The systematic review followed many of the recommendations as
outlined in the Preferred Reporting Items for Systematic Reviews
and Meta-Analyses (PRISMA) statement.6
The PRISMA statement is primarily
aimed at helping authors improve the reporting of reviews of randomised control
trials. However, the PRISMA statement can also be used as a basis
for reporting systematic reviews of other types of research, as
in the present review.
A protocol specifying all aspects of the review method was developed
before initiation of this review. This protocol included criteria
for considering studies for this review, search methods for identification
of studies, data collection and analysis.
The protocol was planned to minimise the effect of author bias
on the review and, in particular, the potential to alter the method
or data analysis based on study findings. The protocol was peer-reviewed
by the Periodontal team at King’s College London (London, United
Kingdom) before commencing the study.
Five electronic databases were searched: MEDLINE, Web of Science,
The Cochrane Library, SCOPUS and EMBASE. All databases were searched
from their earliest records until August 2012. The full search strategy
developed for MEDLINE is shown in Table I. This search strategy
was customised according to the database being used. The searches
were restricted to English language publications.
Search strategy developed for MEDLINE
The bibliographies of all relevant papers and review articles
were manually searched. Unpublished data was not sought.
The search strategy was designed to examine the effect of smoking
on bone healing generally. Spinal studies were excluded from this
review, due to the complicated nature of healing at this site across
intervertebral spaces. This review focused instead only on the non-spinal
Study inclusion and exclusion criteria
Prospective and retrospective clinical studies assessing bone
healing in smokers and non-smokers were included. The studies involved
clinical interventions, ranging from conservative treatment such
as cast immobilisation to surgical interventions, for example internal
fracture fixation. In addition, arms of clinical trials comparing
different interventions that reported results separately for smokers and
non-smokers were included. The following inclusion criteria were
applied: 1) publications written in the English language; 2) human
studies; and 3) studies categorising subjects into at least two
groups (non-smokers and smokers).
Studies were excluded if they contained inadequate data to allow
a clear comparison of bone healing in smokers and non-smokers following
Included studies needed to report one or more of the following
primary outcomes, which were: 1) clinical indicators of bone healing
(including measures of time to clinical union and diagnoses of non-, delayed
or malunion); and 2)radiological indicatorsof
bone healing (including measures of time to radiologically defined
Secondary outcomes were: 1) complications of bone healing (including
diagnoses of infection or
osteomyelitis); and 2) patient based-outcomes (when not used as part
of a clinical indicator for bone healing), which included pain and
Study selection, data collection
In the first phase of study selection, the titles and abstracts
of all identified publications were independently screened by two
reviewers (RAP and RMP). Disagreement between reviewers was resolved
The full texts of all studies of possible relevance were obtained
for independent assessment against the inclusion criteria stated.
Studies rejected at this stage were recorded in a rejection table
with reasons for rejection. Again disagreement between reviewers
was resolved by discussion.
Inter-reviewer agreement was calculated with Cohen’s κ score
for each screening stage. Where possible, authors of studies were
contacted to resolve doubts about study design, patient populations
and to request missing information.
A data extraction form was used to collect information from the
included studies. The data collected included study characteristics,
outcome measures, treatment characteristics, results, quality assessment
data and other general information.
The quality of the included studies was assessed according to:
1) similarity of the smoking and non-smoking groups at baseline;
2) masking of smoking status to the clinician(s) assessing the outcome(s);
3) reproducibility of the outcome measure(s); and 4) completeness
of follow-up and explanations for dropouts. Studies with a retrospective
design were not assessed with reference to completeness of follow-up.
When all criteria were met the risk of bias was estimated as
low. A moderate risk of bias was assigned when one or more criteria
were partly fulfilled and a high risk assigned when one or more
criteria were not met. The quality assessment was not used to exclude
any studies qualifying for the review on the basis of their inclusion
A summary table was constructed using information collected on
the data extraction forms. The pooled data was analysed in a descriptive
format. Studies were analysed for similarities and suitability for meta-analysis.