution) were not eligible to participate in Caple 2004 but were included in the other trials. More than 95% of the participants in Salhotra 2004 were 'small for gestational age'. One third of participants in Karagol 2012 were ELBW infants.
Interventions and comparisons
All trials commenced interval bolus intragastric feeding within one to five days after birth. Infants were randomly allocated to one of two rates of daily increments in enteral feed volume:
- Rayyis 1999 15 versus 35 ml/kg/day;
- Caple 2004 20 versus 35 ml/kg/day;
- Salhotra 2004 15 versus 30 ml/kg/day;
- Krishnamurthy 2010 20 versus 30 ml/kg/day;
- Karagol 2012 20 versus 30 ml/kg/day.
In one trial, only formula-fed infants were eligible to participate (Rayyis 1999). In Caple 2004, Krishnamurthy 2010 and Karagol 2012 infants received either expressed breast milk or formula, or a combination of both. In Salhotra 2004, all participating infants were exclusively fed with expressed breast milk. All of the trial protocols specified indications for interrupting or ceasing enteral feeding such as residual gastric contents of more than about one-third of the previous feed volume, frequent vomiting, abdominal distention, or detection of blood in the stools (including occult blood).
All of the trials reported the incidence of NEC (Bell stage II/III) confirmed radiologically, or at surgery or autopsy. The other reported outcomes included time to regain birth weight, time to establish full enteral feeding, duration of hospital stay and rates of invasive infection.
Two trials were excluded (Book 1976; Berseth 2003) (see Characteristics of excluded studies). In Book 1976, enteral feeding volumes were advanced at 10 ml/kg/day versus 20 ml/kg/day, that is both groups received 'slow' advancement of feed volumes. In Berseth 2003, infants were randomly allocated to either a stable (not progressively increased) trophic feeding volume or to feed volume advancement at 20 ml/kg/day.
Risk of bias in included studies
The methodological quality of the included trials was generally good. In all five trials, methods to ensure adequate allocation concealment were employed and complete or near-complete assessments of the primary outcomes were reported. None of the trials were able to conceal the feeding strategies from parents, caregivers or clinical investigators. The assessment of abdominal radiographs (for diagnosis of NEC) was clearly masked in three studies. In Salhotra 2004 and Karagol 2012 it was unclear whether precautions had been taken to ensure that radiological assessors were blinded to the allocation group.
Effects of interventions
Incidence of necrotising enterocolitis (Outcome 1.1)
Meta-analysis did not detect a statistically significant effect: typical RR 0.97 (95% CI 0.54 to 1.74); typical RD -0.00 (95% CI -0.04 to 0.04) (Figure 1). There was not any statistical evidence of heterogeneity (I² = 0%).
|Figure 1. Forest plot of comparison: 1 Slow versus faster rates of feed advancement, outcome: 1.1 Incidence of necrotising enterocolitis.|
Mortality (Outcome 1.2)
Meta-analysis did not find a statistically significant difference: typical RR 1.41 (95% CI 0.81 to 2.47); typical RD 0.03 (95% CI -0.02 to 0.08) (Figure 2). There was not any statistical evidence of heterogeneity (I² = 0%).
|Figure 2. Forest plot of comparison: 1 Slow versus faster rates of feed advancement, outcome: 1.2 Mortality.|
All five trials reported that it took a statistically significantly longer time to regain birth weight in infants in the slow rate of advancement group:
- Rayyis 1999, median difference 3 (95% CI not given) days;
- Caple 2004, median difference 2 (95% CI 1, 3) days;
- Salhotra 2004, median difference 5 (95% CI not given) days;
- Krishnamurthy 2010, median difference 6 (95% CI not given) days;
- Karagol 2012, mean difference 3.8 (95% CI not given) days.
Longer-term growth parameters were not reported by any of the trials.
None of the trials assessed any neurodevelopmental outcomes.
Time to establish full enteral feeding
All five of the trials reported that it took statistically significantly longer to establish full enteral feeds in infants in the slow rate of advancement group:
- Rayyis 1999, median difference 4 (95% CI not given) days;
- Caple 2004, median difference 3 (95% CI 2 to 3) days;
- Salhotra 2004, mean difference 4.8 (95% CI not given) days;
- Krishnamurthy 2010, median difference 2 (95% CI not given) days;
- Karagol 2012, mean difference 3.2 (95% CI not given) days.
Time to establish full oral feeding
Not reported by any of the included trials.
Feeds intolerance (causing interruption of enteral feeding) (Outcome 1.3)
Meta-analysis of data from Salhotra 2004, Krishnamurthy 2010 and Karagol 2012 did not find a statistically significant difference: typical RR 1.29 (95% CI 0.90 to 1.85); typical RD 0.08 (95% CI -0.03 to 0.19) (Figure 3).
|Figure 3. Forest plot of comparison: 1 Slow versus faster rates of feed advancement, outcome: 1.3 Feeds intolerance (causing interruption of enteral feeding).|
Incidence of invasive infection (Outcome 1.4)
Meta-analysis of data from Krishnamurthy 2010 and Karagol 2012 did not detect a statistically significant effect: typical RR 1.50 (95% CI 0.71 to 3.16); typical RD 0.05 (95% CI -0.04 to 0.15) (Figure 4).
|Figure 4. Forest plot of comparison: 1 Slow versus faster rates of feed advancement, outcome: 1.4 Incidence of invasive infection.|
Duration of hospital stay
Two trials did not detect a statistically significant difference:
- Rayyis 1999, median difference 4 (95% CI not given) days;
- Caple 2004, mMedian difference 5 (95% CI -1 to 8) days.
Two trials reported that the duration of hospital stay was statistically significantly longer in infants in the slow rate of advancement group:
- Krishnamurthy 2010, median difference 1.5 (95% CI not given) days;
- Karagol 2012, mean difference 6 (95% CI not given) days.
1. Exclusively formula-fed infants (Rayyis 1999). No statistically significant differences were detected:
- NEC RR 1.44 (95% CI 0.63 to 3.32); RD 0.04 (95% CI -0.05 to 0.13);
- mortality RR 0.59 (95% CI 0.10 to 3.46); RD -0.01 (95% CI -0.06 to 0.03).
2. Infants at least partially fed with human milk. Subgroup data were not available.
3. ELBW or extremely preterm infants. None of the trials recruited predominantly ELBW or extremely preterm infants.
4. Infants with intrauterine growth restriction (Salhotra 2004). No statistically significant differences:
- NEC RR 0.21 (95% CI 0.01 to 4.12); RD -0.07 (95% CI -0.19 to 0.04) (Figure 1);
- mortality RR 1.78 (95% CI 0.83 to 3.81); RD 0.20 (95% CI -0.05 to 0.46) (Figure 2).
Summary of main results
The currently available trial data do not provide evidence that slowly advancing enteral feed volumes reduces the risk of NEC in very preterm or VLBW infants. The boundaries of the 95% CI for the estimate of risk difference are consistent with either one extra or one fewer case of NEC in every 25 infants who have slow rates of feed advancement. Infants who had slow advancement of feed volumes regained their birth weight two to six days later than infants who had faster rates of advancement. The clinical importance of this effect is unclear as long-term growth or developmental outcomes have not been assessed. Similarly, infants who had feed volumes advanced at a slow rate established full enteral feeding two to five days later than infants who had faster rates of advancement. Whether this is associated with important clinical adverse consequences such as a higher rate of nosocomial infection secondary to prolonged use of parenteral nutrition is not yet known as few studies have reported this outcome. Despite the effect on the establishment of enteral feeding, the included trials did not find consistent evidence of an important effect on the duration of hospital admission.
Overall completeness and applicability of evidence
These findings should be applied with caution for several reasons. None of the studies included predominantly ELBW or extremely preterm infants, known to be at the highest risk of NEC (Luig 2005). One third of the participants in Karagol 2012 were of ELBW but only a minority of infants in the two larger trials weighed < 1000 grams or were < 28 weeks gestation at birth, or had evidence of intrauterine growth restriction (Rayyis 1999; Caple 2004). Infants who had severe respiratory distress requiring oxygen supplementation or ventilatory support were not eligible to participate in three of the trials (Salhotra 2004; Krishnamurthy 2010; Karagol 2012). The findings may not be applicable to these populations, at highest risk of developing feed intolerance or NEC.
Less than one-half of the total number of participating infants were fed with breast milk. Evidence exists that artificial formula feeding increases the risk of feed intolerance and NEC (Quigley 2007). The risk-benefit balance of enteral feeding strategies may differ between human milk-fed and formula-fed very preterm or VLBW infants. It is also unclear whether the findings can be applied to infants who receive continuous infusion of intragastric feeds as all of the infants in the included trials received enteral feeds as interval boluses. Randomised controlled trials have reported conflicting findings about the effect of continuous enteral infusion on feed tolerance in very preterm or VLBW infants (Premji 2011).
Although the finding that slow enteral feed volume advancement delays the establishment of full enteral feeds seems intuitive, it is also plausible that advancing feed volumes faster could have resulted in more feed intolerance and therefore a delay in the establishment of full enteral feeding. The included trials all pre-specified definitions of feed intolerance that mandated interrupting or ceasing feed volume advancement, principally the detection of 'gastric residuals' (the gastric content aspirated prior to a planned gastric tube feed) and abdominal distension. However, the trial reports presented only limited data on the frequency of these outcomes. Furthermore, only limited evidence exists that the volume or colour of gastric residuals is predictive of the risk of NEC for infants whose feed volumes are advanced conservatively (Mihatsch 2002; Cobb 2004; Bertino 2009). Similarly, the clinical importance of abdominal distension or bowel loops visible through the abdominal wall (without other features of intra-abdominal pathology) is unclear, especially in the modern era when early and prolonged use of continuous positive airway pressure results in intestinal gaseous distension.
Quality of the evidence
The included trials were generally of good methodological quality but, in common with other trials of feeding interventions in this population, it was not possible to mask caregivers and clinical assessors to the nature of the intervention (Figure 5). Although the lack of blinding may have resulted in surveillance and ascertainment biases, this is more likely to have caused an overestimation of the incidence of feed intolerance and NEC in infants whose feed volumes were advanced faster. The assessment of abdominal radiographs for signs of NEC was masked in most trials to ensure that the diagnosis of severe NEC (confirmed by the radiological detection of gas in the bowel wall or portal tract) was not prone to bias. However, since the microbial generation of gas in the bowel wall is substrate dependent, infants who received more enteral milk (substrate) may have been more likely to demonstrate this radiological sign than infants with equally severe bowel disease who had less intraluminal substrate. This 'substrate effect' is also more likely to cause over-ascertainment of NEC in the infants who had faster rates of feed volume advancement (Tyson 2007).
|Figure 5. Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies.|
Agreements and disagreements with other studies or reviews
This review specifically focused on the comparison of slow versus faster rates of feed volume advancement and did not compare progressive advancement with enteral fasting or trophic feeding (minimal enteral nutrition). Only one randomised controlled trial has compared trophic feeding with progressive enteral feed volume advancement (at daily increments of 20 ml/kg) (Berseth 2003). Although the trial found the risk of NEC to be statistically significantly higher in the infants whose feed volumes were progressively advanced, this finding should be interpreted cautiously. The trial was stopped early following an interim analysis and therefore the finding of an effect on the incidence of NEC may be spurious (Montori 2005). Caregivers and assessors were not blind to the intervention. As discussed above, this may have resulted in several sources of bias that are likely to cause an over-estimation of the incidence of NEC in infants whose feed volumes are being advanced.
Implications for practice
These data suggest that slowly advancing enteral feed volumes does not reduce the risk of NEC in very preterm or VLBW infants. Increasing the volume of enteral feeds at slow rates (< 24 ml/kg/day) results in several days delay in the time taken to regain birth weight and establish full enteral feeds. The long-term clinical importance of these effects is unclear. Only limited data are available on the effect of this intervention on outcomes for extremely preterm or ELBW infants or infants who are growth restricted. Although current practice tends to favour a conservative approach to enteral feeding in these populations, it also needs to be considered that there are other possible consequences of slowly advancing feed volumes such as prolonging the use of parenteral nutrition that may be associated with important adverse clinical outcomes.
Implications for research
Further randomised controlled trials could provide more precise estimates of the effects of different rates of daily increases in enteral feed volumes on important outcomes for very preterm or VLBW infants. Trials should aim to ensure the participation of ELBW and extremely preterm infants as well as infants with evidence of compromised intrauterine growth so that subgroup analyses can be planned for these populations at high risk of NEC. Masking caregivers and investigators to the nature of this intervention is unlikely to be possible. Since the unblinded evaluation of feed intolerance and NEC is subject to surveillance and ascertainment biases, trials could aim to assess more objective outcomes, principally mortality and long-term growth and development. Furthermore, since conservative feeding strategies may result in other 'competing outcomes', such as invasive infection that may affect long-term survival and neurodisability rates, it is essential that trials are powered and structured to assess these outcomes.
We gratefully acknowledge the contributions of Drs Kennedy, Tyson, Chamnanvanakij and Bombell to previous iterations. We are grateful to Dr Namasivayam Ambalavanan for providing further details and data from his trial (Rayyis 1999).
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Index terms
Last assessed as up-to-date: 28 December 2012.
Protocol first published: Issue 4, 1998
Review first published: Issue 4, 1998
Contributions of authors
Drs Morgan and Young updated the search, independently determined the eligibility of identified studies, assessed the methodological quality of the included trials, and extracted the relevant information and data. All authors completed the final review.
Declarations of interest
Sources of support
- Centre for Reviews and Dissemination, Hull York Medical School, UK.
- National Institute for Health Research (NIHR), UK.Jessie Morgan and Lauren Young are NIHR Academic Clinical Fellows in Child Health.
- Eunice Kennedy Shriver National Institute of Child Health and Human Development National Institutes of Health, Department of Health and Human Services, USA.Editorial support of the Cochrane Neonatal Review Group has been funded with Federal funds from the Eunice Kennedy Shriver National Institute of Child Health and Human Development National Institutes of Health, Department of Health and Human Services, USA, under Contract No. HHSN267200603418C.
Medical Subject Headings (MeSH)
*Infant, Very Low Birth Weight; Enteral Nutrition [*methods]; Enterocolitis, Necrotizing [*prevention & control]; Infant, Low Birth Weight; Infant, Newborn; Infant, Premature; Infant, Premature, Diseases [*prevention & control]; Parenteral Nutrition [adverse effects]; Randomized Controlled Trials as Topic
MeSH check words