- Top of page
Postoperative ileus is common after abdominal surgery, especially after procedures involving the colon1, 2. Symptoms commonly associated with ileus include abdominal distension, pain, nausea, vomiting and intestinal cramps3. Postoperative ileus contributes to prolonged hospital stay and has a major economic impact on healthcare services4, 5. It has been estimated to account for additional hospital costs of US $750 million per year in the USA5, 6. Ileus delays oral food intake and may require treatment, including nasogastric tube decompression, electrolyte replacement and analgesia.
Owing to the significant implications of ileus after colectomy and the lack of effective therapies, surgeons have devised various preventive strategies7. Multimodal approaches used to treat ileus after colorectal surgery include the use of laparoscopy to minimize surgical trauma, an early return to solid food intake and mobilization, avoidance of nasogastric tubes, use of thoracic epidural analgesia and restrictive fluid management8. The aim of these so-called ‘fast-track programmes’ has been to provide pain- and stress-free operations with lower rates of organ dysfunction, thereby reducing morbidity and enhancing recovery9.
The intake of coffee might offer a simple approach to alleviating or preventing ileus after elective colectomy. Coffee is a popular beverage, and its effects on general well-being, the central nervous system and the cardiovascular system are well known10. Although coffee may stimulate bowel function in certain healthy volunteers11, there is limited scientific evidence regarding its effects on gastrointestinal function. In particular, there has been no prospective evaluation of its impact on intestinal function after elective colectomy. The aim of this randomized clinical trial was to assess whether coffee consumption accelerates the recovery of bowel function after elective colectomy.
- Top of page
This multicentre open-label randomized clinical trial was undertaken at one university hospital (University of Heidelberg, Germany, enrolment from March 2010) and two affiliated teaching hospitals (Krankenhaus Salem, Heidelberg, Germany, enrolment from October 2010; GRN Krankenhaus Sinsheim, Sinsheim, Germany, enrolment from May 2010). The departments of surgery of all three centres are headed by M.W.B., and patients are distributed to the centres according to operative complexity. Adult patients (aged at least 18 years) scheduled for elective open or laparoscopic colonic resection for malignant or benign diseases were eligible for inclusion in the study. Written informed consent was obtained on the day before surgery. Preoperative data, including coffee drinking and smoking habits, were obtained from a standardized, self-administered patient questionnaire. Patients were excluded if rectal resection was intended, a stoma was required or multivisceral resection was planned. Further exclusion criteria were known hypersensitivity or distaste for coffee, expected lack of compliance, and an impaired mental state. Randomized patients were excluded if the intended colonic surgery was not performed. The independent ethics committee of the University of Heidelberg approved the study.
Randomization and masking
A randomization list for 80 patients with an allocation ratio of 1:1 was generated with SAS® version 9.1 software (SAS Institute, Cary, North Carolina, USA), without stratification by centre. Allocation assignments were deposited in consecutively numbered, opaque, sealed envelopes and stored in the study office. For allocation the study manager was called the day before surgery after informed consent had been obtained. Patients were informed about the treatment assignment before surgery. Neither the physician nor the patient was blinded to the treatment assignment.
Objectives and endpoints
The primary objective of this trial was to investigate whether postoperative coffee intake reduces the duration of postoperative ileus after elective colectomy. The primary endpoint was time to the first postoperative bowel movement (time from the end of surgery until the first passage of stool recorded by nursing staff) as a surrogate endpoint for the duration of ileus. Secondary endpoints were time to tolerance of solid food (no vomiting), time to first flatus, need for additional laxatives, safety, and length of hospital stay (interval from day of operation until day of discharge). The time to first postoperative bowel movement, tolerance of solid food and first flatus was recorded in hours after the end of the operation. Safety was measured by the rate and degree of postoperative complications12. Anastomotic leakage was defined as the communication between intraluminal and extraluminal compartments owing to a defect in the intestinal wall at the site of the anastomosis, which was confirmed by imaging or relaparotomy13. Perioperative mortality was defined as any death within 30 days after surgery. All study endpoints were predefined in the study protocol.
Surgical technique and perioperative care
After enrolment, the same evidence-based protocol of perioperative management was applied to all patients, following the principles of fast-track surgery7. In brief, patients did not undergo mechanical bowel preparation and no oral antibiotic preparation was used. Patients scheduled for left hemicolectomy or rectosigmoid resection received one enema before surgery. All patients received a single dose of antibiotic prophylaxis consisting of 4 g mezlocillin and 0·5 g metronidazole at the time of anaesthesia induction. Low molecular weight heparin was administered and venous compression stockings were used, starting the night before surgery. Unless the patient refused, participants in both study arms received a thoracic epidural catheter (TEC) that was removed on day 3 after surgery or thereafter. Epidural catheters were filled with 0·2 per cent ropivacaine hydrochloride plus 0·25–0·5 µg/ml sufentanil, administered at a rate of 8–10 ml/h, with the rate reduced as quickly as possible. Peripheral non-steroidal analgesics (metamizole or paracetamol) formed the basis of the analgesic therapy. After the TEC had been removed, non-steroidal analgesics were combined with opioids (2 mg oxycodon intravenously or 10 mg Targin® (Mundipharma Medical, Hamilton, Bermuda) intramuscularly for breakthrough pain if necessary. Postoperative pain was assessed at least three times daily by means of a visual analogue scale and the pain regimen was adapted accordingly.
All operations were performed by, or under the close supervision of, a consultant surgeon, either laparoscopically or by an open approach via a midline laparotomy. A double-layer, handsewn anastomosis was performed in open surgery. During hand-assisted laparoscopic procedures, a circular stapler was used to achieve end-to-end or end-to-side colorectal anastomosis. Nasogastric tubes were removed during extubation.
The postoperative feeding regimen was standardized. In summary, water was offered 6 h after surgery, liquid food from the first day after surgery, and solid food from the second day. The postoperative feeding regimen was determined by the patient's ability and willingness to consume food and was not dependent on intestinal function, such as passage of flatus or bowel movements. Except for oral magnesium (1200 mg/day) and macrogol (polyethylene glycol; 3 bags, equivalent to 39·3 g/day) according to a fixed regimen in both study arms, no further laxative/prokinetic therapy was allowed. In the event of postoperative nausea and vomiting, patients received parenteral fluids and dimenhydrinate (62 mg in 10 ml injection solution) or granisetron (3 mg in 50 ml injection solution). The postoperative mobilization schedule was standardized according to in-house protocols and was the same for both study groups. Criteria for hospital discharge included stable vital signs with no febrile morbidity for at least 24 h, passage of stool, toleration of a regular diet, and the absence of other complications.
Patients in the treatment (coffee) arm had to drink three cups of coffee daily (100 ml at 08.00 12.00 and 16.00 hours), beginning on the morning after surgery. In the control arm coffee was replaced by warm water. Patients randomized to water were not allowed coffee at all until the first bowel movement had occurred. The beverages were served at 50–60 °C, and the patients were asked to drink the entire volume within 10 min under the supervision of a nurse. No additives (such as milk or sugar) were allowed. All patients were instructed not to consume black tea or additional coffee. However, they were free to drink any amount of still mineral water. No further restrictions on food consumption (for example chocolate) or smoking were imposed. Coffee was prepared with a conventional coffee machine using the same brand and type of coffee capsules (Lavazza Blue Espresso Dolce 100 per cent Arabica coffee, 8 g coffee per capsule; Lavazza Germany, Frankfurt, Germany).
For the power calculation, superiority of the primary endpoint, time to first bowel movement, in the treatment arm was assumed (H1 hypothesis). Previous studies on postoperative ileus after colectomy showed a mean time to the first bowel movement of 89·4–93·6 h for control patients, with a standard deviation of 24–36 h14, 15. To detect a clinically relevant absolute difference of 24 h in time to the first bowel movement with a two-sided significance level of 0·05 and a power of 80 per cent, it was calculated that 36 patients per study arm would be required assuming a standard deviation of 36 h. With an estimated dropout rate of 10 per cent (postoperative refusal or protocol violations), a total of 80 patients were needed for the study. No interim analysis was planned or performed, and no early stopping rules were implemented.
Continuous data are presented as mean(s.d.). Time to first bowel movement, time to tolerance of solid food and time to first flatus were assessed by means of t test and ANOVA, as appropriate. Other continuous data were compared using the Mann–Whitney U test, and categorical data with the χ2 test. Ninety-five per cent confidence intervals (c.i.) of proportions were calculated by the Wilson score method. All endpoints were analysed on an intention-to-treat (ITT) basis. For confirmation, per-protocol (PP) analysis was also carried out. In an ancillary analysis, the primary endpoint (mean ranks) was assessed by multivariable regression analysis with backward variable selection. P values were estimated with likelihood ratio tests. Two-sided P < 0·050 was considered statistically significant. Statistical analyses were performed using SPSS® version 17.0 (SPSS, Chicago, Illinois, USA) and the R Project for Statistical Computing version 2.12.2 (http://www.r-project.org).
- Top of page
Patients were recruited between March 2010 and March 2011. Of 103 eligible patients, 80 were randomized, 40 to each study arm (University Hospital Heidelberg 45, Salem 17, Sinsheim 18). Five patients randomized to receive coffee did not receive the allocated intervention: two were unable to drink coffee in the days after surgery and two could drink less than the scheduled amount (all four patients usually drink coffee). One patient received water owing to an allocation error. The mean amount of coffee consumed until the first bowel movement in the coffee group was 497(287) ml (excluding the five patients who did not receive the scheduled amount). One patient in the control arm underwent palliative enteric bypass surgery and was excluded from the study. A further four patients in this group did not comply with the protocol and drank 30–300 ml coffee. For ITT analysis patients were grouped according to the randomization, apart from the excluded patient for whom no primary endpoint was available. For the PP analysis, all patients who followed the protocol were considered (Fig. 1). Patient characteristics (Table 1), surgical procedures, duration of operation and perioperative results did not differ significantly between the two study groups in ITT analysis (Table 2).
Table 1. Patient characteristics
| ||Total (n = 79)||Water (n = 39)||Coffee (n = 40)|
|Sex ratio (M:F)||44:35||19:20||25:15|
|ASA grade|| || || |
| I||8 (10)||3 (8)||5 (13)|
| II||48 (61)||24 (62)||24 (60)|
| III||23 (29)||12 (31)||11 (28)|
|Type of colonic disease|| || || |
| Malignant||45 (57)||22 (56)||23 (58)|
| Benign||34 (43)||17 (44)||17 (43)|
|Diagnosis|| || || |
| Colonic cancer||44 (56)||20 (51)||24 (60)|
| Diverticular disease||22 (28)||11 (28)||11 (28)|
| Inflammatory bowel disease||10 (13)||6 (15)||4 (10)|
| Other||3 (4)||2 (5)||1 (3)|
|Coffee drinker before surgery||69 (87)||34 (87)||35 (88)|
|Preop. coffee intake (ml/day)*†||375(354)||315(220)||433(443)|
|Smoker||14 (18)||9 (23)||5 (13)|
|Co-morbidities|| || || |
| Cardiovascular||36 (46)||18 (46)||18 (45)|
| Pulmonary||10 (13)||4 (10)||6 (15)|
| Renal||5 (6)||1 (3)||4 (10)|
| Diabetes mellitus||14 (18)||5 (13)||9 (23)|
Table 2. Surgical procedures and perioperative results
| ||Total (n = 79)||Water (n = 39)||Coffee (n = 40)||P†|
|Surgical procedure|| || || ||0·392|
| Ileocaecal resection||10 (13)||6 (15)||4 (10)|| |
| Right hemicolectomy||26 (33)||15 (38)||11 (28)|| |
| Left hemicolectomy||9 (11)||5 (13)||4 (10)|| |
| Sigmoid/rectosigmoid||34 (43)||13 (33)||21 (53)|| |
| resection|| || || || |
|Surgical approach|| || || ||0·889|
| Open||48 (61)||24 (62)||24 (60)|| |
| Laparoscopic||31 (39)||15 (38)||16 (40)|| |
|Duration of operation (min)*||179(56)||183(57)||173(56)||0·410‡|
|Epidural analgesia||68 (86)||34 (87)||34 (85)||0·780|
|Blood loss (ml)*||176(162)||202(211)||150(87)||0·487‡|
|Patient received PRBC||10 (13)||7 (18)||3 (8)||0·163|
|ICU admission||15 (19)||8 (21)||7 (18)||0·733|
Reinsertion of a nasogastric tube was required in 16 patients (20 per cent). The time to first bowel movement was significantly shorter in the coffee arm (ITT: 60·4(21·3) versus 74·0(21·6) h, P = 0·006; PP: 62·1(21·5) versus 73·7(22·0) h, P = 0·028) (Tables 3 and 4). The absolute difference was 13·6 (95 per cent c.i. 4·0 to 23·2) h in ITT analysis and 11·6 (1·3 to 21·9) h in PP analysis. Univariable analysis demonstrated that time to first bowel movement was also significantly associated with the American Society of Anesthesiologists grade (Table 5). In multivariable regression analysis, coffee consumption and ASA grade were the only significant co-variables for the time to first bowel movement (Table 5). Sex and type of operation had some influence, but were not statistically significant. Anastomotic leakage had no effect.
Table 3. Outcome data: intention-to-treat analysis
| ||Total (n = 79)||Water (n = 39)||Coffee (n = 40)||Absolute difference†||P‡|
|Time to first bowel movement (h)||67·2(22·4)||74·0(21·6)||60·4(21·3)||13·6 (4·0, 23·2)||0·006§|
|Time to tolerance of solid food (h) (n = 76)||52·5(24·9)||55·8(30·0)||49·2(21·3)||6·6 (−5·4, 18·4)||0·276§|
|Time to first flatus (h) (n = 61)||43·3(18·2)||46·4(20·1)||40·6(16·1)||5·8 (−3·5, 15·2)||0·214§|
|Use of any laxative*||34 (43)||21 (54)||13 (33)|| ||0·055|
| Enema||15 (19)||10 (26)||5 (13)|| ||0·137|
| Prokinetics||5 (6)||3 (8)||2 (5)|| ||0·623|
| Enteral laxative||19 (24)||11 (28)||8 (20)|| ||0·394|
|Length of hospital stay (days)||11·1(4·4)||11·3(4·5)||10·8(4·4)||0·5 (−1·5, 2·6)||0·497¶|
|Reinsertion of nasogastric tube*||16 (20)||10 (26)||6 (15)|| ||0·239|
|Abdominal reoperation*||5 (6)||4 (10)||1 (3)|| ||0·157|
|Total morbidity*||18 (23)||10 (26)||8 (20)|| ||0·550|
|Grade of complications12*|| || || || ||0·344¶|
| I||8 (10)||3 (8)||5 (13)|| || |
| II||4 (5)||3 (8)||1 (3)|| || |
| IIIB||5 (6)||4 (10)||1 (3)|| || |
| IVA||1 (1)||0 (0)||1 (3)|| || |
|Anastomotic leakage*||6 (8)||5 (13)||1 (3)|| ||0·083|
|Grade of anastomotic leakage13*|| || || || ||0·206¶|
| A||2 (3)||1 (3)||1 (3)|| || |
| B||1 (1)||1 (3)||0 (0)|| || |
| C||3 (4)||3 (8)||0 (0)|| || |
Table 4. Outcome data: per-protocol analysis
| ||Total (n = 71)||Water (n = 36)||Coffee (n = 35)||Absolute difference†||P‡|
|Time to first bowel movement (h)||68·0(22·4)||73·7(22·0)||62·1(21·5)||11·6 (1·3, 21·9)||0·028|
|Time to tolerance of solid food (h) (n = 68)||51·4(24·5)||57·3(29·3)||45·1(16·4)||12·3 (0·7, 23·8)||0·038|
|Time to first flatus (h) (n = 55)||43·8(18·4)||47·1(20·5)||40·6(15·9)||6·5 (−3·4, 16·4)||0·191|
|Use of any laxative*||30 (42)||19 (53)||11 (31)|| ||0·069§|
|Length of hospital stay (days)||10·8(4·0)||11·6(4·6)||10·0(3·2)||1·6 (−0·3, 3·5)||0·129¶|
Table 5. Analysis of co-variables for the primary outcome
| ||Univariable analysis||Full regression model¶||Regression model#|
| ||Time to first bowel movement (h)*||P‡||β-coefficient†||P**||β-coefficient†||P**|
|Treatment arm|| ||0·006|| ||0·002|| ||< 0·001|
| Water||74·0(21·6)|| ||Reference|| ||Reference|| |
| Coffee||60·4(21·3)|| ||− 15·2 (−25·7, − 4·7)|| ||− 16·0 (−25·5, − 6·4)|| |
|Sex|| ||0·674|| ||0·068|| ||0·065|
| M||68·1(21·4)|| ||Reference|| ||Reference|| |
| F||66·0(23·9)|| ||− 9·1 (−20·0, 1·9)|| ||− 8·7 (−18·4, 1·1)|| |
|Age (years)|| ||0·163|| ||0·474|| || |
| < 65||63·9(20·1)|| ||Reference|| || || |
| ≥ 65||71·0(24·6)|| ||4·1 (−8·6, 16·8)|| || || |
|ASA grade|| ||0·020§|| ||0·009|| ||0·004|
| I||46·6(16·4)|| ||Reference|| ||Reference|| |
| II||68·9(21·3)|| ||25·0 (7·0, 42·9)|| ||25·0 (8·8, 41·1)|| |
| III||70·7(23·4)|| ||27·2 (6·0, 48·4)|| ||29·2 (10·7, 47·7)|| |
|Smoker|| ||0·702|| ||0·747|| || |
| No||67·6(23·3)|| ||Reference|| || || |
| Yes||65·1(18·4)|| ||− 2·1 (−16·5, 12·3)|| || || |
|Coffee drinker|| ||0·559|| ||0·779|| || |
| No||71·1(20·2)|| ||Reference|| || || |
| Yes||66·6(22·8)|| ||− 2·0 (−17·6, 13·6)|| || || |
|Type of colonic disease|| ||0·701|| ||0·531|| || |
| Malignant||66·3(21·4)|| ||Reference|| || || |
| Benign||68·3(23·9)|| ||3·5 (−8·8, 15·8)|| || || |
|Epidural analgesia|| ||0·750|| ||0·838|| || |
| No||65·1(25·3)|| ||Reference|| || || |
| Yes||67·5(22·1)|| ||− 2·2 (−18·0, 13·0)|| || || |
|Surgical procedure|| ||0·605§|| ||0·131|| ||0·074|
| Ileocaecal resection||59·9(20·3)|| ||Reference|| ||Reference|| |
| Right hemicolectomy||65·6(18·4)|| ||4·4 (−16·7, 25·5)|| ||2·2 (−13·7, 18·1)|| |
| Left hemicolectomy||72·1(25·2)|| ||12·8 (−9·6, 35·2)|| ||12·0 (−7·0, 31·0)|| |
| Sigmoid/rectosigmoid resection||69·2(25·2)|| ||15·0 (−2·2, 32·2)|| ||14·9 (−0·1, 29·9)|| |
|Surgical approach|| ||0·842|| ||0·769|| || |
| Open||66·8(20·2)|| ||Reference|| || || |
| Laparoscopic||67·8(26·0)|| ||1·2 (−11·9, 14·3)|| || || |
|Anastomotic leakage|| ||0·156|| ||0·651|| || |
| No||66·1(21·2)|| ||Reference|| || || |
| Yes||79·7(33·5)|| ||4·2 (−16·3, 24·6)|| || || |
As anastomotic leakage could hypothetically prolong the time to first bowel movement and more patients in the water group experienced an anastomotic leak (Table 3), a subgroup analysis was carried out on patients without a leak. Time to bowel movement remained significantly shorter in the coffee arm (ITT: 60·4 versus 74·0 h; P = 0·004).
The secondary endpoints—time to tolerance of solid food, time to first flatus (reported by 61 patients), need for laxatives and length of hospital stay—were reduced in the coffee arm, but there were no significant differences except for time to tolerance of solid food in PP analysis (P = 0·038) (Tables 3 and 4).
No patient died within 30 days of operation and the overall morbidity rate was 22 per cent, with no significant difference between the two groups. The rate of anastomotic leak (12·8 (95 per cent c.i. 5·6 to 26·7) versus 2·5 (0·4 to 12·9) per cent) and reoperations was higher in the control arm than in the coffee arm, but the differences were not significant.
- Top of page
In this randomized trial the time to first bowel movement was significantly shorter after regular coffee consumption during the postoperative period after elective colectomy. Although about 10 per cent of the patients did not want to drink strong coffee at this time, coffee consumption was well accepted by the patients and no coffee-related complications were noted.
Postoperative ileus is defined as the delayed return of physiological coordinated gastrointestinal motility and commonly occurs after intra-abdominal surgery, but also after retroperitoneal or even extra-abdominal surgery16, 17. It is usually self-resolving. The inhibition of small bowel motility is transient, and the stomach recovers within 24–38 h, whereas colonic function takes 48–72 h to recover6, 18, 19. The mechanisms involved in postoperative ileus include inhibitory sympathetic input, release of hormones, neurotransmitters and other mediators, inflammatory reactions, the effects of analgesics and the effects of surgical manipulation20. A timely return of bowel mobility is highly relevant clinically as a delay causes discomfort to the patient, prolongs hospital stay and increases healthcare expenditure5, 21. Sarawate and colleagues22 reported that hospital stay was on average 3 days longer for patients with postoperative ileus than for those without, leading to increased costs of US $4000–9000 per hospital stay for intra-abdominal operations22.
Multimodal approaches have been developed to reduce the incidence of postoperative ileus and the length of hospital stay, including the use of drugs, epidural anaesthesia, surgical techniques and early feeding23. Laparoscopic colonic surgery has been studied widely in recent years and has been shown consistently to improve many short-term outcomes compared with open surgery24, 25. Carli and co-workers26 reported that thoracic epidural analgesia has distinct advantages over patient-controlled morphine analgesia in providing superior quality of analgesia and shortening the duration of postoperative ileus26. On the other hand, a recently published multicentre trial showed that epidural analgesia and laparoscopy had no influence on the rate of ileus after sigmoid resection for diverticular disease27. These factors were also found to have no effect on time to first bowel movement in the present study, even though all types of elective colonic resection were included, for malignant as well as benign diseases.
Another approach is medical treatment with µ-receptor antagonists (such as alvimopan), particularly to target opioid-induced postoperative ileus28. In a study by Bell and colleagues29, use of alvimopan was associated with a significantly shorter mean time to gastrointestinal recovery (18·8 (95 per cent c.i. 13·2 to 24·4) h) and decrease of 1 full day in hospital stay compared with placebo, saving around US $900 per patient. The mean costs of coffee in the present study (on average five cups to first bowel movement) were US $2·80 for each patient. Compared with the costs of alvimopan (mean 8·9 12-mg doses, US $558 per patient)29, drinking coffee is a considerably cheaper treatment with a similar effect.
New medications for the treatment of postoperative ileus, such as ghrelin receptor agonists, are currently being tested in clinical studies. Another way of stimulating bowel function after colonic resection is sham feeding in the form of gum chewing. Asao and co-workers2 reported an earlier return of bowel function and a trend towards earlier hospital discharge in patients who chewed gum after laparoscopic colectomy. Chewing gum is postulated to activate the cephalic vagal reflex, which is usually enhanced by food, and to increase the production of gastrointestinal hormones that are associated with bowel motility2. However, none of these treatments has gained general acceptance in clinical practice.
Although there is a vast literature on the effects of coffee, little is known about its effect on the bowel. In a study in which healthy volunteers reported their desire to defaecate after consuming coffee, the rectosigmoidal motor responses to coffee consumption were measured using multiport manometry11. The authors found an increase in colonic motor activity 4 min after the ingestion of regular as well as decaffeinated coffee in the responders, but not among non-responders. None of the volunteers showed a response after the ingestion of water. In an investigation of the effects of 240 ml regular coffee, decaffeinated coffee, water and a 1000-kcal (approximately 4000 kJ) meal on colonic motility, Rao and colleagues30 performed ambulatory manometry with a catheter positioned from the rectum up to the mid-transverse colon3. Regular coffee, decaffeinated coffee and a meal induced more colonic activity and more propagated contractions than water. It was concluded that caffeinated coffee stimulates colonic motor activity to a magnitude comparable to that of a high-calorie meal. These previous studies examined the effect of coffee on colonic motility in healthy volunteers, whereas the present trial demonstrated that coffee can stimulate bowel activity in patients after elective colectomy. Thus, coffee consumption offers a simple strategy to shorten the time to first bowel movement.
Neither this nor previous studies can explain the mechanisms by which coffee stimulates intestinal motility. Although the most obvious stimulatory agent is caffeine, in a previous study decaffeinated coffee had an effect on bowel peristalsis30. Thus, components other than caffeine must play a role. Decaffeinated coffee was not used in the present trial, and it would be interesting to evaluate its effect in further studies. Alternatively, the effect of coffee could be physical instead of chemical in nature. For instance, acidity, osmolality or volume load could stimulate the gastrointestinal system. However, water did not show an effect in the present study and the physical properties of water are very similar to those of coffee, so a physical effect is less likely.
The study has several limitations. As there were no previous studies to rely on, the assumptions for the sample size calculation (difference (Δ) 24 h, s.d. 36 h) were rather arbitrary. The study did not reach its intended goal as the difference in time to first bowel movement between the study arms was just under 14 h. However, the standardized effect size (Δ/s.d.) of 0·64 is very close to the anticipated value of 0·67. Considering that coffee consumption is part of the lifestyle of many patients and its side-effects are well known, the effect of 14 h might still be of clinical interest. One might question the surrogate endpoint of time to first bowel movement as an indicator of ileus, but it is commonly used in similar studies27. Use of the incidence of ileus instead would have required a much larger patient cohort, making this initial trial unfeasible. Another limitation was the heterogeneity of the patient population, but this reflects the actual clinical situation. Although bias resulting from heterogeneity cannot be ruled out entirely, randomization and the results of the multivariable and subgroup analyses make this unlikely. The rate of protocol violation (particularly coffee consumption in the water arm) was relatively high, but not when the special circumstance of this study is considered. In contrast to typical pharmacological studies, all patients had free access to the test agent (coffee). Although patients were clearly instructed not to consume coffee if randomized to the control arm, they did not all comply. Furthermore, blinding was not feasible even though it would have been desirable in a randomized trial. Hospital stay in the present study was relatively long for patients in a fast-track programme. However, social and insurance issues made it necessary to prolong hospital stay beyond medical necessity.
Postoperative coffee consumption is a cheap and safe way to activate bowel motility after elective colonic surgery.