INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Pulmonary complications are a significant source of morbidity and mortality in immunocompromised patients. Although infections are often the major concern, these individuals can develop respiratory distress due to other problems, such as hemorrhage, drug toxicity, recurrent malignancy, or a variety of inflammatory disorders (1). A specific diagnosis is often sought to assure that appropriate treatment is given and the toxicity of unneeded therapy is avoided; however, performance of invasive techniques carries some risk. Bronchoscopy with bronchoalveolar lavage (BAL) and transbronchial biopsy has been found to be helpful to identify opportunistic infections and some noninfectious processes and is widely used in this population. The procedure is well tolerated and BAL can be performed safely, even in critically ill thrombocytopenic patients (2, 3). Needle aspiration of the lung is also helpful in patients with localized lesions and adequate hematologic parameters to undergo the procedure (4). In some patients, however, these procedures may not suggest a diagnosis or even if a diagnosis is obtained, there may be no response to therapy and an open lung biopsy is considered.

Although the yield of an open lung biopsy is known to be higher than that of other invasive pulmonary procedures, it is not clear if the benefit justifies the risk of general anesthesia and surgery in all cases (5). Open lung biopsy may identify only a nonspecific pathologic pattern of acute diffuse lung injury, interstitial fibrosis, or chronic pneumonia and not clarify the precipitating cause of the pulmonary process resulting in no change in therapy. Further, even if a specific diagnosis is found, some studies of open lung biopsies in immunocompromised patients have been unable to document that this leads to prolonged survival (5, 8, 11). Finally, the results of prior studies on the yield of open lung biopsy may have only limited applicability to current patients because of the changing spectrum and frequency of pulmonary complications in immunocompromised hosts. Some infections, such as Pneumocystis carinii pneumonia (PCP) and cytomegalovirus (CMV) pneumonia, which were commonly found at open lung biopsy in the past, are now much less frequent because of effective prophylactic or preemptive regimens.

Given these uncertainties, we reviewed the recent experience at Memorial Sloan-Kettering Cancer Center (MSKCC) with open lung biopsies to determine its yield and impact on management. We selected only those with hematologic malignancies because these patients frequently present the most difficult management issues. They historically have the highest incidence of nonspecific diagnoses at biopsy and also have concomitant medical problems that make both invasive procedures and empiric therapies potentially dangerous (5, 14).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The MSKCC Thoracic Surgery database was reviewed for the time period of January 1, 1996 to December 31, 1998. Patients were identified for the study if they (1) had an open lung biopsy by thoracotomy or video-assisted thoracic surgery (VATS) for diagnosis of a pulmonary abnormality of unknown cause, and (2) had a diagnosis of a hematologic malignancy or had undergone an allogeneic or autologous bone marrow transplant (BMT) for a hematologic malignancy. The clinical records were reviewed and data collected, including the type of hematologic malignancy and transplant, chemotherapy and radiation to the thorax given, type of surgical procedure performed, pathologic or microbiologic results of open biopsies and invasive pulmonary diagnostic procedures performed before the biopsy, changes in therapy after open biopsy, complications of the surgery, and date of death. In addition, the radiographic pattern, presence or absence of neutropenia and thrombocytopenia, and need for ventilatory support before and after the open biopsy were assessed.

Definitions

Neutropenia. Absolute neutrophil count of less than 1,000 cells per microliter.

Thrombocytopenia. Platelet count of less than 50,000 platelets per microliter.

Change in therapy. Addition or withdrawal of antimicrobial agents, chemotherapy, or steroids.

Radiographs. Diffuse findings were bilateral infiltrates involving at least 50% of the lung; focal findings were distinct masses, nodules, or localized infiltrates, even if both lungs were involved.

Diagnoses. Specific diagnoses were malignancy, infection, or some inflammatory disorders that were definitive enough to direct the clinician satisfactorily in the need for future treatments. Inflammatory conditions considered specific were bronchiolitis obliterans with organizing pneumonia (BOOP), chemotherapy-induced drug toxicity, and granulomatous inflammation. Additionally, one patient with a nodular density found to be an intraparenchymal lymph node was also considered specific. BOOP was diagnosed when the predominant finding pathologically was plugs of fibroblastic connective tissue within respiratory bronchioles with mononuclear cell infiltrates in the adjacent parenchyma. Drug-induced lung disease was diagnosed if typical changes of hyperplastic type II alveolar lining cells were found with associated interstitial pneumonitis or fibrosis and there was a clear temporal relationship of clinical findings to administration of a drug known to cause pulmonary toxicity.

Nonspecific diagnoses were acute diffuse lung injury, interstitial fibrosis or pneumonitis without the criteria for chemotherapy-induced drug toxicity listed previously, or a chronic pneumonia pattern without organisms. These entities were not definitive enough to establish a cause and clearly direct future therapy.

Chemotherapeutic agents given were divided into potential pulmonary toxins and those without pulmonary toxicity. The first group included bleomycin, busulfan, carmustine, chlorambucil, cyclophosphamide, cytarabine, lomustine, melphalan, 6 mercaptopurine, methotrexate, procarbazine, and retinoic acid. The nonpulmonary toxic agents received included adriamycin, asparaginase, carboplatin, 2 chlorodeoxyadenosine, dacarbazine, etoposide, fludarabine, idarubicin, interferon, hydroxyurea, platinum, ifosfamide, mitoxantrone, thiotepa, vincristine, and vinorelbine.

Biopsies

The decision to pursue an open biopsy was made by the attending physician caring for the patient and the thoracic surgical attending, often in consultation with the pulmonary service. The final decision on type of surgical procedure, location and number of biopsies was made by the Thoracic Surgeon. Surgical procedures were performed under general anesthesia with bronchoscopy performed concurrently. In many cases, a meaningful sharp distinction could not be drawn between a limited thoracotomy, VATS, or a limited thoracotomy with partial visualization of the pleural cavity by a thoracoscope; therefore the procedures were characterized as to whether or not video assistance was employed.

Patients who were thrombocytopenic had platelet transfusions given preoperatively. If the platelet count did not reach 50,000 post-transfusion, additional platelet transfusions were given during the procedure. Biopsies were obtained if possible from two different lobes of the lung and at the interface between apparent normal and pathologic tissue. All specimens were sent fresh and sterile to pathology for frozen and permanent sections analysis and a portion of the lung tissue was sent routinely to microbiology for bacterial, fungal, mycobacterial, and viral cultures; other pathologic and microbiologic examinations were obtained in selected patients.

Surgical complications were defined as death directly related to the procedure, abnormal bleeding postprocedure, returns to the operating room for repair of a complication of the open lung biopsy, prolonged chest tube drainage (longer than 7 d), and inability to extubate a patient postbiopsy after 48 h if the patient did not require mechanical ventilation before the biopsy.

Statistics

Association was evaluated using Fisher's exact statistics with Stata software (Stata Corp., College Station, TX).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients and Clinical Presentation

Sixty-three patients with a mean age of 49 yr (range 18 to 84 yr), undergoing a total of 67 open lung biopsies, met inclusion criteria. Patient characteristics and clinical presentation are shown in Table 1. Active cancer was known to be present in 40% of the patients before the open lung biopsy. Chemotherapy had been received by 93% of all patients at some time before the biopsy. At least one chemotherapy agent capable of causing pulmonary toxicity had been received by 88% of the patients before the biopsy and in 46% it was received within the 6 mo before the biopsy. Thoracic radiation to the thorax had been received in 34% of all patients before undergoing biopsy and in 10% of cases the radiation was within the 6 mo before the biopsy.

Yield

A specific diagnosis was found in 40 of the open lung biopsies for a yield of 62%. Only a nonspecific diagnosis was found in 27 biopsies (38%). In five patients, two concomitant pulmonary conditions were found. Thus, a total of 72 pulmonary processes were identified from the 67 biopsies. Of the 72 causes of pulmonary disease, 45 (62%) were specific and 27 (38%) nonspecific. The pulmonary diagnoses obtained at biopsy in relationship to the underlying disease are shown in Table 2.

Inflammatory disorders were the most common category of pulmonary disease, accounting for 23% of all diagnoses. Nine patients had BOOP; five had granulomatous disorders; and vasculitis, drug toxicity, and an intraparenchymal lymph node were found in one case each. The characteristics of those with BOOP are shown in Table 3. BOOP occurred in patients with all types of hematologic disorders. In four of the nine patients with BOOP, cancer was in remission, in the remainder it was active. Chest radiographic patterns showed either diffuse patchy infiltrates or multiple nodules. In three cases, BOOP was associated with another pulmonary diagnosis.

Infection accounted for 21% of all diagnoses and occurred in patients with various underlying malignancies. Fungi were the most frequent pathogens, being found in eight biopsies (Cryptococcus neoformans in four cases, Aspergillus fumigatus in two, and Coccidioides immitis and Histoplasma capsulatum, each in one case). Mycobacteria were found in four biopsies (Mycobacterium avium-intracellulare in three, and M. tuberculosis in one). There were also two cases of pyogenic bacteria and one case of CMV.

Malignancy accounted for 18% of all diagnoses, almost all of which occurred in patients with lymphoma and represented recurrent disease. Exceptions were two cases, one with a leukemic infiltrate in a patient with leukemia and one with adenocarcinoma in a patient previously diagnosed with lymphoma. There were 27 nonspecific diagnoses with interstitial fibrosis accounting for 17%, diffuse lung injury for 14%, and chronic pneumonia for 7% of all diagnoses. These entities were found associated with all of the hematologic malignancies.

The impact of clinical factors on the chance of finding a specific diagnosis at open biopsy is shown in Table 4. A specific diagnosis was found in 79% of cases with a focal abnormality on chest radiograph compared with 32% with a diffuse pattern (a relative risk [RR] of 2.60, p = 0.0001). No patient who was neutropenic or on mechanical ventilation at the time of the biopsy had a specific diagnosis found, although the number of patients in these categories was small. Analyzing according to the underlying disease, those patients with lymphoma were most likely to have a specific diagnosis and patients post-BMT were least likely but neither of these distinctions reached statistical significance. The presence or absence of known active cancer at the time of the biopsy did not impact the likelihood of finding a specific diagnosis.

Nonspecific lung injury was more likely to occur in those who had received prior pulmonary toxic chemotherapy within 6 mo of the biopsy than those who had not received such therapy (RR 1.60, p = 0.03). Prior radiation therapy had no significant impact on whether a specific or nonspecific diagnosis was found.

Effect of Prior Invasive Procedures

Thirty-five patients had undergone a bronchoscopy before proceeding to open lung biopsy with all undergoing BAL and eight transbronchial biopsies. Forty-six percent of the patients with a prior bronchoscopy had a specific diagnosis found at open lung biopsy compared with 75% in those without a prior bronchoscopy (see Table 4). Of the eight patients who underwent a transbronchial biopsy, four (50%) subsequently had a specific diagnosis made at open biopsy.

The specific diagnoses not found at bronchoscopy but identified at open biopsy were BOOP in five cases, BOOP and vasculitis in one case, lymphoma in two cases, and one case each of adenocarcinoma, granulomatous inflammation M. avium- intracellulare (MAC), As. fumigatus, CMV, C. neoformans, C. immitis, and methicillin-resistant Staphylococcus aureus (MRSA). In two cases, the diagnosis had been suggested at bronchoscopy but was not considered definitive. In one case determined to be lymphoma at open biopsy, a dense lymphoid infiltrate was present on transbronchial biopsy with cell markers suggestive of lymphoma. In another patient postallogeneic BMT with a chronic infiltrate, MRSA was obtained at bronchoscopy. At open biopsy, there was a chronic pneumonia with this organism.

Six patients had undergone fine needle aspiration before open lung biopsy. A specific diagnosis was ultimately found at open biopsy in five of these patients (Table 4). The specific diagnoses obtained from the open lung biopsy in those with a negative needle aspirate were a leukemic infiltrate, granulomatous inflammation MAC, C. neoformans and MRSA, found in one case each. The patient with chronic pneumonia with MRSA discussed previously had had both a bronchoscopy and a needle aspirate. Both less invasive procedures showed this organism.

Time Post-BMT

Because of the well-defined periods of immunologic deficits with associated pulmonary problems post-high-dose chemotherapy with BMT, we analyzed these patients in the traditional time periods of less than 1 mo, between 2 and 6 mo, and greater than 6 mo after the transplant. Of the 15 allogeneic BMT patients, none were biopsied in the first month post-transplant. Seven had biopsies between 1 and 6 mo and eight at longer than 6 mo post-transplant. The yields for specific diagnosis were 43% and 50%, respectively, for these two time periods, which were not significantly different. Of 10 autologous transplant patients, two biopsies were done in the first month, three between 1 and 6 mo, and five at longer than 6 mo post-transplant. The yields for a specific diagnosis in these time periods were 0, 67% and 60%, respectively.

Change of Therapy

Records were adequate in 65 of the 67 procedures to determine if the biopsy results led to a change of therapy. Overall, a change in treatment was made after 37 (57%) of the biopsies. Changes were made in 27 of 39 biopsies (69%) where a specific diagnosis was found and in 10 of 26 biopsies (38%) where a nonspecific diagnosis was found (p < 0.01).

Mortality

Mortality data for all groups and some subsets of patients are shown in Table 5. The overall mortality of the patients at 30 d postbiopsy was 18% and at 90 d was 22%. Those with a specific diagnosis had an improved mortality compared with those with a nonspecific diagnosis at 30 d (5% versus 38%, p < 0.008) and at 90 d (12% versus 38%, p < 0.02). Mortality for all BMT patients was higher being 36% at 30 d. There was a very significant difference in those with a specific diagnosis compared with those without (8% versus 62%, p = 0.01). The pulmonary diagnoses obtained are shown in Table 2. None of the three patients on ventilatory support at the time of the biopsy nor the five patients who required ventilatory support postbiopsy survived 30 d after the open lung biopsy.

Surgical Procedure and Complications

A procedure with video assistance alone was performed in 49% of the cases and a thoracotomy in the other 51%. In 11 patients, pleural adhesions (seven cases) and inadequate access (four cases) led to conversion to a thoracotomy after an initial attempt at a VATS. Complications occurred in 13% of the patients (Table 6) with major complications after two biopsies (3%). One patient had a splenic rupture in the immediate postoperative period and underwent immediate splenectomy. This patient, with a platelet count of 12,000, had not been on a ventilator prior to the procedure, required ventilatory support postoperatively, and ultimately died several weeks latter. Open biopsy showed acute diffuse lung injury. In a second patient, with a platelet count over 100,000, intrathoracic bleeding required a return to the operating room for successful repair.

Prolonged chest tube drainage was necessary in five patients (7%). Five patients (7%) who did not require mechanical ventilation preoperatively required ongoing mechanical ventilation postoperatively; two of these patients also required prolonged chest tube drainage. All of the patients who required postoperative mechanical ventilation died on the ventilator between 3 and 23 d postbiopsy. All five had nonspecific diagnoses found at biopsy with interstitial pneumonitis/fibrosis in two patients and diffuse alveolar damage in three.

Procedures were done in 18 patients with a platelet count less than 50,000 and in five of these patients, the platelet count was less than 20,000. Ten underwent VATS and eight thoracotomy. All thrombocytopenic patients received platelet transfusions according to the protocol indicated in METHODS. The complication rate was 39% in those with less than 50,000 platelets compared with 4% in those with greater than 50,000 but there were few complications related to bleeding and this occurred in a patient with a platelet count of greater than 100,000.

In patients with focal abnormalities, biopsies were done in the area of the abnormality. In the 28 patients with diffuse lesions, biopsies were taken from the right lung in 19 cases (68%) and from the left in nine (32%). Eleven patients (36%) had three separate portions of lung sampled, 15 patients (54%) two, and two patients (7%) had one biopsy. Of the 26 procedures in which more than one biopsy was taken, eight (31%) were in the same lobe and 18 (69%) in different lobes. Of the total of 65 individual biopsies, 14 were from the right upper lobe, 18 from the right middle lobe, 14 from the right lower lobe, 10 from lingula, seven from the left lower lobe, and two from other parts of left upper lobe.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In the immunocompromised patient with pulmonary abnormalities, open lung biopsy is done in the hope that a specific diagnosis can be found for which targeted therapy is available and altered treatment will improve survival. In our study of 63 patients with hematologic malignancies who underwent 67 open lung biopsies, a specific diagnosis was found in 62% of the biopsies and changes in therapy made in 57% of the cases based on the biopsy results. When a specific diagnosis was obtained, changes in therapy were made in 69% of cases.

Reports in the literature on the results of lung biopsies in the immunocompromised patient have been similar with a specific diagnosis being reported in 32 to 82% of patients (7, 11, 15, 16) and changes in therapy postbiopsy in 28 to 69% of cases (9, 12, 15, 17). The differences in these studies are related in part to the type of patients included and also to the definition accepted for specific diagnosis. A study in 1990 by Snyder and coworkers in 87 predominantly pediatric BMT recipients found specific diagnoses in 32% of cases, with most of the diagnoses infectious in nature (12). Another study of 74 immunosuppressed patients, including 36 with hematologic malignancies, found a specific diagnosis in 54% of the patients, with infections accounting for 46% of all diagnoses (13). A recent study from Israel of 39 immunocompromised patients, including 30 with hematologic malignancies, reported a specific diagnosis in 82% of patients (15). These reports, along with our own, indicate that open lung biopsy will yield a specific answer in a significant percentage of immunosuppressed patients with hematologic malignancy and unexplained pulmonary processes. Nonspecific pathologic findings of acute lung injury, however, will be found in some patients and this represents an ongoing diagnostic and management problem in the immunocompromised patient.

In our study, finding a specific diagnosis was significantly associated with improved survival. Ninety-five percent of patients with a specific diagnosis were alive at 90 d compared with 62% of those without a specific diagnosis. This was particularly dramatic in the recipients of BMTs, where a specific diagnosis was associated with an 8% 90-d mortality compared with 62% in those with a nonspecific diagnosis. Given that 69% of all patients with a specific diagnosis had a change in therapy postbiopsy, it seems reasonable to assume that knowledge of the specific diagnosis with the ability to treat effectively and eliminate unnecessary medication may have contributed to the good outcome, although it is possible that the improved prognosis is related only to the nature of the underlying problem itself. The benefit of making a nonspecific diagnosis is less clear as changes in therapy occurred in only 31% of these patients. These data also point out the significant mortality associated with the development of nonspecific lung injury patterns in the vulnerable lungs of the immunocompromised patient and the need for strategies to determine which factors precipitate this reaction and how to prevent it.

One of the major considerations in performing an open lung biopsy in immunocompromised patients has been whether the patient's overall prognosis is acceptable to proceed to surgery. Several studies have noted mortality rates of 30% to 60% in the perioperative period in immunocompromised patients who required open lung biopsy, although death was not clearly attributable to the surgery (9, 12, 15). The overall mortality in our patients was 18% at 30 d and 22% at 90 d. Even in the bone marrow population, believed to be the highest risk group, 90-d mortality was only 36%. Our results suggest that not all patients with hematologic malignancy and undiagnosed pulmonary problems have a poor prognosis per se, although it must be recognized that the patients in this study are a select group. Although there is no formal policy on selection of patients for open lung biopsy at our hospital, many clinicians tend to avoid surgery in patients with multiorgan system failure, neutropenic patients, particularly in the first month after BMT or those already in respiratory failure. The patients in our study reflect that bias with few patients falling in those categories.

We looked for factors predictive of a specific diagnosis being obtained at open lung biopsy. The presence of focal masses, nodules, or localized infiltrates on radiograph rather than a diffuse infiltrate, was the factor most predictive of finding a specific diagnosis. This increased yield with a focal abnormality has previously been noted in those with Hodgkin's disease and pulmonary infiltrates; specific diagnoses were found in 79% of those with focal masses but in 36% of those with other radiographic patterns in one study (18). Factors predictive of obtaining a nonspecific diagnosis in our study were presence of neutropenia, the need for mechanical ventilation, or receipt of pulmonary toxic chemotherapy within the 6 mo before the biopsy. None of the seven patients who were either neutropenic or on a ventilator had a specific diagnosis made and the 30-d mortality for this group was 71%. Although these numbers are small, the low yield and high overall mortality has made us very wary about sending such patients to open lung biopsies. Others have also reported that neutropenic patients have a high likelihood of having nonspecific pathologic findings (6, 11). Our study is not able to determine if those biopsied while on a ventilator might have had a better outcome with a biopsy if the procedure had been done earlier in their course. We believe, however, that this is unlikely because most immunocompromised patients at our institution are treated rapidly and aggressively for all pulmonary problems and progression to respiratory failure tends to be rapid.

The cause of nonspecific lung injury in immunocompromised patients is unknown but clinicians have often suspected that prior pulmonary toxic agents or thoracic radiation may contribute to the vulnerability of the lung to subsequent injury. Our data support a possible role of recent pulmonary toxic chemotherapy increasing the likelihood of nonspecific injury. We saw no effect of radiation therapy on the type of pulmonary diagnoses found. Patients who had a prior bronchoscopy had a slightly decreased likelihood of finding a specific diagnosis at open lung biopsy, compared with those without the procedure, but the overall yield for a specific diagnosis remained high enough that this was not clinically significant.

A notable difference in our study compared with prior reports is the high incidence of inflammatory disorders found. These entities, rather than infections, were the most common diagnosis at open lung biopsy. This may be attributable to the widespread use of bronchoscopy to diagnose many opportunistic infections as well as current preventive and preemptive strategies to prevent infection. Fungal and bacterial infections, including mycobacteria, were the infections found in our study. The classic opportunistic infections often found commonly in past studies, such as PCP and CMV were rare. Only one case of CMV and no cases of PCP were identified in our study. The applicability of these results to other patient groups will depend on the prophylactic strategies which are routinely employed during periods of immunosuppression and the types of empiric therapy given.

BOOP was a frequent diagnosis in our patients and occurred in association with all types of hematologic malignancy. This disorder is an inflammatory pathologic reaction of the lung characterized by patchy interstitial inflammation and organizing granulation tissue in the small airways and alveoli (19). Clinically, it may be asymptomatic or present as pneumonia. The radiographic findings include nodules, multiple patchy infiltrates, or diffuse interstitial infiltrates, and it can be very difficult to distinguish clinically from other entities (20). The cause of BOOP is unknown but it is believed to be immune-mediated and usually responds to corticosteroids in those cases where treatment is needed. An increased frequency of BOOP is seen in association with a variety of disorders, including connective tissue disorders, drugs, organ transplantation, radiotherapy, and hematologic malignancy (21). The relationship of the disorder to hematologic malignancy is unknown. In our patients cancer was in remission in approximately half of the cases and unlikely to be the direct precipitating factor. In one patient with lymphoma, the occurrence of BOOP may have heralded the recurrence of the lymphoma, which is a phenomenon that has been noted in one prior case (21). Other precipitating factors for BOOP may have been viral or other infections.

In our series, open lung biopsy had an overall complication rate of 13% with major complications occurring in only 3% and one death attributable to the surgical procedure. Complication rates in the literature have been reported from 10% to 33%, with only rare deaths (9, 12, 15, 17, 18). We were able to identify a higher risk group of patients with eight of 18 patients with less than 50,000 platelets experiencing complications, although these were not related to bleeding. One of the concerns in sending an ill patient to open lung biopsy is that it will precipitate respiratory failure and the need for mechanical ventilation. This was the case in 7% of our patients. All had nonspecific findings on biopsy and ultimately died, suggesting that the underlying disease caused the respiratory failure, but this can not be proved and it should be anticipated that some immunocompromised patients who were not in respiratory failure will require ongoing ventilatory support after open lung biopsy.

VATS has been used in some patients who require open lung biopsy with the hope that with a more limited chest wall incision, there may be reduced pain and improved lung function in the postoperative period. To date, reports have failed to document a significant decrease in pain or length of stay, although there may be some better preservation of shoulder girdle motion (24). As a practical matter, the distinction between thoracotomy (rib spread to allow palpation and manipulation of lung) and VATS (multiple 1- to 2-cm incisions allowing at best the entry of a single digit into the chest cavity) is being blurred with thoracic surgeons performing procedures that blend the two techniques. The results achieved in our series suggest that carefully selected techniques, whether VATS or thoracotomy, allow the procedure to be performed with acceptable morbidity and mortality with a significant diagnostic yield.

In summary, we have confirmed that open lung biopsy can have a beneficial role in patients with hematologic malignancy and pulmonary disease and can be performed safely in this population. Changes in therapy will occur in a significant number of patients, particularly when a specific diagnosis is found and those with a specific diagnosis have an improved survival compared with those with a nonspecific diagnosis. In selecting patients for biopsy, those with focal disease are likely to have a specific diagnosis found. The presence of neutropenia and mechanical ventilation identifies a group based on our study and prior reports, with a high risk and low yield.

Several new points emerged from our study. Infection is no longer the most common problem found, likely owing to the widespread use of bronchoscopy and other diagnostic techniques as well as preemptive, prophylactic, and empiric strategies. The classic opportunistic infections previously found at open lung biopsy, such as PCP and CMV are now rare. There is a high frequency of inflammatory disorders found at open biopsy and new diagnostic and preventive approaches need to be developed directed to these types of problems. The high mortality associated with nonspecific lung injury on biopsy suggests that future improvement in survival from pulmonary disease in this population will need to come from understanding more of the nature of the precipitating causes of this type of injury and ways to prevent it. Our study provides evidence suggesting a link between this type of injury and prior recent pulmonary toxic therapy. Our study also shows that VATS is a safe alternative to open thoracotomy in some patients with hematologic malignancies but thrombocytopenia is a marker for increased postoperative complications.