Primary total hip arthroplasty versus hemiarthroplasty for displaced intracapsular hip fractures in older patients: systematic review
BMJ 2010; 340 doi: https://doi.org/10.1136/bmj.c2332 (Published 11 June 2010) Cite this as: BMJ 2010;340:c2332
All rapid responses
Dear Sir
We read with interest the letter submitted by Dr Carroll and
colleagues and would like to take the opportunity to respond to their
comments. The presentation of their arguments as outright errors of data
extraction or methodology on our part is incorrect. Our approach to the
review was to include data when appropriate and to take a conservative
approach about the classification of events, particularly towards THA, as
this is a change from the standard of care. We were diligent in
extracting information from the papers as described in the review, and the
use of data reflects the conscious decisions we made and are consistent
with the principles we adopted at the outset of the study. We do
recognise that different interpretations of the data may also be valid,
which introduces significant complexity, but should highlight that there
are some suggested figures and interpretations proposed by Dr Caroll that
we do not agree with.
Firstly, we acknowledge our error in the number of patients in the
arms of the trial reported by Ravikumar et al. The number of patients
with HA and THA are indeed 91 and 89 respectively, not 89 and 91 as
described in our paper.(1,2) We regret that such an error has been
incorporated into the analysis and are fortunate that it does not
substantially influence the meaning of the review. We have reviewed our
data extraction tables and found that the error was introduced when the
data was transposed into the analysis package.
In the study by Mouzopoulos, a number of patients were defined as
exclusions from the study for the following reasons; they had a previous
contralateral hip fracture; they died prior to one year following surgery;
they were revised prior to one year following surgery; or they were
lost.(3) In our analysis we performed a worst case analysis, classifying
these patients as dead at one year. The two patients with HA who were
revised prior to 1 year were not counted in the meta-analysis of re-
operations, which is conservative towards THA. In our view this is a
legitimate method of treating these patients and is consistent with our
general conservative approach. We accept that there may be other ways to
classify the mortality data presented by Mozoupoulos, but do not agree
with the numbers suggested by Carroll et al, 6 dead in each group, as they
ignore the two patients with lost data. Notwithstanding this, Dr Caroll's
group would also need to make a decision on how to treat the patients
excluded from the study with previous contralateral hip fracture and we
find that there is more than one way to do this.
Our classification of re-operations is clearly defined as any event
where the author has indicated that a re-operation has taken place,
including revisions, re-operations for dislocation and other non revision
re-operations. Understanding the absolute risk of further exposure to
surgery is clinically relevant in this elderly population and this is a
pragmatic classification that avoids ambiguity about the inclusion of data
and subjective decisions on classifying revisions verses re-operation. We
applied this definition consistently to all studies, and found that all re
-operations appeared to be listed in each paper and that in most the only
re-operations were revisions. Furthermore, doing this is consistent with
our approach as it includes all data and is conservative towards THA, as
the incidence of re-operation reported following dislocation was higher
with THA than with HA. In line with this methodology we included all re-
operations reported by Blomfeldt and we have definitely not mistakenly
analysed mortality data.(4) The figures we use include re-operations for
trauma of the lower limb, of which there were three in each group. The
figure suggested by Dr Carroll appears only to include one of these, a
peri prosthetic fracture due to a fall in the THA group. The remaining
five appear to have been excluded on the assumption they were on the
contra-lateral side and not related to the implant. To be consistent
they should also have excluded the peri-prosthetic fracture as this was
due to a fall, not a failure of the implant, and as such should also be
classified as not implant related. We ourselves did not assume that re-
operations due to trauma are unrelated to the type of arthroplasty. There
is an association between having a hip fracture and a higher risk of
further trauma, so these re-operations may not be completely unrelated to
the primary arthroplasty for fracture.(5) Also, we are unsure whether the
risk of further trauma could be related to the choice of arthroplasty.
Patients with THA may have a higher risk of dislocation, suggesting
greater joint instability and a greater propensity to fall. Conversely
they appear to have better function which may lead to a lower risk of
having a fall.(6) To detect a difference in the risk of re-operation due
to further trauma would probably require a very large sample size and
although we did not take this approach ourselves there is an argument for
excluding these re-operations. To do so correctly we suggest that the
number of re-operations should be 1/60 verses 0/60, not 2/60 verses 0/60
as suggested. If all the re-operations due to post-operative trauma are
included, as we have done, the numbers are 4/60 verses 3/60, the same
absolute difference in number of re-operations.
In our sensitivity analyses we stratified the data according to the
average length of follow up of the study not according to whether the
study had discreet data for less than 2 years. Consequently the 2 studies
with 483 patients reporting dislocation at less than 2 years are by
Blomfeldt and Eyssel.(4,7) We do not agree with classifying the
dislocation data reported by Macauly as discreet data for less than 2
years.(7) The manuscript indicates that there was only "one dislocation
in the THA group occurring 5 months after surgery", and reports no
dislocation with HA. In our view, this clearly means that in the whole
study there was only one dislocation in either group out to final follow
up. Additionally, our interpretation of the article by Baker is that
throughout the study there were no dislocations in the HA group, and that
there were 3 in the THA group, all of which occurred in the immediate post
-operative period.(8) Consequently we believe inclusion of these two
studies, (Baker and Macauly) in the strata with 2 years or longer follow-
up is appropriate.(8,9)
Our inclusion of data from the study by Keating is in line with our
general approach as we have included all the data we believed to be
appropriate and have also taken a conservative approach with respect to
THA.(10) There is an argument that only the data on the 69 HA patients in
the head to head comparison should be used. However, we understand that
this is a subset of all 111 patients in the study, not a separate trial
arm. In the remaining 42 HA patients at the centres outside the head to
head comparison the incidence of re-operation and dislocation, (1/42), was
lower than in the 69 HA patients within it, (5/69). We chose to include
all 111 HA patients as it reported a lower incidence of re-operation with
HA and was hard to ignore considering that THA is not the current standard
of care. This may have reduced the internal validity of the data, but
may have increased the external validity through the inclusion of a
greater number of centres. The study by Dorr reported no difference in
mortality and that out to one year there were seven deaths across both
groups.(11) On this basis we concluded that the most likely distribution
of death out to 1 year was 3/39 (7.6%) and 4/50 (8.0%).
We do agree that correct data extraction is a challenge in systematic
reviewing. Despite our best efforts and good practice we adopted, we
concede that we have made a mistake inadvertently swapping the sample
sizes for Ravikumar, and again express our disappointment and regret that
this has occurred. However, we contest the implication of Dr Carroll and
colleagues that we did not conduct the review in the manner we described
and that there are a series of errors in data extraction and
interpretation, which we find disappointing. The purpose of our response
is to emphasise that data extraction has a subjective element and that
decisions on how to treat data can lead to differences in the data
extracted and analyses performed. This is particularly relevant to
reviews of papers such as these, which were not written with meta-analysis
in mind, and in which the information was not presented in a consistent
way. This is highlighted by the differing views being discussed about our
own figures, and by the ongoing disagreement we have with some of the
alternative suggestions being made. We would also like to draw attention
to the fact, that with this study, the differences in interpretation and
analysis do not lead to any substantial difference in the meaning of the
analyses. As Dr Carroll has indicated with dislocations and mortality, if
we apply all the changes suggested to the analysis on re-operations there
is only a small difference in favour of THA in the risk of re-operation
across all the studies, RR 0.52 (95% CI 0.29 - 0.91) with the changes
compared with RR 0.57 ( 95% CI 0.34 to 0.96) reported originally.
Finally, we accept that there are limitations with the review, as the
data set used is imperfect, but believe we have mitigated in our subgroup
analyses and discussions, and are more than happy to discuss any comments
respectfully in the spirit of peer review and scientific discussion. We
believe we have adequately responded to each observation and wish to
assure the readers and editorial board of the BMJ that the methodology we
have adopted is entirely justifiable and has been implemented fully and
properly throughout this study.
Yours faithfully
Colin Hopley, Dirk Stengel, Michael Wich, Axel Ekkernkamp
References
1. Ravikumar KJ, Marsh G. Internal fixation versus hemiarthroplasty
versus total hip arthroplasty for displaced subcapital fractures of femur-
-13 year results of a prospective randomised study. Injury 2000;31:793-7.
2. Hopley C, Stengel D, Ekkernkamp A, Wich M. Primary total hip
arthroplasty versus hemiarthroplasty for displaced intracapsular hip
fractures in older patients: systematic review. BMJ 2010;340:c2332
3. Mouzopoulos G, Stamatakos M, Arabatzi H, Vasiliadis G, Batanis G,
Tsembeli A et al. The four-year functional result after a displaced
subcapital hip fracture treated with three different surgical options. Int
Orthop 2008;32:367-73.
4. Blomfeldt R, Tornkvist H, Eriksson K, Soderqvist A, Ponzer S, Tidermark
J. A randomised controlled trial comparing bipolar hemiarthroplasty with
total hip replacement for displaced intracapsular fractures of the femoral
neck in elderly patients. J Bone Joint Surg Br 2007;89:160-5.
5. Kanis J, Johnell O, Oden A, Jonsson B, Dawson A, W Dere. Risk of Hip
Fracture Derived from Relative Risks: An Analysis Applied to the
Population of Sweden. Osetoporos Int (2000)11:120-127
6. Dargent-Molina P, Favier F, Grandjean H, Baudin C, Schott A, Hausherr
E, Meunier P, Breart G, The Lancet 1996;348:145-149
7. Eyssel M, Schwenk W, Badke A, Krebs S, Stock W. [Total endoprosthesis
or dual head prosthesis in endoprosthetic management of femoral neck
fractures?]. Unfallchirurg 1994;97:347-52
8. Macaulay W, Nellans KW, Garvin KL, Iorio R, Healy WL, Rosenwasser MP.
Prospective randomized clinical trial comparing hemiarthroplasty to total
hip arthroplasty in the treatment of displaced femoral neck fractures:
winner of the Dorr Award. J Arthroplasty 2008;23:2-8.
9. Baker RP, Squires B, Gargan MF, Bannister GC. Total hip arthroplasty
and hemiarthroplasty in mobile, independent patients with a displaced
intracapsular fracture of the femoral neck. A randomized, controlled
trial. J Bone Joint Surg Am 2006;88:2583-9.
10. Keating JF, Grant A, Masson M, Scott NW, Forbes JF. Randomized
comparison of reduction and fixation, bipolar hemiarthroplasty, and total
hip arthroplasty. Treatment of displaced intracapsular hip fractures in
healthy older patients. J Bone Joint Surg Am 2006;88:249-60
11. Dorr LD, Glousman R, Hoy AL, Vanis R, Chandler R. Treatment of femoral
neck fractures with total hip replacement versus cemented and noncemented
hemiarthroplasty. J Arthroplasty 1986;1:21-8
Competing interests: Colin Hopley: Employee of DePuy International Ltd, a manufacturer of total hip and hemi-arthroplasty prostheses that may have a financial interest in the results of this review.
Dear Editors,
We have read the review by Hopley and colleagues1 on total hip
arthroplasty (THA) versus hemiarthroplasty (HA) with great interest as we
are currently engaged in completing a similar review on the same topic for
the UK NIHR Evaluation, Trials and Studies co-ordinating Centre (NETSCC).2
We have some methodological concerns with the conduct of the review that
we feel are worthy of comment. We draw attention to these issues to
highlight broader generic concerns around the conduct of peer-reviewed
systematic reviews.
The review states that "Two reviewers (CH and DS) independently
abstracted data in duplicate, including ... The reviewers also extracted
and entered into an electronic database event rates with nominators and
denominators for different end points". And later, "The researchers were
diligent about extracting ... information ... from the available papers".
However, we feel that there are a number of errors in the extracted data
being analysed, as well as some issues of interpretation over other data
used. We highlight these for the outcomes of dislocation, "reoperations"
and 1-year mortality, as we have conducted the same analyses for our
review.
In terms of errors, the number of participants analysed by Hopley and
colleagues in the THA arm for the trial reported by Ravikumar et al3 are
actually the number of participants in the HA arm, and vice versa. This
error occurs for all of the three outcomes of interest. For the 1-year
mortality analysis, the number of events, i.e. deaths, attributed to each
arm for the trial by Mouzopoulos et al 4(THA vs HA: 10 vs 13) actually
include revisions, exclusions due to previous fracture, and individuals
for whom the data were lost, as well as deaths. The correct event rates
are six deaths in each arm. For the "reoperations" outcome, Hopley and
colleagues analyse the one-year mortality data for Blomfeldt et al5 rather
than the actual "reoperations" data (THA vs HA): 4/60 vs 3/60 rather than
2/60 vs 0/60. The systematic review method is explicitly designed to guard
against errors such as these, as long as the method is being followed. The
"reoperations" outcome is not clearly defined and, in terms of numbers of
events analysed, appears to consist of revisions only, in most cases, and
all reoperations involving a second anaesthetic (eg. including closed
reductions and other non-revision reoperations) in other cases, eg. for
the trials of Dorr et al6 and Keating et al7. We propose that a more
robust analysis would have combined, firstly, only data relating to
revisions, where those data were available, and secondly, data on any
reoperation, also where available, taking care to avoid double-counting.
Some elements of the sensitivity analyses are also neither transparent nor
accurate. For example, in Table 6 on dislocations, Hopley and colleagues
record only two studies, which are unnamed, with 483 participants as
having discrete data for less than two years, but Blomfeldt et al,
Macaulay et al8and Baker et al9 (all included in the review) only report
dislocation data for less than two years. It is therefore unclear how
these studies and their data have been categorised.
Other data used may be more open to question or interpretation. One
example of this is the denominator for the HA group from the Keating et al
trial7: Hopley and colleagues use data from a separate trial arm (with 111
participants) in this study, but these data arguably should not be
included because different eligibility criteria were being applied for
this arm (i.e. the surgeons and centres involved were either unwilling or
unable to have participants randomised to THA). The head-to-head trial
data (n=69) should have been used, as in the Cochrane review by Parker et
al10 and another relevant, recent review by Liang and colleagues.11 The
difference is not small; for example, for the "reoperations" outcome (THA
vs HA), it is 6/69 vs 6/111 rather than 6/69 vs 5/69. Furthermore, the one
-year mortality data analysed for the trial reported by Dorr et al are not
published; if these are unpublished data then the authors should report
this.
Using data only from the RCTs common to the reviews of Hopley and
colleagues and ourselves (i.e. not including a recently published trial by
van den Bekerom et al12), meta-analysis of the different data reported by
Hopley and colleagues and ourselves produces little difference in relative
risk (RR) of patients experiencing dislocations: RR 1.88 (95% CI 1.08,
3.26), p=0.03 (for Hopley and colleagues' data) compared to 1.90 (95% CI
1.09, 3.32), p=0.02. The same applies for the outcome of mortality within
one year: RR 0.92 (0.70, 1.21), p=0.85 (Hopley and colleagues) compared
to: RR 0.91 (0.65, 1.29), p=0.60. Data extraction errors in systematic
reviews are not uncommon.13;14 Double-data extraction should offer a
possibly more reliable method.15 It should certainly prevent outright
errors. The full and proper implementation of the methodology as it has
been developed and tested surely offers the optimal means of achieving
accuracy and the highest standards. Indeed, it has been pointed-out
previously how far "Correct data extraction is fundamental to the
integrity of meta-analysis".14 Therefore, although the aforementioned
errors do not induce a clinically or statistically different conclusion
here, this may not always be the case.
Finally, we do not find convincing Hopley and colleagues arguments
for the inclusion of retrospective cohort studies. The case for including
data from non-randomised studies is not strong here, as these studies are
inherently more biased and their outcomes more questionable.16 Their
inclusion must be based, firstly, on the prior clinical belief that the
relative risk does not differ between the two types of study (RCT and
retrospective cohort) and, secondly, on the performance of an appropriate
statistical test to indicate that the hypothesis cannot be rejected.
Neither is reported as being the case here.
Once again, we wish to stress that we draw attention to these issues
only in order to highlight broader concerns surrounding the conduct of
peer-reviewed systematic reviews. These concern extraction errors and the
lack of transparency surrounding reported analyses, and the limitations of
the peer-review process: peer-reviewers must necessarily place a sizeable
degree of trust in the reported methods authors claim they have used.
Dr Christopher Carroll, Senior Lecturer in Health Technology
Assessment
Dr Matt Stevenson, Reader in Health Technology Assessment
Dr Eva Kaltenthaler, Senior Research Fellow
Dr Alison Scope, Research Associate
Reference List
(1) Hopley C, Stengel D, Ekkernkamp A, Wich M. Primary total hip
arthroplasty versus hemiarthroplasty for displaced intracapsular hip
fractures in older patients: systematic review. British Medical Journal
2010; 340.
(2) Carroll C, Stevenson M, Scope A, Evans P, Buckley S.
Hemiarthroplasty and total-hip arthroplasty for treating primary
intracapsular fracture of the hip: A systematic review and cost
effectiveness analyses. Health Technology Assessment 2011;
15(Forthcoming).
(3) Ravikumar KJ, Marsh G. Internal fixation versus hemiarthroplasty
versus total hip arthroplasty for displaced subcapital fractures of femur-
-13 year results of a prospective randomised study. Injury 2000; 31:793-
797.
(4) Mouzopoulos G, Stamatakos M, Arabatzi H, Vasiliadis G, Batanis
G, Tsembeli A et al. The four-year functional result after a displaced
subcapital hip fracture treated with three different surgical options.
International Orthopaedics 2008; 32(3):367-373.
(5) Blomfeldt R, T?rnkvist H, Eriksson K, S?derqvist A, Ponzer S,
Tidermark J. A randomised controlled trial comparing bipolar
hemiarthroplasty with total hip replacement for displaced intracapsular
fractures of the femoral neck in elderly patients. The Journal of Bone and
Joint Surgery British volume 2007; 89:160-165.
(6) Dorr LD, Glousman R, Hoy AL, Vanis R, Chandler R. Treatment of
femoral neck fractures with total hip replacement versus cemented and
noncemented hemiarthroplasty. The Journal of Arthroplasty 1986; 1:21-28.
(7) Keating JF, Grant A, Masson M, Scott NW, Forbes JF. Randomized
comparison of reduction and fixation, bipolar hemiarthroplasty, and total
hip arthroplasty. Treatment of displaced intracapsular hip fractures in
healthy older patients. The Journal of Bone and Joint Surgery American
volume 2006; 88:249-260.
(8) Macaulay W, Nellans KW, Garvin KL, Iorio R, Healy WL,
Rosenwasser MP et al. Prospective randomized clinical trial comparing
hemiarthroplasty to total hip arthroplasty in the treatment of displaced
femoral neck fractures: winner of the Dorr Award. The Journal of
Arthroplasty 2008; 23:2-8.
(9) Baker RP, Squires B, Gargan MF, Bannister GC. Total hip
arthroplasty and hemiarthroplasty in mobile, independent patients with a
displaced intracapsular fracture of the femoral neck. A randomized,
controlled trial. The Journal of Bone and Joint Surgery American volume
2006; 88:2583-2589.
(10) Parker MJ, Gurusamy KS, Azegami S, Parker MJ, Gurusamy KS,
Azegami S. Arthroplasties (with and without bone cement) for proximal
femoral fractures in adults. [Update of Cochrane Database Syst Rev.
2006;3:CD001706; PMID: 16855974]. Cochrane Database of Systematic Reviews
2010; 6:CD001706.
(11) Liang LZ. Meta analysis of total hip arthroplasty versus
hemiarthroplasty for displaced femoral neck fractures in elderly patients.
Journal of Clinical Rehabilitative Tissue Engineering Research 2010;
14(22):3991-3995.
(12) van den Bekerom P, Hilverdink E, Sierevelt I, Reuling E,
Schnater J, Bonke H et al. A comparison of hemiarthroplasty with total hip
replacement for displaced intracapsular fracture of the femoral neck.
Journal of Bone and Joint Surgery British Volume 2010; 92:1422-1428.
(13) Horton J, Vandermeer B, Hartling L, Tjosvold L, Klassen T,
Buscemi N. Systematic review data extraction: cross-sectional study showed
that experience did not increase accuracy. Journal of Clinical
Epidemiology 2010; 63:289-298.
(14) Jones A, Remmington T, Williamson P, Ashley D, Smyth R. High
prevalence but low impact of data extraction and reporting errors were
found in cochrane systematic reviews. Journal of Clinical Epidemiology
2011; 58:741-742.
(15) Buscemi N, Hartling L, Vandermeer B, Tjosvold L, Klassen T.
Single data extraction generated more errors than double data extraction
in systematic reviews. Journal of Clinical Epidemiology 2006; 59:697-703.
(16) Bhandari M, Tornetta P, Ellis T, Audige L, Sprague S, Kuo JC et
al. Hierarchy of evidence: differences in results between non-randomized
studies and randomized trials in patients with femoral neck fractures.
Archives of Orthopaedic and Trauma Surgery 2004; 124(1):10-16.
Competing interests: No competing interests
Sir
We thank Mr Hunter and Mr Westerman for their interest in our article
and their observations on revision rates and function, and we are more
than happy to discuss the points they have raised and hope to provide some
further insight into the reasoning for our conclusions.
We agree that the use of cementless hemi-arthroplasty (HA), such as
the Austin Moore, has been shown to have inferior outcomes compared to
cemented hemi-arthroplasty.1 This was the reason why we had explored this
in detail by sensitivity analysis and a test for interaction, presented in
table 5 of our article.2 We feel there is still sufficient evidence for a
qualified conclusion that there could be a difference in the re-operation
rate while indicating that there are several variables influencing the
results, including the use of cementless HA. Excluding studies with
cementless HA does not completely eliminate the treatment effect. We
employed random-effects models for all analyses for the benefit of
consistency. The relative risk (RR) of re-operation was 0.88 (95%
confidence interval 0.44 to 1.73) when comparing THA to all-cemented
hemiarthroplasties. This translates to a risk difference (RD) of 2.2% (-
2.5 to 6.9%). The I2 index of heterogeneity would have justified to
compute a fixed-effects model, with an RR of 0.78 (0.44 to 1.38) and an RD
of 1.4% (-1.6 to 4.5%). All these studies used modern cemented mono-block
or bi-polar prostheses and represent a comparison of contemporary
prostheses.3-8
Studies with cementless hemi-arthroplasty were included in the review
because there could still be a significant incidence of failure related to
acetabular wear as well as failure related to the stem. Although it is not
possible to determine the incidence of revision for stem loosening, the
maximum number is 24 of the 54 revisions recorded with cementless HA, less
than half. Only two could be definitively defined as revisions for
loosening4,5, five were in uncemented hemi-arthroplasties for undefined
pain6, either associated with the fixation or the acetabular, and 16 were
for either erosion or loosening.3
It seems unlikely that all the revisions that cannot be classified
are all for stem loosening, particularly in the study by Ravikumar. The
incidence of revision in the HA group in this study is much higher than
might be expected if the only reason for the difference was stem loosening
due to the use of the Austin Moore prosthesis. After 13 years of follow-
up, the incidence of revision in the THA group was 6.5%.3 The Australian
Joint Registry has reported 1.9 revisions per component year with Austin-
Moore devices compared to 1.1 revisions per component year with Thompson's
cemented hemi-arthroplasty, a crude RR of 1.8.1 If the revision rate of
THA was no different to that of a cemented HA then one could expect to see
a 12% to 15% incidence of revision at 13 years in the Austin-Moore-HA
group of the study by Ravikumar, not 25% as reported.3 Even taking the
most extreme case in which all the revisions that could not be classified
were assumed to be stem related and excluded from the analysis along with
those classified as stem related, there would still be a lower risk of
revision with THA in the studies by Gebhard, Dorr and Squires, but not by
Ravikumar. It may be convincing if one would exclude Ravikumar's study
from the aggregated analysis. This reduces I2 to only 18%, and still gives
a fixed-effects RR of 0.59 (0.38 to 0.90) in favour of THA.
Mr Hunter and Mr Westerman argue that a 5.4 point difference in the
Harris Hip Score (HHS) is less than the difference between an occasional
ache and mild pain not affecting activities. However, there is still
reason to believe that the difference is clinically important. Trying to
interpret the difference by analysis of the individual domains is
problematic as with hip specific outcomes measures such as the HHS and the
Oxford Hip Score (OHS) many of the domains in each instrument are
correlated and post-operative scores are not normally distributed, but
skewed.9,10 With hip replacement, a large proportion of patients have the
highest or near perfect scores while most of the remaining patients have
substantially lower scores.11 Therefore, the difference in HSS
identified is not a measure that can be applied to all patients but means
that fewer patients reported low scores with THA than with HA. This
difference was large enough to increase the mean score by a significant
amount using standard statistical approaches. Although this cannot be
properly quantified with the available information, it could mean that
some patients have substantially better function with THA than they would
have with HA.
The skewed nature of HHS data is one of the reasons why it has been
difficult to determine a minimum clinically important difference MCID. A
potential estimate for the MCID difference in OHS has been suggested to be
about half the standard deviation of the change.12 Estimating from the
confidence interval, the standard deviation of the difference in HHS in
the random effects model is 4.2 and applying this approach suggests that
the 5 point difference in means may be greater than the MCID. Using this
measure it is possible to estimate that the difference in EQ-5D at 24
months reported by Keating, is also clinically important.13 Also Baker
reports a rather small difference of 3 OHS points in favour of THA.14 This
is considered to be clinically important by the developer of the
instrument, and is apparent in the greater walker distance of the THA
patients in this study which was almost twice as much (2.1 miles) than
that of the HA patients (1.3 miles).
Regarding conflicts of interest, the purpose of declaring these, as
we have done, is to enable the reader to draw their own conclusion.
Notwithstanding this, we believe that we have used a robust scientific
approach to draw appropriate conclusions and have presented this in a
transparent way in our manuscript. Within the author group, with one
exception, all the authors are free from any conflicts, and have received
no industrial funding for this piece of academic work. This systematic
review was no remittance work. The idea for this study was developed by
the first author. Specifically, the contributing methodological and
statistical expert is affiliated to an independent academic unit, member
of both the Cochrane Injuries Group and the Cochrane Musculoskeletal
Group, and advisor to many independent scientific and professional bodies,
specifically the German Association of Trauma Surgeons. This has
effectively safeguarded the integrity of the work, which has been
rigorously tested in the peer review process and is in the long-term best
interest of all the authors.
The research question we have attempted to answer is an important
public health issue and is of legitimate interest to industry as a
significant stakeholder in the provision of orthopaedic and trauma
surgery. The article is not arguing for the wholesale migration of all
displaced intracapsular hip fracture patients from HA to THA, but that
there is a sub-population of healthier, mentally fit patients that may
benefit from THA. The Scottish Hip Fracture Audit has reported that 34%
of hip fracture patients are independently living prior to fracture.15,16
Many of these patients may be more suitable for THA, but The National Hip
Fracture Database has reported that less than 5% of operations for
displaced intracapsular hip fracture are treated with THA.
Mr Hunter and Mr Westerman state that "more expensive total hip
arthroplasties will only generate significant benefits for companies such
as DePuy". While it is acknowledged that many designs of THA will have a
higher cost, the direct cost of some cemented polyethylene cups may be
lower than some designs of bipolar hemi-arthroplasty. Also, in the UK,
the STARS trial found that the hip related healthcare costs of previously
independently living hip fracture patients treated with THA was lower than
that for patients treated with HA out to 24 months and that the cost of
the acute phase of treatment was similar.12 In the German Diagnosis-
Related Group (DRG) environment, managing femoral neck fractures with a
hemi-arthroplasty still generates a similar or even higher reimbursement
than primary total hip replacement.
Whilst conducting the review, we have systematically tested for
confounders, considered the clinical importance of the findings, and have
moderated our conclusions to reflect this. With the available data it is
not possible to draw definitive conclusions about any potential benefits
of total hip replacement, but likewise we believe that these studies do
provide some evidence of a treatment effect that should not be completely
discounted. We feel the data from our investigation strongly support an
equipoise assumption. This gives an important argument not to declare any
treatment option the "standard of care" until the results of the ongoing
HEALTH trial are available, which we hope will answer the unresolved
questions.
Colin Hopley, Dirk Stengel, Axel Ekkernkamp, Michael Wich
1. Australian Orthopaedic Association National Joint Replacement
Registry. Annual Report. Adelaide: AOA; 2009 available at URL:
http://www.dmac.adelaide.edu.au/aoanjrr/ accessed 22/07/10
2. Hopley C, Stengel D, Ekkernkamp A, Wich M. Primary total hip
arthroplasty versus hemiarthroplasty for displaced intracapsular hip
fractures in older patients: systematic review. BMJ 2010;340:c2332.
3. Ravikumar KJ, Marsh G. Internal fixation versus hemiarthroplasty
versus total hip arthroplasty for displaced subcapital fractures of femur-
-13 year results of a prospective randomised study. Injury 2000;31:793-7.
4. Dorr LD, Glousman R, Hoy AL, Vanis R, Chandler R. Treatment of
femoral neck fractures with total hip replacement versus cemented and
noncemented hemiarthroplasty. J Arthroplasty 986;1:21-8.
5. Gebhard JS, Amstutz HC, Zinar DM, Dorey FJ. A comparison of total
hip arthroplasty and hemiarthroplasty for treatment of acute fracture of
the femoral neck. Clin Orthop Relat Res 1992;123-31.
6. Squires B, Bannister G. Displaced intracapsular neck of femur
fractures in mobile independent patients: total hip replacement or
hemiarthroplasty? Injury 1999;30:345-8.
7. Macaulay W, Nellans KW, Garvin KL, Iorio R, Healy WL, Rosenwasser
MP. Prospective randomized clinical trial comparing hemiarthroplasty to
total hip arthroplasty in the treatment of displaced femoral neck
fractures: winner of the Dorr Award. J Arthroplasty 2008;23:2-8.
8. Narayan KK, George T. Functional outcome of fracture neck of femur
treated with total hip replacement versus bipolar arthroplasty in a South
Asian population. Arch Orthop Trauma Surg 2006;126:545-8.
9. Marx RG, Jones EC, Atwan NC, Closkey R, Salvati E, Sculco T.
Measuring improvement following total hip and knee arthroplasty using
patient-based measures of outcome. J Bone Joint Surg Am 2005;87-A:1999-
2005.
10. Murray D. Outcome Measures in Orthopaedics, The Hip. 1993
Butterworth Heinman.
11. Murray D, Fitzpatrick R, Rogers K, Panidt H, Beard D, Carr A,
Dawson J. The use of the Oxford hip and knee scores. J Bone Joint Surg Br
2007;89B:1010-4.
12. Keating JF, Grant A, Masson M, Scott NW, Forbes JF. Randomized
comparison of reduction and fixation, bipolar hemiarthroplasty, and total
hip arthroplasty. Treatment of displaced intracapsular hip fractures in
healthy older patients. J Bone Joint Surg Am 2006;88:249-60.
13. Baker RP, Squires B, Gargan MF, Bannister GC. Total hip
arthroplasty and hemiarthroplasty in mobile, independent patients with a
displaced intracapsular fracture of the femoral neck. A randomized,
controlled trial. J Bone Joint Surg Am 2006;88:2583-9.
14. Scottish Hip Fracture Audit. Scottish Hip Fracture Report 2006:
Available from URL: http://www.shfa.scot.nhs.uk/AnnualReport/Main.htm
accessed 22/07/10
Competing interests: Colin Hopley: Employee of DePuy International Ltd, a manufacturer of total hip and hemi-arthroplasty prostheses that may have a financial interest in the results of this review
We read this systematic review and the linked editorial with
interest, but are concerned about the inferences drawn. Hemiarthoplasty
prostheses have evolved from uncemented monoblocks, such as the Austin
Moore prosthesis, to modern cemented monoblocks and bipolar components.
The attempt to group hemiarthroplasty outcomes has given rise to
misleading conclusions.
We have reconsidered the outcomes from this review, by focusing on
the papers that specify the uncemented and cemented component fixation.
The uncemented outcomes have significantly skewed the overall results:
• Austin Moore hemiarthroplasty(1); 13% dislocations, 25% reoperation
and 3% deep infection. General complications are not recorded.
• Cemented hemiarthroplasty(2,3,4,5,6,7) : 2% dislocations, 7%
reoperation, 2% deep infection and 30% general complications.
• Total Hip replacement: 6% dislocations, 5% reoperation, 2% deep
infection and 33% general complications.
The review describes a statistically significant but clinically
unimportant difference in Harris Hip score of 5.4, which should be put in
context; the difference between an occasional ache and a mild pain not
affecting average activity alone carries a Harris score difference of 10.
Although the first author’s affiliations have been declared, it
should be borne in mind that any widespread change of practice towards the
use of more expensive total hip arthroplasties will generate significant
benefits for companies such as Depuy; indeed given our calculations they
will be the only beneficiary.
Cemented hemiarthroplasties have a similar outcome to total hip
arthroplasty for intracapsular neck of femur fractures, and offer
considerable health economic benefits.
The sub-heading of the editorial was misleading, and indeed much
stronger than Madsen’s conclusion. We would go further; total hip
arthroplasty for fractured neck of femur should be reserved for defined
indications such as co-existent rheumatoid arthritis. Hemiarthroplasty
should remain the standard procedure.
References:
1 Ravikumar KJ, Marsh G. Internal fixation versus hemiarthroplasty versus
total hip arthroplasty for displaced subcapital fractures of femur—13 year
results of a prospective randomised study. Injury 2000;31:793-7.
2 Keating JF, Grant A, Masson M, Scott NW, Forbes JF. Randomized
comparison of reduction and fixation, bipolar hemiarthroplasty, and total
hip arthroplasty. Treatment of displaced intracapsular hip fractures in
healthy older patients. J Bone Joint Surg Am 2006;88:249-60.
3 Baker RP, Squires B, GarganMF, Bannister GC. Total hip arthroplasty
and hemiarthroplasty in mobile, independent patients with a displaced
intracapsular fracture of the femoral neck: a randomized, controlled
trial. J Bone Joint Surg Am 2006;88:2583-9.
4 Blomfeldt R, Tornkvist H, Eriksson K, Soderqvist A, Ponzer S,
Tidermark J. A randomised controlled trial comparing bipolar
hemiarthroplasty with total hip replacement for displaced intracapsular
fractures of the femoral neck in elderly patients. J Bone Joint Surg Br
2007;89:160-5.
5 Eyssel M, Schwenk W, Badke A, Krebs S, Stock W. [Total
endoprosthesis or dual head prosthesis in endoprosthetic management of
femoral neck fractures?] Unfallchirurg 1994;97:347-52.
6 Healy WL, Iorio R. Total hip arthroplasty: optimal treatment for
displaced femoral neck fractures in elderly patients. ClinOrthop Relat Res
2004;43-8.
7 Xu X, Liu Y, Liu J, Li Y. Prosthetic replacement in treatment of
subcapital femoral neck fractures in the elderly. Chin J Traumatol
2002;5:28-31.
Competing interests:
None declared
Competing interests: No competing interests
Dear Sir,
We read with interest the above article and broadly agree with the
findings.
In terms of generalising the findings to the District general hospital
population in the UK the paper has looked mainly at mobile patients with a
normal MTS. Many would agree that this is not the average patient group in
a DGH setting here.
Whilst we agree that in younger patients in the above mentioned groups THA
would be the preferred treatment choice, does this hold true throughout
the broader patient population?
Secondly the 5 point improvement in the Harris hip score and the
medium improvement in the overall hip function evaluation scores are
difficult to correlate with clinical function and satisfaction. Is there
any way of assessing this better in the future?
Thanks to the authors for their hard work.
Competing interests:
None declared
Competing interests: No competing interests
I was a little preplexed by this statement just before the
conclusions in the review article(1):
"Under these circumstances, total hip replacement may still be the
dominant treatment strategy, as dislocation can be managed on an
outpatient basis and is a less serious event than reoperation."
Dislocation is usually a rather urgent in-patient problem and
recurrent dislocations may well result in the need for reoperation. As
the review makes clear it is becoming clear that in high demand younger
patients total hip replacement gives better functional outcomes, the
tricky decisions are when patients fall into the grey area between the
young fit patients and the extremely unfit elderly patients.
1. Colin Hopley et al. Primary total hip arthroplasty versus
hemiarthroplasty for displaced intracapsular hip fractures in older
patients: systematic review. BMJ 2010;340:c2332.
Competing interests:
None declared
Competing interests: No competing interests
As a Foundation Trainee in Trauma and Orthopaedics I have come to
realise that each of my 6 Consultants have preferences, dislikes and
methods that are worlds apart from the other. When admitting patients I
always try to predict the Consultants preference on how he or she will
manage the intra-capsular fracture to make sure I am spared my blushes
when theatre space and equipment is booked. My experience (although very
limited) supports the conclusion that outright support for one method over
the other is still a long way off.
In addition, I have quickly come to
realise the pressures that Consultants are under to reduce costs of
surgery. Orthopaedic devices are a gold mine and the kits, screws even
cement make my brain shiver at the numbers. If a unipolar or un-cemented
hemi is cheaper than a bipolar or cemented prosthesis that will be used
even if the latter may be better or more suitable for the patient. At
times, consultants have to battle and justify every procedure because the
costs seem too high, even if the clinical outcomes are better. In the
current economic climate I fear that the cheapest option will be deemed
the best option and not the most clinically effective method.
Competing interests:
None declared
Competing interests: No competing interests
Re:Extraction and analysis errors and queries: Author Response
Dear Sir,
We would like to clarify some issues in relation to the reply by Hopley
and colleagues to our rapid response. The correspondence relates to their
review on the relative effectiveness of total hip arthroplasty (THA) and
hemiarthroplasty (HA) for displaced intracapsular fracture.1
Firstly, we were not challenging the findings of the paper. We
specified that any differences in outcomes produced by differences in our
data were not clinically or statistically significant. Nor did we raise
any point about the review's approach being weighted to favouring THA,
which Hopley et al seem anxious to point out in their response. This was
raised by Westerman and Hunter in a previous rapid response.2
Secondly, we wish to stress that our letter was written entirely in
"the spirit of peer review and scientific discussion". We emphasised twice
that our rapid response was not a 'point-scoring' exercise, but aimed to
raise broader issues about the conduct and, especially, the reporting of
systematic reviews. Perhaps the title should have been "Extraction and
analysis errors and queries?". The addition of the question mark may have
better represented our views.
The explanations provided by Hopley and colleagues do go some way to
satisfying our queries. However, they do also highlight one of our two
original points: that what was being done in the review was not clearly
described. This is important because of the scope for different
interpretations.
With regard to specific points. Hopley and colleagues respond that
they applied a worse case analysis scenario to some data from the
Mouzopoulous study.3 Nowhere was this stated in the original paper. Surely
the burden is not on the reader to identify issues with the data and make
assumptions about the analyses being performed? There is indeed an issue
with how one treats the possible exclusions in this case. The original
inclusion of those with a prior fracture was an implementation error by
those applying the inclusion/exclusion criteria and thus these can be
justifiably excluded from the analysis.4 We explain this decision in our
own report.5 This affects the demonimator only, not the numerator (number
of events).
Another example of interpretation concerns the inclusion of two
studies only (unnamed in the original paper) providing data for <2
years in the sensitivity analyses. Hopley and colleagues respond that they
categorised studies by "average length of follow-up" rather than using
discrete data from within that 2-year timeframe where they were available.
They therefore included Macaulay et al6 and Baker et al7 in the studies
for >2 years. If this was the approach, then there may be a consistency
issue. Hopley and colleagues do not appear to apply the "average length of
follow-up" criterion when using Macaulay et al's 6-month data mortality
data in their 1-year analysis. Macaulay et al report mortality rates of 1
vs 5 (THA vs HA) at 6 months, and 5 vs 9 at 2 years, but it is not clear
when these additional deaths occurred: some may have taken place in the
period 6-12 months. We hope Hopley and colleagues may therefore appreciate
our point regarding the reporting of what is actually taking place in an
analysis.
On the reported 7 deaths across both groups in Dorr et al8, Hopley
and colleagues state, "On this basis we concluded that the most likely
distribution of death out to 1 year was 3/39 (7.6%) and 4/50 (8.0%)." This
essentially sums-up our principal point about reporting. As reviewers, we
had examined the Dorr study for our own research and those data (THA 4 vs
HA 3) were not reported. Is it so unreasonable to seek an explanation for
the source of the data being analysed in a review when those data do not
appear in the original study? Clarity of reporting would have pre-empted
any such query.
With further reference to the Mouzopolous data, we are not convinced
that patients excluded for having lost data, or having had a revision or
previous fracture, can credibly be treated as deaths, even in a "worse
case analysis", which seems to be the argument proposed by Hopley and
colleagues in their response. The number we cite (6 in each group) is the
number of deaths actually reported in the primary study and reported and
analysed in another review also9, so the burden is surely on the authors
to clarify the data they have used, especially as these are different from
the 'same' data being reported and analysed elsewhere? We remain
unconvinced that these exclusions represent valid mortality data.
Regarding the numbers of "reoperations" reported by Blomfeldt et
al10. We apologise for our original comment as we have clearly
misinterpreted the data in the original review as mortality data. We did
this because the number of events was the same and the source of these
figures was otherwise not apparent. We could not apprehend the difference
between these data and those both in the review by Parker et al9 and
ourselves. This was because the different criteria being applied (any re-
operation to either hip) were not explained. We also accept the point
about the peri-postoperative nature of the fall causing a revision in the
Blomfeldt study, which was not due to the original implant. We have
accordingly revised our own report to specify that the analysis describes
reoperation events relating to the operated hip only. We seek only to
clarify the exact nature of the data being analysed. The burden is often
on reviewers to explain and justify decisions because the data are not
always well-reported in the primary studies. The broader inclusion
criteria applied also reinforces our previous point that combination in a
meta-analysis of the revision data only surely offers greater internal
validity and comparability of data than including additional, potentially
non-comparable data. Again, the approach taken may be defended, but it is
difficult for those assessing the evidence to reach conclusions without
transparency regarding the data being analysed.
Finally, the combining of different study designs in a meta-analysis
is an interesting point worthy of debate as, once again, it holds
relevance for the conduct of systematic reviews more broadly. There may be
biases in the case of the non-randomised studies; the parameters being
estimated in these trials may be different from the parameters being
estimated in the randomised trials. We would be very interested to
understand more fully the rationale being applied by Hopley et al.
In summary, we strongly agree with Hopley and colleagues that data
extraction has a "subjective element", in the sense that it is at times a
matter of interpretation (as we stated previously), and for this reason we
continue to make the point that explanation and clarity is vital. Indeed,
this was the principal point of our initial letter. Data extraction is a
fundamental but under-researched stage of systematic review methodology.
Our conduct of a review comparable to several others only recently
published, including that of Hopley and colleagues, offered an interesting
opportunity to identify and seek explainations for differences, due to
either interpretation or implementation of the method.
Different authors may disagree over approaches used, which is a vital
and stimulating element of research, but this is only possible if the
approach taken is made transparent. We hope Hopley and colleagues accept
these points in the spirit of enquiry and the advancement of methods in
which they are intended.
Dr Christopher Carroll, Senior Lecturer in Health Technology
Assessment
Dr Matt Stevenson, Reader in Health Technology Assessment
Dr Eva Kaltenthaler, Senior Research Fellow
Dr Alison Scope, Research Associate
Reference List
(1) Hopley C, Stengel D, Ekkernkamp A, Wich M. Primary total hip
arthroplasty versus hemiarthroplasty for displaced intracapsular hip
fractures in older patients: systematic review. British Medical Journal
2010; 340.
(2) Westerman R, Hunter J. Misleading conclusions from systematic
analysis? British Medical Journal 2010;
http://www.bmj.com/content/340/bmj.c2332.full/reply#bmj_el_250474(Rapid
response).
(3) Mouzopoulos G, Stamatakos M, Arabatzi H, Vasiliadis G, Batanis
G, Tsembeli A et al. The four-year functional result after a displaced
subcapital hip fracture treated with three different surgical options.
International Orthopaedics 2008; 32(3):367-373.
(4) Fergusson D, Aaron S, Guyatt GH, Hebert P. Post-randomisation
exclusions: the intention to treat principle and excluding patients from
analysis. British Medical Journal 2002; 325:652-654.
(5) Carroll C, Stevenson M, Scope A, Evans P, Buckley S.
Hemiarthroplasty and total-hip arthroplasty for treating primary
intracapsular fracture of the hip: A systematic review and cost
effectiveness analyses. Health Technology Assessment 2011;
15(Forthcoming).
(6) Macaulay W, Nellans KW, Garvin KL, Iorio R, Healy WL,
Rosenwasser MP et al. Prospective randomized clinical trial comparing
hemiarthroplasty to total hip arthroplasty in the treatment of displaced
femoral neck fractures: winner of the Dorr Award. The Journal of
Arthroplasty 2008; 23:2-8.
(7) Baker RP, Squires B, Gargan MF, Bannister GC. Total hip
arthroplasty and hemiarthroplasty in mobile, independent patients with a
displaced intracapsular fracture of the femoral neck. A randomized,
controlled trial. The Journal of Bone and Joint Surgery - American Volume
2006; 88:2583-2589.
(8) Dorr LD, Glousman R, Hoy AL, Vanis R, Chandler R. Treatment of
femoral neck fractures with total hip replacement versus cemented and
noncemented hemiarthroplasty. The Journal of Arthroplasty 1986; 1:21-28.
(9) Parker MJ, Gurusamy KS, Azegami S. Arthroplasties (with and
without bone cement) for proximal femoral fractures in adults. [Update of
Cochrane Database Syst Rev. 2006;3:CD001706; PMID: 16855974]. Cochrane
Database of Systematic Reviews 2010; 6:CD001706.
(10) Blomfeldt R, Tornkvist H, Eriksson K, Soderqvist A, Ponzer S,
Tidermark J. A randomised controlled trial comparing bipolar
hemiarthroplasty with total hip replacement for displaced intracapsular
fractures of the femoral neck in elderly patients. The Journal of Bone and
Joint Surgery - British volume 2007; 89:160-165.
Competing interests: No competing interests