European Spine Journal 2003 (Apr); 12 (2): 149–165 ~ FULL TEXT
Lise Hestbaek, Charlotte Leboeuf-Yde, and Claus Manniche
Odense University Hospital,
5950 Ringe, Denmark.
It is often claimed that up to 90% of low back pain (LBP) episodes resolve spontaneously within 1 month. However, the literature in this area is confusing due to considerable variations regarding the exact definitions of LBP as well as recovery. Therefore, the claim - attractive as it might be to some - may not reflect reality.
In order to investigate the long-term course of incident and prevalent cases of LBP, a systematic and critical literature review was undertaken. A comprehensive search of the topic was carried out utilizing both Medline and EMBASE databases. The Cochrane Library and the Danish Article Base were also screened.
Journal articles following the course of LBP without any known intervention were included, regardless of study type. However, the population had to be representative of the general patient population and a follow-up of at least 12 months was a requirement. Data were extracted independently by two reviewers using a standard check list. The included articles were also independently assessed for quality by the same two reviewers before they were studied in relation to the course of LBP using various definitions of recovery. Thirty-six articles were included.
The results of the review showed that the reported proportion of patients who still experienced pain after 12 months was 62% on average (range 42-75%), the percentage of patients sick-listed 6 months after inclusion into the study was 16% (range 3-40%), the percentage who experienced relapses of pain was 60% (range 44-78%), and the percentage who had relapses of work absence was 33% (range 26-37%).
The mean reported prevalence of LBP in cases with previous episodes was 56% (range 14-93%), which compared with 22% (range 7-39%) for those without a prior history of LBP. The risk of LBP was consistently about twice as high for those with a history of LBP. The results of the review show that, despite the methodological variations and the lack of comparable definitions, the overall picture is that LBP does not resolve itself when ignored. Future research should include subgroup analyses and strive for a consensus regarding the precise definitions of LBP.
From the FULL TEXT Article:
The natural history of a disease relates to its development
in the absence of clinical intervention, whereas the clinical
course is defined as the development subsequent to diagnosis
and the initiation of treatment. Obviously, without
a thorough understanding of the natural history of a disease,
the background for evaluating the clinical course is
lacking, and therapeutic interventions cannot be assessed in a rational manner. In fact, inadequate understanding of
the course of a disease can lead to false conclusions about
the need for, as well as the benefit of, therapeutic interventions.
Presently, the literature in the area is confusing and inconclusive.
The most obvious reason for this confusion is
the lack of distinction between outcome parameters. For
example, one study, which seems to be partly responsible
for the widely accepted belief that 90% of low back pain
(LBP) patients recover within 1 month, in fact showed
that 90% LBP patients stopped consulting their medical
practitioner within 1 month.  Furthermore, Waddell
has been cited for postulating, that 80–90% of LBP attacks
resolve within 6 weeks , but in fact he refers to
return to work – not cessation of pain.  Another study
that has had an impact on the spontaneous recovery belief,
also studied return to work and found that approximately
75% of sick-listed LBP patients returned to work within
1 month.  However, return to work provides an incomplete
picture of the natural course of LBP, because chronic
pain patients may “move” in and out of employment, return
to work at physically less demanding jobs, or reduce
their workload.  In contrast, Croft et al. demonstrated
that 75% of LBP patients from general practice still experienced
pain 1 year later.  Obviously, return to work or
cessation of medical consultations does not necessarily
correlate with the cessation of symptoms. Although the
various outcome measures (pain, disability, sick leave and
medical consultations) are related, they should not be considered
Additionally, the choice of cohort represents a problem
when studying the natural course of a disease. In classical
epidemiologic study designs (such as cross-sectional or
longitudinal surveys) the cohort is made up of prevalent
cases, including subjects at different stages of the disease,
which results in an “apples-and-pears cohort”.
The present confusion may also be partly explained by
a lack of distinction between the short-term and long-term
prognosis. LBP is characterized by variation and change,
rather than absolute recovery.  Thus, concentration on
the short-term development might present the condition as
cured, whereas long-term follow-up may reveal a more recurrent
scenario. Therefore, this review will concentrate
on the long-term course of LBP.
Although this area has been extensively studied, it remains
difficult to gain a clear overview. Therefore, we
conducted a systematic critical review of the epidemiologic
literature to improve our understanding of the natural
course of LBP and, in particular, to investigate
whether there is evidence to support the popular claim of
80–90% spontaneous recovery within 1 month.
Materials and methods
The literature search was modified from the comprehensive search
strategy recommended by the Back Review Group of the Cochrane
The MEDLINE database was searched from the beginning of
the database (1992 via PubMed) to June 1999. The decision to
use the more easily accessible database from 1992 was made
because the study quality was presumed to be better in the
1990s and the up-to-date literature more relevant. The search
combined the terms “low back pain” (MeSH)/ “back pain” (free
text)/ “low back” (free text) with one or more of the following
free text words: “epidemiology”, “natural history”, “natural
course”, “prospective”, “longitudinal”, “follow-up” or “prognos*”
or the MeSH terms: “prognosis” or “survival analysis”.
An additional Medline search specifically for randomised controlled
trials did not reveal additional relevant studies.
A similar search, modified as necessary, was run in EMBASE
(the terms “prognosis” and “survival analysis” were not included
The Cochrane Library was screened for reviews on the topic.
Relevant systematic reviews and their references were screened.
Den Danske Artikelbase (the Danish Article Base) was searched
for “low back pain”/ “back pain”.
Article selection was based on 1948 titles, and abstracts were
screened for suitability by the first author. In addition to epidemiologic
studies, randomized controlled trials were included if they
contained a control group that received only sham treatment or
treatment from a general practitioner. In studies where there was a
statistically significant difference in treatment results for the intervention
as compared to the control group, only data from the control
group were considered. Otherwise all relevant data were included.
Eighty-four articles were found and screened for inclusion
and exclusion criteria.
Criteria for consideration
The inclusion criteria were:
Original journal articles from the Western world
Articles written in English, Danish, Norwegian or Swedish
A sample size of 50 or more (in the case of randomised controlled
trails this applies to the control group) was arbitrarily
Follow-up period of at least 12 months
The exclusion criteria were:
Articles relating to chronic LBP (absence from work for a minimum
of 6 months), because this is usually considered one of the
possible end-points of back pain and because a population of
this type is not representative of the general population
Studies based on a specific population such as a specific occupational
group or pregnant women
Studies of LBP due to acute injury
This selection procedure identified 36 articles, which were included
in this review.
All included articles were reviewed for relevant information using
a standard check list (Appendix 1). This was done independently
by two reviewers (L.H. and C.L-Y.) and disagreements were reolved by consensus. Data on study populations, study design and
outcome measures (pain, sick leave, disability, recurrences and
consultations) were noted and, finally, information was retrieved
in relation to nationality, age and gender.
All the studies were independently assessed for methodological
quality as it relates to natural history by two reviewers (L.H. and
C.L-Y.) using a standard check list. Where disagreement occurred,
the matter was discussed and consensus reached. No existing standard
criteria list was found suitable, since following the course of
an event does not require the same method as a randomised controlled
trial. In contrast to cause-effect research, in which internal
validity is of utmost importance, representativeness and generalization
are more important in descriptive epidemiological studies.  A list of specific criteria was adapted from Von Korff  to
suit the requirements of the subject, including both descriptive (external
validity) and methodological (internal validity) criteria.
Thus, the general quality of the studies was not assessed, but only
quality as it relates to natural history, and the assigned quality score
does not necessarily reflect the quality of the study as a whole. The
criteria for obtaining a maximum score are listed in Appendix 1.
Based on these, a quality score was assigned to each study and the
results are presented in Table 1. The full quality assessment can be
obtained from the authors.
It is possible that the results differ in relation to the definition of
recovery, in such a way that the consequences of LBP (e.g. medical
consultations and absence from work) would result in a seemingly
quicker recovery than actual symptoms. Therefore, the various
outcomes, such as sick leave, recurrence of sick leave, consultations,
disability, pain and recurrence of pain, were studied separately.
Furthermore, as sick leave and consultations may depend on
legislation, which varies between countries, national differences in
relation to sick leave were also analysed. We also attempted to investigate
the course of LBP as it relates to age, gender, and a previous
history of LBP.
Twenty-eight observational studies and eight randomised
controlled trials fulfilled our inclusion criteria. Information
regarding these 36 studies is presented in Table 1.
Studies are listed in alphabetical order according to the
name of the first author.
Quality of data
The overall quality was generally good, but the following
concerns are noteworthy:
In 42% (13/31) of the relevant articles, comparison of
responders and non-responders was missing.
The exact anatomical demarcation of LBP was not defined
in 33% (12/36) of the studies.
In 8% (3/36) of the studies, data had not been collected
in the preferred manner, i.e. sick leave data from administrative
sources and symptom data from interviews
or questionnaires. All other criteria were fulfilled,
and no studies scored below 67%. It was therefore
decided not to exclude any of the studies on the
basis of the quality assessment.
Number and type of studies
The 36 included studies were published between 1981 and
1999 (October). Only four studies were published in the
1980s. [1, 2, 27, 37] Seven studies were randomised controlled
trials [6, 16, 17, 25, 34, 35, 39], five were retrospective
observational studies [2, 19, 20, 21, 46], and the
remaining 24 were prospective observational studies. No
difference in outcome was noted between these three
types of design.
The majority of studies had a population size between 100
and 500, with a range of 62  to 89,190.  The exact
numbers can be seen in Table 1.
Study populations were drawn from several sources:
the army , schools [5, 21, 29, 32], the general population [28, 30] workers receiving compensation [1, 2, 20, 25, 27, 35, 37] and clinical populations. [4, 6–9, 12, 13, 16, 17, 19, 22–24, 31, 33, 34, 36, 39, 42, 44–46]
There were two inception cohorts [28, 44] (first onset
of disease) and the rest were either consecutive (included
as they appear at the study site) or prevalent (all cases with
LBP at a certain point in time) cases. With only two inception
studies, it is not possible to determine whether the
results from such cohorts differ from those of other types.
Description of LBP
The gluteal folds were commonly defined as the lower
border in the definition of LBP [16, 17, 22–24, 28, 31, 36, 45], whereas the upper border varied from the scapula  to the first lumbar vertebra.  In several studies
the only description provided was “back pain” or “low
back pain”. Patients with radiating pain were specifically
excluded in only one study.  In 14 studies [2, 4, 6, 16, 20–23, 27, 34, 35, 37, 39, 46], both patients with
and those without leg pain were included, and in the remaining
21 studies there was no mention of radiating pain
The duration of symptoms at baseline was mentioned
in only one-third of the studies. [6–8, 16, 17, 19, 25, 31, 33, 34, 39, 44, 45] Because of this lack of homogeneity in
relation to LBP definitions, time of inception and followup
periods, it is difficult to compare results and to reach
definitive conclusions. This heterogeneity is illustrated in
Table 2 and Table 3.
In two studies [30, 32], the only outcome measure was
“pain”, another two (authored by the same group and based
on the same sample) [12, 13] measured only “disability”
and in four studies [1, 2, 20, 25] “return to work” was the
only outcome measure. In the remaining 28 studies, different
combinations of “pain”, “disability”, “recurrence”,
“sick leave”, and/or “consultations” were reported. In addition,
the duration of episodes was analysed in three of the
studies. [19, 24, 45]
The definitions of decreased pain varied greatly (from
“completely better” [4, 33] to “no longer disabling LBP” ). Figure 1 illustrates the course of LBP over time as
measured by pain. It was not possible to compare disability
rates over time due to the large variety of ways in
which disability was reported. Likewise, the definitions of
“recurrence” were difficult to compare, since most authors
failed to define what constituted a recurrence. Nevertheless,
in a large number of studies, previous LBP was
found to be an important prognostic factor for the development
of a new episode. [5, 9, 10, 16, 21, 27–29, 30, 32, 33, 36, 37] Figure 2 shows the incidence of recurrence.
Figure 3 shows the risk of having LBP at follow-up
in individuals with LBP at baseline and in those without
LBP at baseline. It should be noted that, in this context,
there is no distinction between recurrences and entirely
In five studies [6, 9, 27, 37, 39], consultations were
recorded. Two of these were based on the same population,
made up of sick-listed industrial workers [27, 32],
and the others from medical practitioners’ practices. Not
surprisingly, the sick-listed workers seemed to consult
more (49% the 1st year and 32% the 2nd year) than the
consecutive office patients (8% , 40%  and 42% 
in the 1st year). The results from these four studies indicated
a high degree of persistence or recurrence.
Figure 4 illustrates the natural course of LBP in relation
to sick leave. This is based on two Norwegian, one
Swedish, one Danish and one Dutch study. [2, 20, 33, 35, 45] The Norwegian studies demonstrated the highest persisting absence. [20, 35] Looking at the levels of recurrence
of sick leave in Fig. 2, Norway also had the highest level of recurrence within 1 year , but otherwise no difference between countries was detected.
Age and gender
Most populations consisted of people of working age, but three populations consisted of children or adolescents. [5, 29, 32] The latter showed a steady increase in the point prevalence of LBP, from about 3% around age 10 to 13%
at age 15. Apart from these, the age-specific prevalence of LBP was reported in only two studies: 26% at age 30  and 19% at age 28. 
There were no major differences in results between
studies involving predominantly male subjects [1, 2, 10, 27, 30, 37] and those with a mixed population (Fig. 2, Fig. 3, Fig. 4). However, the results are very widespread, so differences
could well be hidden within the general variation.
Summary of results
The reviewed studies were not sufficiently homogeneous
to make meta-analyses possible. Ranges of study estimates
are therefore employed to illustrate the extent of
persistent or recurring symptoms of LBP.
Between 42 and 75% of subjects still experience pain
after 12 months (Fig. 1) and between 3 and 40% are still
sick listed 6 months after inclusion in a study (Fig. 4).
Between 44 and 78% of subjects experience relapses of
pain, and for relapses of work absence the estimates
range between 26 and 37% (Fig. 2).
The point prevalence rates of LBP in persons with one
or more previous episodes of LBP range from 14 to
93%, whereas the corresponding rates for those without
a prior history of LBP are from 7 to 39%, with the risk
of LBP being consistently about twice as high for those
with a history of LBP (Fig. 3).
When interpreting the results of this review, the selection
process must be kept in mind. Including only articles written
in the English and Scandinavian languages might introduce
some bias. It has been proposed that positive results
from non-English speaking countries are more likely
to be published in English and negative results in the authors’
native language.  However, in this case, we do
not believe this to be a serious problem, as negative or
positive results are not defined in descriptive studies, and
in the randomised controlled studies only control groups
were studied for the purpose of this review. On the other
hand, there may be national differences in pain perception,
reimbursement policy, etc. This means that the results
from this review may not be transferable to countries
outside the English and Scandinavian language regions.
Disability pension, worker’s compensation and absence
from work all depend on legislation, and care seeking is
also influenced by the system of payment. Therefore, national
differences in legislation and the level of reimbursement/sickness
benefit must be considered before comparing
results between countries. With regard to sick leave,
the figures from Norwegian studies are higher than the
others. This could very well be related to the fact that Norway
has a very generous reimbursement system. However,
due to the different study populations, this is not certain to
be a result of national differences – whether legislative,
cultural or otherwise – but might be attributable to differences
in LBP status at inclusion. There were no Norwegian
studies reporting persistence of symptoms, hence it
was not possible to determine whether such figures would
be correspondingly high compared to other countries.
Additionally, the type of work and the question of
whether the person has no option but to return to the same
function and/or hours obviously has a large influence on the
length of sick leave. This aspect is nevertheless most often
ignored.  It could be argued that return to work is
merely a manifestation of both the extent to which an individual’s
job can be adapted in order to avoid them being
forced to resign, and the extent to which monetary necessity
may force them to stay in an unsuitable job despite the pain.
With these arguments in mind, pain and disability may
be more suitable parameters – at least for the individual,
although not necessarily for society. Although they may
not be sufficiently objective, an individual’s perception of
their own pain and functional ability is of paramount importance
for the way the problem impacts on the quality of
daily life and should not be ignored. These are also measures
recommended by Deyo et al. in an effort to promote
the standardization of outcome: pain, function, well-being,
disability and satisfaction with care.  With regard to
pain and disability, no national differences were detected.
Recurrence rates and risk ratios for developing LBP in
case of previous LBP are measures that clearly illustrate
the recurring pattern of LBP. This pattern questions the
value of short-term “recovery” as a valid outcome measure. Long-term prevention of recurrences may be a more
The choice of cohort also requires careful consideration,
as it may limit generalisability. Obviously, the optimal
method for studying the natural course of LBP would
be to study the general population in a lifelong prospective
study. As this is impossible, prospective study cohorts
most often consist of consecutive cases from clinical settings
who are followed for a limited period of time. When
studying clinical populations, some selection bias cannot
be avoided, as care-seeking in itself and the choice of
provider constitute a selection process. This must be considered
when extrapolating results to the general population.
Among others, Borghouts et al.  consider an inception
cohort (included at onset of first episode) to be of
optimal value when studying the course of a disease.
However, bearing in mind the early onset of LBP , if
adult cases are selected this might bias the selection
against patients with chronic back pain, as some of them
may have had problems since childhood. The results of
the von Korff and Saunders study  did not indicate
that recent onset of symptoms was an important prognostic
variable. Therefore, except for studies including young
populations, cohorts made up of prevalent or consecutive
cases might provide a better picture of the diversity of the
problem, but great care must be taken to record and analyse
the different characteristics of the individuals’ LBP (such
as previous LBP, duration of present period, disability,
anatomical extent of LBP and intensity of pain).
It would be useful to divide LBP patients into subgroups
in relation to symptoms, which might follow different
patterns of recovery. In particular, the presence or
absence of leg pain has been reported to be an important
prognostic factor [27, 30, 36, 37, 45], and the duration of
symptoms at baseline also seems to influence the course
of LBP. [[22, 27, 37, 46] Nevertheless, the magnitude and
duration of symptoms are poorly defined in the majority
of studies. Only four studies [16, 17, 31, 45] describe both
the anatomical demarcation of pain and the duration of
pain at baseline, and they do so in very different ways.
Therefore, it is not possible, presently, to analyse data as
they relate to symptomatic subgroups.
The only sub-categorization of subjects that this material
allows relates to age. Here the high point prevalence of
LBP in children is noteworthy, especially considering the
high risk of recurrence. Mikkelson et al.  and Salminen
et al.  showed that 52 and 93% of teenage subjects
respectively had LBP at follow-up in case of LBP at baseline,
as compared to 12 and 39% of those who did not have
LBP at baseline (Fig.3), and Burton  found that the proportion
of children with LBP who reported their trouble to
be recurrent rose from 44% at age 11 to 59% at age 15.
Recommendations for future studies
In order to further clarify issues relating to the course of
LBP, future studies should:
Provide a clear definition of LBP
Provide subsets of data for various LBP-subgroups
Where relevant, report clearly what constitutes a “recurrence”
If possible, report raw data
Where relevant, discuss limitations of the chosen cohort
and choice of outcome measures
Due to the methodological variations and the lack of clear
definitions in the included articles, no firm conclusions
regarding the natural course of LBP can be reached. However,
despite the large heterogeneity, the overall picture is
clearly that LBP is not a self-limiting condition. There is
no evidence supporting the claim that 80–90% of LBP patients
become pain free within 1 month.
Unfortunately, it was not possible to study the outcome
of various types of LBP from the retrieved material, but it
is strongly recommended that researchers emphasize this.
For this purpose, it is essential first to reach a consensus
regarding the definition of LBP, as this will allow subgroup
analysis. Furthermore, a greater degree of homogeneity
of outcome measures is warranted. If such homogeneity
is not reached, we will continue to gather bits and
pieces of scattered evidence, and overall conclusions will
not be firmly founded.