Professional Documents
Culture Documents
Yvonne Wu, M.D., M.P.H. BACKGROUND: Risk factors for ventriculoperitoneal (VP) shunt complications have
Departments of Neurology not been assessed with population cohort data since the advent of modern surgical
and Pediatrics,
techniques. We examined demographic factors and VP shunt complications in a pop-
University of California, San Francisco,
San Francisco, California ulation-based retrospective cohort study of all nonfederal California hospital admis-
sions between 1990 and 2000.
Nella L. Green, B.A. METHODS: We identified all admissions in a statewide administrative hospital dis-
Department of Neurology, charge database that indicated a VP shunt insertion. Repeat hospital admissions within
University of California, San Francisco,
San Francisco, California
California generating a discharge diagnosis of surgical shunt complication, defined as
a shunt replacement, revision, removal, or exploration, represented the primary out-
Margaret R. Wrensch, Ph.D. come of interest. We performed Kaplan-Meier survival analyses and Cox proportional
Department of Neurological Surgery, hazards modeling to analyze sex, age, ethnicity, socioeconomic status based on payer
University of California, San Francisco, status, and type of hydrocephalus in relation to the risk of shunt complications.
San Francisco, California
RESULTS: The annual population incidence of VP shunt placement was 5.5 per 100,000.
Shoujun Zhao, M.D. Among 14,455 individuals with 65,040 person-years of follow-up, the cumulative
Department of Neurology,
complication rate at 5 years was 32%. Children demonstrated a higher rate of shunt
University of California, San Francisco, complications than did adults at 5 years (48 versus 27%, P ⬍ 0.0001). The following
San Francisco, California demographic factors were independently associated with increased risk of shunt com-
plications: male sex (hazard ratio [HR], 1.1; 95% confidence interval [CI], 1.03–1.2),
Nalin Gupta, M.D., Ph.D. low socioeconomic status (HR, 1.2; 95% CI, 1.1–1.3), and age younger than 19 years
Departments of Pediatrics
(HR, 1.6; 95% CI, 1.4–1.7). Compared with communicating hydrocephalus, obstruc-
and Neurological Surgery,
University of California, San Francisco, tive hydrocephalus was also associated with increased risk of shunt complications
San Francisco, California (HR, 1.7; 95% CI, 1.5–1.9).
CONCLUSION: VP shunt complications are common, especially in children. Further
Reprint requests:
Yvonne Wu, M.D., M.P.H., study is needed to explain the associations between demographic factors and elevated
Department of Neurology, shunt complication rates.
University of California, San Francisco,
350 Parnassus Avenue, Suite 609, KEY WORDS: Epidemiology, Hydrocephalus, Ventriculoperitoneal shunt
San Francisco, CA 94143-0137.
Email: wuy@neuropeds.ucsf.edu Neurosurgery 61:557–563, 2007 DOI: 10.1227/01.NEU.0000280021.61888.C8 www.neurosurgery-online.com
H
Accepted, April 29, 2007. ydrocephalus is a common condition shunt malfunction include obstruction, mech-
accounting for 69,000 hospital dis- anical disconnection or breakage, infection,
charges each year in the United States and overdrainage. In pediatric surgical series,
(1). Congenital hydrocephalus is the second shunt failures occur in 14% of patients just
most common congenital brain malformation within the first month after shunt placement
after spina bifida (22), and there are approxi- (13), and 40 to 50% of shunts will fail within
mately 30,000 cerebrospinal fluid (CSF) shunt the first year (12, 19). Adults also experience a
procedures performed annually in the United relatively high (29%) shunt failure rate within
States (1, 16), amounting to $94 million of med- the first year (7). Long-term studies suggest
ical costs per year (1). that 45 to 59% of all patients, regardless of age,
CSF shunt procedures have dramatically will require a shunt revision (2, 7). Multiple
reduced the morbidity and mortality of hydro- shunt failures (11, 19, 21) and infections (10)
cephalus, but they have potential complica- are common, and revisions account for 48% of
tions that may require multiple surgical proce- all shunt-related procedures performed in the
dures during a patient’s lifetime. Causes of United States (1).
Studies of CSF shunt complications based on experiences In a survival analysis modeling time to first shunt complication, we
from a single hospital (2, 6, 9, 11, 13, 14, 17, 19, 21) have shown considered the period at risk to begin at the time of VP shunt place-
that younger age, prematurity, increasing number of previous ment. Because the exact date of procedure was not available to us, we
revisions, and shorter time to first revision increase the risk of estimated the beginning of the risk period as the 15th of the month
shunt complications (13, 17, 19). Specific causes of hydro- when the VP shunt was placed. The failure event was defined as the
first surgical shunt complication. Patients were censored or removed
cephalus, such as intraventricular hemorrhage and brain
from analysis at the time of death during any hospitalization. Patients
tumor, have also been reported to increase the risk of shunt fail- who were not identified as deceased in the dataset were assumed to
ure (7, 11, 19), although not all studies have supported this have been followed until the end of the study period and, thus, were
association (13, 17). censored in December of 2000. Because we were unable to account for
Few population-based studies of CSF shunts have been patients who left the state of California or to identify those who died
reported in the United States. One study predates the 1980s outside of a California hospital, our methods overestimate the true at-
(8), whereas the other two are cross-sectional studies that lack risk period. Thus, our study provides a conservative or lower-bound
follow-up data (1, 16). Thus, the demographic factors that estimate of the rate of VP shunt complication.
influence the risk of shunt complications have yet to be deter- We constructed Kaplan-Meier curves to model complication-free
mined in an unselected population. We analyzed California shunt survivals and performed log-rank tests to identify factors that are
associated with lower shunt survivals in a univariate analysis. We then
hospital discharge data over an 11-year period to examine
constructed multivariate models using Cox proportional hazard regres-
ventriculoperitoneal (VP) shunt complication rates and to sion analysis and calculated hazard ratios (HRs) and 95% confidence
determine factors that may be associated with increased risk intervals (CIs). All individual factors that were significant predictors in
of shunt complications. the univariate analysis were included in the multivariate model.
Statistical analyses were performed using the Stata statistical software
package (v. 7; Stata Press, College Station, TX).
PATIENTS AND METHODS
The Office of Statewide Health Planning and Development collects RESULTS
administrative data regarding discharge diagnoses and procedures for
all hospital admissions to nonfederal hospitals in California. Using this
The incidence rate of new VP shunt placements was 5.5 per
database, we identified all individuals admitted to a California hospi-
tal during the years 1990 to 2000 who received a discharge procedure
100,000 per year, based on the 19,284 patients identified during
code (International Classification of Diseases, 9th Revision-Clinical the study period. Children represented one-third of all patients
Modification [ICD-9-CM], 02.34) indicating insertion of a VP shunt. who received a VP shunt during the study period (Table 1).
Patients who had a revision, replacement, or removal of a CSF shunt Obstructive hydrocephalus (49%) was the most common
were only included in this study if they also carried a discharge proce- accompanying diagnosis, followed by congenital hydro-
dure code indicating the placement of a new VP shunt during a previ- cephalus, including spina bifida (17%) and communicating
ous hospitalization within the study period. The incidence rate of CSF hydrocephalus (14%).
shunt insertion was calculated as the total number of new placements The rate of shunt complications could only be determined in
per year, divided by the California population for that given year (4). the 75% of patients who had a social security number and who
Patients whose zip codes indicated residence outside the state of
could, thus, be followed over time. Infants aged younger than
California were excluded, as were patients with an age of greater than
100 years when their CSF shunts were placed (n ⫽ 9).
1 year were more likely to lack a social security number and,
Patients who have a social security number are given a unique iden- thus, to be excluded from follow-up analyses than were
tifier in the Office of Statewide Health Planning and Development data- patients aged more than 1 year at the time of initial shunt inser-
base and, thus, could be followed over subsequent hospitalizations tion (67 versus 14%, P ⬍ 0.0001). Hispanics were also more
within California during the 10-year study period. However, patients likely than non-Hispanics to lack a social security number (40
without a unique identifier could not be tracked over time and were, versus 20%, P ⬍ 0.001).
therefore, excluded from analyses regarding shunt complications.
We analyzed the following individual characteristics in relation to
Shunt Complications
shunt complication rates: sex, ethnicity, age at the time of the first
shunt placement (⬍1 month [neonate], 1 to 12 months [infant], 1 to 18 The remaining analyses are restricted to the 14,455 patients
years [child], ⬎18 years [adult]), hydrocephalus type (spina bifida, (of whom 80% were aged ⬎ 18 yr) with a VP shunt whose
congenital excluding spina bifida, communicating, obstructive), and subsequent hospitalizations could be identified. The mean
socioeconomic status (SES) based on source of payment (high SES length of follow-up was 54 months (standard deviation, 38
indicates private insurance or self-pay; low SES indicates Medi- mo), and children were followed on average longer than
care/Medicaid or indigent). adults (57.4 versus 53.1 mo, P ⬍ 0.001). Three percent of
Our primary outcome of interest was the time to the first surgical
patients died during the initial hospitalization when the VP
shunt complication. A surgical shunt complication was defined as the
presence of at least one of the following procedure codes in a subse-
shunt was placed. Another 11% of patients died during a sub-
quent hospital discharge record: replacement or revision of ventricular sequent hospitalization within California. Children were less
shunt (ICD-9-CM, 2.42), removal of ventricular shunt (ICD-9-CM, likely to die in a hospital during the study period than were
02.43), exploration of ventricular shunt (ICD-9-CM, 02.41), or distal adults (6.2 versus 16.7%, P ⬍ 0.0001). Among those who died
shunt revision (ICD-9-CM, 54.95). during a subsequent hospitalization, the median time from
TABLE 1. Characteristics of 19,284 patients who received a ven- TABLE 2. Surgical shunt complication rates among 14,455 chil-
triculoperineal shunt in California between 1990 and 2000 dren and adults who received a ventriculoperineal shunt in
California between 1990 and 2000a
%
All patients (%) Children (%) Adults (%)
Male 52.4
n ⴝ 14,455 n ⴝ 2905 n ⴝ 11,550
Age at shunt placement
Any shunt complication 29.0 46.4 24.6
Neonate (<1 mo) 7.9
Shunt replacement 21.6 36.4 17.9
Infant (1 mo-1 y) 12.2
Shunt removal 11.0 20.8 8.6
Child (1–18 y) 13.8 Distal shunt complication 5.2 7.9 5.9
Adult (>19 y) 66.2 Shunt exploration 0.7 1.1 0.5
Ethnicity
a
Note that all types of shunt complications were statistically more common in
Caucasian 60.1 children than in adults (P ⬍ 0.0001).
Hispanic 18.7
African-American 7.2
Asian 5.9 adults (26 versus 12% in the first month, 39 versus 21% at
Native American 0.4 1 year, 48 versus 27% at 5 years, and 55 versus 29% at 10 years,
Unknown 7.8 P ⬍ 0.0001, log-rank test). Compared with adults, neonates
(univariate HR, 2.5; 95% CI, 2.2–2.9) had the worst outcome
Low socioeconomic status 50.2
(Fig. 1A). Infants (HR, 2.0; 95% CI, 1.8–2.2) and older children
Hydrocephalus type (HR, 2.0; 95% CI, 1.9–2.2) also exhibited an increased risk of
Obstructive only 48.5 complications when compared with adults.
Communicating only 14.3 Patients with obstructive hydrocephalus (Fig. 1B) exhibited a
Communicating and obstructive 11.5 higher shunt complication rate than patients with communicat-
ing hydrocephalus (HR, 1.9; 95% CI, 1.7–2.1). The shunt failure
Congenital, not spina bifida 11.5
rates did not differ significantly between patients with spina
Spina bifida 5.7
bifida and those with other types of congenital hydrocephalus
None of the above 8.5 (P ⫽ 0.45). Male sex (HR, 1.14; 95% CI, 1.1–1.2) and lower SES
(HR, 1.5; 95% CI, 1.4–1.6) were also associated with a higher
rate of shunt complications.
shunt placement to in-hospital death was 8 months (inter- Caucasians and Asians (Fig. 1C) exhibited similar survival
quartile range, 2–30 mo). curves that were not statistically different (P ⫽ 0.38), and sur-
Among patients with a VP shunt who could be followed vival curves for African-Americans and Hispanics also were
over time, 35% experienced at least one surgical shunt compli- not statistically different (P ⫽ 0.71). In univariate analysis,
cation during the study period, and 11% had multiple (two or African-Americans and Hispanics together experienced an
more) shunt complications. Patients with multiple complica- increased risk of VP shunt complications when compared with
tions had a mean of three complications each (standard devia- Caucasians and Asians (HR, 1.3; 95% CI, 1.2–1.4). The 54 Native
tion, 1.8), and 16 patients underwent 10 to 23 hospitalizations Americans identified in our cohort had the highest complica-
for shunt complications during the follow-up period. tion rate compared with Caucasians (HR, 1.7; 95% CI, 1.2–2.6).
Shunt replacement or removal occurred in 95% of patients
with a shunt complication, whereas distal shunt complica- Multivariate Analysis
tions and shunt explorations were less common (Table 2). All
We included the following variables in a Cox proportional
forms of shunt complications were significantly more frequent
hazards model: sex, ethnicity, age group, type of hydro-
in children.
cephalus, and SES. The type of hydrocephalus was the
strongest independent risk factor for VP shunt complications
Time to First Shunt Complication (Table 3). Compared with patients with communicating hydro-
The shunt survival analysis included 11,550 adults and 2905 cephalus only, those with congenital hydrocephalus (HR, 3.1;
children who received a VP shunt, with 51,150 and 13,889 per- 95% CI, 2.6–3.6) or obstructive hydrocephalus (HR, 1.7; 95% CI,
son-years of follow-up, respectively. Children consisted of 1.5–1.9) were at increased risk of shunt complications, as were
patients aged 1 to 18 years (56%), infants aged 1 to 12 months patients with both communicating and obstructive hydro-
(32%), and neonates aged less than 1 month (12%). cephalus (HR, 2.4; 95% CI, 2.1–2.7). Male sex (HR, 1.1; 95% CI,
Overall, the cumulative VP shunt complication rate was 15% 1.03–1.2) and low SES (HR, 1.2; 95% CI, 1.1–1.3) were also asso-
in the first month, 25% at 1 year, 32% at 5 years, and 34% at 10 ciated with increased risk of shunt complications, whereas eth-
years. Children had higher rates of shunt complications than nicity no longer demonstrated a significant effect on outcome
A B
after adjustment for other factors such as SES. Compared with tions than Caucasians and Asians in univariate analyses, eth-
adults, children of all age groups also demonstrated an nic differences were no longer apparent after adjusting for SES
increased risk of shunt complications, after adjustment for in the multivariate analysis. Male sex and type of hydro-
potential confounders (neonate: HR, 1.4; 95% CI, 1.1–1.7; infant: cephalus also independently increased the risk of surgical
HR, 1.3; 95% CI, 1.2–1.5; and child: HR, 1.7; 95% CI, 1.5–1.8). shunt complications.
Of the patients who had a first shunt complication, 31% went This study is subject to a number of important limitations.
on to experience at least one more complication during the Only patients with a social security number were able to be fol-
study period. A shorter time to first shunt complication (⬍6 lowed over time, leading to potential selection bias and an
mo) was associated with a greater probability of having a sub- underrepresentation of infants and Hispanics. We lacked data
sequent shunt complication in children (odds ratio, 1.3; 95% CI, regarding mortality outside of California hospitals during the
1.02–1.7) but not in adults (odds ratio, 0.85; 95% CI, 0.71–1.03). study period, as well as information regarding which patients
might have left the state, again leading to potential bias if
DISCUSSION patients remaining in California differed systematically from
those who were unknowingly lost to follow-up. Furthermore,
In this first longitudinal study of VP shunts within a United our shunt complication and mortality rates represent lower-
States population since the 1970s, we found that the overall bound estimates of the true population rates, given our lack of
rate of surgical shunt complications was 35% during a mean knowledge of how many patients left the state. We relied on
follow-up period of 4.5 years. Children, and especially neo- hospital discharge codes that may not have been recorded reli-
nates, experienced the highest complication rates. Almost half ably. For instance, the accuracy with which communicating
of the children required a repeat surgical procedure by 5 years and obstructive hydrocephalus was coded is unknown. These
of follow-up. Although African-Americans and Hispanics study limitations are offset by a number of advantages, includ-
demonstrated a significantly higher rate of VP shunt complica- ing the large, ethnically diverse population base, a relatively
3. Casey AT, Kimmings EJ, Kleinlugtebeld AD, Taylor WA, Harkness WF, inroads into the epidemiological problems associated with shunts
Hayward RD: The long-term outlook for hydrocephalus in childhood. A ten- over the past 25 years.
year cohort study of 155 patients. Pediatr Neurosurg 27:63–70, 1997.
4. CDC Wonder. http://wonder.cdc.gov/Census.html. Accessed July 20, 2006. Paul H. Chapman
5. Chumas P, Tyagi A, Livingston J: Hydrocephalus—What’s new? Arch Dis Boston, Massachusetts
Child Fetal Neonatal Ed 85:F149–F154, 2001.
6. Davis SE, Levy ML, McComb JG, Masri-Lavine L: Does age or other factors
influence the incidence of ventriculoperitoneal shunt infections? Pediatr
Neurosurg 30:253–257, 1999.
T his is a population-based survey performed over a 10-year period
in California examining the outcomes of 14,455 shunt placements.
Via uni- and multivariate analysis, the authors found that younger
7. Di Rocco C, Marchese E, Velardi F: A survey of the first complication of newly
age, male sex, low socioeconomic status, and obstructive hydro-
implanted CSF shunt devices for the treatment of nontumoral hydrocephalus.
Cooperative survey of the 1991–1992 Education Committee of the ISPN. cephalus were independent risk factors for shunt complications. As
Childs Nerv Syst 10:321–327, 1994. the authors discuss, there are a number of inaccuracies related to
8. Hardie NA, Molgaard CA, Laws ER, O’Fallon WM, Kurland LT: Incidence inaccurate coding, missing patients (patients not entered because of
and effectiveness of cerebrospinal fluid shunts in Olmsted County, lack of social security number), patients lost to follow-up or moving
Minnesota, 1956–1981. Neuroepidemiology 5:95–104, 1986. out of state, patients dying, etc. in large data sets such as this. These
9. Kulkarni AV, Drake JM, Lamberti-Pasculli M: Cerebrospinal fluid shunt infec- inaccuracies are somewhat compensated for by large numbers; thus,
tion: A prospective study of risk factors. J Neurosurg 94:195–201, 2001.
the estimates for the patients identified are reasonably accurate and
10. Kulkarni AV, Rabin D, Lamberti-Pasculli M, Drake JM: Repeat cerebrospinal
are likely lower estimates of complications. Suffice it to say that
fluid shunt infection in children. Pediatr Neurosurg 35:66–71, 2001.
11. Lazareff JA, Peacock W, Holly L, Ver Halen J, Wong A, Olmstead C: shunt complications remain 32% overall and 48% in children and
Multiple shunt failures: An analysis of relevant factors. Childs Nerv Syst have likely not changed much over the previous several decades.
14:271–275, 1998. Studies such as this remind us that the success of treatment of hydro-
12. Liptak GS, McDonald JV: Ventriculoperitoneal shunts in children: Factors cephalus with shunts is not satisfactory and is worthy of every effort
affecting shunt survival. Pediatr Neurosci 12:289–293, 1985. to improve things.
13. McGirt MJ, Leveque JC, Wellons JC 3rd, Villavicencio AT, Hopkins JS, Fuchs
HE, George TM: Cerebrospinal fluid shunt survival and etiology of failures: James M. Drake
A seven-year institutional experience. Pediatr Neurosurg 36:248–255, 2002. Toronto, Canada
14. McGirt MJ, Zaas A, Fuchs HE, George TM, Kaye K, Sexton DJ: Risk factors for
T
pediatric ventriculoperitoneal shunt infection and predictors of infectious his is a review of adult and pediatric shunts performed in California
pathogens. Clin Infect Dis 36:858–862, 2003. over a 10-year period, with data on reoperation for shunt complica-
15. Metzemaekers JD, Beks JW, van Popta JS: Cerebrospinal fluid shunting for
tions obtained from a state registry. The results are not surprising and
hydrocephalus: A retrospective analysis. Acta Neurochir (Wien) 88:75–78, 1987.
16. Patwardhan RV, Nanda A: Implanted ventricular shunts in the United States: basically confirm earlier and smaller single institution studies. Children
The billion-dollar-a-year cost of hydrocephalus treatment. Neurosurgery had a higher rate of shunt problems than adults, and lower socioeco-
56:139–145, 2005. nomic status was an independent risk factor for shunt complications.
17. Piatt JH Jr, Carlson CV: A search for determinants of cerebrospinal fluid shunt The study probably underestimates the true incidence of problems
survival: Retrospective analysis of a 14-year institutional experience. Pediatr because patients who died or moved out of state were lost to follow-up
Neurosurg 19:233–242, 1993. and infants were underreported because they often lacked social secu-
18. Smith ER, Butler WE, Barker FG 2nd: In-hospital mortality rates after ven-
rity numbers. The analysis of obstructive versus communicating hydro-
triculoperitoneal shunt procedures in the United States, 1998 to 2000: Relation
to hospital and surgeon volume of care. J Neurosurg Spine 100:90–97, 2004.
cephalus is probably erroneous because there was no way to really
19. Tuli S, Drake J, Lawless J, Wigg M, Lamberti-Pasculli M: Risk factors for distinguish these from each other in a study of this type.
repeated cerebrospinal shunt failures in pediatric patients with hydro- Despite the hope that endoscopic third ventriculostomy would elim-
cephalus. J Neurosurg 92:31–38, 2000. inate the need for shunts and shunt complications, shunts with all of
20. Tuli S, Tuli J, Drake J, Spears J: Predictors of death in pediatric patients requir- their attendant problems remain. The fact that patients of lower socioe-
ing cerebrospinal fluid shunts. J Neurosurg 100:442–446, 2004. conomic status had more problems is unexplained. I suspect this is
21. Vinchon M, Fichten A, Delestret I, Dhellemmes P: Shunt revision for asymp- because this group of patients had more complex forms of hydro-
tomatic failure: Surgical and clinical results. Neurosurgery 52:347–356, 2003.
cephalus, such as postinfectious, posttraumatic, and posthemorrhagic
22. Wiswell TE, Tuttle DJ, Northam RS, Simonds GR: Major congenital neurologic
malformations. A 17-year survey. Am J Dis Child 144:61–67, 1990. types, which tend to lead to multicompartmental ventricles, infections,
and other risk factors.
Acknowledgments Leslie N. Sutton
We thank Heather J. Fullerton for her assistance with the data analysis and for Philadelphia, Pennsylvania
critically reviewing this article. Yvonne Wu, M.D., Ph.D., is supported by
T
NS35902. Nalin Gupta, M.D., Ph.D., is supported by K08 NS055061 to 01. his study examines a number of factors related to the incidence of
Margaret R. Wrensch, Ph.D., is supported by R01CA52689 and P50CA097257. ventriculoperitoneal shunt complications in the state of California
from 1990 to 2000, thereby providing a large cohort to analyze. As the
COMMENTS authors note, there were a number of limitations to the data available,
including the fact that 25% of the patients had no social security num-
obstructive hydrocephalus leaves much room for question. How is it American? To be considered as such, does one have to be 100% Native
possible to accurately separate communicating and obstructive hydro- American or will 51% suffice? Presumably, the selection was made by
cephalus from hydrocephalus that is only communicating versus hydro- the individual checking off a box on a questionaire. Were genealogy
cephalus that is only obstructive? This is pure fiction. As a side note, all studies performed for confirmation? It would take but few complica-
hydrocephalus are obstructive with the exception of that related to cere- tions to skew the outcomes. To draw conclusions based on 54 Native
brospinal fluid overproduction associated with a choroid plexus papil- Americans seems suspect.
loma. A neonate born with hydrocephalus will be correctly coded as This study does put into broad overview the complication rates asso-
having congenital hydrocephalus. On subsequent admissions, this dis- ciated with ventriculoperitoneal shunts. As long as hydrocephalus is
tinction is often lost. Even in our office records, I note discrepencies for treated with mechanical devices, there will be complications. To say
older children as to whether the hydrocephalus is properly coded as that the current treatment of hydrocephalus is unsatisfactory for a large
congenital versus acquired. proportion of patients is a bit harsh. One has only to remember the
The incidence of hydrocephalus associated with an open neural tube prognosis for patients with hydrocephalus before shunts became avail-
defect (spina bifida/myelomeningocele) has dropped markedly. A able. The complication rate has slowly diminished, but do we need to
study covering the time interval from 2000 to 2010 will have but a frac- do better? Yes, much better.
tion of the 5.7% incidence noted in the present study.
Fifty-four of the 14,455 patients in the study group were classified as J. Gordon McComb
Native American. What are the criteria to be classified as Native Los Angeles, California
19th century English whaling gear, illustrated by William Scoresby, Jr. From: Matthews, LH: The Whale. New York, Simon & Schuster,
1968.