Use of antibodies to carcinoembryonic antigen and by alicejenny


									                   Downloaded from on September 16, 2012 - Published by

                                                                                 J Clin Pathol 1984;37:1215-1221

Use of antibodies to carcinoembryonic antigen and
human milk fat globule to distinguish carcinoma,
mesothelioma, and reactive mesothelium
From the University Department ofHistopathology, Southampton General Hospital, Southampton

SUMMARY   Antibodies raised against human milk fat globule (HMFG 1 and 2) and carcinoem-
bryonic antigen were used in an immunoperoxidase technique to differentiate mesothelioma,
carcinoma, and benign, reactive mesothelium. Sixteen mesotheliomas, 27 lung carcinomas, and
13 specimens of reactive mesothelium were examined. Staining for carcinoembryonic antigen was
not seen in reactive mesothelium or mesothelioma but was present in 22 of 27 carcinomas.
Mesothelioma and carcinoma usually stained with HMFG 1 and 2; reactive mesothelium did not.
These three antibodies may help to distinguish carcinoma, mesothelioma, and reactive

Distinguishing mesothelioma from carcinoma is a             malignant mesothelioma was obtained from
well recognised problem.' Histochemistry and elec-          pleuro-pneumonectomy and pleurectomy specimens
tron microscopy may help to make this distinction           or from biopsies taken at thoracoscopy or
but do not always give a definitive answer. Anti-           thoracotomy (Table 1). Adequate material was
bodies to carcinoembryonic antigen (CEA) and                available in all cases for definitive diagnosis to be
keratin have been assessed with conflicting                 made of biphasic (seven cases), epithelial (eight
results.) It is equally difficult to distinguish benign     cases), or mesenchymal (one case) malignant
from malignant mesothelium. Morphometry6 and                mesothelioma using morphological criteria. All
the use of histiocytic markers in an immunoperoxid-         cases were stained with periodic acid Schiff after
ase technique8 have been advocated for this pur-            diastase digestion and were negative. The presence
pose.                                                       of acid mucopolysaccharide sensitive to hyaluronid-
  Anti-CEA and HMFG 2 have been used previ-                 ase digestion was confirmed by staining with Alcian
ously to diagnose malignancy in cytological prepara-
tions of pleural fluids.9 '0 Their usefulness in this
respect depends on a knowledge of the staining               Table 1 Malignant mesotheliomas
behaviour of reactive mesothelium and of the pat-
tern of staining of benign and malignant cells in his-       Case Type of specimen         Histological Presence of
                                                             no                            type         hyaluronidase
tological preparations.                                                                                 sensitive
  In this study we have evaluated three antibodies,                                                     mucopolysaccharide
anti-CEA and HMFG 1 and 2, to see if they would                  1   Pleuropneumonectomy   Epithelial   +
distinguish between carcinoma, mesothelioma, and                 2   Pleuropneumonectomy   Epithelial
                                                                 3   Pleuropneumonectomy   Mixed        +
benign, reactive mesothelium.                                    4   Pleurectomy           Epithelial   +
                                                                 5   Thoracotomy           Mixed
Material and methods                                             6   Thoracoscopy          Mixed        +
                                                                 7   Thoracoscopy          Mixed        +
                                                                 8   Thoracoscopy          Mixed
Tissue was examined from 16 mesotheliomas, 27                    9   Thoracoscopy          Mixed        +
                                                             10      Thoracoscopy          Mixed        +
lung carcinomas, and 13 specimens in which reactive          11      Thoracoscopy          Epithelial
mesothelium was present. Tissue diagnosed as                 12      Thoracoscopy          Epithelial +
                                                             13      Thoracoscopy          Epithelial +
                                                             14      Thoracoscopy          Epithelial
                                                             15      Thoracoscopy          Epithelial +
                                                             16      Thoracoscopy          Mesenchymal-
Accepted for publication 24 July 1984
             Downloaded from on September 16, 2012 - Published by

1216                                                                                Marshall, Herbert, Braye, Jones
blue at pH I or 2-5 in the cytoplasm or acini of 10 of       Table 3 Results
the 15 cases with an epithelial component.                                             No   HMFG 1 HMFG 2 CEA
   The carcinomas were obtained from lobectomy,
pneumonectomy, or lung biopsy specimens. Eight               Mesotheliomas
were squamous carcinomas, eight oat cell, four large           Epithelial               6    +        +
                                                                                        2    +        _        _
cell, and seven adenocarcinomas, including one                Biphasic                  6    +        +
bronchioloalveolar cell carcinoma (Table 2).                   Mesenchymal              1
   Reactive mesothelium was examined in eight                Carcinomas
pleurae removed after pneumothorax (Table 2).                  Squamous                 7
Three pericardial and two pleural cases of mesothel-           Oat cell                 4    +        +        +
ial reaction to tumour infiltration were also                                           2    +        +        _
examined.                                                                               1    -        +        _
                                                                                        1    -        +        +
   Tissue was fixed in 10% neutral buffered formol-            Large cell               3    +        +        +
saline, routinely processed, and embedded in                                            1    -        +        +
                                                               Adeno                    6    +        +        +
paraffin wax. Sections (5 ,um) were cut and stained                                     1    +        +        _
with Mayer's haematoxylin and eosin. Further sec-            Reactive mesothelium      10
                                                                                        2    +        -        _
tions were examined with polyclonal rabbit antihu-                                      1    -        +        _
man CEA (Code No Al 15, Dakopatt Ab, Sweden),
used at a dilution of 1/500, using the method                though no consistent difference in staining pattern
described by Mepham et al. " Two monoclonal anti-            was seen with these two antibodies. Bronchial
bodies, HMFG 1 and 2, (Seward Laboratory, Lon-               reserve cells were prominent in seven cases and
don) were used at a dilution of 1/3 in an indirect           were stained with HMFG 1 and 2 (Fig. 1).
immunoperoxidase technique. These antibodies are               Normal alveolar epithelium was seen in 10 cases
directed against determinants in the membranes of            and in nine stained strongly with HMFG 2. HMFG 1
human milk fat globules.'2 They react with a variety         stained with varying intensity in five cases.
of carcinomas and normal glandular epithelia.'3 14             Reactive alveolar epithelium, present in 10 cases,
   Sections in which the first stage antibody was            stained strongly with HMFG 2 in nine cases and
replaced by Tris buffered saline served as negative          variably with HMFG 1 in seven cases.
controls and showed no staining in any of the cases
studied. Sections of breast carcinoma and colonic            Anti-CEA
carcinoma were used as positive controls for HMFG            The anti-CEA antibody stained bronchi and bron-
1 and 2 and CEA respectively.                                chioles with varying intensity in eight cases. Reserve
                                                             cells were stained in three of the seven cases in a
Results (Table 3)                                            pattern similar to that seen with the HMFG anti-
                                                             bodies. Normal alveolar epithelium stained strongly
NORMAL     BRONCHIAL                  AND         ALVEOLAR   in nine of 10 cases, as with HMFG 2. The pattern of
EPITHELIUM                                                   staining of reactive alveolar epithelium was variable
HMFG I and 2                                                 and closely paralleled HMFG 1.
Bronchial epithelium was present in 14 of the cases
examined. HMFG 1 gave focal staining of the cilia            REACTIVE MESOTHELIUM
and basal plate region of bronchial epithelial cells.        HMFG I and 2
HMFG 2 stained the epithelium of respiratory bron-           Ten of the 13 specimens of reactive mesothelium
chioles with a similar pattern but was generally             were negative with HMFG 1 and 2. One case
negative in the larger bronchioles and bronchi,              stained with HMFG 2 and two cases with HMFG 1.
                                                             Staining was very focal and seen on the free border
Table 2 Carcinomas and reactive mesothelium                  of the cells.
                                                                Care was needed to distinguish reactive pneumo-
                                                             cytes from mesothelium in excised bullous cysts, but
                                                             identification was made easier by the presence of
Carcinoma                                             27     lung tissue. Pneumocytes then stained strongly with
  Squamous                                             8
  Oat cell                                             8     all three antibodies, while mesothelium was gener-
  Large cell
                                                             ally negative (Fig. 2).
Reactive                                              13
  Pleurectomy following pneumothorax                   8     Anti-CEA
  Pericardectomy for malignant infiltration            3     This antibody did not stain any reactive mesothelial
  Pleural biopsies in suspected malignant effusions    2
                Downloaded from on September 16, 2012 - Published by

Use of antibodies to carcinoembryonic antigen and human milk fat globule
                                                  .....<a S:



                 Fig. 1 Bronchus stained for HMFG 1. The cilia and basal plate region are
                 pbsitive. Reserve cells stain on the surface adjacent to the respiratory epithelium
                 (arrow). PAP x 350.

                              o.X:._,                                                                                  149                               I*
                                         *Iso Vt                        4.
                                                                                                                                           I   f       :...9

                                             WlJ. A.A:,                                                                                    I.
                                                                                      4...                                                 %%
                  'r,   -.. Apo                         4                                             "4:,
                                                                                                                                               4       i%
                                  I.P.                                                                   %kk.

                                                                                                6 Jp

                                                                                                                 I.0             i4   5 .9I
                                                                                                                                      ,l   ^

                                                                                                                                  A        .
                             a                                                                                                    0
                           to1                                          4
                                                               MD                                                            .

                                                                ".. e

                                                                             j:*4.    A

                 Fig. 2 Bullous cyst stained for HMFG 2. The respiratory epithelium on the inner
                 surface is positive. Mesothelial cells do not stain. PAP x 250.
       Downloaded from on September 16, 2012 - Published by

1218                                                                                              Marshall, Herbert, Braye, Jones

                                                                                9/?                V5

                    A*Wa. 3p..


                                                                           .;. }

             Fig. 3 Mesothelioma stained for HMFG 1. The solid clump of cells (top centre)
             stains weakly. Staining is denser where lumina are formed. PAP x 300.



                                                                      e ..:-

                                                         .S.:   QR


              Fig. 4 Spindle cell element of a biphasic mesothelioma stained for HMFG 1.
              PAP x 400.
        _*.eEi0,';._t/ s § B! .
                                      Downloaded from on September 16, 2012 - Published by

Use of antibodies to carcinoembryonic antigen and human milk fat globule                                                                                                                                                                 1219

                                            4444ta ,f..

                             444                                                                                                                                                            'I

                            ;,                              /                                     W 'J,                                                                             i                        4?   1.

                   g s j
                   sEX                                                          #t                    >>8
                                                                                                       A                                    }            %X*
                                   tAd                                                                                                                                           ~~~~~~~~~~~~~~~~~~~~~~~t

                                                                                                                                                                        IV' ?-
                                                                                                                                                 .4. O

                                                                                                                                                  .q            ,

                                                                                                                                        I           -.

                       ..      X; w*

         .s.       s           v
                                     Sst 0        S   ...
                                                                s              .sey .h
                                                                               6: f                                .'

                                                                                                                             v /                         lC o
                                                                                         z        t
                                                                                                        *5                              v         wiPt. t{.
                                                                                                                                                                              Fig. 5(a) Squamous carcinoma stained for HMFG
                                                                                                                              4 4s <t                               e         2. There is dense staining of the cell surface in better
                                                                                                  _S.                        )                                                differentiated areas. PAP x 400. (b) Oat cell
                                                                8<e                      e    v   *%'?-o

                                                                                                                    2:                      tW
                                                                                                                                            2 *^ ';B:                         carcinoma stained for HMFG 2. PAP x 400. (c)
                              §               §         *
                                                                                         e        v
                                                                                                                                                                          *   Poorly differentiated adenocarcinoma stained for
                                                                <                            vY
                                                                                                      s .:5
                                                                                                                         $   se
                                                                                                                                                                w           a HMFG 1. PAP x 400.
                                                                                                               ^4.S ^
                              Je              *             Y               w
                       *      fiiF

         ^ .Xk 9 + x ..i
               r             s+ s                      e i
                                   'S5                                     $

MESOTHELIOMAS                                                                                                                                                                    of 27 were positive with HMFG 1. (Table 3). All the
HMFG I and 2                                                                                                                                                                     squamous carcinomas showed focal staining with
Twelve of the 16 mesotheliomas were positive with                                                                                                                                both antibodies. There were individual differences
HMFG 1 and 2 (Table 3). Staining was present in                                                                                                                                  in the staining seen with each antibody but no con-
the cytoplasm of some cases but was more com-                                                                                                                                    stant pattern emerged. Staining was cytoplasmic
monly present at the cell surface. Solid clumps of                                                                                                                               with surface accentuation in more differentiated
epithelial cells did not stain, but where clefts formed                                                                                                                          areas (Fig. 5a). Four oat cell carcinomas stained
within the tumour, the luminal surfaces of the cells                                                                                                                             strongly with HMFG 1 and 2 (Fig. 5b).
became strongly positive (Fig. 3). Three of the seven                                                                                                                               Two did not stain with HMFG 1 and two others
biphasic tumours showed strong, granular, cyto-                                                                                                                                  stained only weakly; these four carcinomas also
plasmic staining of the spindle cell element (Fig. 4).                                                                                                                           stained weakly with HMFG 2. The four large cell
Anti-CEA                                                                                                                                                                         anaplastic tumours showed strong staining with
Anti-CEA did not stain any of the cases of                                                                                                                                       HMFG 2. Two of them showed similar staining with
mesothelioma. Staining in occasional areas of nec-                                                                                                                               HMFG 1, one stained weakly with this antibody,
rosis was interpreted as negative.                                                                                                                                               and one not at all. All seven adenocarcinomas
                                                                                                                                                                                 stained with both antibodies (Fig. Sc), staining
CARCINOMAS                                                                                                                                                                       strongly in all but one case. Variations in staining
HMFG I and 2                                                                                                                                                                     with the two antibodies were noted but showed no
All carcinomas were positive with HMFG 2, and 24                                                                                                                                 constant pattern.
             Downloaded from on September 16, 2012 - Published by

1220                                                                             Marshall, Herbert, Braye, Jones
Anti-CEA                                                 regarded as proved.
Twenty two of 27 carcinomas stained strongly with           Two previous studies have examined reactive
the anti-CEA antibody. One poorly differentiated         mesothelium. In a histological study all three cases
squamous carcinoma, three oat cell, and one poorly       studied were negative.3 A further study examined
differentiated adenocarcinoma were negative.             cytological preparations of "active mesothelial
Those squamous carcinomas which stained posi-            cells" from 22 benign effusions, and again all were
tively with this antibody showed cytoplasmic stain-      negative.9 Our results agree with these observations.
ing, strongest at the cell membrane in more differen-       In contrast, all but one of the adenocarcinomas
tiated areas. Staining in oat cell carcinomas was        stained with the anti-CEA antibody; five of them,
cytoplasmic, dense in three cases and weak in two.       including two poorly differentiated, stained strongly.
The large cell tumours had a focal, granular,            This observation is in agreement with previous
intracytoplasmic staining pattern with surface accen-    studies,2-4 '5 which have found most pulmonary
tuation in one case. One poorly differentiated           adenocarcinomas to stain strongly for CEA. This
adenocarcinoma showed focal, weak, cytoplasmic           contrasts with adenocarcinomas at other sites (large
staining. The remaining five adenocarcinomas-two         bowel carcinoma excluded), where fewer tend to
well, one moderately, and two poorly                     stain for CEA.'6 The other histological types of lung
differentiated-showed strong, granular, cyto-            carcinoma were also generally positive with anti-
plasmic staining which was more pronounced at the        CEA (Table 3). When a problem exists in differen-
luminal surface. The single bronchioloalveolar cell      tiating mesothelioma from carcinoma, it is usually
carcinoma stained strongly at the luminal surface of     adenocarcinoma which is under consideration. Our
the cells. Areas of necrosis were commoner in car-       findings suggest that where a tumour stains with an
cinoma; staining in these areas was interpreted as       anti-CEA antibody, it favours the diagnosis of
negative.                                                adenocarcinoma. Absence of staining favours the
                                                         diagnosis of mesothelioma. The presence of CEA
Discussion                                               also distinguishes metastatic carcinoma from reac-
                                                         tive mesothelium. It does not distinguish reactive
In this study we have used an immunoperoxidase           mesothelium from mesothelioma, nor does it distin-
technique to evaluate three antibodies-anti-CEA          guish between the histological types of lung car-
and HMFG 1 and 2-as aids to the diagnosis of             cinoma.
mesothelioma, carcinoma, and reactive mesothelial           Twelve of 16 mesotheliomas stained with both
proliferation.                                           HMFG 1 and 2. Only two mesotheliomas were
   CEA was absent in all mesotheliomas and cases of      negative with both antibodies. In contrast, 10 of 13
reactive mesothelium. Previous studies of                cases of benign mesothelium did not stain with
mesothelioma employing this antibody have differed       either antibody. The other three stained with
in their findings2 9 (Table 4). When present, posi-      HMFG 1 or 2 but never both. Twenty four of the 27
tive staining has been attributed to the increased       carcinomas were positive with both antibodies; the
sensitivity of the immunoperoxidase technique            other three stained with HMFG 2 but not HMFG 1.
compared with immunofluorescence3 but the largest        Differences were noted in strength or pattern of
series,26 which examined a total of 77 cases with an     staining in those tumours which did stain with both
immunoperoxidase technique, found them all nega-         antibodies. Similar differences have been noted in
tive. The most recent study7 found positive staining     breast carcinomas.'7
in two cases. The source of their antibody to CEA           Only a single case of mesothelioma, in a cytologi-
differed from other studies and digestion with tryp-     cal preparation, has been studied previously, using
sin was carried out for 1 h. Most other studies have     HMFG 2 alone.9 Tumour cells were strongly posi-
omitted the use of trypsin. In our study tissues were
trypsinised for 15-30 min;trypsinisation for 1 h         Table 4 Results ofprevious studies of mesotheliomas
tended to produce non-specific staining. Two other       using anti-carcinoembryonic antigen
studies have found weakly positive staining in his-      Study             No of cases   No positive   No negative
tological preparations. In one of these,5 the method                       examined
of fixation differed from other studies and the other3
                                                         Whitaker et a12    40            0             40
used an indirect immunoperoxidase technique as           Corson et all      20            9             11
opposed to the peroxidase-antiperoxidase method.         Wanget aP          12            0             12
Clearly, differences in reagents and techniques may      Said et al5         8            2              6
                                                         Kwee et alt        37            0             37
explain the differing findings but the detection of      Holden et all      22            8             14
positive staining with antibodies to CEA in some         Ghosh et at9        1            1              0
                                                         Total             140           20            120
studies means that the case for its absence cannot be
                    Downloaded from on September 16, 2012 - Published by

Use of antibodies to carcinoembryonic antigen and human milk fat globule                                              1221
tive. Previous histological studies of reactive               study of 20 cases and comparison with pulmonary adenocar-
mesothelium have not been performed with these                cinomas. Am J Pathol 1982; 108:80-7.
                                                            Wang N, Huang S, Gold P. Absence of carcinoembryonic
antibodies, and cytological studies, using only               antigen-like material in mesothelioma. An immunohistochem-
HMFG 2, have differed in their findings. One                  ical differentiation from other lung cancers. Cancer
reported benign mesothelium as negative'0; another            1979;44:937-43.
found weak staining in seven and strong staining in         Said JW, Nash G, Tepper G, Banks-Schlegel S. Keratin proteins
                                                              and carcinoembryonic antigen in lung carcinoma: An
another seven of 22 cases.9 This discrepancy was              immunoperoxidase study of fifty-four cases, with ultrastruc-
explained by differences in fixation and staining             tural correlations. Hum Pathol 1983; 14:70-6.
technique.                                                  Kwee WS, Veldhuizen RW, Golding RP, et al. Histologic distinc-
   Our findings suggest that histological sections            tion between malignant mesothelioma, benign pleural lesion
                                                              and carcinoma metastasis. Evaluation of the application of
which do not stain with either antibody are more              morphometry combined with histochemistry and immuno-
likely to be benign than malignant mesothelium.               staining. Virchows Arch (Pathol Anat) 1982; 397:287-99.
Positive staining with both antibodies is evidence in 7 Holden J, Churg A. Immunohistochemical staining for keratin
favour of malignancy. Positive staining with these            and carcinoembryonic antigen in the diagnosis of malignant
                                                              mesothelioma. Am Surg Pathol 1984;8:277-9.
antibodies does not help to differentiate 8 Herbert A, Gallagher JPJ. Interpretation of pleural biopsy speci-
mesothelioma from carcinoma; nor does the loca-               mens and aspirates with the immunoperoxidase technique.
tion of the staining within malignant cells help to           Thorax 1982;37:822-7.
make this distinction, since staining of the cell sur-      Ghosh AK, Spriggs AI, Taylor-Papadimitriou J, Mason DY.
                                                              Immunocytochemical staining of cells in pleural and peritoneal
face, particularly where it borders a lumen or cleft, is      effusions with a panel of monoclonal antibodies. J Clin Pathol
seen in both adenocarcinoma and mesothelioma.                  1983;36: 1154-64.
   Focal staining was commonly seen with all three '° Epenetos AA, Canti G, Taylor-Papadimitriou J, Curling M,
antibodies. In two specimens of mesothelioma and              Bodmer WF. Use of two epithelium-specific monoclonal anti-
                                                              bodies for diagnosis of malignancy in serous effusions. Lancet
four of carcinoma areas of positive staining were so           1982;ii: 1004-6.
widely separated that a small biopsy specimen might         Mepham BL, Frater W, Michell BS. The use of proteolytic
well have appeared negative. Clearly this affects the         enzymes to improve immunoglobulin staining by the PAP
                                                              technique. Histochem 1979; 11:
value of these antibodies when used on biopsy 2 Taylor-Papadimitriou J,J Peterson 345-58.
                                                                                                 JA, Arklie J, Burchell J,
specimens.                                                    Ceriani RL, Bodmer WF. Monoclonal antibodies to
   Staining for keratin has been advocated to distin-         epithelium-specific components of the human milk fat globule
guish mesothelioma from adenocarcinoma of the                 membrane: production and reaction with cells in culture. Int J
lung3: the former is usually positive, and the latter         Cancer 1981;28:17-21.
                                                            Arklie J, Taylor-Papadimitriou J, Bodmer WF, Egan M, Millis
negative. A combination of an anti-keratin and                R. Differentiation antigens expressed by epithelial cells in the
anti-CEA antibody would help greatly to distinguish           lactating breast are also detectable in breast cancers. Int J
between these two. Anti-CEA antibodies are of no              Cancer 1981;28:23-9.
                                                                                                    et al. Use
use in differentiating reactive mesothelium from 4 Gatter KC, Abdulaziz Z, Beverley P, diagnosis ofof monoclonal
                                                              antibodies for the histopathological             human malig-
mesothelioma, while the monoclonal antibodies to              nancy. J Clin Pathol 1982;35: 1253-67.
milk fat globule do help to make this distinction. 'Pascal RR, Mesa-Tejada R, Bennett SJ, Garces A, Fenoglio CM.
They generally indicate malignancy in the tissue we           Carcinoembryonic antigen: immunohistochemical identifi-
have examined but are not specific for either car-            cation in invasive and intraepithelial carcinomas of the lung.
                                                              Arch Pathol Lab Med 1977;101:568-71.
cinoma or mesothelioma.                                  16 Goldenberg DM, Sharkey RM, Primus FJ. Immunocytochemical
                                                                     detection of carcinoembryonic antigen in conventional his-
We thank the staff of the University Histology                       topathology specimens. Cancer 1978;42: 1546-53.
Department and Miss Margaret Harris for typing                     Burchell J, Durbin H, Taylor-Papadimitriou J. Complexity of
                                                                     expression of antigenic determinants, recognized by mono-
the manuscript.                                                      clonal antibodies HMFG 1 and HMFG 2, in normal and
References                                                           malignant human mammary epithelial cells. J Immunol (in
   Kannerstein M, Churg J, McCaughey WTE. Asbestos and
     mesothelioma: a review. Pathol Ann 1978; 13:81-129.         Requests for reprints to: Dr RJ Marshall, Department of
 2 Whitaker D, Shilkin KB. Carcinoembryonic antigen in tissue
     diagnosis of malignant mesothelioma. Lancet 1981;i: 1369.   Histopathology, Level E, South Block, Southampton Gen-
 3Corson JM, Pinkus GS. Mesothelioma: Profile of keratin pro-    eral Hospital, Tremona Road, Southampton, 509 4XY,
     teins and carcinoembryonic antigen. An immunoperoxidase     England.
       Downloaded from on September 16, 2012 - Published by

                                  Use of antibodies to
                                  carcinoembryonic antigen and
                                  human milk fat globule to
                                  distinguish carcinoma,
                                  mesothelioma, and reactive
                                  R J Marshall, A Herbert, S G Braye, et al.

                                  J Clin Pathol 1984 37: 1215-1221
                                  doi: 10.1136/jcp.37.11.1215

                                  Updated information and services can be found at:

                                  These include:
         References               Article cited in:

     Email alerting               Receive free email alerts when new articles cite this
           service                article. Sign up in the box at the top right corner of the
                                  online article.


To request permissions go to:

To order reprints go to:

To subscribe to BMJ go to:

To top