Oral celecoxib versus oral diclofenac for post-perineal repair analgesia after spontaneous vaginal birth: A randomised trial


: Dr Peng Chiong Tan, Department of Obstetrics and Gynaecology, University of Malaya, Lembah Pantai, 50603 Kuala Lumpur, Malaysia. Email: pctan@um.edu.my


Objective:  To compare oral celecoxib with oral diclofenac as pain reliever after perineal repair following normal vaginal birth.

Methods:  One hundred and sixty-four and 165 women, respectively, were randomised to 200 mg celecoxib or to 100 mg diclofenac orally 12 hourly for 24 h after perineal repair. A ten-point visual analog scale (VAS) for pain recorded at one, two, four, eight, 12 and 24 hours at rest and when mobilising was the primary outcome.

Results:  Repeated measures analysis of variance showed a larger reduction of VAS pain score at rest with celecoxib compared to diclofenac (P = 0.044). Mean VAS pain score at rest was 2.2 versus 2.7, 2.1 versus 2.5, 1.8 versus 2.2, 1.6 versus 1.6, 1.3 versus 1.4 at one, two, four, eight, 12 and 24 hours, respectively, for celecoxib versus diclofenac. The difference in pain score when mobilising was not significant (P = 0.75). Univariate analyses with the Student's t-test indicated significantly lower pain score at rest only at one, two and four hours with celecoxib.

Randomisation to celecoxib was associated with less upper gastrointestinal symptoms reported: 38 of 163 (23.3%) versus 57 of 165 (34.5%) (relative risk 0.67 95% CI 0.48–0.96: P = 0.029) but additional analgesia for breakthrough pain was not significantly different eight of 161 (5.0%) versus 18 of 165 (10.9%) (relative risk 0.46 95% CI 0.20–1.01: P = 0.065).

Conclusion:  Celecoxib was associated with a slightly lower VAS pain score at rest and less upper gastrointestinal symptoms were reported when compared to diclofenac.


Perineal pain on the first day following vaginal delivery occurs in 97% of women with episiotomies, 95% of women with first to second degree tears and 75% of women with intact perinea.1 Meta-analysis of randomised trials of episiotomy indicated that 71.5% of women assigned to a restrictive policy would sustained posterior perineal trauma compared to 81.6% assigned to a liberal policy indicating that posterior perineal trauma during vaginal birth is common and unavoidable.2

Suturing of perineal injury improves healing but does not alleviate pain.3 Continuous subcuticular suturing compared to placing interrupted sutures is associated with less short-term perineal pain.4 A similar reduction in short-term perineal pain has also been reported with the use of synthetic suture material as opposed to chromic catgut when performing perineal repairs.5

Oral diclofenac sodium, a non-steroidal anti-inflammatory drug (NSAID) is an established and effective analgesic for moderate to severe postoperative pain.6 Celecoxib, a more selective cyclooxygenase 2 inhibitor (COX-2 inhibitor) is a newer analgesic that is also effective for postoperative pain.7 Celecoxib and diclofenac have similar analgesic effect in chronic arthritic pain,8–10 and celecoxib is better tolerated with less adverse upper gastrointestinal events.8–11 Celecoxib and diclofenac also have similar analgesic effect in acute pain caused by ankle sprain.12 Both agents appeared to be safe in short-term use even in the elderly.13

The use of diclofenac and celecoxib is considered to be safe during lactation.14–16 Post-caesarean analgesia with epidural bupivacaine results in more effective breastfeeding and greater neonatal weight gain with less diclofenac needed for breakthrough pain indicating that effective analgesia can influence breast feeding.17

Diclofenac administered orally or rectally as pain reliever has been studied for episiotomy or perineal tear repair.18–20 We performed a PubMed search in all languages on 20 June 2007 (via http://www.ncbi.nlm.nih.gov/entrez/query.fcgi) using search terms celecoxib and episiotomy or perineal repair without any article being highlighted.

As perineal pain is very common and analgesia is usually required after repair of episiotomy or spontaneous perineal tear following vaginal delivery, we sought to compare celecoxib and diclofenac for their analgesic effect, tolerability and effect on breast feeding in a randomised study as this information is useful to guide analgesia use after normal vaginal birth. Celecoxib and diclofenac were widely available and frequently used for the relief of acute pain. They were appropriate choices for a randomised study.


A randomised trial was performed to compare the effect of oral celecoxib and oral diclofenac sodium following episiotomy or perineal tear repair after spontaneous vaginal birth. Ethical approval for the study was granted by the Ministry of Health of Malaysia Ethics Committee. Written informed consent was obtained from each participant. This study was performed at a fully state-funded public maternity hospital which also served as an affiliated teaching hospital in Penang, Malaysia, that had conducted 4632 deliveries in 2005.

Women who had spontaneously delivered a singleton fetus and sustained perineal damage that required repair were recruited and randomised as soon as possible after completion of suturing.

We excluded women who had known allergy to NSAIDs, epidural anaesthesia in labour, third or higher degree perineal tears, instrumental vaginal delivery, a history of peptic ulcer, asthma, thrombocytopaenia, renal impairment or severe post-partum haemorrhage of > 1500 mL. Recruitment was usually carried out by the provider who had conducted the delivery. Recruitment took place between January and June 2006.

Randomisation was by means of opening a sealed opaque envelope indicating treatment allocation A or B. Randomisation sequence was created in blocks of ten and the randomised treatment allocation was placed in numbered sealed opaque envelopes by an author (SL). A numbered envelope was assigned to a study woman according to sequence of recruitment. Sequential allocation of the numbered envelopes was confirmed by spot checks. The envelopes were kept in a box on the delivery suite.

We used commercially available celecoxib and diclofenac in our study. The drugs were kept in individually prepared boxes labelled A (containing celecoxib) or B (containing diclofenac sodium) – each box contained three 200 mg celecoxib doses and three 100 mg diclofenac doses with instruction to the ward staff for the drugs to be administered orally immediately, 12 h and 24 h after randomisation. Women were not told of their allocated treatment.

Women in our hospital who were not involved in our study who had perineal pain after childbirth were prescribed mefenamic acid as standard first-line analgesic. We opted not to use mefenamic acid as a study drug as we wanted an analgesic not in routine use to reduce potential bias of comparing a ‘new’ to an ‘old’ drug among ward staff.

Study women were informed that further analgesia was available if pain relief was inadequate. We allocated pethidine 50 mg i.m. as rescue analgesia. They were asked to note the time of commencement of their first successful attempt at breast feeding. Babies were breast feed as soon as the mothers were considered to be ready.

Subjects were given oral and written instructions on how to complete as primary outcomes, the ten-point visual analog scale (VAS) pain scores (0 for no pain to 10 for unbearable pain) at rest and if they had mobilised, when they were walking at one, two, four, eight, 12 and 24 hours. Ward nurses were briefed to remind women to record their VAS. They were also instructed in the marking of the overall VAS pain relief satisfaction score (0 for total dissatisfaction to 10 for complete satisfaction) and to complete the adverse symptoms questionnaire (yes or no answers) of symptoms over the 24-h study period as secondary outcome measures.

Episiotomy (all mediolateral) and second degree perineal tears were repaired in layers by the provider following local infiltration of 1% lignocaine (typically 10 mL volume) with chromic catgut suture. The perineal skin is typically closed with interrupted sutures. Study women received standard postnatal care and were mobilised as soon as could be tolerated.

We calculated that to detect a 0.5-point difference on a ten-point VAS for pain using the t-test, assuming standard deviation of 1.5, alpha of 0.05 and power of 80%, would require 142 women in each drug arm. Assuming a 10% dropout rate, a minimum of 316 women would need to be recruited.

Analysis is by intention to treat. Data were entered into a statistical software package spss version 14 (SPSS Inc., Chicago, IL, USA) and GraphPad Instat and QuickCalcs software (GraphPad Software Inc., San Diego, CA, USA) were also used for data analysis. Repeat measures analysis of variance was used on VAS pain score series at rest and when mobilising. The t-test was used to compare means, Fisher's exact test for categorical 2 &times; 2 datasets, χ2 for larger categorical datasets and relative risks (RR) and 95% confidence interval (CI) calculated using GraphPad Instat program. Numbers needed to treat (NNT) and its 95% CI was generated using GraphPad Quickcalc. P < 0.05 in any test was considered statistically significant and two-sided results were used for all tests.


We randomised 329 women –164 to celecoxib and 165 to diclofenac. Thirty women did not complete their VAS score sheets (15 randomised to celecoxib and 15 to diclofenac sodium) but some scores were available. In keeping with the principle of analysis by intention to treat, these 30 women were included. Three hundred and twenty-eight women were available for final analysis – trial flow chart in Figure 1.

Figure 1.

Recruitment flow chart for randomised trial of oral celecoxib or oral diclofenac sodium as pain reliever after perineal repair following vaginal birth.

The characteristics of study women were shown in Table 1. There was no significant difference between the randomised groups in any characteristic.

Table 1.  Characteristics of study women stratified according to treatment allocation to celecoxib or diclofenac orally for post-perineal repair analgesia after spontaneous vaginal birth
 Celecoxib n = 164Diclofenac n = 165
  1. Mean ± standard deviation, median (interquartile range) and number [%].

  2. Analyses by t-test for comparison of means and Fisher's exact test (for 2 &times; 2 datasets) or χ2 test for larger categorical datasets.

Age (years)    26.7 ± 4.3    26.8 ± 4.8
Primiparous    91 [55.8]   85 [51.5]
Malay  121 [74.2]  120 [72.7]
Chinese   26 [16.0]   22 [13.3]
Indian   9 [5.5]  15 [9.1]
Others   7 [4.3]   8 [4.8]
Induced labour   20 [12.3]  12 [7.3]
Estimated post-partum blood loss   214 ± 145    215 ± 113
≥ 500 mL   4 [2.5]   8 [4.8]
Blood transfusion   1 [0.6]   2 [1.2]
Perineal injury
Episiotomy  159 [97.5]  160 [97.0]
Second degree tear   4 [2.5]   5 [3.0]
Apgar score (5 min)  9 (1) 9 (1)
Apgar < 7 (5 min)   1 [0.6] 0 [0]
Birthweight (kg)       3.02 ± 0.47       3.06 ± 0.41
Medical problems in pregnancy
None  136 [84.0]  144 [87.3]
Diabetes   7 [4.3]   5 [3.0]
Hypertension   7 [4.3]   5 [3.0]
Anaemia  12 [7.4]  11 [6.7]

Repeated measures analysis of variance show the effect of the interventions did not change over time on VAS score at rest (P = 0.58) or when mobilised (P = 0.39). VAS pain score declined over time at rest (P < 0.001) and when mobilised (P < 0.001). The reduction in VAS score at rest was greater with celecoxib (F = 4.1, P = 0.044) but there was no difference (F = 0.1, P = 0.75) when mobilised.

On crude analysis with the t-test, women randomised to celecoxib reported significantly lower VAS pain score at rest one, two and four but not at eight, 12 and 24 hours. There was a 0.4–0.5 point difference on the ten-point VAS pain score at one, two and four hours in favour of celecoxib. VAS pain score when mobilised was not different at any time points (Table 2).

Table 2.  Visual analog scale* (VAS) for pain at rest and when mobilising at specified time intervals after celecoxib or diclofenac orally for post-perineal repair analgesia. Mean ± standard deviation
TimeVAS pain score at rest*P-valueVAS pain score when mobilised*P-value
n = 163n = 165n = 163n = 165
  • Analyses using t-test for comparing means.

  • *

    VAS ranging from 0 (representing no pain) to 10 (representing unbearable pain).

  • Some data missing as some women did not complete their VAS scores sheets.

  • Missing data as some women had not yet mobilized (during the earlier hours after randomisation). Also other women did not complete their VAS score sheets.

1 h2.2 ± 1.82.7 ± 2.0P = 0.0142.6 ± 2.12.8 ± 1.9P = 0.51
n = 159n = 163 n = 125n = 128 
2 h2.1 ± 1.62.5 ± 1.8P = 0.0322.6 ± 1.92.7 ± 1.9P = 0.71
n = 156n = 158 n = 128n = 133 
4 h1.8 ± 1.52.2 ± 1.5 P  = 0.0142.2 ± 1.62.5 ± 1.9P= 0.09
n = 155n = 158 n = 145n = 146 
8 h1.6 ± 1.41.8 ± 1.6P = 0.141.9 ± 1.62.1 ± 1.7P = 0.30
n = 157n = 161 n = 153n = 156 
12 h1.5 ± 1.41.6 ± 1.6P = 0.421.7 ± 1.41.9 ± 1.6P = 0.32
n = 156n = 159 n = 155n = 157 
24 h1.3 ± 1.41.4 ± 1.4P = 0.441.5 ± 1.41.6 ± 1.4 P  = 0.54
n = 157n = 163 n = 154n = 161 

Overall VAS satisfaction score for pain relief over the 24-h study period also favoured celecoxib with a mean score of 6.2 versus 5.7 on a ten-point VAS (Table 3). Requirement for rescue analgesia was not significantly different: celecoxib 5.0% versus diclofenac 10.9% (RR 0.46, 95% CI 0.2–1.0, P = 0.065). No woman needed more than one dose of rescue analgesia. Interval to commencement of breast feeding was not different (Table 3).

Table 3.  Visual analog scale (VAS) for satisfaction with pain relief over 24 h study period, rescue analgesia for break through pain and effect on commencement of breast feeding after celecoxib or diclofenac orally for post-perineal repair analgesia. Mean ± standard deviation and number (%)
n = 163
n = 165
  • Analyses by t-test to compare means and χ2 test for categorical data.

  • *

    VAS with 0 representing total dissatisfaction to 10 representing complete satisfaction with analgesia.

  • Rescue analgesia with 50 mg of pethidine intramuscularly.

  • Two women had missing data on rescue analgesia.

  • §

    Missing data due to neonatal admission immediately after birth which precluded normal attempt to breast feed (nine women – six allocated to celecoxib and three allocated to diclofenac), and in two women, data were not submitted.

VAS* for overall satisfaction with analgesia at 24 h6.2 ± 1.55.7 ± 1.7P = 0.010
n = 163n = 165 
Requirement for rescue analgesia8/161 (5.0)18/165 (10.9)P = 0.065
Commencement of breast feeding (min)49 ± 3653 ± 35P = 0.25
n = 157§n = 160§ 

Women allocated to celecoxib reported less minor adverse upper gastrointestinal symptoms (Table 4): number needed to be treated to benefit (NNTb) = 9 (95%CI 5–67) to avoid minor upper gastrointestinal symptoms. Serious side-effects like overt gastrointestinal bleeding were not seen.

Table 4.  Adverse effects reported at 24 h after celecoxib or diclofenac orally for post-perineal repair analgesia. Number (%)
n = 163
n = 165
Relative risk (RR) (95% confidence interval (CI))P-value
  • Analysis was by Fisher's exact test.

  • *

    At least one of nausea, vomiting, epigastric pain or gastrointestinal bleeding.

  • Defined as frank haematemesis, coffee ground vomitus or melaena.

  • RR and P value not calculated as two empty cells.

  • §

    Data missing for one woman.

Upper gastrointestinal symptoms*38 (23.3)57 (34.5)RR 0.67 (95% CI 0.47–0.96)P = 0.029
Nausea35 (21.5)51 (30.9)RR 0.69 (95% CI 0.48–1.01)P = 0.06
Vomiting3 (1.8)8 (4.8)RR 0.38 (95% CI 0.10–1.41)P = 0.22
Epigastric pain4 (2.5)13 (7.9)RR 0.31 (95% CI 0.10–0.94)P = 0.043
Upper gastrointestinal bleeding0 (0)0 (0)
Diarrhoea2 (1.2)2 (1.2)RR 1.01 (95% CI 0.14–7.10)P = 1.0
Constipation3 (1.8)5 (3.0)RR 0.61 (95% CI 0.15–2.50)P = 0.72
Flatulence3 (1.9)1 (0.6)RR 3.03 (95% CI 0.32–28.9)P = 0.37
Headache14 (8.6)11 (6.7)
n = 164§
RR 1.28 (95% CI 0.60–2.74)P = 0.54
Pruritis1 (0.6)3 (1.8)RR 0.34 (95% CI 0.04–3.21)P = 0.62
Tinnitus2 (1.2)3 (1.8)RR 0.67 (95% CI 0.11–3.99)P = 1.0
Rash0 (0)0 (0)


Women randomised to celecoxib reported lower VAS pain scores when at rest compared to diclofenac but the difference was fairly small – not more than a half point on a ten-point VAS. Although statistical significance is achieved, the clinical significance of a half point difference in VAS is likely to be insignificant. The effect was only significant at one, two and four hours. These earlier differences were unlikely to be due to differential absorption as most women would have taken their drug allocation on an empty stomach (feeding with solids was not permitted while in active labour in our setting) and peak plasma levels were typically reached with celecoxib at 2.8 h versus 2.3 h with diclofenac.21

Our study celecoxib dose of 200 mg was less than the recommended 400 mg initial dose for acute pain but similar to the dose used in previous trials for acute pain after dental and orthopaedic surgery which have demonstrated that 200 mg celecoxib was effective for moderate to severe pain.7 Our study diclofenac sodium dose of 100 mg was at the upper end of doses used in previous studies for acute pain.6 Our finding of lower VAS pain score at rest for celecoxib was therefore unlikely to be the result of an inadequate dose of diclofenac and may also indicate that an initial celecoxib dose of 200 mg was adequate for post-perineal repair analgesia after normal birth.

Our study contained relatively few women with spontaneous perineal tears. This is mainly because of the liberal policy on episiotomy prevalent in our setting at the time with mediolateral episiotomy commonly performed in nulliparas and with low threshold episiotomy in multiparas to avoid perineal tears. Randomised trials of episiotomy have also reported episiotomy rates of 81–100%22,23 in liberal use of episiotomy arms.

We opted to study the analgesic effect of our study drugs over the first 24 h only as post-repair perineal pain is most intense in the first 24 h and a randomised study of diclofenac administered rectally has shown that by 48 h diclofenac analgesic effect is no better than to placebo.19 In addition, a large majority of study eligible women would be discharged 24 h after birth in our setting.

Our study women reported high rates (29.0% overall rate) of minor adverse upper gastrointestinal symptoms predominantly of nausea; actual vomiting was uncommon (3.4% overall rate). The high incidence of nausea may in part be attributable to intramuscular ergometrine (500 µg) and oxytocin (5 IU) routinely given for post-partum haemorrhage prophylaxis except for the 3.7% with hypertension. A large randomised trial of oxytocin–ergometrine versus oxytocin for prophylaxis of post-partum haemorrhage has a nausea rate of 28% in the oxytocin–ergometrine arm.24

Celecoxib was associated with significantly less composite upper gastrointestinal symptoms and specifically of epigastric pain compared to diclofenac (Table 4). We could not exclude severe side-effects as our sample size was not adequately powered. A possible confounding factor to the reporting of composite adverse upper GI symptoms would be the use of intramuscular pethidine as rescue analgesia as pethidine might have caused nausea and vomiting. After excluding women who had received pethidine, the finding was similar: 21.9% versus 33.3% (P = 0.029) in favour of celecoxib.

Overall analgesia satisfaction score over the 24-h period was also slightly better with celecoxib compared to diclofenac (Table 3).

As commercial preparations of celecoxib and diclofenac were used, concealment was not ideal and reporting bias was possible with regard to our findings. We did not reveal drug allocation to women but we did not evaluate the effectiveness of this measure in keeping women blinded to drug allocation.

Celecoxib was more expensive than generic diclofenac sodium in our setting by a factor of more than 10. Diclofenac and celecoxib were effective painkillers as mean VAS pain score was less than three out of a maximum of ten at all assessment points in both arms of our trial. The cost difference between them should be considered when making decisions at the local level on the choice of painkillers for post-partum perineal repair pain. In resource poor settings, diclofenac remains a reasonable choice for perineal repair analgesia.

Based on the results of our study, oral celecoxib had slightly better analgesic effects at rest and minor upper gastrointestinal adverse effects were less frequently reported for post-perineal repair pain relief after normal birth but celecoxib cost more at present compared to diclofenac sodium. Further research on first line analgesia for mild to moderate perineal pain should focus on comparing with a cheap, effective and well-tolerated analgesic such as paracetamol.


We acknowledged the assistance of biostatistician Dr K.F. Quek of Monash University, Sunway Campus, Malaysia, in the statistical analysis.