1.1.1. Cardiopulmonary complications (Analysis 1.1)

Two trials (140 participants) reported cardiopulmonary complications (Aitola 1998; Tsereteli 2002). The cardiopulmonary complication rate was 5.7% in the nitrous oxide group and 2.9% in the carbon dioxide group. There was no evidence of a difference in cardiopulmonary complications between the groups (RR 2.00, 95% CI 0.38 to 10.43; low‐quality evidence; Analysis 1.1). There was clinical heterogeneity because the two trials performed quite different laparoscopic operations (cholecystectomy versus foregut surgery). This finding was downgraded to very low quality due to very serious study limitations (incomplete outcome data and selective reporting) and serious imprecision (wide CIs and small sample size).

The TSA graph showed that the cumulative Z‐curve did not cross the naive 5% statistical boundaries (Figure 4). The analysis showed a diversity‐adjusted required information size of 3781 participants (the number of participants needed to reach firm evidence of an intervention effect of 20% RRR). The number of participants included corresponded to only a small fraction (3.7%) of the diversity‐adjusted required information size; therefore, the trial sequential boundaries could not be drawn. Accordingly, we lack evidence to conclude equivalence of nitrous oxide and carbon dioxide pneumoperitoneum.

Figure 4

---

---

Trial sequential analysis of nitrous oxide pneumoperitoneum versus carbon dioxide pneumoperitoneum for cardiopulmonary complications. Analysis was performed with an event rate of 2.9% (Pc) in the control group, a risk ratio reduction of 20%, alpha 5%, beta 20%, and observed diversity 0%. The accrued sample size was so small that the trial sequential boundaries could not be drawn. The cumulative Z‐curve did not cross the naive 5% statistical boundaries (red horizontal lines). The results showed that the observed diversity‐adjusted required information size was 3781 participants, corresponding to 3.7% of the total sample size in the included trials. Accordingly, the meta‐analysis did not support or refute an intervention effect as data were too few.

1.1.2. Procedure‐related general complications (surgical morbidity) (Analysis 1.2)

Two trials (143 participants) reported surgical morbidity (Aitola 1998; Tsereteli 2002). The surgical morbidity was 2.8% in the nitrous oxide group versus 2.8% in the carbon dioxide group. There was no evidence of a difference in the surgical morbidity (procedure‐related general complications) between the groups (RR 1.01, 95% 0.18 to 5.71; very‐low‐quality evidence; Analysis 1.2). There was clinical heterogeneity because the two trials performed quite different laparoscopic operations (cholecystectomy versus foregut surgery). This finding was downgraded to very low quality due to very serious study limitations (incomplete outcome data and selective reporting) and serious imprecision (wide CIs and small sample size).

The TSA graph showed that the cumulative Z‐curve did not cross the naive 5% statistical boundaries (Figure 5). The analysis showed a required information size of 3919 participants (the number of participants needed to reach firm evidence of an intervention effect of 20% RRR). The number of participants included corresponded to only a small fraction (3.6%) of the diversity‐adjusted required information size; therefore, the trial sequential boundaries could not be drawn. Accordingly, we lack evidence to conclude equivalence of nitrous oxide and carbon dioxide pneumoperitoneum.

Figure 5

---

---

Trial sequential analysis of nitrous oxide pneumoperitoneum versus carbon dioxide pneumoperitoneum for surgical morbidity. Analysis was performed with an event rate of 2.8% (Pc) in the control group, a risk ratio reduction of 20%, alpha 5%, beta 20%, and observed diversity 0%. The cumulative Z‐curve did not cross the naive 5% statistical boundaries (red horizontal lines). The results showed that the observed diversity adjusted required information size was 3919 participants, corresponding to 3.6% of the total sample size in the included trials. Accordingly, the meta‐analysis did not support or refute an intervention effect as data were too few.

1.1.3. Pneumoperitoneum‐related serious adverse events

None of the trials reported any pneumoperitoneum‐related serious adverse events. This finding was downgraded to low quality of evidence due to serious study limitations (incomplete outcome data) and serious imprecision (small sample size for such a rare outcome).

1.2.1. Mortality

None of the trials reported any deaths. This finding was downgraded to low quality of evidence due to serious study limitations (incomplete outcome data) and serious imprecision (small sample size for such a rare outcome).

1.2.2. Quality of life

None of the trials reported quality of life.

1.2.3. Pain scores

Two trials (140 participants) reported pain scores (Aitola 1998; Tsereteli 2002). Both reported lower pain scores (about 1 cm on a VAS scale of 1 cm to 10 cm with lower numbers indicating less pain) in the nitrous oxide group compared with the carbon dioxide group at various time points on the first postoperative day. However, as neither trial reported the standard deviation (SD) for pain scores, we did not perform a meta‐analysis. There was clinical heterogeneity because the trials performed quite different laparoscopic operations (cholecystectomy versus foregut surgery). This finding was downgraded to very low quality of evidence due to study limitations (incomplete outcome data), indirectness and serious imprecision (small sample size).

Another trial reported pain scores using McGill pain questionnaire, but was not considered for this outcome, because it did not use the VAS scale (Lipscomb 1993, 53 participants undergoing laparoscopic tubal ligation).

1.2.4. Analgesia requirements (Analysis 1.3)

Three trials (193 participants) reported analgesia requirements. The trials used different measurement scales (milligrams versus tablets per 24 hours), therefore we calculated a SMD. The fixed‐effect model showed less analgesic consumption (oxycodone or ibuprofen) in the nitrous oxide group compared with the carbon dioxide group (SMD ‐0.79, 95% CI ‐1.09 to ‐0.49; very low quality evidence). The statistical heterogeneity was substantial (I² = 82%) and applying the random‐effects model did not show any evidence of difference in analgesic consumption between the groups (SMD ‐0.69, 95% CI ‐1.42 to 0.04; very‐low‐quality evidence; Analysis 1.3). In addition, there was clinical heterogeneity because the three trials performed quite different laparoscopic operations (cholecystectomy, foregut surgery, tubal ligation). Consequently, this finding was downgraded to very low quality of evidence due to study limitations (incomplete outcome data), serious imprecision (small sample size), and serious inconsistency.

1.2.5. Costs

None of the trials reported costs.

1.2.6. Cardiopulmonary changes (Analysis 1.4)

One trial (100 participants) reported cardiopulmonary changes (Tsereteli 2002). There was no evidence of a difference in the following cardiopulmonary parameter changes between the groups: heart rate (MD ‐0.60 beats/minute, 95% CI ‐4.13 to 2.93; very‐low‐quality evidence), mean arterial pressure (MD ‐3.80 mmHg, 95% CI ‐7.90 to 0.30; very‐low‐quality evidence), oxygen saturation (MD 0%, 95% CI ‐0.39 to 0.39; very low quality of evidence), and peak airway pressure (MD ‐0.30 cmH₂O, 95% CI ‐2.17 to 1.57; very low quality of evidence) (Analysis 1.4). None of the other cardiopulmonary changes were reported. These findings were downgraded to very low quality of evidence due to study limitations (incomplete outcome data), serious imprecision (small sample size), and indirectness of the outcome (surrogate outcome).

2.1.1. Cardiopulmonary complications (Analysis 2.1)

Three trials (128 participants) reported cardiopulmonary complications (Bongard 1993; Neuhaus 2001; O'Boyle 2002). The cardiopulmonary complication rate was 4.4% in the helium group and 3.0% in the carbon dioxide group. There was no evidence of a difference in cardiopulmonary complications between the groups (RR 1.46, 95% CI 0.35 to 6.12; very‐low‐quality evidence; Analysis 2.1). There was clinical heterogeneity because the three trials performed quite different laparoscopic operations (e.g. cholecystectomy, fundoplication, and gastrointestinal surgery). This finding was downgraded to very low quality of evidence due to serious study limitations (lack of blinding and selective reporting) and very serious imprecision (wide CIs and small sample size).

The TSA graph showed that the cumulative Z‐curve did not cross the naive 5% statistical boundaries (Figure 6). The analysis showed a diversity‐adjusted required information size of 3651 participants (the number of participants needed to reach firm evidence of an intervention effect of 20% RRR). The number of participants included corresponded to only a small fraction (3.5%) of the diversity‐adjusted required information size; therefore, the trial sequential boundaries could not be drawn. Accordingly, we lack evidence to conclude equivalence of helium and carbon dioxide pneumoperitoneum.

Figure 6

---

---

Trial sequential analysis of helium pneumoperitoneum versus carbon dioxide pneumoperitoneum for cardiopulmonary complications. Analysis was performed with an event rate of 3.0% (Pc) in the control group, a risk ratio reduction of 20%, alpha 5%, beta 20%, and observed diversity 0%. The cumulative Z‐curve did not cross the naive 5% statistical boundaries (red horizontal lines). The results showed that the observed diversity adjusted required information size was 3651 participants, corresponding to 3.5% of the total sample size in the included trials. Accordingly, the meta‐analysis did not support or refute an intervention effect as data were too few.

2.1.2. Procedure‐related general complications (surgical morbidity)

None of the trials reported any significant procedure‐related general complications. This finding was downgraded to very low quality of evidence due to very serious study limitations (lack of blinding, incomplete outcome data, and other bias) and serious imprecision (small sample size).

2.1.3. Pneumoperitoneum‐related serious adverse events (Analysis 2.2)

Three trials (128 participants) reported serious adverse events (Bongard 1993; Neuhaus 2001; O'Boyle 2002). There were three serious adverse events (subcutaneous emphysema) related to helium pneumoperitoneum; the serious adverse event rate was 4.9% in the helium pneumoperitoneum group and 0% in the carbon dioxide pneumoperitoneum group. There was no evidence of a difference in the Peto OR for pneumoperitoneum‐related serious adverse events between groups (Peto OR 8.28, 95% CI 0.86 to 80.03; very low quality of evidence; Analysis 2.2). There was clinical heterogeneity because the three trials performed quite different laparoscopic operations. This finding was downgraded to very low quality evidence due to study limitations (lack of blinding and selective reporting), indirectness and serious imprecision (small sample size for such a rare outcome).

The TSA graph showed that the cumulative Z‐curve did not cross the naive 5% statistical boundaries (Figure 7). The analysis showed a diversity‐adjusted required information size of 4793 participants (the number of participants needed to reach firm evidence of an intervention effect of 20% RRR). The number of participants included corresponded to only 2.7% of the diversity‐adjusted required information size; therefore, the trial sequential boundaries could not be drawn. Accordingly, we lack evidence to conclude equivalence of helium and carbon dioxide pneumoperitoneum.

Figure 7

---

---

Trial sequential analysis of helium pneumoperitoneum versus carbon dioxide pneumoperitoneum for serious adverse events. Analysis was performed with an event rate of 2.3% (Pc) in the control group, a risk ratio reduction of 20%, alpha 5%, beta 20%, and observed diversity 0%. The cumulative Z‐curve did not cross the naive 5% statistical boundaries (red horizontal lines). The results showed that the observed diversity adjusted required information size was 4793 participants, corresponding to 2.7% of the total sample size in the included trials. Accordingly, the meta‐analysis did not support or refute an intervention effect as data were too few.

2.2.1. Mortality

None of the trials reported any deaths. This finding was downgraded to low quality of evidence due to serious risk of bias and imprecision (small sample size for such a rare outcome).

2.2.2. Quality of life

None of the trials reported quality of life.

2.2.3. Pain scores (shoulder or abdominal pain) (Analysis 2.3)

Two trials (108 participants) reported pain scores (Neuhaus 2001; O'Boyle 2002). There was no evidence of a difference in the first postoperative day pain scores (graded by VAS on a scale of 1 cm to 10 cm, with lower numbers indicating less pain) between the groups (MD 0.49 cm, 95% CI ‐0.28 to 1.26; very low quality evidence; Analysis 2.3). There was clinical heterogeneity because the two trials performed quite different laparoscopic operations. This finding was downgraded to very low quality of evidence due to study limitations (random sequence generation was at unclear risk), indirectness and serious imprecision (small sample size).

2.2.4. Analgesia requirements (Analysis 2.4; Analysis 2.5)

Two trials (108 participants) reported analgesia requirements (Neuhaus 2001; O'Boyle 2002). One trial reported the amount of analgesia consumed (O'Boyle 2002). The overall analgesic (morphine) consumption was higher in the helium group than the carbon dioxide group (MD 12.00 mg, 95% CI 4.44 to 19.56; very‐low‐quality evidence; Analysis 2.4). One trial reported the number of participants requiring analgesia (Neuhaus 2001). There was no evidence of a difference in analgesia (morphine) requirements between the helium group (3/8; 37.5%) and carbon dioxide group (9/10; 90%) (Analysis 2.5). However, the trial was underpowered with only 18 participants (very low quality of evidence). There was clinical heterogeneity because the two trials performed quite different laparoscopic operations. This finding was downgraded to very low quality of evidence due to study limitations (random sequence generation was at unclear risk), indirectness and serious imprecision (small sample size).

2.2.5. Costs

None of the trials reported costs.

2.2.6. Cardiopulmonary changes (Analysis 2.6)

Two trials (34 participants) reported blood pH (Bongard 1993; Naude 1996). There was no evidence of a differences between the groups in blood pH at the start of pneumoperitoneum (MD 0.01, 95% CI ‐0.01 to 0.04; very‐low‐quality evidence) or the middle of pneumoperitoneum (MD ‐0, 95% CI ‐0.03 to 0.02; very‐low‐quality evidence). However, the blood pH was higher in the helium group compared with the carbon dioxide group at the end of pneumoperitoneum (MD 0.10, 95% CI ‐0.06 to ‐0.14; very‐low‐quality evidence). Three trials (52 participants) reported partial pressure of carbon dioxide (Bongard 1993; Naude 1996; Neuhaus 2001). There was no evidence of differences between the groups in partial pressure of carbon dioxide at the start of pneumoperitoneum (MD 0.31 mmHg, 95% CI ‐1.79 to 2.40; very‐low‐quality evidence) or the middle of pneumoperitoneum (MD 0.84 mmHg, 95% CI ‐2.02 to 3.70; very‐low‐quality evidence). However, the partial pressure of carbon dioxide was lower in the helium group than the carbon dioxide group at the end of pneumoperitoneum (MD ‐12.78 mmHg, 95% CI ‐16.78 to ‐8.77; very‐low‐quality evidence). There was clinical heterogeneity because the included trials performed quite different laparoscopic operations. These findings were downgraded to very low quality of evidence due to study limitations (lack of blinding, incomplete outcome data, and other bias), serious imprecision (small sample size), and indirectness of the outcome (surrogate outcome).

3.1.1. Cardiopulmonary complications

The trial did not report any cardiopulmonary complications. This finding was downgraded to very low quality of evidence due to very serious study limitations (allocation and blinding were unclear) and serious imprecision (small sample size).

3.1.2. Procedure‐related general complications (surgical morbidity)

The trial did not report surgical morbidity.

3.1.3. Pneumoperitoneum‐related serious adverse events

The trial did not report any pneumoperitoneum‐related serious adverse events. This finding was downgraded to very low quality of evidence due to very serious study limitations (allocation and blinding were unclear) and serious imprecision (small sample size for such a rare outcome).

3.2.1. Mortality

The trial did not report any deaths. This finding was downgraded to low quality of evidence due to study limitations (allocation and blinding were unclear) and serious imprecision (small sample size for such a rare outcome).

3.2.2. Quality of life

The trial did not report quality of life.

3.2.3. Pain scores (Analysis 3.3)

The first postoperative day pain scores (graded by VAS on a scale of 1 cm to 10 cm with lower numbers indicating less pain) were lower in the room air group than in the carbon dioxide group (MD ‐0.80 cm, 95% CI ‐1.15 to ‐0.45; very low quality evidence). This finding was downgraded to very low quality of evidence due to very serious study limitations (allocation and blinding were unclear) and serious imprecision (small sample size).

3.2.4. Analgesia requirements

The trial did not report analgesia requirements.

3.2.5. Costs (Analysis 3.4)

The total hospital costs were lower in the room air group than in the carbon dioxide group (MD ‐CNY2667.00, 95% CI ‐3275.68 to ‐2058.32; very‐low‐quality evidence; equivalent to approximately USD300 to USD475 in November 2016). This finding was downgraded to very low quality of evidence due to very serious study limitations (allocation and blinding were unclear) and serious imprecision (small sample size).

3.2.6. Cardiopulmonary changes (Analysis 3.5)

There was no evidence of a difference between groups in heart rate at the start of pneumoperitoneum (MD ‐0.10 beats/minute, 95% CI ‐3.11 to 2.91; very‐low‐quality evidence). However, heart rate was lower in the room air group compared with the carbon dioxide group in the middle of pneumoperitoneum (MD ‐7.30 beats/minute, 95% CI ‐9.78 to ‐4.82; very low quality evidence) and the end of pneumoperitoneum (MD ‐8.70 beats/minutes, 95% CI ‐11.72 to ‐5.68; very low quality evidence) of pneumoperitoneum (Analysis 3.5).

There was no evidence of differences between groups in blood systolic pressure or partial pressure of carbon dioxide at the start, middle, or end of pneumoperitoneum (all very low quality evidence).

All these findings were downgraded to very low quality of evidence due to study limitations (allocation and blinding were unclear), serious imprecision (small sample size), and indirectness of the outcome (surrogate outcome).