Вы находитесь на странице: 1из 6

Radiology

R adiology Stuart A. Taylor, MRCP, FRCR Steve Halligan, MD, MRCP, FRCR Clive I. Bartram, FRCR,

Stuart A. Taylor, MRCP, FRCR Steve Halligan, MD, MRCP, FRCR Clive I. Bartram, FRCR, FRCP, FRCS

Index terms:

Anus, abnormalities, 757.245 Anus, MR, 757.121413 Pilonidal sinus, 33.244

Published online before print

10.1148/radiol.2263011758

Radiology 2003; 226:662–667

Abbreviation:

STIR short inversion time inversion recovery

1 From the Intestinal Imaging Centre, Level 4V, St Mark’s Hospital, Watford Rd, Northwick Park, Harrow, Mid- dlesex HA1 3UJ, England. Received October 30, 2001; revision requested January 18, 2002; final revision re- ceived June 13; accepted July 3. Ad- dress correspondence to S.H. (e-mail:

s.halligan@ic.ac.uk).

Author contributions:

Guarantor of integrity of entire study, S.H.; study concepts and design, S.H.; literature research, S.A.T., S.H.; clinical studies, S.H., C.I.B., S.A.T.; data acquisi- tion and analysis/interpretation, S.A.T., S.H.; statistical analysis, S.A.T., S.H.; man- uscript preparation, S.A.T.; manuscript definition of intellectual content and fi- nal version approval, S.A.T., S.H., C.I.B.; manuscript editing and revision/review, S.H., C.I.B.

© RSNA, 2003

662

Pilonidal Sinus Disease: MR Imaging Distinction from Fistula in Ano 1

PURPOSE: To describe magnetic resonance (MR) imaging features in patients with proved pilonidal sinus disease and to compare these features with those in a matched group of patients with proved fistula in ano to determine the accuracy with which MR imaging can be used to distinguish between the two diseases.

MATERIALS AND METHODS: Seven patients with pilonidal sinus disease under- went MR imaging with a body coil. The site and morphology of sepsis were noted, with particular reference to natal cleft sepsis and deep-seated sepsis, including intersphincteric anal canal sepsis and any enteric communication. Comparison was made with 14 age- and sex-matched patients with fistula in ano. Categoric frequen- cies were compared to calculate differences between the groups and sensitivities, specificities, and predictive values.

RESULTS: All patients with pilonidal sinus had natal cleft sepsis, but five (71%) had sepsis at deep-seated sites more characteristic of fistula in ano. Eight patients with fistula in ano (57%) had natal cleft sepsis that was thought characteristic of pilonidal sinus. No patient with pilonidal sinus had intersphincteric sepsis or an enteric communication, in contrast to all patients with fistula in ano having both (P .001). Natal cleft sepsis reached the subcutaneous tissues overlying the coccyx and sacrum in only one patient with fistula (7%), in contrast to six (86%) with pilonidal sinus (P .001). MR imaging had a sensitivity of 86% (six of seven), specificity of 100% (14 of 14), positive predictive value of 100% (six of six), and negative predictive value of 93% (14 of 15) for diagnosis of pilonidal sinus disease.

CONCLUSION: MR imaging features of perianal and deep-seated sepsis, character- istic of fistula in ano, are also found in patients with pilonidal sinus, but the absence of intersphincteric sepsis or enteric opening allows reliable MR imaging distinction between the two.

© RSNA, 2003

Pilonidal sinus disease is characterized by natal cleft suppuration due to hair follicle infection. The disease results from chronic infection of hair follicles and subsequent formation of a subcutaneous abscess because of persistent folliculitis. Hair then enters the abscess cavity and provokes a foreign body tissue reaction; chronic suppuration and discharge through a midline sinus follow (1). Although a midline natal cleft sinus is characteristic, clinical diagnosis may occasionally be difficult because of clinical overlap with fistula in ano. This is especially the case when a pilonidal sinus is complicated by secondary tracts and extensions, which are also characteristic of anal fistula (2). Differen- tiation between the two is vital because both have a tendency to recur if surgery has been inappropriate or inadequate (3). The central difference is that pilonidal sinus disease originates subcutaneously, whereas anal fistulas have an enteric communication within the anal canal or within the rectum in extrasphincteric fistulas (2,3). Magnetic resonance (MR) imaging has become the reference standard for diagnosis and classification of fistula in ano (4), surpassing even surgical exploration (5), and is now well established in this role. MR imaging is able to depict the fistula tract and any extensions in surgically important planes and can also be used to determine the relationship of the tract to the anal sphincter complex to predict the likelihood of postoperative inconti- nence.

Radiology

Surprisingly, to our knowledge, there are no descriptions of the use of MR im- aging for detecting pilonidal sinus dis- ease or of any MR imaging features that help distinguish pilonidal sinus from s- tula in ano. We hypothesized that MR imaging should be able to depict the presence and distribution of pilonidal sepsis with a high degree of accuracy, similar to that for stula in ano. More importantly, we also hypothesized that MR imaging should be able to be used reliably to differentiate between pilo- nidal sinus and stula in ano because intersphincteric sepsis is present only in the latter. The purpose of this article was therefore to describe MR imaging fea- tures in patients with proved pilonidal sinus disease and to compare these fea- tures with those in a matched group of patients with proved stula in ano to de- termine the accuracy with which MR im- aging can be used distinguish between the two diseases.

MATERIALS AND METHODS

Patients

A retrospective review of all patients undergoing MR imaging for the evalua- tion of perianal sepsis in our department between January 1998 and December 2000 yielded seven male patients (me- dian age, 24 years; age range, 1534 years) with pilonidal sinus disease, as compared with approximately 1,100 patients who underwent imaging for stula in ano. All had a surgically proved diagnosis of pi- lonidal sinus disease. Two patients pre- sented with a midline natal cleft abscess, and the remaining ve had recurrent dis- ease after previous treatment for a known diagnosis of pilonidal sinus. All patients underwent denitive surgery after MR imaging, at which time the diagnosis was conrmed. Our institutional review board did not require its approval or patient con- sent for our retrospective review. Each patient with pilonidal sinus dis- ease was matched to two subjects under- going MR imaging for idiopathic cryp- toglandular stula in ano. Subjects with stula in ano were selected by choosing the next two patients who underwent imaging after the study patient, who sat- ised these criteria, and who were sex- and age-matched to within 5 years to each study patient. All of the 14 patients (median age, 26 years; age range, 19 32 years) selected had a surgically and pathologically proved diagnosis of cryp- toglandular stula in ano either at the time of surgery or subsequently; 10 had

recurrent disease, and four presented for the rst time.

Imaging

MR imaging was performed by using a 1.0-T superconducting magnet (Gyroscan T10-NT; Philips Medical Systems, Reigate, Surrey, England) and body coil, with pa- tients in the supine position. Imaging was supervised by an attending radiolo- gist, and all patients underwent imaging according to a well-established depart- mental protocol for MR imaging of peri- anal sepsis (3). The long axis of the anal canal was identied by using a midline sagittal localizer image. By using this image for guidance, transverse and coronal short inversion time inversion-recovery (STIR) images were aligned with the longitudinal axis of the anal canal and were obtained by using the following parameters: 1,500/15 (repetition time msec/echo time msec), inversion delay of 140 msec, 375-mm eld of view, 256 256 matrix, 4-mm section thickness, 1-mm intersection gap, and four signals acquired. Sagittal T2-weighted images were sometimes also obtained at the discretion of the super- vising radiologist by using the following parameters: 4,563/150, 350-mm eld of view, 256 256 matrix, 6-mm section thickness, and 0.6-mm intersection gap.

Image Review

All images were reviewed by an experi- enced radiologist (S.H.) with a subspe- cialty interest in gastrointestinal radiol- ogy using a dedicated MR imaging workstation (Easyvision; Philips Medical Systems). The assessing radiologist was unaware of the underlying diagnosis in each case although was aware that the diagnosis was either pilonidal sinus dis- ease or stula in ano. Images from the seven patients with pilonidal sinus dis- ease and the 14 patients with stula in ano were reviewed at the same session but in a randomized order. Sepsis was dened by using established MR imaging criteria for inammation on STIR sequencesnamely, areas of high signal intensity within the soft tissues of the natal cleft, perineum, and ischioanal fossae, which were distinct from the high signal intensity normally returned by blood vessels (24). The presence of sep- sis was noted and its distribution re- corded. Particular note was made of natal cleft sepsis, which was suggestive of pilonidal sinus disease. The natal cleft was dened

as the midline tissues and adjacent but- tock 2 cm or more posterior to the anal orice. Note was made if natal cleft sepsis extended so far cranially as to involve the subcutaneous tissues overlying the bony coccyx and/or sacrum. Note was also made of sepsis involving the intersphinc- teric plane (ie, the plane between the ex- ternal and internal anal sphincters), which was suggestive of an anal canal enteric communication and therefore an underlying diagnosis of stula in ano (3). The site and level of any enteric opening were recorded if present. Perianal, ischioanal, and supralevator sepsis, all suggestive of stula in ano, were also sought and the site noted. The perianal region was dened as the super- cial area within 2 cm of the anal orice. The ischioanal fossa was dened as the tissues bounded by the inferior aspect of the levator plates superiorly, the sphinc- ter apparatus medially, and the level of the inferior border of the subcutaneous external sphincter inferiorly. Supraleva- tor sepsis was dened as any sepsis lying at a level above the levator plate. Where sepsis was identied, its mor- phology was classied as either a tract, dened as a longitudinal tubular struc- ture, or as an abscess, dened as a local- ized spherical collection. Whether tracts or abscesses were single or multiple was also determined for each site, as was whether sepsis crossed striated muscle of the external sphincter, to form a trans- sphincteric tract, or the levator plate di- rectly. After these observations were made, the observer ascribed a nal diag- nosis of either pilonidal sinus or stula in ano on the basis of his overall impression of the imaging ndings.

Statistical Analysis

Imaging ndings from patients with pilonidal sinus disease and stula in ano were subsequently compared to deter- mine any differences between the two. Categoric frequencies for the ndings as- sessed were calculated as simple percent- ages for patients with pilonidal sinus dis- ease and stula in ano, and signicant differences in proportions were calcu- lated by using the Fisher exact test. Sta- tistical signicance was assigned to a probability value of .05. The disease cat- egorypilonidal sinus disease or stula in anoultimately ascribed by the as- sessing radiologist was compared with the true category to calculate the sensi- tivity, specicity, positive predictive value, and negative predictive value of MR imaging.

Radiology

RESULTS

The MR imaging ndings are summa- rized in the Table. There was MR imaging evidence of active sepsis in all patients in the study. All patients with pilonidal sinus had natal cleft sepsis, and in six (86%) pa- tients this extended so far as to reach the subcutaneous tissues overlying the coc- cyx and/or sacrum (Fig 1). However, three (43%) patients also had perianal sepsis (Fig 2), four (57%) had ischioanal sepsis (Fig 3), and one (14%) had supra- levator sepsis (Fig 4). Thus, only patients 4 and 7 (29%) had typical features of pi- lonidal sinus diseasenamely, a subcuta- neous natal cleft sinus without evidence of perianal, ischioanal, or supralevator disease to cause confusion with stula in ano (Fig 1, Table). Striated muscle was crossed by tracts in two patients; the le- vator plate was directly penetrated by a tract in one (Fig 4), and the most inferior aspect of the subcutaneous external anal

664 Radiology March 2003

sphincter was crossed in another (Fig 5). There was no evidence of intersphinc- teric sepsis in any patient with pilonidal sinus disease. Of the patients with stula in ano, three had intersphincteric stula, 10 had transsphincteric stula, and one had both intersphincteric and transsphinc- teric stula. In contrast to those with pi- lonidal sinus, all 14 patients with stula in ano had ndings of intersphincteric sepsis at MR imaging (Fig 6, Table), and an enteric communication was accu- rately predicted from this in all 14 sub- jects. As expected, all patients with stula in ano had MR imaging evidence of peri- anal sepsis. There was additional ischio- anal sepsis in six subjects (43%) with s- tula and supralevator sepsis in two (14%), both of which were extensions from complicated intersphincteric stulas. Al- though eight patients with stula in ano had natal cleft sepsis (57%), this sepsis reached the subcutaneous tissues overly-

ing the coccyx and sacrum in only one patient (7%). Thus, patients with pilonidal sinus dis- ease had signicantly more disease over- lying the coccyx (P .001) and less in- tersphincteric sepsis (P .001), enteric communication (P .001), and perianal sepsis (P .006) than did patients with stula in ano. There was no signicant difference between the two groups with respect to ischioanal sepsis (P .659) or supralevator sepsis (P .99). The in- creased frequency of natal cleft sepsis in patients with pilonidal sinus disease was not statistically signicant (P .061). Using MR imaging, the blinded ob- server correctly placed all 14 patients with stula in ano and six of the seven patients with pilonidal sinus disease into the correct category. For stula, these re- sults gave MR imaging a sensitivity of 100% (14 of 14), a specicity of 86% (six of seven), a positive predictive value of 93% (14 of 15), and a negative predictive value of 100% (six of six), for a preva-

Taylor et al

Radiology
Radiology

Figure 1. Sagittal T2-weighted MR image (4,563/150, 350-mm eld of view, 256 256 matrix, 6-mm section thickness, 0.6-mm inter- section gap) in a patient with pilonidal disease shows a subcutaneous sinus (arrow) to the right of the midline at the level of the coccyx. P posterior.

lence of abnormality of 67% (14 pa- tients). For pilonidal sinus disease, these results gave MR imaging a sensitivity of 86% (six of seven), a specicity of 100% (14 of 14), a positive predictive value of 100% (six of six), and a negative predic- tive value of 93% (14 of 15), for a preva- lence of abnormality of 33% (seven pa- tients). These data are suggestive of excellent discrimination between the two diseases on the basis of MR imaging ndings. The only misclassied case was in patient 5, who had pilonidal sinus in which the pilonidal tract crossed the inferior aspect of the subcutaneous external anal sphincter, which raised the possibility of a transsphincteric tract and therefore s- tula in ano (Fig 5, Table).

DISCUSSION

Described by Mayo in 1833 (6), pilonidal sinus disease was initially thought likely to be congenital in origin, with abnormal persistence of the embryologic neural tube the most favored cause (7). During World War II, the disease attracted a great deal of attention because of its high prevalence among soldiers. Soon after the end of the war, the majority of work- ers concluded the disease was most likely acquired and a result of hair follicle in- fection, perhaps because of skin stretch- ing in adolescence (8). It is thought that

of skin stretch- ing in adolescence (8). It is thought that Figure 2. Transverse STIR MR

Figure 2. Transverse STIR MR image (1,500/15, 375-mm eld of view, 256 256 matrix, 4-mm section thickness, 1-mm intersection gap, four signals acquired) in a patient with pilonidal sinus disease shows subcutaneous perianal sepsis (arrow).

the intersphincteric plane between the internal and external anal sphincters (13). The ensuing intersphincteric ab- scess then reaches the perineum through a variety of possible routes (2,3,13). Be- cause the infected gland communicates directly with the anal canal lumen, usu- ally at the level of the dentate line, the result is a stula, as opposed to a sinus. Recurrence is well recognized and trou- blesome, both for the patient and sur- geon, and is usually because of additional tracts and abscesses unsuspected and thus undrained at the time of surgery (3). MR imaging has revolutionized preoper- ative classication of stula in ano and has repeatedly been proved more sensi- tive than surgical exploration in detec- tion of complicating tracts and abscesses

(4,5).

However, not all perineal sepsis is due to stula in ano, and pilonidal sinus dis- ease, although less common, is an impor- tant differential diagnosis. Most obvi- ously, the presence of secondary tracts and extensions can cause clinical overlap with stula in ano, especially if they reach the perianal region; 43% of the pa- tients with pilonidal sinus in our study had perianal sepsis (Fig 2). Furthermore, although pilonidal disease is subcutane- ous in origin, the sinus may eventually penetrate deep into the ischioanal fossa, again causing clinical confusion; 71% of the subjects with pilonidal sinus in our study had deep ischioanal tracts that

surrounding hairs are sucked into the di- lated follicle by normal movement of the contiguous surfaces of the buttocks at the natal cleft. These hairs incite a foreign body reaction that results in a subcutane- ous sinus lined with granulation tissue, squamous epithelium, and epithelial de- bris (1). Pilonidal sinus occurs mainly in young adults, and predisposing factors include male sex, obesity, and family history (9). Most of those affected are in their twen- ties, with the disease being rare in people older than 40 years old (10). Supporting this explanation, all of the patients in our study were male, and the oldest was 34 years old. Although a single midline si- nus posterior to the anus is characteristic, additional sinuses are frequent, often with lateral openings (11). Approximately half of the patients affected present with acute abscesses, while the others have chronic infection with discomfort and a foul-smelling discharge (10). Conservative treatment with meticulous hygiene and shaving may be adequate for some pa- tients, but denitive treatment is often sur- gical. Like surgery for stula in ano, the aim is to remove the whole of the infected sinus and its ramications. Also echoing stula disease, authors of some series (12) report recurrence rates as high as 30%. In contrast to pilonidal sinus, in which the origin of infection is supercial, most stulas in ano arise from infection of the cryptoglandular anal glands, which lie in

Radiology
Radiology

Figure 3. Transverse STIR MR image (1,500/15, 375-mm eld of view, 256 256 matrix, 4-mm section thickness, 1-mm intersection gap, four signals acquired) in a patient with pilonidal disease shows the sinus tract (straight arrow) reaching the lateral margin of the external anal sphincter muscle (curved arrow).

contacted the external anal sphincter or levator plate. In one patient, the pi- lonidal tract actually penetrated the leva- tor muscles to terminate within the su- pralevator space. We hypothesized that MR imaging of pilonidal sinus would be able to demon- strate the presence, nature, and distribu- tion of sepsis as effectively as it does for stula in ano. Our study results suggest that this may be the case, with high MR imaging sensitivities and specicities for both diseases, combined with excellent positive and negative predictive values. We had thought accurate distinction be- tween the two diseases would be simple with the use of MR imaging but were surprised by the frequency of perianal, ischioanal, and even supralevator sepsis in patients with pilonidal sinus, all fea- tures that we had hoped would reliably distinguish them from stula in ano. Furthermore, sepsis reached the natal cleft in 57% of patients with stula in ano, which perhaps is suggestive of pilo- nidal sinus disease. However, this reached the subcutaneous tissues overlying the coc- cyx and sacrum in only one patient with stula (7%), in contrast to six (86%) of those with pilonidal sinus, which suggests that this feature raises the possibility of pilonidal disease if present at MR imaging. The most reliable distinction was based on the presence of intersphincteric sepsis and an inferred enteric communication

sepsis and an inferred enteric communication Figure 4. Sagittal T2-weighted MR image (4,563/150, 350-mm

Figure 4. Sagittal T2-weighted MR image (4,563/150, 350-mm eld of view, 256 256 matrix, 6-mm section thickness, 0.6-mm inter- section gap) in a patient with pilonidal disease shows the sinus tract (straight arrow) coursing posterior to the coccyx (arrowhead) and penetrating the posterior aspect of the levator muscle to reach the supralevator space (curved arrow).

current sepsis, and it is in this group that accurate preoperative exclusion of an un- derlying stula is vital. This nding ex- plains the relatively small numbers of pa- tients with pilonidal sinus disease, as compared with those with stula in ano, recruited during the study, which is a potential weakness. Similarly, 10 of the patients with stula in ano also had re- current disease, since MR imaging is par- ticularly valuable in this group. Preoperative distinction between the two causes is important because the sur- gical approach is different and both have a tendency to recur if surgery is inade- quate (12,13). Surgery for both is based on laying the infected tracts open so that sepsis is adequately drained. By deni- tion, stula in ano involves the anal sphincter muscles so that surgery neces- sitates a variable degree of sphincter in- cision, the extent of which depends on the precise geography of the stula (2,3); this incision risks postoperative inconti- nence. In contrast, surgery for pilonidal sinus should not risk incontinence. The value of correct preoperative distinction between the two lies with demonstrating an unsuspected enteric communication in a patient believed to have pilonidal sinus disease. Some surgeons (14,15) also advocate closed procedures for pilonidal sinus, which would be inappropriate if the underlying diagnosis were actually stula in ano.

rather than on deep-seated or perianal sepsis. By denition, patients with cryp- toglandular stula disease will have inter- sphincteric sepsis because the disease originates in the space between the inter- nal and external sphincters (2,3). In con- trast, pilonidal sinus would not be ex- pected to penetrate the sphincters and involve the anus. These contrasting causes were supported because no patient with pilonidal sinus had any MR imaging evi- dence of intersphincteric sepsis, whereas all patients with stula did. The blinded observer misclassied only one case in this study in a patient with pilonidal sinus in which the sinus involved part of the subcutaneous anal sphincter. However, by denition, this was below the level of the termination of the internal sphincter, which terminates at the level of the supercial external anal sphincter, and there was therefore no intersphincteric sepsis. In retrospect, this should have alerted the observer to the possibility of pilonidal sinus disease rather than stula in ano. It would also have been preferable to use more than one observer to determine agreement. It should also be borne in mind that patients with pilonidal sinus were not specially recruited for this study and were therefore patients in whom there was some clinical doubt as to the cause or extent of disease: Five of the patients with pilonidal sinus in our study had re-

666 Radiology March 2003

Taylor et al

Radiology
Radiology

Figure 5. Transverse STIR MR image (1,500/15, 375-mm eld of view, 256 256 matrix, 4-mm section thickness, 1-mm intersection gap, four signals acquired) in a patient with pilonidal sinus mistak- enly classied as stula in ano shows that the tract (straight arrow) apparently crosses the muscles of the external anal sphincter (curved arrow). L left.

In summary, results of this study sug- gest that MR imaging may be able to demonstrate the site and nature of sepsis in patients with pilonidal sinus with the same accuracy as that for stula in ano. MR imaging features of perianal and deep-seated sepsis, characteristic of s- tula in ano, are also found in patients with pilonidal sinus, and natal cleft sep- sis, typical of pilonidal sinus, is also found in patients with stula in ano. However, natal cleft sepsis that reaches the subcutaneous tissues overlying the coccyx and sacrum, and the absence of intersphincteric sepsis or enteric open- ing, are suggestive of pilonidal sinus dis- ease rather than stula in ano.

References

1. Nivatvongs S. Pilonidal disease. In: Nich- olls RJ, Dozois RR, eds. Surgery of the

disease. In: Nich- olls RJ, Dozois RR, eds. Surgery of the Figure 6. Transverse STIR MR

Figure 6. Transverse STIR MR image (1,500/15, 375-mm eld of view, 256 256 matrix, 4-mm section thickness, 1-mm intersection gap, four signals acquired) in a patient with stula in ano shows an ischioanal collection (straight arrow) and a tract crossing the external anal sphincter to reach the intersphincteric plane and a posterior anal canal opening (curved arrow).

9. Sondenaa K, Anderson E, Nesvik I, Sore-

ide A. Patient characteristics and symp- toms in chronic pilonidal sinus disease. Int J Colorectal Dis 1995; 10:39 42.

10. Jones D. Pilonidal sinus. BMJ 1992; 305:

410 412.

11. da Silva H. Pilonidal cyst. Dis Colon Rec- tum 2000; 43:1146 1156.

12. Bisset IP, Isbister WH. The management of patients with pilonidal disease: a com- parative study. AustNZJ Surg 1987; 57:

939 942.

13. Parks AG, Gordon PH, Hardcastle JD. A classication of stula-in-ano. Br J Surg 1976; 63:112.

14. McLaren CA. Partial closure and other

techniques in pilonidal surgery: an assess- ment of 157 cases. Br J Surg 1984; 71:561

562.

15. Spivak H, Brooks VL, Nussbaum M, Fried-

man I. Treatment of chronic pilonidal disease. Dis Colon Rectum 1996; 39:1136

1139.

colon and rectum. New York, NY:

Churchill Livingstone, 1997; 245254.

2. Morris J, Spencer JA, Ambrose NS. MR imaging classication of perianal stulas and its implications for patient manage- ment. RadioGraphics 2000; 20:623635.

3. Halligan S. Imaging stula-in-ano. Clin Radiol 1998; 53:8595.

4. Lunniss PJ, Armstrong P, Barker PG, Reznek RH, Phillips RKS. Magnetic reso- nance imaging of anal stulae. Lancet 1992; 340:394 396.

5. Chapple KS, Spencer JA, Windsor AC, Wilson D, Ward J, Ambrose NS. Prognos- tic value of magnetic resonance imaging in the management of stula-in-ano. Dis Colon Rectum 2000; 43:511516.

6. Mayo OH. Observations on injuries and diseases of the rectum. London, England:

Burgess & Hill, 1833; 4546.

7. Mallory FB. Sacro-coccygeal dimples, si- nuses and cysts. Am J Med Sci 1892; 103:

263277.

8. Patey DH, Scarff RW. Pathology of post- anal pilonidal sinus: its bearing on treat- ment. Lancet 1946; 2:484 486.