Clindamycin vs. first-generation cephalosporins for acute osteoarticular infections of childhood—a prospective quasi-randomized controlled trial


Corresponding author: H. Peltola, Children’s Hospital, Helsinki University Central Hospital, PO Box 281 (11 Stenbäck St) 00029 HUS, Helsinki, Finland


Clin Microbiol Infect 2012; 18: 582–589


No sufficiently powered trial has examined two antimicrobials in acute osteoarticular infections of childhood. We conducted a prospective, multicentre, quasi-randomized trial in Finland, comparing clindamycin with first-generation cephalosporins. The age of patients ranged between 3 months and 15 years, and all cases were culture-positive. We assigned antibiotic treatment intravenously for the first 2–4 days, and continued oral treatment with clindamycin 40 mg/kg/24 h or first-generation cephalosporin 150 mg/kg/24 h in four doses. Surgery was kept to a minimum. Subsiding symptoms and signs and normalization of C-reactive protein (CRP) level were preconditions for the discontinuation of antimicrobials. The main outcome was full recovery without further antimicrobials because of an osteoarticular indication during 12 months after therapy. The intention-to-treat analysis comprised 252 children, 169 of whom were analysed per-protocol: 82 cases of osteomyelitis, 80 of septic arthritis, and seven of osteomyelitis–arthritis. Staphylococcus aureus strains (all methicillin-sensitive) caused 84% of the cases. Except for one non-serious sequela during convalescence in both groups, and two late infections caused by dissimilar agents in one child, all patients recovered. The entire courses (medians) of clindamycin and cephalosporin lasted for 23 and 24 days, respectively. CRP normalized in both groups in 9 days. The patients were discharged, on average, on day 10. Loose stools were reported less often (1%) in the clindamycin group than in the cephalosporin group (7%), but two clindamycin recipients developed rash. Clindamycin or a first-generation cephalosporin, administered mostly orally, perform equally well in childhood osteoarticular infections, provided that high doses and administration four times daily are used. As most methicillin-resistant staphylococci remain clindamycin-sensitive, clindamycin remains an option instead of costly alternatives.


No comparative data deriving from a sufficiently powered trial on antimicrobials have previously been available. Clinicians have based the treatment of acute childhood osteoarticular infections—haematogenous osteomyelitis (OM), septic arthritis (SA), and the osteomyelitis–arthritis (OM–SA) combination—on retrospective case series, reviews, and their own experience [1–6]. The two largest series on OM comprised only 80 [7] and 64 [8] subjects. Also for SA, most information has been obtained from retrospective, although sometimes large [4–6], series.

Realizing the need for a large comparative study, we set up such a trial in Finland, knowing that it would take a long time because of the rarity of these diseases in an industrialized country. We aimed at identifying a safe, moderately priced and effective antimicrobial that, after a few days of intravenous treatment, could be administrated orally without serum assays; thus, the entire treatment process would be simplified and costly hospital stays would be shortened. Clindamycin and first-generation cephalosporins fulfilled our requirements [9–11].

Patients and Methods

Study design and diagnosis

We treated children born on odd days with clindamycin, and those born on even days with cephalosporin. A concurrent randomization assigned the patients to a short (OM, 20 days; SA, 10 days) or long (30 days regardless of the diagnosis) course of antimicrobial [12,13]. The study was a multicentre (seven referral hospitals; see Appendix 2), quasi-randomized, open-label, parallel-group equivalence trial, which we carried out in Finland in 1983–2005 [12,13]. The study protocol was approved by each institution’s ethical committee, and the child was included only if consent was provided by a legal guardian. The same form also explained the set-up of the study. We designed, conducted and analysed the trial independently of any medical companies or manufacturers, and registered the trial in an international register (ISRCTN 38832979).

We included previously healthy children aged 3 months to 15 years who presented with an acute osteoarticular infection—fever, and a painful, tender, warm and swollen bone or joint area without trauma. We analysed only culture-positive cases. OM was defined as a case with an agent isolated from the site, or, if identified from blood culture, with documentation of bone involvement by local pus or compatible signs on radiography or magnetic resonance imaging. In SA and OM–SA, bacteria were isolated from the site, or, if they were only isolated from blood, radiological proof of joint effusion was required.

We kept surgery to a minimum, apart from percutaneous aspiration or surgical incision to obtain a representative sample for bacteriology. We did not recommend large corticotomy in OM, or repeated aspirations or arthrotomy in SA, not even in hip [14] or shoulder arthritis, although the attending orthopaedic or paediatric surgeon had the final decision. Blood culture was performed in all cases.


Most osteoarticular infections in industrialized countries are caused by Gram-positive agents, so we used clindamycin 40 mg/kg/24 h in four doses or a first-generation cephalosporin (see below) at 150 mg/kg/24 h in four doses [3,10–13]. In cases of allergy, we assigned the alternative. As SA during the early study years was often caused by Haemophilus influenzae type b [15], children aged 0–4 years initially also received ampicillin or amoxycillin, both at 200 mg/kg/24 h in four doses. Once the aetiology was determined, the course was completed with one agent only. After H. influenzae type b was eliminated by vaccination, we abandoned the use of ampicillin/amoxycillin in 1997 [16].

Our first choice among first-generation cephalosporins was cephradine [9], because it was available for parenteral and oral use. Its later withdrawal from all of Scandinavia forced us to change to intravenous cephalothin, and oral cephalexin or cephadroxil. We did not deem this to be critical to the study, because all first-generation cephalosporins perform rather similarly, and the dosing is the same. Before the new alternatives were approved by all participating centres, we used clindamycin; thereafter, randomization continued until both arms completed the sample size requirement.

All medications were started intravenously four times daily, but after 2–4 days they were switched to oral admistration with the same high doses. We did not assay serum or joint fluid concentrations, or use adjuvant dexamethasone [17]. The attending clinician (paediatrician, or paediatric or orthopaedic surgeon) gave non-steroidal anti-inflammatory drugs at his or her discretion.

The attending clinician discontinued the antimicrobial once the clinical response was found to be good—the patient had defervesced, and most local symptoms and signs were subsiding—and the serum C-reactive protein (CRP) level had fallen to <20 mg/L [18–21]. If the symptoms and signs were still prominent, or CRP remained elevated or re-increased notably, the antimicrobial was continued until two normal values (extra checks) were obtained.


The preset laboratory and radiographical investigations in hospital and after discharge comprised a plain radiograph on admission, and on days 10 and 19, and basic blood analysis at presentation and on days 5 and 10. We measured CRP [18–21] and erythrocyte sedimentation rate (ESR) [3,18–21] throughout the follow-up. A CRP level exceeding 20 mg/L [18,22] or an ESR of 20 mm/h [3,18] was considered to be elevated. Computerized tomography and magnetic resonance imaging were performed on demand. The data were recorded in special forms, and then computerized and analysed in Helsinki. A researchers’ meeting was held once a year.

Osteoarticular infections sometimes re-occur [4,23,24], and long-term sequelae may develop slowly [25]. We scheduled control visits at 2 weeks and 3 and 12 months post-hospitalization, and organized extra visits if needed. The liaison person, a paediatrician or surgeon, performed all examinations, paying special attention to potential sequelae.

Outcome measures and statistical analysis

The primary endpoint in both the intention-to-treat (ITT) and per-protocol (PP) analyses was full recovery; that is, the patient was free of the symptoms or signs of an osteoarticular infection, with no antimicrobials being re-administered for this indication after the treatment course during the 12-month follow-up. Secondary outcomes included a comparison of the time to normalization of the laboratory indices between the clindamycin and cephalosporin recipients, and hospital stay.

Power calculations were performed on the assumption of a success rate of 95%. In all, 59 patients per group were required for a power of 80% for detection of a 10% difference (considered to be clinically meaningful) between groups. The Newcombe method was applied to calculate the 95% CIs of the difference between proportions. p-Values were two-sided and not adjusted for multiple analysis. The level of statistical significance was set at α = 0.05.


The series, localization, and causative agents

Two hundred and sixty-five cases were bacteriologically positive (Fig. 1). Thirteen were excluded because the clinician had commenced a non-protocol antimicrobial. Thus, 252 children were included in the ITT analysis (Table 1). Both antimicrobials were administered intravenously for 3 days (interquartile range (IR) for a median of 3–4 days). In total, the patients received clindamycin for 23 days (IR 20–30 days) and cephalosporin for 24 days (IR 13–30 days). As 83 children were treated entirely with ampicillin/amoxycillin or with a non-recommended agent, the final analysis comprised 169 patients, all of whom had received a full course of clindamycin or a first-generation cephalosporin. Also included were the 14 children (six in the clindamycin group and eight in the cephalosporin group) who, for the first few days, had also been treated with ampicillin. In all, there were 82 children with OM (the three main sites being the femur (24), tibia (21), and calcaneum (12)), 80 with SA (hip in 27, knee in 22, and ankle in 20), and seven with OM–SA (affecting mostly the hip and femur in four).

Figure 1.

 Flow chart. Randomization was performed by contacting a special ward of the Children’s Hospital (Helsinki, Finland). A computer-generated number was given by telephone. Quasi-randomization was performed by birth date. Children born on an odd day were given clindamycin; children with an even birthday received first-generation cephalosporin.

Table 1.   Intention-to treat analysis
N = 252
Randomized to:
N = 141
N = 111
  1. CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; IQR, interquartile range; SE, standard error of the mean.

  2. aStreptococcus group G, Meningococcus group B.

  3. bMeningococcus W 135.

  4. cWith compatible symptoms and signs, and objective evidence of an osteoarticular infection.

Full recovery at the 12-month check-up, no. (%)249 (99)139 (99)110 (99)
Gender, male/female154/9881/6073/38
Age (years), median (IQR)8 (4–11)9 (5–12)7 (3–10)
First symptoms (days), median (IQR)3 (2–5)3 (2–5)3 (1–4)
Type of infection, no. (%)
 Osteomyelitis100 (40)58 (41)42 (38)
 Septic arthritis130 (52)69 (49)61 (55)
 Osteomyelitis + septic arthritis22 (9)14 (10)8 (7)
Causative agent, no. (%)
 Staphylococcus aureus189 (75)108 (43)81 (73)
 Streptococcus pyogenes25 (10)14 (10)11 (10)
 Haemophilus influenzae type b24 (10)14 (10)10 (9)
 Streptococcus pneumoniae11 (4)3 (1)8 (7)
 Other3 (1)2a (1)1b (1)
Site from which agent cultured, no. (%)
 Joint and blood41 (16)23 (16)18 (16)
 Joint only60 (24)36 (26)24 (22)
 Bone and blood33 (13)17 (12)16 (14)
 Bone only40 (16)21 (15)19 (17)
 Blood onlyc78 (31)49 (35)29 (26)
Initial value, mean ± SEM
 CRP (mg/L)87 ± 488 ± 485 ± 6
 ESR (mm/h)51 ± 254 ± 246 ± 3
 Blood leukocyte count (per mm3)12 500 ± 40012 400 ± 50012 700 ± 700
Treatment completed with ampicillin/amoxycillin or antibiotic outside protocol, no. (%)83 (33)42 (30)41 (37)
Antimicrobials given (days), median (IQR)
 Intravenous administration3 (3–4)3 (3–4)3 (3–5)
 Total course24 (17–30)21 (20–30)25 (15–30)
Any complaint at the 12-month check-up (see text), no.321

The two groups were comparable (Tables 1 and 2). There was a slight preponderance of males, most children were of school age, medical advice was sought mostly within 3 days, and SA was slightly more common than OM. OM–SA was found in <10% of cases.

Table 2.   Per-protocol analysis
N = 169
Randomized to:
N = 99
N = 70
  1. CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; IQR, interquartile range; SEM, standard error of the mean.

  2. aStreptococcus group G.

  3. bNeisseria meningitidis group B, Haemophilus influenzae type b.

  4. cWith compatible symptoms and signs, and objective evidence of an osteoarticular infection.

  5. dAmpicillin/amoxycillin, clindamycin or cephalosporin discontinued once agent had been identified. Children with a full course of ampicillin/amoxycillin (H. influenzae type b disease) and those with other medications excluded from analysis.

Gender, male/female102/6753/4649/21
Age (years), median (IQR)9 (6–11)10 (6–12)8 (4–11)
 First symptoms (days), median (IQR)3 (2–5)3 (2–5)3 (2–4)
Type of infection, no. (%)
 Osteomyelitis82 (49)50 (51)32 (46)
 Septic arthritis80 (47)44 (44)36 (51)
 Osteomyelitis + septic arthritis7 (4)5 (5)2 (3)
Causative agent, no. (%)
 Staphylococcus aureus142 (84)88 (89)54 (77)
 Streptococcus pyogenes15 (9)7 (7)8 (11)
 Streptococcus pneumoniae9 (5)3 (3)6 (9)
 Other3 (2)1a (1)2b(3)
Site from which agent cultured, no. (% of total)
 Joint and blood22 (13)13 (13)9 (13)
 Joint only41 (24)23 (23)18 (25)
 Bone and blood17 (10)8 (8)9 (13)
 Bone only26 (16)13 (13)13 (19)
 Blood onlyc63 (37)42 (43)21 (30)
Initial value, mean ± SEM
 CRP (mg/L)78 ± 482 ± 573 ± 7
 ESR (mm/h)48 ± 251 ± 342 ± 3
 Blood leukocyte count (per mm3;)10 900 ± 40011 000 ± 50010 800 ± 700
Treatment initiated with ampicillin/amoxycillind, no. (%)14 (8)6 (6)8 (11)
Antimicrobials (days), median (IQR)
 Intravenous administration3 (3–4)3 (3–3)3 (3–4)
 Total course24 (20–30)23 (20–30)24 (13–30)
Any complaint at the 12-month check-up (see text)211

Staphylococcusaureus, which was invariably susceptible to methicillin and clindamycin, caused 84% of cases, whereas Streptococcus pyogenes caused 9% and Streptococcus pneumoniae caused 5%. Comparable increases in CRP, ESR and leukocyte count (Tables 1 and 2) suggested that the disease severity was similar in the clindamycin and cephalosporin groups.

Recovery and outcome

The main result of the whole trial was that all 99 of the 99 (95% CI 96.3–100) clindamycin recipients and—if the child with late re-infections is included—69 of the 70 (98.6%; 95% CI 92.3–99.8) cephalosporin recipients recovered with the primary treatment. There was one non-serious sequela in each group.

Table 3 summarizes the results for the 252 patients included in the ITT analysis (2 × 2 factorial design). Regarding the clinical response, no major difference was observed between clindamycin and cephalosporin. In PP analysis, the CRP peak was reached 1 day sooner in the clindamycin group (Fig. 2), but the difference remained insignificant (p 0.13 on day 3). The groups did not differ significantly in the frequency of surgical procedures (Table 3). Furthermore, CRP and ESR normalized almost identically, as levels of 20 mg/L and 20 mm/h were reached on day 9 and day 29, respectively (p 0.47 and p 0.45; Fig. 2). Blood leukocytes normalized in 5 days. The laboratory indices remained unchanged after the discontinuation of the antimicrobial (Fig. 2).

Table 3.   Results of the intention-to-treat analysis (N = 252)
 ClindamycinCephalosporin (first generation)Either agent
  1. aData given as no. (%, 95% CI).

  2. bSurgical procedures apart from diagnostic sampling (lavation, arthrotomy, or trepanation of bone).

  3. cExcluded from final analysis (ampicillin/amoxycillin or antibiotic outside protocol).

  4. dIncluded in the final (per-protocol) analysis.

Short treatment   
 Re-infection0 (0, 0–0)a0 (0, 0–0)0 (0, 0–0)
 Permanent sequelae0 (0, 0–0)1 (0.4, 0–2)1 (0.4, 0–2)
 Invasive surgeryb23 (33, 23–45)13 (23, 14–36)36 (29, 22–37)
 Excludedc21 (8, 6–12)20 (8, 5–12)41 (16, 12–21)
 Per-protocold48 (19, 15–24)36 (14, 10–19)84 (33, 28–39)
Long treatment   
 Re-infection0 (0, 0–0)1 (0.4, 0–2)1 (0.4, 0–2)
 Permanent sequelae2 (1, 0–3)0 (0, 0–0)2 (1, 0–3)
 Invasive surgery18 (25, 16–36)9 (16, 9–28)27 (21, 15–29)
 Excluded21 (8, 6–12)21 (8, 6–12)42 (17, 13–22)
 Per-protocol51 (20, 16–26)34 (13, 10–18)85 (34, 28–40)
Short or long treatment   
 Re-infection0 (0, 0–0)1 (0.4, 0–2)
 Permanent sequelae2 (1, 0–3)1 (0.4, 0–2)
 Invasive surgery41 (29, 22–37)29 (26, 19–35)
 Excluded42 (17, 13–22)41 (16, 12–21)
 Per-protocol99 (39, 33–45)70 (28, 23–34)
Figure 2.

 Normalization of the serum C-reactive protein (CRP) level (± standard error of the mean) and erythrocyte sedimentation rate (ESR) and of white blood cell (WBC) count occurred almost identically in the clindamycin and cephalosporin groups. The small difference in the CRP levels on day 3 (105 mg/L in the cephalosporin group and 87 mg/L in the clindamycin group) remained non-significant (p 0.13 by t-test).

The symptoms and signs subsided gradually, and the mean hospital stay was similar: in ITT analysis, 11 days (±0.5 days) in the clindamycin group and 12 days (±0.6 days) in the cephalosporin group (p 0.10). In PP analysis, hospital stay was the same: 10 days (±0.5 days) for both groups (p 0.75).

Two weeks after discharge, 38% (95% CI 29–48) of clindamycin recipients and 34% (95% CI 24–46) of cephalosporin recipients had some complaints, but there was no difference between the groups. At 3 months, three patients showed some minor symptoms, all of which disappeared in the next 12 months. At the 1-year check-up, one child in each group had developed symptoms: an 11-year-old boy with perinatal Erb’s palsy and elbow arthritis showed a mild extension deficit (probably related to previous palsy), and a 14-year-old boy with hip arthritis had limb shortening of 1 cm. The palsy remained, but the extension deficit and the shortening healed within 12 months.

The only child with sequelae in the ITT analysis was a 12-year-old boy with ankle OM–SA who was treated with intravenous clindamycin for 3 days and oral penicillin for 77 days; this major prolongation of medication was a result of slow clinical recovery. The patient was excluded from the PP analysis because of the use of penicillin. At the 1-year check-up, he complained of pain during exercise, and radiography identified joint destruction in the tibiotalar joint.

A 10-year-old boy’s staphylococcal tibiotalar arthritis had been treated with cephradine for 28 days [12]. Seventeen months later, the same joint was affected by S. aureus. Cephradine was re-instituted, but a suboptimal response led to a change to clindamycin. Again, recovery seemed uneventful, until coagulase-negative staphylococci re-infected the same ankle 8 months later. A 30-day course of clindamycin was successful, and since 1990 the child has remained well. No surgery besides aspiration was carried out. A scintigram was normal, and no immunodeficiency was found. Also, no cephradine or clindamycin resistance was encountered.

Drug tolerability

High doses were tolerated well with both agents. Unexpectedly, loose stools were reported slightly less often in the clindamycin group than in the cephalosporin group (1% (95% CI 0–4) vs. 7% (95% CI 4–14), respectively). Two clindamycin recipients developed rash.


Our prospective quasi-randomized study on 169 PP-treated bacteriologically confirmed cases of acute osteoarticular infections of childhood is the largest to date, and addresses the call for a randomized equivalence trial [26]. It also answers several questions that have been raised over the years [1,2,5,6].

All patients recovered with an approximately 3-week, mostly oral, course of clindamycin or a first-generation cephalosporin. This finding is especially relevant to osteoarticular infections, in which up to one-third of prolonged courses are interrupted because of adverse events [27].

Clinically, both antimicrobials performed similarly. They also proved to be safe. Notably, diarrhoea was not a problem, even in the clindamycin group. Good activity against the relevant bacteria, and sufficient absorption and tissue penetration [9–11], allow oral administration, provided that exceptionally large doses are used [11]—we believe that this and the administration four times daily of these time-dependent antimicrobials are pivotal. No serum assays are needed, but good adherence to treatment is mandatory. Palatability, which is an important issue in paediatrics, is especially good with cephalosporins. The oral route simplifies treatment and benefits the patients and personnel. Most oral agents are also considerably cheaper than their parenteral forms. The initial intravenous phase can be very short [26] (if needed at all), and successful treatment is soon reflected by declining CRP (and ESR).

As most methicillin-resistant S. aureus strains have remained susceptible to clindamycin [28], its good efficacy in methicillin-susceptible OM–SA is welcome news in these days of active promotion of costly alternatives, which also cause more adverse events without providing better effectiveness [7,29,30]. In this trial, two clindamycin recipients developed rash, but even if a causal association is assumed, the 2% incidence rate is still tolerable in light of all the advantages of this moderately priced old remedy.

Our study also had limitations. Unexpected withdrawal of cephradine led to a temporary overuse of clindamycin. Enrolment of patients lasted for a long time, because the incidence of acute OM at age 0–14 years in Finland is merely 4.5 per 100 000 per year [13], and SA is even rarer [12]. We had to choose between diagnostic accuracy—requiring culture positivity in all cases—or inclusion of less well-documented cases. We chose the first alternative, but had to pay for our decision with a long enrolment time. However, the researchers’ yearly meetings kept standards the same over the years.

We did not detect cases caused by Kingella kingae, which is compatible with the epidemiology in Finland. If K. kingae is common, clindamycin monotherapy cannot be used empirically.

Some patients were treated with antimicrobials that are not recommended in the protocol, but, importantly, this never occurred because initial clindamycin or cephalosporin had failed. We could not always see indication for the deviations, but agree with the ethical view that the clinician responsible for the patient has the final decision. We used birthdays for quasi-randomization instead of computer-generated randomization. We do not believe that this has introduced bias, given the similarly of the treatment groups at baseline.

One patient developed two late infections but, importantly, he was not a treatment failure. The first re-infection occurred after 17 months—95% of relapses occur within 12 months [23]—and the causative agents differed. Once infected, an osteoarticular site remains prone to further infections for some time [24]; an analogy with endocarditis is evident. No evidence exists that any prolonged medication would prevent all of these late re-infections [2].

In summary, clindamycin and first-generation cephalosporins were shown to be equally effective and safe agents for treatment of osteoarticular infections of childhood, local resistance permitting. Even if methicillin-resistant S. aureus is prevalent, most strains are still clindamycin-susceptible [28], and clindamycin remains an alternative to newer, but very expensive, agents [7,29,30]. However, large doses (≥40 mg/kg daily in four doses, intravenously, then orally) and good compliance are needed. Normalization of CRP usually allows discontinuation of antimicrobials. We believe that the treatment of acute osteoarticular infections of childhood can and should be simplified, but always taking into account the specific needs of an individual patient.


We are indebted to all members of the OM-SA Study Group (listed below), who, over the years, conducted this laborious trial in their institutions. S. Sarna contributed to the statistical analysis, and J. Ajanki drew the picture.

Transparency Declaration

No author had a conflict of interest.


Appendix 1

The OM-SA Study Group

Besides the main authors, the OM-SA Study Group comprised the following persons: K. Aalto, Aurora Hospital, and Children’s Hospital, Helsinki University Central Hospital, and University of Helsinki; I. Anttolainen, Päijät-Häme Central Hospital, Lahti, Finland; M. Heikkinen, Kuopio University Hospital, Kuopio, Finland; A. Hiippala, Etelä-Saimaa Central Hospital, Lappeenranta, Finland; U. Kaski (deceased), Seinäjoki Central Hospital, Seinäjoki, Finland; N. Kojo, Etelä-Saimaa Central Hospital, Lappeenranta, Finland; P. Lautala, Päijät-Häme Central Hospital, Lahti, Finland; J. Merikanto, Children’s Hospital, Helsinki University Central Hospital, and University of Helsinki; P. Ojajärvi, Jorvi Hospital, Espoo, Finland; and E. Salo, Aurora Hospital, and Children’s Hospital, Helsinki University Central Hospital, and University of Helsinki.

Appendix 2

Aurora Hospital, Helsinki, Finland; Children’s Hospital, Helsinki University Central Hospital, Helsinki, Finland; Etelä-Saimaa Central Hospital, Lappeenranta, Finland; Jorvi Hospital, Espoo, Finland; Kuopio University Hospital, Kuopio, Finland; Päijät-Häme Central Hospital, Lahti, Finland; and Seinäjoki Central Hospital, Seinäjoki, Finland.