R e s e a r c h A r t i c l e T h e E f f e c t s o f M a n u a l H y p e r f l a t i o n U s i n g S e l f - I n f l a t i n g M a n u a l R e s u s c i t a t i o n B a g s o n A r t e r i a l O x y g e n T e n s i o n s a n d L u n g C o m p l i a n c e - A META-ANALYSIS OF THE LITERATURE This research was submitted to the University of the Witwatersrand in partial fulfillment of the degree MSc (Physiotherapy) A B S T R A C T : The self-inflating manual resuscitation bag (MRB) is a modality which is commonly used by physiotherapists to manually hyperinflate the lungs o f mechanically ventilated patients. There is limited scientific evidence to support its therapeutic use and the literature is not in agreement as to the effects o f m anual hyper­ inflation. A meta-analysis o f the current research on humans has been conducted to investigate the effects o f this m odality on arterial oxygen tensions a nd lung compliance. A ll published studies evaluating the effects o f m anual hyperinflation (or bagging) on arterial oxygen tensions and/or lung compliance on mechanically ventilated patients have been retrieved. Only studies which reported results in terms o f mean values and standard deviation or standard error o f the mean could be used in this analysis. Eleven studies were identified between the time p erio d 1968 -1995. Seven o f these studies fitte d the inclusion criteria. The mean a nd stan­ dard error o f the mean values f o r arterial oxygen tensions (P a 0 2) and lung compliance (C L) have been used to calculate the 95% confidence intervals and these results were plo tted on a graph. A comparative analysis has been perform ed on the results o f the seven studies.'A generally non-significant association between bagging and the P a 0 2 / and CL values was demonstrated. Great discrepancies were identified in the designs o f the seven included studies. Since the seven studies included in this meta-analysis show an overall non-significant association, it is reasonable to assume that the therapeutic value o f the self-inflating m anual resuscitation bag is questionable. The studies presented such divergent designs that they do not offer conclusive evidence. More standardized, multi-centre studies are required to clarify the therapeutic value o f this modality. Other methods o f recruiting the lungs o f critically ill p atients during and after physiotherapy intervention, need to be explored. KEYWORDS: META-ANALYSIS, M ANUAL HYPERINFLATION, PHYSIOTHERAPY, M ECH ANICAL VENTILATION BARKER M, MSc (Physio), MCSP, ICU PHYSIOTHERAPY SPECIALIST' EALES a , PhD2 'G u y's and S t Thomas' Hospitals, United Kingdom Physiotherapy Departm ent U niversity o f the W itw atersrand INTRODUCTION M anual hyperinflation is a technique traditionally used by physiotherapists in the treatm ent o f m echanically ventilated patients. The use o f this technique is based m ore on clinical experience than on scientific evidence. Few experimental trials have addressed the ability o f m a­ nual resuscitators per se to achieve; increased secretion clearance and recruit- f m ent o f atelectatic lung u n its/ O f this small number of published trials, a greater proportion have to do with self-inflating manual resuscitation bags (M RBs) and for this reason this analysis has focused on the self-inflating type o f MRB. J T he literature which is available pre­ sents conflicting evidence to the reader. This is particularly true when each of these studies are viewed in isolation and not as part o f a whole. The results o f the few experim ental trials do not allow standard guidelines to be adopted on w hen the MRB should be used and when not (Reiterer, 1993). The literature has thus failed to provide a firm research consen su s to support or refute the continued use o f the self-inflating MRB in the treatm ent o f m echanically venti­ lated patients in the intensive care unit. The aim o f this m eta-analysis is to evaluate the available trials on the thera­ peutic value o f the MRB quantitatively and qualitatively. The results will offer useful inform ation to both the researcher and the practising physiotherapist. A m eta-analysis will also aid in identifying m ethodological errors in the existing literature and so help to guide future research projects in the field. An analy­ sis such as this can help to bring physio­ therapists closer to the truths about the techniques w hich they use. This is espe­ cially im portant in to d a y ’s need for evidence based practice. C O RRESPO N D EN C E: M ichael B arker IC U Physiotherapy Specialist, A dult ICU , G u y ’s and St T h o m as’s NH S Trust L ondon SE1 U nited Kingdom SA J o u r n a l o f Physiotherapy 2000 V o l 56 No 1 7 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 3. ) METHOD This meta - analysis was conducted in the D epartm ent o f Physiotherapy at the U niversity o f the W itw atersrand. An online com puter search o f the MEDLINE (National Library o f Medicine, Bethesda, M d) database was perform ed to isolate all the relevant literature on self-inflating manual resuscitation bags. The following terms were used to search the titles and abstracts o f indexed articles: Respiration, A rtificial/M ethods or Standards; In su f­ flation/M ethod or Instrum entation; R es­ piratory Therapy/M ethod; R espiratory Insufficiency/Therapy; Positive Pressure Respiration/M ethod. The search dated back from 1995 to 1968. As M ED LIN E may not have all the relevant articles indexed on its data­ base and, therefore, has lim itations as a retrieval system (Dickerson et al, 1985), the referen ce lists o f tho se articles selected from the com puter search were scanned for additional published reports. The reference lists o f two recognised textbooks in the field (M ackenzie, 1989; Scanlan, 1990) were also scrutinised to id en tify any fu rth er literatu re. T he authors have also recognised the exis­ tence o f publication bias in retrieving “all” the current literature for a m eta­ analysis. For inclusion into the m eta-analysis, the studies had to fit all o f the following criteria. They had to: 1. Be random ised controlled trials. 2. Be experim ental, clinical trials using only hum an subjects who m echani­ cally ventilated via either an endotra­ cheal or tracheostom y tube. 3. Have m ade use o f self-inflating m a­ nual resuscitation bags as m ethod of achieving pulm onary hyperinflation. 4. Have m easured either arterial oxygen tensions ( P a 0 2) and/or the lung com ­ pliance (CL) as the end-points o f effect of the M RB. If the research article reported other independent observa­ tions (e.g. A lveolar-arterial oxygen difference (A -a D 0 2), cardiac output etc.) as well as P a 0 2 and C L these studies were still included but only the latter two param eters w ere used in the analysis. 5. Have presented their results as mean values and standard deviations or stan­ dard error o f the m ean values in order that the 95% confidence intervals (C.I.) could be recalculated. If a study included other physiotherapy or nursing procedures in their m ethods in addition to manual hyperinflation, then only the results o f the groups, w hich w ere m anually hyperinflated, were included in the m eta - analysis. Inclusion into the m eta - analysis was thus based on m ethodology and not on the outcom e o f the trials. Studies were excluded from the m eta - analysis if they: 1. Were laboratory studies conducted on m annequins, test-lungs or animals. 2. M ade use o f a ventilator to deliver hyperinflation volumes to their sub­ jects. 3. A pplied the m anual hyperinflation via a facem ask and not an endotra­ cheal o r tracheostom y tube. 4. D id not m easure P a 0 2 and/or C L. 5. Presented their results graphically or as a percentage increase/decrease in the m easured p aram eters, w ithout tabulating m ean values and standard deviations or standard error o f the m ean values. m ents in the studies w hich fitted all of the above stated inclusion criteria. The mean values (x) and standard deviations (SD) o f P a 0 2 and/or C L values before com m encem ent o f the intervention i.e. the baseline m easurem ent (xb ; SD b) and the m ean and standard deviation values at the end o f the observed period o f the intervention i.e. the final m easure­ m ent (xf ; SD f) w ere identified in each study. W here standard error o f the mean (SEM ) values w ere given instead o f SD, the SD was calculated using the formula: SD = SEM . V n w here n = sam ple size. F or each study, the xf - xb (or xf.b) was determ ined to give the difference o f the m eans between the final and baseline readings for P a 0 2 and/or C L. The SD f and SD b were pooled by applying the formula: Pooled SD = V [(SDf)2 + (SD b)2 ] /2 From this value the pooled SEM for P a 0 2 and /or C L for each study was cal­ culated. The 95% C.I. w ere then deter­ mined using the follow ing relationship: 95% C.I. = xf.b ± ( 1 .9 6 . SEM) The follow ing data w ere extracted from the studies, w hich fitted the inclu­ sion criteria: i) sample (size, patient pathology, selec­ tion criteria, m ean age and whether or not a control group was used) ii) patient position (during treatm ent and during m onitoring) iii)m onitoring (end-points, length of m onitoring and stages of m onitoring) iv)m anual hyperinflation protocol (type of MRB used and the num ber o f com pressions per session/treatm ent and the fractional delivered oxygen concentration (FD02s) pre-treatm ent, during treatm ent and post-treatm ent). A num ber was assigned to each study for ease o f reference eg. Study 1 (S I) and Study 2 (S2). Funding for this research was granted by the Medical Faculty Research Endow ­ m ent Fund o f the U niversity o f the W it­ watersrand, Johannesburg, South Africa. STATISTICAL ANALYSES The following statistical m ethods were applied to the P a 0 2 and/or C L m easure­ The x f_b and C.I. for P a 0 2 (Fig 1) and C L (Fig 2) for each study w ere then plo t­ ted on a graph. Studies were plotted by ranking them in order o f time o f final m easurem ent. This was done to assess any effect of the tim e period over w hich monitoring was conducted on the o ut­ com e o f the two m easured end-points. A separate plot (Fig 3) was done using the values obtained from calculations o f the ‘o ld ’ and ‘new ’ C.I.’s done on Study 3. A com parative analysis o f the studies was then carried out. s ' Note on 'pooled' SD vs. SDf.b and 'n e w ' > C.I. calculations {or Study 3 Given the data supplied in the research reports, pooling the SDs o f P a 0 2 and/ or C L in each study was the only statis­ tical m ethod for calculating the upper and lower bounds o f the m eans o f the sam ples o f the individual studies. The ‘p ooled’ SD value, however, artificially treats the baseline and final m easure­ m ents on each patient as if two separate groups o f patients w ere being com pared i.e. as if betw een patient differences/ 8 SA J o u r n a l o f Physiotherapy 2000 V o l 56 No 1 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 3. ) FIGURE 1: X f . | , and C.l. and C.l of Pa02 values S I, S2, S3, S4, S5 and S6 3 6 - 3 2 .8 3 W h e r e 3 2 - _ 3 1 .3 5 T 2 8 - T 1 1 1 S I = C h u la y , 1 9 9 8 o7l/> <■ 2 4 - 1 1 1 1 S 2 = C u b b e rly , 1 9 9 3 2 0 - 1 1 1 1 S 3 = Eales e t a l, 1 9 9 4 O 1 6 - 1 1Q q' 12- * 14 1 S 4 = N o v a k et a l, 1 9 8 71 1 1 0 .1 4OD 8 - 1 1 | t 9 1 1 7 .1 3 T 1 I S 5 = Eales, 1 9 8 9 4 - 1 1 1 T 1 1T 1.61 S 6 = G o o d n o u g h , 1 9 8 5 N O FFFFCT 0 1 1 I = M in u te s 1 NVw/ L I 1 L \» 1 1 1 1 1 m in _L -3 .3 5 1 - 2 .5 6 I 1 z - 4 - 1 | 1 1 1 - 3 .9 8 1 _L - 6 .9 2 C A B G = C o r o n a r y a r te r y(D CQ Q - 8- 1 1 | 1 1 -9 1 4 - 9 .9 9 I bypass g ra ft < " - 12- 1 | -1 1 . 8 7 1 1 Resp = R es p ira to ry<0 > - 1 6 - _ L - 1 4 . 8 3 1 | _i_ 16 t/> i/>O n - 20- 1 1 1 X f - b = D iffe re n c e b e tw e e n fin a l q ' f O* r U - 2 4 - J_ -2 1 .1 8 1 1 -2 5 .1 3 a n d b a s e lin e m e a n values - 2 8 - 1 C .l. = C o n fid e n c e intervals S I S2 S5 S4 S2 S3 P a 0 2 = A r te r ia l o x y g e n tensions TIME: 4 m in 5 m in 15 m in 3 0 m in 6 0 m in 6 0 m in SAMPLE: 3 2 2 8 18 16 11 11 Adults Adults Adults Adults Adults Adults (CABG) (Cardiac Surgery) (General) (Resp. Failure (Trauma] (Cardiac Surgery) FIGURE 2: x f _ b and C.l. and Ĉ of PAO2 values for S2, S3, S4 and S7 3 6 - | W h e r e 3 2 - | S 2 = C u b b e rly , 1 9 9 3 2 8 -“ OO 2 4 - 1 S 3 = Eales et a l, 1 9 9 4 < ’ <0 > 2 0 - | S 4 = N o v a k et a l, 1 9 8 7 0n 1 6 - | 1 3 .4 8 S 7 = R eiterer et a l, 1 9 9 3 Q n - To' 3 8 - 1 4 - | 3 .9 2 T T * m in Resp = M in u te s = R e s p ira to ry m q F F F F r T 0 , 1 -0 .3 5 1 0 .3 5 0 .2 7 _ = D iffe re n c e b e tw e e n final1 1V—/ L 1 1 L \» 1 1 -0 .5 1 1 X f - b .4- 1 T -3 1 - 0 .6 7 ] _ -3 .4 3 a n d b a s e lin e m e a n values z CQ -8- I I 1 C .l. = C o n fid e n c e intervals Q _ -9.92 < ' > 12“ 1 - ! 6- _i_ -12.94 c L = Lung c o m p lia n c e OQ -2 0 1 o' D -2 4 - | - 2 8 - , S4 S 7 S2 S3 TIME: 3 0 m in 3 0 m in 6 0 min 6 0 min SAMPLE: 16 2 0 11 11 Adults Neonates Adults Adults (Resp. (Trauma) Cardiac Failure) Surgery) SA J o u r n a l o f Physiotherapy 2000 V o l 56 No 1 9 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 3. ) FIG U R E 3 : X f_b a n d o ld a n d n e w C .I. f o r S 3 , P a 0 2 a n d N O EFFECT COQ a> > 3 6 - 3 2 - 2 8 - 2 4 - 2 0 - 1 6 - 1 2 - 8- 4- _0_ - 4 - - 8- - 12- - 1 6 - -2 0 - - 2 4 - - 2 8 - . 1 0 .1 4 1.61 . -6 .9 2 T 6-52 * 1 .1 6 - -3 .3 1 P a 0 2 O ld C.I. = i-........ ^ N e w C.I. - |--------- 1 . 1 3 .4 8 0 .2 7 • 1 2 .9 4 T _ 5 .0 9 0 .2 7 -4 .5 5 C l W h e r e S 3 = Eales et a l, 1 9 9 4 Xf-b C.I. P a 0 2 C L = D iffe re n c e b e tw e e n fin a l a n d b a s e lin e m e a n valu es = C o n fid e n c e intervals = A r te r ia l o x y g e n tensions = Lung c o m p lia n c e deviations w ere being considered. The values obtained from these calculations are generally overestim ated. Similarly, SDf - SD b considers the baseline and final m easurem ents as tw o separate groups and when determ ining this value negative standard deviations (statistical non-entities) are likely. The correct SD to have used in the calculation o f the 95% C.I. would have been the standard deviation of the differences i.e SD t_b. This value considers within patient dif­ ferences/deviations (Galpin, 1994). Having the raw data available for only one of the studies included in the meta - analysis viz. Study 3 (Eales et al, 1994), the SDf_b was calculated for P a 0 2 and for C L. The 95% C.I. for Study 3 were then recalculated using the SD f.b to derive a “new” SEM. The “new” C.I. were then plotted on a third graph together w ith the “old” C.I. for the P a 0 2 and C L m easurem ents in Study 3. Knowing that the SD f_b is often less than the “pooled” SD, the “pooled” SD was then divided by the SD t.b to ascertain the factor by which the “pooled” SD was greater than the S D f_b. It was then hypothesised that this factor by which the “pooled” SD was greater than the SD|_b (for both P a 0 2 and C L) could be used in the other trials to estim ate their SD f_b values. This would facilitate a better look at the data of the other trials. RESULTS Having conducted the literature search, 11 reports were identified w hich docu­ mented the effects o f self-inflating m a­ nual resu scitatio n bags (M R B s), on either arterial oxygen tensions (P a 0 2) and/or lung com pliance (C L). O f the 11 studies, only seven fitted all the inclu­ sion criteria. The studies that fitted the criteria are listed in Table 1 and the descriptive statistics for each o f these studies are presented in Table 3. Three o f the seven m easured P a 0 2 only, one study m easured C L only and three m ea­ sured both P a 0 2 and C L. Three studies were excluded from the meta - analysis as they presented their results graphically with no description o f mean values or standard deviations in table form or in the text. O ne study (Tweed et al, 1991) w as ex clu d ed based on a com plex research design w hich did not render it com parable to the rest o f the trials. A synopsis o f the studies w hich w ere ex clu d ed from the m eta-an aly sis is given in Table 2. RESULTS OF ARTERIAL OXYGEN TENSION (PA02) VALUES - FIG 1 Study 2 and Study 5 showed signifi­ cantly negative associations betw een the use o f the M RBs and the effect on P a 0 2. A com parison betw een the m ethods used in Study 2 and Study 5 shows sim ilarities in that both studies used the A m bu-R uben M RB. P atien ts w ere bagged in both studies until they were clinically clear o f secretions. Both stu­ dies investigated the effects o f bagging on patients with established respiratory failure and who were paralysed and sedated for the duration of the trial. The differences between these studies, how ­ ever, include the times o f final m easure­ m ent and the inconsistencies in the flow rate settings o f the M RBs used. Study 2 used an oxygen flow rate o f 15 L/min achieving a FD 02 o f >0.8, w hile Study 5 reports a flow rate o f 8 L/min and a 10 SA J o u r n a l o f Physiotherapy 2000 V o l 56 No 1 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 3. ) TABLE 1: Summarizing the Methods used in the Studies Included in the Meta-Analysis ie Study 1 Through Study 7 Study Sample Size Patient Type Patient Position MRB and Settings Measured Indices Time Final* S I - C h u la y 1 9 8 8 3 2 A d u lts - w ith in 2 4 hrs post C A B G S u p in e PMR 2 a t 1 5 L /m in P a 0 2 4 min S 2 - C u b b e rle y 1 9 9 4 11 A d u lts - T ra u m a post 4 8 hrs IPPV A SL - 3 0 ° h e a d up A m b u b a g a t 1 5 L / m in P a 0 2 a n d C L 6 0 m in S 3 - Eales et al 1 9 9 5 11 A d u lts - 2 4 h rs post C A B G a n d M V R S u p in e - 2 0 ° h e a d up A m b u b a g a t 1 5 L /m in P a 0 2 a n d C L 6 0 m in S 4 - N o v a k et al 1 9 8 7 1 6 A d u lts - R e s p ira to ry fa ilu re A SL Laerdal 3 ( F D 0 2 = 0 . 8 ) P a 0 2 a n d C L 3 0 m in S 5 - Eales 1 9 8 9 1 8 A d u lts - G e n e r a l S u p in e A m b u b a g a t 8 L /m in ( F D 0 2 = 0 . 6 4 ) P a 0 2 1 5 min S 6 - G o o d n o u q h 1 9 8 5 2 8 A dults - w ith in 4 - 6 hrs post c a r d ia c su rg e ry N o t stated Test b a g system ( F D 0 2 = 1 .0 ) P a 0 2 5 m in S 7 - R eiterer et al 1 9 9 3 2 0 N e o n a te s N o t stated M a rq u e s t system 6 - 8 L /m in C L 3 0 m in (M R B = m a n u a l resuscitation b a g , * = tim e o f fin a l m ea s u re m e n t in ea c h trial , hrs. = hours, C A B G = c o r o n a r y a r te r y bypass g ra ft, PM R = Puritan M a n u a l Resuscitator, m in = m inutes, IPPV = interm itten t positive pressure v e n tila tio n , P a 0 2 = a rte ria l o x y g e n tensions, C L = lung c o m p lia n c e , A S L = a lte rn a te side ly in g , M V R = m itral v a lv e re p la c e m e n t, F D 0 2 = fra c tio n a l d e liv e re d o x y g e n p e rc e n ta g e ). TABLE 2: Summarizing the Methods Used in the Studies Excluded from the Meta-Analysis Study Sample Size Patients Studied Position Indices Measured MRB Protocol Tw eed et al ( 1 9 9 1 ) 2 4 A dults Low er a b d . su rg e ry - intra o p e ra tiv e ly T re n d ellen b u rg position ( A - a ) D 0 2 fro m A B G M H I 3 - 4 tim es to TLC a t 3 0 c m H 20 O k k e n et al ( 1 9 8 7 ) 1 5 P re-term infants In tu b ated - on n asal CPAP N o t stated T c P 0 2 M H I fo r 5 m in e v e ry 2 0 - 3 0 m in F ox et al ( 1 9 7 8 ) 1 3 N e o n a te s In tu b ated - S p o n tan eo u s b re a th in g S u p in e D y n a m ic C L + A B G M H I w ith 1 0 b re a th s Jones et al ( 1 9 9 2 ) 2 0 A dults Fully ve n tila te d A SL Static C L + S a 0 2 N o t c le a rly stated (A b d . = a b d o m in a l, CPAP = continuos positive a ir w a y pressure, A S L = a lte rn a te sid e ly in g , ( A - a ) D 0 2 = a lv e o la r -a r te r ia l o x y g e n d iffe re n c e , A B G = a r te r ia l b lo o d g a s , T c P 0 2 = tra n c u ta n e u o s o x y g e n tensions m e a s u re d continuously b y a trancutaneuos o xy g e n electrode, C L = lung c o m p lia n c e , S a 0 2 = a rte ria l o x y g e n saturations, M H I = m a n u a l h y p e rin fla tio n , TLC = total lung ca p a c ity , m in = minutes) FDC>2 o f 0.64. Study 2 also describes positioning o f their subjects in alternate side lying, whereas Study 5 maintained their subjects in the supine position throughout the trial. Study 1, Study 4 and Study 6 dis­ played the widest C.l.s for P a 0 2 across all the studies. Study 1 and Study 6 were similar in their study designs in that they both used post-operative cardiac surgery patients and m onitored their subjects for short periods o f time (four minutes for Study 1 and five minutes for Study 6). There is, however, little evidence from the data to suggest m uch similarity between these two trials and Study 4. On the con­ trary, differences are apparent in the study design o f Study 4 in terms of the condition o f the patients used, the time o f m onitoring and the sample number. Study 4 had a population consisting of patients who required mechanical venti­ lation for respiratory failure and had the smallest sample size (n=16) o f the three studies. This is exactly half that of Study 1 (n = 32) and ju st over h alf that of Study 6 (n = 28). Study 1 and Study 6 displayed positive xf_b values for P a 0 2 o f 14 and 9, respectively, w hile the x r_b for P a 0 2 in Study 4 was -9. Study 1 and Study 2 were the only two studies, which were significantly differ­ ent from each other with regard to the m easurem ents o f Pa02- Considering the original C.I values which were calcu­ lated from the data given in the reports, SA J o u r n a l o f Physiotherapy 2000 V o l 56 No 1 11 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 3. ) TABLE 3: Table Showing Calculated Statistics for P a02 and CL Values of Study 1 Through Study 7 P a 0 2 Study X f-b Pooled SD Pooled SEM 95% C.l. Estimated SEM* Estimated 95% C.i S I 1 4 5 0 . 0 7 8 . 8 5 [ - 3 . 3 5 ; 3 1 . 3 5 ] 5 . 1 4 [ 3 . 9 3 ; 2 4 . 0 7 ] S 2 - 9 . 9 9 1 0 . 6 4 3 . 0 7 [ - 1 6 ; - 3 . 9 8 ] 1 .7 8 [ - 1 3 . 4 7 ; - 6 . 5 9 ] S 3 1 .6 1 1 6 . 8 7 4 . 3 5 [ - 6 . 9 2 ; 1 0 . 1 4 ] 3 . 4 6 [ - 3 . 3 1 ; 6 . 5 2 ] S 4 - 9 3 2 . 9 0 8 . 2 3 [ - 2 5 . 1 3 ; 7 . 1 3 ] 4 . 7 8 [ - 1 8 . 3 6 ; 0 . 3 6 ] S 5 - 1 1 . 8 7 2 0 . 1 5 4 . 7 5 [ - 2 1 . 1 8 ; - 2 . 5 6 ] 2 . 7 6 [ - 1 7 . 3 0 ; - 6 . 4 6 ] S 6 9 6 4 . 3 4 1 2 . 1 5 [ 1 4 . 8 3 ; 3 2 . 8 3 ] 7 . 0 7 [ 4 . 8 5 ; 2 4 . 8 5 ] CL S 2 0 . 3 5 6 . 7 0 1 .9 3 [ - 3 . 4 3 ; 4 . 1 3 ] 0 . 7 3 [ - 1 . 0 8 ; 1 .7 8 ] S 3 0 . 2 7 2 6 . 1 3 6 . 7 4 [ - 1 2 . 9 4 ; 1 3 . 4 8 ] 2 . 4 6 [ - 4 . 5 5 ; 5 . 1 0 ] S 4 -3 1 4 . 1 4 3 . 5 3 [ - 9 . 9 2 ; 3 . 9 2 ] 1 .3 3 [ - 2 . 3 4 ; 2 . 8 8 ] S 7 -0 .5 1 0 . 3 8 0 . 0 8 [ - 0 . 5 7 ; - 0 . 3 5 ] 0 . 0 3 [ 0 . 5 7 ; - 0 . 4 5 ] The valu es in d ic a te d b y the asterisk (* ) w e r e estim ated b y d iv id in g the fa c to r o f 1 . 7 2 a n d 2 . 6 4 fo r P a 0 2 a n d C L , respectively into the p o o le d S E M a n d c a lc u la tin g n e w 9 5 % C .l. G r e y s h a d e d a re a s = the a c tu a l v alu es re c a lc u la te d from the r a w d a ta o f S 3 . (X f-b = the d iffe re n c e b e tw e e n fin a l a n d b a s e lin e m e a n v a lu e s , S D = s ta n d a rd d e v ia tio n , S E M = s ta n d a rd e rro r o f the m e a n , C .l. = co n fid en ce in terval, P a 0 2 = a rte ria l o xyg en tension values, C L = lung co m p lian ce). the upper bound on the C .l for Study 2 (-3.98) did not overlap w ith the lower bound on the C .l for Study 1 (-3.35). The extent to w hich these two studies were significantly different from each other w as enhanced w hen the “new ” C.I.s w ere calculated, resulting in a sig­ nificantly positive outcom e for Study 1 (Table 3). T he tim e o f final m e a su re ­ m ent and the condition o f the patients were the salient features w hich distin­ guished Study 1 from Study 2 (Table 1). A com parison betw een Study 2 and Study 3 revealed a significantly negative outcom e in Study 2 and no statistically significant outcom e in Study 3 (Fig 1). This result is despite a sim ilar protocol adopted in both studies. T he positions in which the subjects were treated in these two trials differed, however the most notable feature separating Study 2 and Study 3 was the study sample used. Study 2 used traum a patients after 48 hours o f m echanical ventilation, while Study 3 used patients after 18 hours of m echanical ventilation who had under­ gone cardiac surgery. As expected from the recalculation o f the 95 % C .l for P a 0 2 in Study 3 using the SD t_b (Table 3), the range becam e m uch smaller. The factor by w hich ‘p ooled’ SD was greater than SD f_b for P a 0 2 in Study 3 was 1.72. The ‘n ew ’ C .l reflects the true w ithin patient differ­ ences betw een the baseline and final m easurem ents. T he new values only altered the significance o f Study 1. W here the original C .l calculated from the data supplied in the report suggested no significant difference, calculation o f the new C.l revealed a positively signi­ ficant C .l for Study ] [3.92; 24.07]. RESULTS OF LUNG COMPLIANCE (CL) VALUES - FIG 2 Lung com pliance was m easured in four o f the studies, w hich m et the inclu­ sion criteria o f the m eta-analysis (Fig 2). O f the four studies, only Study 7 dem onstrated a statistically significant result (Fig 2). A lthough this was a nega­ tive result, Study 7 showed the sm allest variance in the C .I’s values and also reflected the largest sample size (n=20) when com pared to the other studies w hich m easured changes in pulm onary com pliance (Study 4, Study 2 and Study 3). Study 7 was the only study w hich exam ined neonates prior to extubation and w ho were recovering from respira­ tory distress. The other three studies all considered adult populations. The study showing the widest variance with regard to com pliance m easurem ents was Study 3. W hereas Study 2 , Study 4 and Study 7 in clu d ed p atients w ith established respiratory failure, Study 3 was the only study, w hich investigated the changes in C L in post-operative cardiac surgery patients. The differences in the com pliance values obtained in Study 3 and Study 2 are quite m arked. T hese two studies, as m entioned above, follow ed a sim ilar protocol, the only dif­ ference being the pathologies o f the patients used in the study (Table 1). Fig 3 shows the ‘o ld’ and ‘new ’ C.l. for CL m easurem ents in Study 3. The range o f the ‘n ew ’ C.L is expectedly less than that o f the ‘o ld ’ C.L The factor by w hich the ‘pooled’ SD was greater than the SD f_b for C L was 2.64. Dividing this factor into the ‘o ld’ SEM values for Study 2, Study 4 and Study 7 did not significantly alter the results o f any o f these studies. DISCUSSION The thrust o f this analysis is to question the use of the self-inflating manual resus­ citation bag (M RB) by physiotherapists in the treatm ent o f patients requiring m echanical ventilation. The value o f the 12 SA J o u r n a l o f Physiotherapy 2000 V o l 56 No 1 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 3. ) MRB in the resuscitation situation, or in the case where temporary ventilation of an intubated patient is required is not being questioned. The literature reviewed in this m eta­ analysis does not provide convincing evidence to support the use o f the s e lf^ inflating MRB by physiotherapists for attaining therapeutic goals.''Furthermore, it has been described as being potentially hazardous when used on critically ill patients (M iller and Hamilton, 1969; Jumper et al, 1983; Arellano et al, 1987). It is questionable w hether the self- inflating MRB should continue to b e / recognised as an efficacious modality in intensive care respiratory therapy. ' W hen treating patients in the intensive care unit (IC U ) the physiotherapist needs to ask if the objectives of the res­ piratory therapy can best be met by using a M RB. T hese objectives include the prevention o f hypoxaem ia induced by endotracheal (ET) suction, the m o­ bilising and removal of retained p ul­ m onary secretions and the recruitm ent o f co llap sed perip h eral lung units. Alterations in m easured param eters such as the arterial oxygen tensions ( P a 0 2) and/or the lung compliance (CL) are usu­ ally observed to assess whether these objectives have been attained. Hyper- oxygenation and hyperinflation breaths before, during and after ET suction form part o f a series o f techniques w hich are used in order to achieve the above stated objectives. H yperoxygenation im plies ^ that the patient is offered a fractional ", inspired oxygen concentration ( F i0 2) above the b aseline F i 0 2. Sim ilarly, hyperinflation refers to delivering a tidal volum e (VT) 1.5 tim es greater than the fi p atient’s baseline volume./The MRB has traditionally been used to produce these h y p eroxygenation and h y p erinflation breaths. The literature, however, suggests that this has not been accom plished successfully. HYPEROXYGENATION Stone (1990) noted that when ventilators / are used to deliver hyperoxygenation < breaths before ET suction, higher o r ) equivalent P a 0 2 levels were attained when com pared to those attained by using a M RB. Such findings ought to b reak th e m indset am ongst p h y sio ­ therapists and nurses who w ork in the ICU that the MRB is always the most appropriate or the only method o f hyper- oxygenating patients before ET suction. / > HYPERINFLATION T he inspiratory volum es, w hich are d elivered from a M R B , d ep en d on factors such as the com pliance o f the patient’s lungs and thorax, the actual volum e of the bag and the technique used to com press the bag viz. a one- handed, a tw o-handed technique or any ^ a lte rn a tiv e tech n iq u e such as hand- against forearm. Com pressing a MRB .with two hands has consistently been -) shown to be associated with higher tidal ̂ volumes (VTs) when com pared to a one handed technique (Carelen and Hughes, 1975; Lebouef, 1980; Eaton, 1984; Hess and Goff, 1987; A ugustine et al, 1987; K issoon et al, 1992; Glass et al; 1993). Glass et al (1993) showed that when using one hand to compress the bag, critical care nurses delivered a m ean VT o f 17% less than the pre-set volum e on the ventilator. Hypoinflation, rather than hyp erin flatio n is th erefo re likely t o / occur when the M RB is com pressed with one hand./ Authors o f trials conducted on the therapeutic effects o f M RBs seldom /S tate the technique w hich was used to /'compress the bag. This was certainly the ^case in the studies included in the meta- yanalysis. T his in fo rm atio n becom es im portant to the reader and the m eta­ analyst so that factors w hich govern the ^outcom e o f the trial can be discerned.