Etiological diagnosis of pneumonia: A goal worth pursuing?

pneumonia: A goal worth pursuing? P NEU MO NIA IS A COM MON DIS EASE, AND IDEN TI FI CA TION OF re spon si ble patho gens should lead to a re duc tion of still un ac cepta bly high mor tal ity and mor bid ity rates. How ever, the ex tent of di ag nos tic ef forts should be tai lored to the se ver ity of ill ness. In view of the high in ci dence of com mu nity ac quired pneu mo nia (CAP) and its at ten dant mor bid ity and mor tal ity (1 to 5% in outpatients and 25% in inpatients), the im por tance of a cor rect etio logi cal di ag no sis and ap pro pri ate an ti mi cro bial ther apy be comes evi dent (1). Rapid in ves ti ga tions, with high posi tive and nega tive pre dic tive val ues, are nec es sary (2). De spite ex ten sive, costly in ves ti ga tion in care fully per formed pro spec tive stud ies, an etio logi cal di ag no sis is made in only 50% of cases (3). The ne ces sity of ex haus tive at tempts to di ag nose CAP has been ques tioned (1). Clini cal acu men has long been touted as criti cal to mak ing a bac te rio logi cal di ag no sis of pneu mo nia, es pe cially in ma jor texts (4,5). Fang et al (6), ana lyz ing 22 com mon symp toms, signs and labo ra tory find ings in pneu mo nia, did not find any fea tures that helped dis tin guish among the five most com mon patho gens (Strep to coc cus pneu mo niae, Hae mo phi lus in flu en zae, Le gionella spe cies, Chla my dia pneu mo niae or Gramnegative rods). The socalled ‘atyp ical’ pres en ta tion of pneu mo nia (non pro duc tive cough, vi ral pro drome, lack of con soli da tion) was not unique to the ‘atyp ical’ patho gens. The over lap of clini cal pres en ta tion of these patho gens pre cluded pre cise rec om men da tions for spe cific an ti mi cro bial ther apy. The mani fes ta tions of pneu mo nia are more a re flec tion of the host re sponse than of the in cit ing mi cro or gan ism. Wood head et al (2) found that the re sults of rou tine tests (such as spu tum Gram stain and cul ture, blood cul tures and se rol ogy) changed man age ment in only 8% of pa tients. Since 30% of pa tients can not pro duce spu tum, an other 30% have pre vi ously been treated with an ti bi ot ics and 25% have or gan isms that are not eas ily cul tured, the use of spu tum Gram stain and cul ture as di ag nos tic tools is lim ited (7). Un der ideal cir cum stances, a sen si tiv ity of 62% and a speci fic ity of 85% have been re ported; how ever, un der less than ideal cir cum stances (the ‘real world’), these val ues can be as low as 50% and 12%, re spec tively (8). Spu tum cul tures are po ten tially mis lead ing in that or gan isms iso lated may be patho gens or colo niz ers (3). In proven pneu mo coc cal pneu mo nia, spu tum cul tures have a 45% false nega tive re sult (9). How ever, spu tum Gram stain and cul ture can di ag nose or gan isms not nor mally pres ent in the res pi ra tory tract, such as My co bac te rium tu ber cu lo sis, Pneu mo cys tis car inii and histo plasma, and can pos si bly iden tify re sis tant or gan isms (1). Blood cul tures are posi tive in only 7 to 25% of pneu mo nias, most com monly in those of pneu mo coc cal ori gin (7,10,11). Rou tine blood tests (white blood cell count, elec tro lytes, liver and re nal func tion tests) pro vide im por tant prog nos tic in for ma tion but are not help ful in a di ag nos tic sense (1). The util ity of chest roent ge no grams has also been dis ap point ing (12). No dis tinc tive roent ge no graphic pat tern can con sis tently dis tin guish among pneu mo coc cal, le gionella, chla my dial and myco plas mal pneu mo nias. Rather than iden ti fy ing a causa tive or gan ism, chest roent ge no grams should be used to as sess the se ver ity of dis ease and com pli ca tions such as ab scesses, ob struc tion, ef fu sions, etc (1). Se ro logi cal an ti body tests were ini tially viewed as a pana cea to the cli ni ci an’s di ag nos tic woes. On the con trary, they are of lim ited avail abil ity, costly, have low yield (25%) and, be cause of proc ess ing de lay, can not af fect pa tient treat ment in the criti cal early pe riod (3). Their use should be lim ited to spe cial situa tions, such as epi de mi ol ogi cal sur veys of pneu mo nia. For pneu mo coc cal in fec tion, mi cro bial an ti gens, such as pneu mo coc cal C poly sac cha ride and cap su lar poly sac cha ride, can be found in spu tum, se rum and urine (10). How ever, these tests are ex pen sive, have vary ing sen si tivi ties and speci fici ties ac cord ing to the source ex am ined and to the di ag nos tic mo dal ity used (ELISA or coun ter cur rent im mu noe lec tro pho re sis, for ex am ple), and re sults are not im me di ately avail able. Fur ther more, se rum and urine an ti gen posi tiv ity per sists up to six months af ter an in fec tion, fur ther low er ing its speci fic ity. Hence, these in ves ti ga tions have not been widely ac cepted. In va sive pro ce dures, such as tran stra cheal as pi ra tion, pro tected brush, pro tected bron choal veo lar lavage (BAL), nee dle lung as pi ra tion and open bi opsy are not in di cated in CAP un less the pa tient is se verely ill, be cause these pro ce dures rarely aid rapid di ag no sis and ex pose the pa tient to un nec es sary risk (1). EDI TO RIAL

P NEU MO NIA IS A COM MON DIS EASE, AND IDEN TI FI CA TION OF re spon si ble patho gens should lead to a re duc tion of still un ac cepta bly high mor tal ity and mor bid ity rates. How ever, the ex tent of di ag nos tic ef forts should be tai lored to the se verity of ill ness. In view of the high in ci dence of com mu nity acquired pneu mo nia (CAP) and its at ten dant mor bid ity and mor tal ity (1 to 5% in out-patients and 25% in in-patients), the im por tance of a cor rect etio logi cal di ag no sis and ap pro pri ate an ti mi cro bial ther apy be comes evi dent (1). Rapid in ves ti gations, with high posi tive and nega tive pre dic tive val ues, are nec es sary (2). De spite ex ten sive, costly in ves ti ga tion in carefully per formed pro spec tive stud ies, an etio logi cal di ag no sis is made in only 50% of cases (3). The ne ces sity of ex haus tive at tempts to di ag nose CAP has been ques tioned (1).
Clini cal acu men has long been touted as criti cal to mak ing a bac te rio logi cal di ag no sis of pneu mo nia, es pe cially in ma jor texts (4,5). Fang et al (6), ana lyz ing 22 com mon symp toms, signs and labo ra tory find ings in pneu mo nia, did not find any fea tures that helped dis tin guish among the five most com mon patho gens (Strep to coc cus pneu mo niae, Hae mo phi lus in fluen zae, Le gionella spe cies, Chla my dia pneu mo niae or Gramnegative rods). The so-called 'atyp ical' pres en ta tion of pneumo nia (non pro duc tive cough, vi ral pro drome, lack of con solida tion) was not unique to the 'atyp ical' patho gens. The over lap of clini cal pres en ta tion of these patho gens pre cluded pre cise rec om men da tions for spe cific an ti mi cro bial ther apy. The mani fes ta tions of pneu mo nia are more a re flec tion of the host re sponse than of the in cit ing mi cro or gan ism.
Wood head et al (2) found that the re sults of rou tine tests (such as spu tum Gram stain and cul ture, blood cul tures and se rol ogy) changed man age ment in only 8% of pa tients. Since 30% of pa tients can not pro duce spu tum, an other 30% have pre vi ously been treated with an ti bi ot ics and 25% have or ganisms that are not eas ily cul tured, the use of spu tum Gram stain and cul ture as di ag nos tic tools is lim ited (7). Un der ideal cir cum stances, a sen si tiv ity of 62% and a speci fic ity of 85% have been re ported; how ever, un der less than ideal cir cumstances (the 'real world'), these val ues can be as low as 50% and 12%, re spec tively (8). Spu tum cul tures are po ten tially mis lead ing in that or gan isms iso lated may be patho gens or colo niz ers (3). In proven pneu mo coc cal pneu mo nia, spu tum cul tures have a 45% false nega tive re sult (9). How ever, sputum Gram stain and cul ture can di ag nose or gan isms not normally pres ent in the res pi ra tory tract, such as My co bac te rium tu ber cu lo sis, Pneu mo cys tis car inii and histo plasma, and can pos si bly iden tify re sis tant or gan isms (1).
Blood cul tures are posi tive in only 7 to 25% of pneu monias, most com monly in those of pneu mo coc cal ori gin (7,10,11). Rou tine blood tests (white blood cell count, elec trolytes, liver and re nal func tion tests) pro vide im por tant prognos tic in for ma tion but are not help ful in a di ag nos tic sense (1).
The util ity of chest roent ge no grams has also been dis appoint ing (12). No dis tinc tive roent ge no graphic pat tern can con sis tently dis tin guish among pneu mo coc cal, le gionella, chla my dial and myco plas mal pneu mo nias. Rather than identi fy ing a causa tive or gan ism, chest roent ge no grams should be used to as sess the se ver ity of dis ease and com pli ca tions such as ab scesses, ob struc tion, ef fu sions, etc (1).
Se ro logi cal an ti body tests were ini tially viewed as a panacea to the cli ni ci an's di ag nos tic woes. On the con trary, they are of lim ited avail abil ity, costly, have low yield (25%) and, because of proc ess ing de lay, can not af fect pa tient treat ment in the criti cal early pe riod (3). Their use should be lim ited to special situa tions, such as epi de mi ol ogi cal sur veys of pneu monia.
For pneu mo coc cal in fec tion, mi cro bial an ti gens, such as pneu mo coc cal C poly sac cha ride and cap su lar poly sac charide, can be found in spu tum, se rum and urine (10). How ever, these tests are ex pen sive, have vary ing sen si tivi ties and speci fici ties ac cord ing to the source ex am ined and to the diag nos tic mo dal ity used (ELISA or coun ter cur rent im mu noe lectro pho re sis, for ex am ple), and re sults are not im me di ately avail able. Fur ther more, se rum and urine an ti gen posi tiv ity persists up to six months af ter an in fec tion, fur ther low er ing its speci fic ity. Hence, these in ves ti ga tions have not been widely accepted.
In va sive pro ce dures, such as tran stra cheal as pi ra tion, pro tected brush, pro tected bron choal veo lar lavage (BAL), nee dle lung as pi ra tion and open bi opsy are not in di cated in CAP un less the pa tient is se verely ill, be cause these pro cedures rarely aid rapid di ag no sis and ex pose the pa tient to unnec es sary risk (1).
De spite much ef fort, few in ves ti ga tions or clini cal pearls have been of de fini tive value in mak ing an etio logi cal di ag nosis of CAP and in tai lor ing an ti bi otic ther apy ap pro pri ately.
Pa tients should be treated em piri cally, there fore, ac cording to the most likely cause of in fec tion, de pend ing on their age, co-morbid dis eases, se ver ity of ill ness and need for hospi tali za tion (1,3). Ex ten sive work-up should be lim ited to sick non re spond ers. Fur ther prog ress in di ag nos tic test ing is needed, and per haps new tech nolo gies, such as the polymerase chain re ac tion, may help fill this void. While ac cepting the role of em piri cal ther apy, one should not es pouse a com pla cent at ti tude to wards it. Rec og ni tion of the evolv ing spec trum of com mu nity patho gens should lead to con tinu ing sur veil lance epi de mi ol ogi cal stud ies. This in fers that prac tice guide lines should be re vis ited fre quently to al low phy si cians to keep abreast of the changes in epi de mi ol ogi cal pat terns.
Im mu no com pro mised pa tients form a sub group in whom spe cific rather than em piri cal ther apy is pre ferred. There is a 40% in ci dence of op por tun is tic in fec tions in these hosts, leading to a 50% mor tal ity rate (13). In va sive in ves ti ga tions are thus im por tant for defi ni tive di ag no sis. The sen si tiv ity and speci fic ity of these pro ce dures vary ac cord ing to the type of im mu no sup pres sion and the pro phy lac tic or em piri cal therapy al ready re ceived (14). Lo cal ized pro cesses can be treated em piri cally ini tially, whereas dif fuse dis ease gen er ally ne ces si tates a di ag nos tic pro ce dure (13,14). Wil liams et al (13) found fi bre op tic bron cho scopy (FOB) with brush ings, lavage and trans bron chial bi opsy to be di ag nos tic in 58% of pa tients, a value simi lar to that ob tained by Crocket et al in this is sue of the Jour nal (pages 286-290). FOB was 90% sensi tive in di ag nos ing an in fec tious pro cess; al though transbron chial bi opsy does not raise this sen si tiv ity more than brush ings alone, it is nec es sary to rule out non in fec tious processes (13). The in di vid ual yields of brush, bi opsy and lavage are about 30 to 55%; it is in their com bined use that the di agnos tic yield of bron cho scopy can rise to 80 to 95% (14). In Williams and col leagues' (13) study, the high nega tive pre dic tive value of FOB im plied that a sub se quent open lung bi opsy had a low prob abil ity of iden ti fy ing an in fec tious dis ease.
Non-intensive care as so ci ated no so comial pneu mo nia occurs in 1% of hos pi tal ized pa tients and is the third most common hos pi tal ac quired in fec tion, with a mor tal ity rang ing from 16 to 37% (15). As in CAP, the eti ol ogy of no so comial pneu monia can not be in ferred from clini cal or ra dio logi cal grounds, and em piri cal ther apy should be ini ti ated, with in va sive workup be ing done in non re spond ers (15).
Ven ti la tor ac quired pneu mo nia (VAP) is de fined as that devel op ing more than 72 h af ter in tu ba tion (16). It oc curs in 10 to 20% of pa tients with res pi ra tory fail ure and in up to 70% of adult res pi ra tory dis tress syn drome (ARDS) pa tients (16). Its mor tal ity of 50 to 80% (17) fur ther un der lines the sig nifi cance of this dis ease in criti cally ill pa tients and the im por tance of prompt, ef fec tive an ti bi otic ther apy. The con se quences of not treat ing VAP are evi dent; those of over treat ment, less so. How ever, ex ces sive use of an ti bi ot ics, be sides in cur ring unnec es sary costs and ex pos ing the pa tient to po ten tial side effects and tox ici ties, leads to colo ni za tion and growth of po ten tially re sis tant or gan isms; sub se quent su per in fec tion re sults in higher mor tal ity than that seen in pa tients not pre viously on an ti bi ot ics (16,17).
There are many dif fi cul ties in the clini cal di ag no sis of VAP. Rec og ni tion of pneu mo nia is usu ally based on the pres ence of fe ver, ele vated white blood cell count, pu ru lent se cre tions and chest roent ge no graphic ab nor mali ties, cri te ria that, in the criti cally ill, can be mis lead ing (16,17). In the in ten sive care unit set ting, many dis or ders, of ten not in fec tive in na ture, can lead to fe ver and an ele vated white blood cell count; spu tum can re sult from leak age of oro pha ryn geal se cre tions around the en do tra cheal tube; chest roent ge no graphic changes can be ex plained by pul mo nary edema, atelec ta sis, pul mo nary em bo lism or ARDS; even posi tive spu tum cul tures may sim ply be from colo ni za tion of the oro pha rynx or from tra cheo bronchi tis (16,17). Clearly, many pa tients in whom there is a strong sus pi cion of pneu mo nia have an other dis or der. In a study by Fagon et al (18), the use of 16 clini cal vari ables for the di ag no sis of pneu mo nia led to a 30 to 40% di ag nos tic error and over treat ment in 16% of pa tients. More de fini tive means of di ag nos ing pneu mo nia are evi dently needed.
The an swer does not lie with non in va sive tests, such as spu tum analy sis or tra cheal as pi ra tion. No cor re la tion has been found be tween cul tures from spu tum and from un contami nated lo ca tions (blood, pleu ral fluid) (16). Tra cheal as pirates have a false posi tive rate of 21%, due to colo ni za tion (19). How ever, in a study by Sa lata et al (20), the pres ence of elas tin fi bres and a high bac te rial grad ing on Gram stain in tracheal se cre tions had a 100% posi tive pre dic tive value and a sen si tiv ity rang ing from 52 to 67%, de pend ing on the di ag nostic cri te ria used for pneu mo nia. This test holds prom ise, but needs to be vali dated in a larger clini cal study.
Chest roent ge no grams are of lim ited use in the di ag no sis of VAP: they are of ten port able, lack a lat eral pro jec tion and are im prop erly po si tioned (21). Their di ag nos tic ef fi ciency is less than 70% and the only use ful sign is that of air bron chograms, which pre dict 64% of pneu mo nias (21).
The role of in va sive test ing for di ag no sis of pneu mo nia is more clear in pa tients with VAP. The tech nique of Wim ber ley et al (22) of pro tected speci men brush has a sen si tiv ity of 65 to 96% and a posi tive pre dic tive value of 75%, sig nifi cant cultures be ing those with more than 10 3 col ony form ing units (CFU)/mL (16,18). The nega tive pre dic tive value ap proaches 100%; how ever, the false nega tive rate is about 5 to 10% (23). Draw backs to its use in clude the small sam ple size ob tained, which may miss an ad ja cent area of in volved lung; the pre cise meth od ol ogy (no use of suc tion or lido caine) needed to maximize re turn; the false posi tive rate of 9 to 25%; and the one-to two-day de lay in ob tain ing fi nal re sults (24)(25)(26).
Even quan ti ta tive cul tures of BAL fluid, con sid ered sig nificant with a growth of more than 10 5 CFU/mL, are fre quently con tami nated from the up per air way (27). In an at tempt to mini mize this, speci mens are ex cluded from analy sis if they con tain more than 1% squa mous epithe lial cells; this pur portedly raises the sen si tiv ity to 88% and speci fic ity close to 100% (16). Gram stain can also be done on BAL fluid to quan -tify in tra cel lu lar or gan isms and thus guide pre limi nary ther apy (27).
The op ti mal di ag nos tic strat egy in pa tients with sus pected VAP is un known. There is dis cor dance among tech niques, includ ing such cru cial vari ables as the defi ni tion of a sig nifi cant quan ti ta tive cul ture, and there are in con sis tency and disagree ment re gard ing the di ag nos tic gold stan dard for pneumo nia, thus ren der ing com pari son be tween stud ies dif fi cult (23). The an swer lies in a well de signed pro spec tive, ran domized study com par ing em piri cal ther apy for VAP, with in ter vention oc cur ring if there is a de te rio ra tion or lack of im prove ment of the clini cal course with a rapid in va sive in ves ti ga tion strategy, in which treat ment is re served for those with a proven bac te rio logi cal eti ol ogy. This should in cor po rate both short term and long term out comes and re source use, a fac tor that is of grow ing im por tance in to day's medi cal prac tice. Un til this study is per formed, de bate will con tinue to rage over the value of early in ter ven tion with in va sive di ag nos tic tests.

Maria An to niou MD Ron ald F Gross man MD FRCPC FACP
Di vi sion of Res pi ra tory Medi cine Mount Si nai Hos pi tal To ronto, On tario