Nasogastric Intubation After Abdominal Surgery: A Meta-analysis of Recent Literature

Objective  To determine whether refraining from nasogastric intubation (NGI) in patients after abdominal surgery will result in the same therapeutic effectiveness as using NGI.

Data Source  We identified randomized trials from the Cochrane Central Register of Controlled Trials published between January 1990 and January 2005.

Study Selection  Two of us independently selected trials based on randomization, abdominal surgery in patients, early vs late removal of the NGI, and reporting at least 1 of the following end points: hospital stay, gastrointestinal function, and postoperative complications.

Data Extraction  Two of us independently performed trial quality assessment and data extraction. Trials were judged using a structured list that included factors relating to internal and external validity. Data were entered and analyzed by means of dedicated software from the Cochrane Collaboration.

Data Synthesis  Seventeen randomized trials met the inclusion criteria. Meta-analysis showed that NGI does not offer any clinically relevant benefits for patients after abdominal surgery, such as recovery of gastrointestinal function or reduction of postoperative complications (relative risk, 1.18; 95% confidence interval, 0.98-1.42). Moreover, NGI showed some undesired effects, such as discomfort (in 60% of the NGI patients) and a later return to a liquid diet (weighted mean difference, 0.65 days; 95% confidence interval, 0.38-0.92 days) or a regular diet, whereas hospital stay was not shortened.

Conclusions  Routine NGI seems to serve no beneficial purpose and may even be harmful in patients after modern abdominal surgery; also, it is uncomfortable. Therefore, NGI is recommended only as a therapeutic approach.

Prophylactic nasogastric intubation (NGI) after abdominal surgery until gastrointestinal (GI) function returns is a routine postoperative procedure in many medical centers. There is widespread acceptance that NGI during abdominal surgery allows for better exposure to the operating field.¹ However, during the 1980s and 1990s, the need for routine NGI was repeatedly challenged on the basis of various trials.^2-6In 1995, Cheatham et al⁶ concluded in a meta-analysis of the literature published up to 1993 that the routine use of NGI after elective surgery did not reduce the risk of ileus and aspiration. To avoid nausea, vomiting, or abdominal distention in 1 patient, at least 20 patients had to be treated with routine NGI. If NGI was not used routinely but only selectively, patients experienced substantially fewer postoperative complications, such as pneumonia, atelectasis, and fever. In addition, NGI could even prolong patient recovery time and hospital stay.⁶ Nevertheless, NGI still seems to be a common procedure.

During the past decade, more randomized controlled trials (RCTs) have been published on this subject. Furthermore, the present anesthetic (eg, epidural analgesic methods)^7,8 and surgical techniques (eg, laparoscopic^9,10 and fast-track^11,12 surgery) are different from those in the period from which the previous meta-analysis dated. These modern techniques have led to an acceleration of patient recovery and a reduction in the hospital stay, which has become an important goal in health care policy. Hence, the aim of this study is to systematically review the recently available literature by using the modern techniques of meta-analysis and by weighing the number of patients included in each trial. This should provide updated evidence to guide surgeons (and medical decision makers) as to the indication for NGI.

To identify trials of potential relevance and adequate design, we built a sensitive search strategy for the Cochrane Central Register of Controlled Trials, presently the largest database of RCTs. The search was limited to RCTs published in the past 15 years to avoid any effect of older surgical and anesthetic techniques, which are likely to have progressed. The search was not limited by language or publication status.

Last execution of the search strategy was January 2005. Two of us (H.V. and M.N.S.-V.) independently screened the titles and abstracts of references identified by the search on potential relevance and design (eg, no RCT and no comparison of early- vs late-removal NGI). Full versions of articles were obtained and checked independently to identify those that met the inclusion criteria. All RCTs that evaluated the effectiveness of early or late removal of the NGI after abdominal surgery were studied. We decided beforehand to include RCTs only.¹³ Trials were also included if the feasibility, safety, and tolerance of early feeding in patients undergoing surgery was studied after early removal of the NGI in 1 of the 2 groups. Trial data published in duplicate were included only once. Cases of disagreement or doubt about eligibility were resolved by discussion, and, if necessary, a third author (D.T.U.) was consulted.

Subsequently, the methodological quality of each trial was assessed systematically and independently (H.V. and M.N.S.-V.) according to the Dutch Cochrane Centre list of factors relating to internal and external validity. Any disagreement was referred to a third author (D.T.U.) for adjudication.

One of us (H.V.) extracted and summarized the details of the trials: design, type of surgery, number of participants, application of NGI in the intervention and control groups, trial end points, and results. This was checked by a second, independent author (D.T.U.) to avoid data input errors. The 2 primary and patient-relevant outcomes were duration of the hospital stay and time to first regular diet. Secondary outcomes were GI function, time to first peristaltic movements, passage of flatus, passage of stools, liquid and regular diet, postoperative complications (especially aspiration, atelectasis, and fever because these are probably NGI related), vomiting, nausea, gastric or abdominal distention, reinsertion of a nasogastric tube, specific NGI-related complications (ie, ulcers of the nose or perforation due to the tube), and discomfort.

Quantitative data were entered into the computer by one of us (H.V.) and analyzed using the Cochrane Collaboration’s associated software (RevMan Analyses 1.0.2, Oxford, England). Data entry and analysis were checked by a second author (D.T.U.). For each outcome, a meta-analysis was performed, and summary estimates of treatment effect (with 95% confidence intervals [CIs]) were calculated for every comparison. For continuous outcomes, weighted mean differences (WMDs), and for dichotomous outcomes, relative risks (RRs), were calculated.¹⁴ Statistical heterogeneity was tested using χ² and I² tests. No totals were calculated if trial heterogeneity was large (I²>60%). Random-effects models were used when statistical heterogeneity among the trials was moderate, that is, when I² was 30% to 60%, and fixed models were used when I² was less than 30%. Further analyses were performed to test for publication bias (funnel plots) and subgroup effects among the trials.¹⁵ Sensitivity analysis was performed by deselecting trials with extreme findings to test the robustness of the results.

The search identified 208 studies, of which 188 were found to be ineligible. Two studies^16,17 were excluded after discussion: one¹⁶ in which only late complications (incisional hernia) of NGI were studied and another¹⁷ in which NGI seemed to be stopped immediately after surgery in both groups. Eighteen trials met the inclusion criteria. Because Koukouras et al¹⁸ apparently had copied the trial published earlier by Montalto et al,¹⁹ we excluded the former trial, leaving 17 trials for analysis. We did not contact the authors or manufacturers and distributors of nasogastric tubes for details of unpublished or ongoing trials.

Allocation was by means of randomization (ie, sealed envelopes) in 1 trial,⁴ by day and month in 1 trial,²⁰ and by tables of random numbers in 2 trials^3,21; the method of randomization was not stated in 10 trials.^1,19,22-29 Randomization was based on a computerized randomization table only in 3 trials.^30-32 Therefore, concealment of allocation was ensured only in these 3 trials. The nature of the intervention made blinding impossible for patients and medical professionals. In 10 trials, the articles were unclear on the blinding of the outcome assessor, which is also difficult for this intervention.

Most trials showed comparable characteristics in their patient groups at baseline (Table 1). The criterion of similar treatment apart from intervention was not met in the trial by Montalto et al¹⁹ because the selective NGI group received standard medication after removing the tube (metoclopramide and ranitidine). None of the trials reported a priori calculation of the required sample size.

The included trials were published between 1990 and 2002. Trial sizes varied from 74 to 1000 (median, 109) patients (N = 2911). All 17 included trials measured at least 1 end point of interest. Four of the 17 trials comprised patients with non-GI surgery,^25,30-32 3 involved gastric surgery,^1,21,26 6 contained a mix of GI and non-GI surgery,^{3,19,20,24,28,29} 3 focused on colon surgery,^22,23,27 and 1 included small- and large-bowel surgery⁴ (Table 2).

In all the trials, early vs late removal of NGI was studied. In the early-removal group, NGI was terminated at the end of the operation in 53% (9/17) of the trials.^{3,4,20,23,26-29,32} In 4 trials,^19,25,30,31 NGI was discontinued within 24 hours after surgery. In the remaining 4 trials,^1,21,22,24 patients received no NGI at all. In 14 trials, the routinely placed NGI was removed “late,” when first flatus passed,^{1,4,19,21,25,26,30} bowel sounds returned,^3,27,31 GI functions returned,^24,32 peristalsis returned,²⁰ or the ileus seemed to be resolved.² In 3 trials, the NGI was removed after at least 24 hours²⁸ or after 3 days.^22,29Table 3 provides an overview of the results of the meta-analysis for each outcome.

In 3 trials,^3,19,23 no standard deviations were reported with their measured outcomes; therefore, these trials could not be incorporated into the meta-analysis. We performed a meta-analysis and calculated an overall effect size for only 2 outcomes: passage of feces and return to a liquid diet, which occurred a half day earlier in the no-NGI group (WMD, 0.65 days; 95% CI, 0.38-0.92 days). No substantial differences were reported for the other outcomes. Type of surgery did not affect the results differentially.

In 2 trials,^4,29 a significantly earlier return of peristaltic movements in favor of no NGI was reported: WMDs of 0.10 days (95% CI, 0.09-0.11 days) and 0.50 days (95% CI, 0.41-0.59 days), respectively. Both studies included patients with a mix of GI and non-GI surgery. This positive effect for no NGI was also suggested by another trial,¹ in which patients with gastric resections were studied (WMD, 0.10 days; 95% CI, −0.12 to 0.32 days).

In 4 trials,^1,4,29,32 earlier return of flatus passage in favor of no NGI was reported, with a mean of a half day. In 1 of these trials,¹ this difference was not significant (0.30 days; 95% CI, −0.02 to 0.62 days). In 2^4,29 of these 4 trials, a mix of GI and non-GI surgery was studied, in 1 trial³² only non-GI surgery, and in 1 trial¹ patients with gastric resections. Three trials^4,24,29 could be incorporated into the meta-analysis for passage of feces (Figure 1). The WMD (random effect) showed a tendency in favor of no NGI, with a WMD of 0.20 days (95% CI, −0.07 to 0.48 days) for earlier passage of feces. In all 3 trials, a mix of patients with GI and non-GI surgery was studied.

Three trials^1,25,32 were included in the meta-analysis of liquid diet (Figure 2). The WMD (fixed effect) showed a significant effect in favor of no NGI, with a 0.65 days (95% CI, 0.38-0.92 days) earlier return to a liquid diet. These 3 trials included patients with non-GI surgery^25,32 or patients with gastric resections.¹ Return to a regular diet was studied in 5 trials.^{1,24,25,29,32} In 2 trials,^1,29 a significantly earlier return to a regular diet was reported of nearly a half day in favor of no NGI: 0.40 days (95% CI, 0.32-0.48 days) and 0.60 days (95% CI, 0.24-0.96 days), respectively. In 1 trial,²⁴ the opposite was reported, that is, an earlier return to a regular diet of −0.45 days (95% CI, −0.79 to −0.11 days) in the NGI group. In the other 2 trials,^25,32 no significant differences were reported: −0.10 days (95% CI, −0.80 to 0.60 days) and 0.20 days (95% CI, 0.24-0.96 days), respectively. In 2 of the 5 trials, patients with a mix of GI and non-GI surgery were studied,^25,29 2 other trials studied patients with non-GI surgery,^25,32 and 1 trial included patients with gastric resections.¹

Meta-analysis was performed for all complication outcomes mentioned separately. Complication rates did not significantly differ between groups, except for discomfort, which was significantly higher in the NGI group (RR, 70.57; 95% CI, 14.2-349.6). Twelve trials were included in the postoperative complications meta-analysis (Figure 3). These trials were divided into 3 subgroups: no NGI at all,^1,21,22,24 NGI removed at the end of the operation,^{3,4,20,23,27,28} and NGI removed within 24 hours.^19,25 No overall and subgroup differences were observed.

Of the 7 included trials^{1,4,20,26,27,29,32} that studied nausea, only one¹ reported a statistically significant positive effect in the no-NGI group for nausea (RR, 2.77; 95% CI, 1.43-5.35). Vomiting tended to occur more in the NGI group. A statistically significant effect in favor of this complication in the NGI group was reported in only 1 trial²⁴ (RR, 0.06; 95% CI, 0.01-0.43). In this trial, the patients did not undergo NGI.

There was no significant difference between the groups in the number of reinsertions of the NGI. Reinsertion NGI was required significantly more often in the no-NGI group in only 2 trials.^23,24 This means that 20 (95% CI, 11-100) patients need to receive routine NGI to avoid 1 extra reinsertion for patients in whom the tube was removed early. In 1 trial,¹⁹ significantly more abdominal distention was reported in the no-NGI group (RR, 1.56; 95% CI, 1.11-2.18). In 5 other trials,^{4,21,24,27,32} no significant differences were reported. Discomfort was studied in only 3 trials.^19,22,29 Overall, 60% of the patients who underwent NGI reported discomfort.

In 9 trials, the duration of the hospital stay was studied. However, in 5 trials,^3,4,19,20,31 standard deviations were not reported. Meta-analysis could not be performed owing to large trial heterogeneity. Only 1 trial²⁴ reported a significant hospitalization difference in favor of no NGI (WMD, 3.63 days; 95% CI, 3.14-4.12 days). Two trials^1,32 reported a tendency toward shorter hospitalization when no NGI was used: WMD of 0.40 days (95% CI, −0.62 to 1.42 days) and 0.80 days (95% CI, −0.22 to 1.82 days), respectively. In 1 trial,²⁵ a trend toward longer hospitalization in the no-NGI group was reported (WMD, −0.90 days; 95% CI, −2.63 to 0.83 days).

Funnel plots were used to detect any publication bias for outcomes on which at least 5 trials were available, that is, postoperative complications and reinsertion NGI (Figure 4 and Figure 5). Plot asymmetry was reported for none of the outcomes. One of 9 trials reporting on vomiting showed extreme results in favor of NGI. This trial did not substantially affect the results after exclusion from the meta-analysis.

Subgroup analysis was performed for surgery type (non-GI surgery, gastric surgery, mix of GI and non-GI surgery, and bowel surgery) and moment of NGI removal (no NGI, NGI removed after surgery, and NGI removed within 24 hours). Subdivision for surgery type yielded no additional information. Postoperative complications tended to be diminished with earlier removal of the NGI (Figure 3). We also performed sensitivity analyses for various postoperative complications. Excluding the trial by Savassi-Rocha et al²⁰ did not change the overall effect size. The same was true for deselecting the trial by Montgomery et al²⁴ for the outcome of vomiting.

This meta-analysis of the recent literature shows that routine postoperative NGI does not offer any clinically relevant benefits for patients after abdominal surgery but merely results in some undesired patient-relevant outcomes, such as substantial (and often neglected) discomfort and return to a liquid or regular diet, whereas hospital stay is not shortened by using NGI. In patients who were not routinely treated with NGI, we demonstrated an earlier return of GI functions. Therefore, the routine use of NGI after abdominal surgery should be abandoned. This holds for GI, gynecologic, and other abdominal surgical procedures. Abandoning routine NGI also fits into modern postoperative recovery strategies, including fast-track surgery.^7,8

Similar conclusions were also drawn by Cheatham et al⁶ in their earlier review and recently by Nelson et al.³³ However, there are several differences compared with this meta-analysis. First, the present meta-analysis included only recent RCTs (published after 1990) because anesthetic^9,10 and surgical (eg, laparoscopic^11,12) techniques and recovery programs (eg, fast-track surgery^7,8) have improved surgical practice since then. These changes were made to accelerate postoperative recovery and to reduce the hospital stay, which has indeed been achieved. Second, we selected patient-relevant outcomes, which are likely to be momentous to professionals making health care decisions. Moreover, we used modern statistical techniques for meta-analysis and for detecting publication bias (using funnel plots) and weighing of the number of patients included in each trial to reach a higher standard of transparency and accuracy in our meta-analysis.

Despite the accumulating evidence against its use, we can foresee that surgeons will still prefer NGI under certain circumstances in some patient categories, for example, in abdominal surgical procedures with a high risk of postoperative gastric dilatation or ileus (eg, esophageal and gastric surgery, pancreaticoduodenectomy [Whipple procedure], and ileoanal pouch procedures). However, these indications are based solely on “gut feeling,” and scientific evidence is lacking to support the use of NGI. Therefore, even in high-risk patients, a strategy of selective or therapeutic use of NGI seems more appropriate than a routine or prophylactic use. To avoid aspiration pneumonia, one should also be aware of the possibility that NGI may hamper glottis function and compromise the airway³⁴ and thus could increase the chance of aspiration.

If one should still consider NGI, it should be used as a dependent system (gravity drainage) without suction, for a limited period, to avoid any harmful effects and discomfort. There is no need to wait until flatus or GI functions return to discontinue the NGI because our subgroup analysis indicated that the earlier the NGI is removed, the fewer postoperative complications will occur. Nevertheless, NGI still seems to be common practice in many medical centers on the basis of old-fashioned habits. However, given the present trend in surgery toward more evidence-based practice to improve health care quality,³⁵ the additional evidence found in this review should incite surgeons involved to give more attention to abandoning the routine use of NGI.

The strength of the conclusions of this review might be limited by some factors. We only searched the Cochrane Central Register of Controlled Trials. Hence, we theoretically might have missed some other recent studies. However, this register is the largest database containing RCTs, which makes the overlooking of important trials unlikely. Furthermore, some of the trials did not include standard deviations for all of their outcome estimates in their “Results” sections; therefore, these could not be pooled in our meta-analysis. However, these omissions did not seem to have led to substantial bias because the mean values of these outcomes generally pointed in the same direction. Finally, the methodological quality of the trials included in this review was fairly good, except for the inability to blind patients and investigators, which is virtually impossible, and the lack of a priori sample size calculations. These factors may contribute to the reason why trials produced equivocal results.

In conclusion, the routine use of NGI after abdominal surgery should be abandoned because it does not lead to any benefits but only to substantial discomfort and possibly even harm of the patient.

Correspondence: Hester Vermeulen, RN, MSc, Department of Surgery, G4-233, Academic Medical Center, Meibergdreef 9, PO Box 22700, 1100 DE Amsterdam, the Netherlands (H.Vermeulen@amc.uva.nl).

Accepted for Publication: May 11, 2005.